Bug 693057 - Upgrade libvpx decoder to v0.9.7-p1, r=cpearce,khuey

--HG--
rename : media/libvpx/vp8/common/filter_c.c => media/libvpx/vp8/common/filter.c
rename : media/libvpx/vpx_scale/generic/scalesystemdependant.c => media/libvpx/vpx_scale/generic/scalesystemdependent.c

configure.in

@@ -5356,16 +5356,41 @@ if test -n "$MOZ_WEBM"; then
         fi
         CFLAGS="-I${LIBVPX_DIR}/include $CFLAGS"
         LDFLAGS="-L${LIBVPX_DIR}/lib $LDFLAGS"
+        MOZ_NATIVE_LIBVPX_DEC_TEST=
         MOZ_CHECK_HEADER(vpx/vpx_decoder.h,
             [if test ! -f "${LIBVPX_DIR}/include/vpx/vpx_decoder.h"; then
              AC_MSG_ERROR([vpx/vpx_decoder.h found, but is not in ${LIBVPX_DIR}/include])
             fi],
             AC_MSG_ERROR([--with-system-libvpx requested but vpx/vpx_decoder.h not found]))
         AC_CHECK_LIB(vpx, vpx_codec_dec_init_ver,
-                     [MOZ_NATIVE_LIBVPX=1
-                      MOZ_LIBVPX_INCLUDES="-I${LIBVPX_DIR}/include"
-                      MOZ_LIBVPX_LIBS="-L${LIBVPX_DIR}/lib -lvpx"],
+                     [MOZ_NATIVE_LIBVPX_DEC_TEST=1],
                      ([--with-system-libvpx requested but symbol vpx_codec_dec_init_ver not found]))
+        if test -n "$MOZ_NATIVE_LIBVPX_DEC_TEST" ; then
+            AC_MSG_CHECKING([for libvpx version >= v0.9.7])
+            dnl We need at least v0.9.7 to fix several crash bugs (for which we
+            dnl had local patches prior to v0.9.7).
+            dnl
+            dnl This is a terrible test for the library version, but we don't
+            dnl have a good one. There is no version number in a public header,
+            dnl and testing the headers still doesn't guarantee we link against
+            dnl the right version. While we could call vpx_codec_version() at
+            dnl run-time, that would break cross-compiling. There are no
+            dnl additional exported symbols between the v0.9.7 release and the
+            dnl v0.9.6 one to check for.
+            AC_TRY_COMPILE([
+                #include <vpx/vpx_decoder.h>
+                #if !defined(VPX_CODEC_USE_INPUT_PARTITION)
+                    #error "test failed."
+                #endif
+                ],
+                [return 0;],
+                [AC_MSG_RESULT([yes])
+                 MOZ_NATIVE_LIBVPX=1
+                 MOZ_LIBVPX_INCLUDES="-I${LIBVPX_DIR}/include"
+                 MOZ_LIBVPX_LIBS="-L${LIBVPX_DIR}/lib -lvpx"],
+                [AC_MSG_RESULT([no])
+                 AC_MSG_ERROR([--with-system-libvpx requested but it is not v0.9.7 or later])])
+        fi
         CFLAGS=$_SAVE_CFLAGS
         LDFLAGS=$_SAVE_LDFLAGS
         LIBS=$_SAVE_LIBS

media/libvpx/Makefile.in

@@ -130,12 +130,13 @@ CSRCS += \
   alloccommon.c \
   blockd.c \
   debugmodes.c \
+  defaultcoefcounts.c \
   dsystemdependent.c \
   entropy.c \
   entropymode.c \
   entropymv.c \
   extend.c \
-  filter_c.c \
+  filter.c \
   findnearmv.c \
   idctllm.c \
   invtrans.c \
@@ -145,7 +146,6 @@ CSRCS += \
   modecont.c \
   modecontext.c \
   postproc.c \
-  predictdc.c \
   quant_common.c \
   recon.c \
   reconinter.c \
@@ -173,7 +173,7 @@ CSRCS += \
   vpx_mem.c \
   gen_scalers.c \
   vpxscale.c \
-  scalesystemdependant.c \
+  scalesystemdependent.c \
   yv12config.c \
   yv12extend.c \
   $(NULL)
@@ -186,6 +186,7 @@ CSRCS += \
   idct_blk_mmx.c \
   idct_blk_sse2.c \
   loopfilter_x86.c \
+  recon_wrapper_sse2.c \
   vp8_asm_stubs.c \
   x86_systemdependent.c \
   x86_dsystemdependent.c \
@@ -284,11 +285,6 @@ VPX_ASFILES = \
   dequantizeb_neon.asm \
   $(NULL)
 
-# The ARM asm needs to extract the offsets of various C struct members.
-# We need a program that runs on the host to pull them out of a .o file.
-HOST_CSRCS = obj_int_extract.c
-HOST_PROGRAM = host_obj_int_extract$(HOST_BIN_SUFFIX)
-
 ifdef VPX_AS_CONVERSION
 # The ARM asm is written in ARM RVCT syntax, but we actually build it with
 # gas using GNU syntax. Add some rules to perform the conversion.
@@ -303,8 +299,6 @@ else
 ASFILES += $(VPX_ASFILES)
 endif
 
-GARBAGE += vpx_asm_offsets.$(OBJ_SUFFIX) vpx_asm_offsets.asm
-
 endif
 
 include $(topsrcdir)/config/rules.mk
@@ -318,8 +312,3 @@ filter_c.o: filter_c.c Makefile.in
 	$(CC) -o $@ -c $(patsubst -xO[45],-xO3,$(COMPILE_CFLAGS)) $<
 endif
 endif
-
-ifdef VPX_ARM_ASM
-vpx_asm_offsets.asm: vpx_asm_offsets.$(OBJ_SUFFIX) $(HOST_PROGRAM)
-	./$(HOST_PROGRAM) rvds $< $(if $(VPX_AS_CONVERSION),| $(VPX_AS_CONVERSION)) > $@
-endif

media/libvpx/build/make/ads2gas.pl

@@ -21,8 +21,14 @@ print "@ This file was created from a .asm file\n";
 print "@  using the ads2gas.pl script.\n";
 print "\t.equ DO1STROUNDING, 0\n";
 
+# Stack of procedure names.
+@proc_stack = ();
+
 while (<STDIN>)
 {
+    # Load and store alignment
+    s/@/,:/g;
+
     # Comment character
     s/;/@/g;
 
@@ -117,8 +123,8 @@ while (<STDIN>)
     # put the colon at the end of the line in the macro
     s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
 
-    # Strip ALIGN
-    s/\sALIGN/@ ALIGN/g;
+    # ALIGN directive
+    s/ALIGN/.balign/g;
 
     # Strip ARM
     s/\sARM/@ ARM/g;
@@ -130,9 +136,23 @@ while (<STDIN>)
     # Strip PRESERVE8
     s/\sPRESERVE8/@ PRESERVE8/g;
 
-    # Strip PROC and ENDPROC
-    s/\sPROC/@/g;
-    s/\sENDP/@/g;
+    # Use PROC and ENDP to give the symbols a .size directive.
+    # This makes them show up properly in debugging tools like gdb and valgrind.
+    if (/\bPROC\b/)
+    {
+        my $proc;
+        /^_([\.0-9A-Z_a-z]\w+)\b/;
+        $proc = $1;
+        push(@proc_stack, $proc) if ($proc);
+        s/\bPROC\b/@ $&/;
+    }
+    if (/\bENDP\b/)
+    {
+        my $proc;
+        s/\bENDP\b/@ $&/;
+        $proc = pop(@proc_stack);
+        $_ = "\t.size $proc, .-$proc".$_ if ($proc);
+    }
 
     # EQU directive
     s/(.*)EQU(.*)/.equ $1, $2/;
@@ -151,3 +171,6 @@ while (<STDIN>)
     next if /^\s*END\s*$/;
     print;
 }
+
+# Mark that this object doesn't need an executable stack.
+printf ("\t.section\t.note.GNU-stack,\"\",\%\%progbits\n");

media/libvpx/compile_errors.patch

@@ -0,0 +1,39 @@
+diff --git a/media/libvpx/vpx/vp8.h b/media/libvpx/vpx/vp8.h
+--- a/media/libvpx/vpx/vp8.h
++++ b/media/libvpx/vpx/vp8.h
+@@ -41,33 +41,33 @@ enum vp8_com_control_id
+     VP8_SET_REFERENCE           = 1,    /**< pass in an external frame into decoder to be used as reference frame */
+     VP8_COPY_REFERENCE          = 2,    /**< get a copy of reference frame from the decoder */
+     VP8_SET_POSTPROC            = 3,    /**< set the decoder's post processing settings  */
+     VP8_SET_DBG_COLOR_REF_FRAME = 4,    /**< set the reference frames to color for each macroblock */
+     VP8_SET_DBG_COLOR_MB_MODES  = 5,    /**< set which macro block modes to color */
+     VP8_SET_DBG_COLOR_B_MODES   = 6,    /**< set which blocks modes to color */
+     VP8_SET_DBG_DISPLAY_MV      = 7,    /**< set which motion vector modes to draw */
+     VP8_COMMON_CTRL_ID_MAX,
+-    VP8_DECODER_CTRL_ID_START   = 256,
++    VP8_DECODER_CTRL_ID_START   = 256
+ };
+ 
+ /*!\brief post process flags
+  *
+  * The set of macros define VP8 decoder post processing flags
+  */
+ enum vp8_postproc_level
+ {
+     VP8_NOFILTERING             = 0,
+     VP8_DEBLOCK                 = 1<<0,
+     VP8_DEMACROBLOCK            = 1<<1,
+     VP8_ADDNOISE                = 1<<2,
+     VP8_DEBUG_TXT_FRAME_INFO    = 1<<3, /**< print frame information */
+     VP8_DEBUG_TXT_MBLK_MODES    = 1<<4, /**< print macro block modes over each macro block */
+     VP8_DEBUG_TXT_DC_DIFF       = 1<<5, /**< print dc diff for each macro block */
+-    VP8_DEBUG_TXT_RATE_INFO     = 1<<6, /**< print video rate info (encoder only) */
++    VP8_DEBUG_TXT_RATE_INFO     = 1<<6  /**< print video rate info (encoder only) */
+ };
+ 
+ /*!\brief post process flags
+  *
+  * This define a structure that describe the post processing settings. For
+  * the best objective measure (using the PSNR metric) set post_proc_flag
+  * to VP8_DEBLOCK and deblocking_level to 1.
+  */

media/libvpx/solaris.patch

@@ -16,69 +16,6 @@ diff --git a/media/libvpx/vp8/common/loopfilter_filters.c b/media/libvpx/vp8/com
  static __inline signed char vp8_signed_char_clamp(int t)
  {
      t = (t < -128 ? -128 : t);
-diff --git a/media/libvpx/vpx/internal/vpx_codec_internal.h b/media/libvpx/vpx/internal/vpx_codec_internal.h
---- a/media/libvpx/vpx/internal/vpx_codec_internal.h
-+++ b/media/libvpx/vpx/internal/vpx_codec_internal.h
-@@ -316,17 +316,17 @@ struct vpx_codec_iface
- 
- /*!\brief Callback function pointer / user data pair storage */
- typedef struct vpx_codec_priv_cb_pair
- {
-     union
-     {
-         vpx_codec_put_frame_cb_fn_t    put_frame;
-         vpx_codec_put_slice_cb_fn_t    put_slice;
--    };
-+    } fn;
-     void                            *user_priv;
- } vpx_codec_priv_cb_pair_t;
- 
- 
- /*!\brief Instance private storage
-  *
-  * This structure is allocated by the algorithm's init function. It can be
-  * extended in one of two ways. First, a second, algorithm specific structure
-diff --git a/media/libvpx/vpx/src/vpx_decoder.c b/media/libvpx/vpx/src/vpx_decoder.c
---- a/media/libvpx/vpx/src/vpx_decoder.c
-+++ b/media/libvpx/vpx/src/vpx_decoder.c
-@@ -165,17 +165,17 @@ vpx_codec_err_t vpx_codec_register_put_f
- 
-     if (!ctx || !cb)
-         res = VPX_CODEC_INVALID_PARAM;
-     else if (!ctx->iface || !ctx->priv
-              || !(ctx->iface->caps & VPX_CODEC_CAP_PUT_FRAME))
-         res = VPX_CODEC_ERROR;
-     else
-     {
--        ctx->priv->dec.put_frame_cb.put_frame = cb;
-+        ctx->priv->dec.put_frame_cb.fn.put_frame = cb;
-         ctx->priv->dec.put_frame_cb.user_priv = user_priv;
-         res = VPX_CODEC_OK;
-     }
- 
-     return SAVE_STATUS(ctx, res);
- }
- 
- 
-@@ -187,17 +187,17 @@ vpx_codec_err_t vpx_codec_register_put_s
- 
-     if (!ctx || !cb)
-         res = VPX_CODEC_INVALID_PARAM;
-     else if (!ctx->iface || !ctx->priv
-              || !(ctx->iface->caps & VPX_CODEC_CAP_PUT_FRAME))
-         res = VPX_CODEC_ERROR;
-     else
-     {
--        ctx->priv->dec.put_slice_cb.put_slice = cb;
-+        ctx->priv->dec.put_slice_cb.fn.put_slice = cb;
-         ctx->priv->dec.put_slice_cb.user_priv = user_priv;
-         res = VPX_CODEC_OK;
-     }
- 
-     return SAVE_STATUS(ctx, res);
- }
- 
- 
 diff --git a/media/libvpx/vpx_ports/mem.h b/media/libvpx/vpx_ports/mem.h
 --- a/media/libvpx/vpx_ports/mem.h
 +++ b/media/libvpx/vpx_ports/mem.h
@@ -172,8 +109,8 @@ diff --git a/media/libvpx/vpx_ports/x86.h b/media/libvpx/vpx_ports/x86.h
 +    asm volatile ("pause \n\t")
  #else
  #if ARCH_X86_64
- /* No pause intrinsic for windows x64 */
- #define x86_pause_hint()
+ #define x86_pause_hint()\
+     _mm_pause();
  #else
  #define x86_pause_hint()\
      __asm pause

media/libvpx/update.sh

@@ -51,16 +51,17 @@ fi
 
 # These are relative to SDK source dir.
 commonFiles=(
+  vp8/vp8_dx_iface.c
   vp8/common/alloccommon.c
   vp8/common/blockd.c
   vp8/common/debugmodes.c
+  vp8/common/defaultcoefcounts.c
   vp8/common/entropy.c
   vp8/common/entropymode.c
   vp8/common/entropymv.c
   vp8/common/extend.c
-  vp8/common/filter_c.c
+  vp8/common/filter.c
   vp8/common/findnearmv.c
-  vp8/common/generic/systemdependent.c
   vp8/common/idctllm.c
   vp8/common/invtrans.c
   vp8/common/loopfilter.c
@@ -69,7 +70,6 @@ commonFiles=(
   vp8/common/modecont.c
   vp8/common/modecontext.c
   vp8/common/postproc.c
-  vp8/common/predictdc.c
   vp8/common/quant_common.c
   vp8/common/recon.c
   vp8/common/reconinter.c
@@ -84,9 +84,10 @@ commonFiles=(
   vp8/common/arm/filter_arm.c
   vp8/common/arm/loopfilter_arm.c
   vp8/common/arm/reconintra_arm.c
-  vp8/common/arm/vpx_asm_offsets.c
   vp8/common/arm/neon/recon_neon.c
+  vp8/common/generic/systemdependent.c
   vp8/common/x86/loopfilter_x86.c
+  vp8/common/x86/recon_wrapper_sse2.c
   vp8/common/x86/vp8_asm_stubs.c
   vp8/common/x86/x86_systemdependent.c
   vp8/decoder/dboolhuff.c
@@ -94,19 +95,18 @@ commonFiles=(
   vp8/decoder/decodframe.c
   vp8/decoder/dequantize.c
   vp8/decoder/detokenize.c
-  vp8/decoder/reconintra_mt.c
-  vp8/decoder/generic/dsystemdependent.c
   vp8/decoder/idct_blk.c
   vp8/decoder/onyxd_if.c
+  vp8/decoder/reconintra_mt.c
   vp8/decoder/threading.c
   vp8/decoder/arm/arm_dsystemdependent.c
   vp8/decoder/arm/dequantize_arm.c
   vp8/decoder/arm/armv6/idct_blk_v6.c
   vp8/decoder/arm/neon/idct_blk_neon.c
+  vp8/decoder/generic/dsystemdependent.c
   vp8/decoder/x86/idct_blk_mmx.c
   vp8/decoder/x86/idct_blk_sse2.c
   vp8/decoder/x86/x86_dsystemdependent.c
-  vp8/vp8_dx_iface.c
   vpx/src/vpx_codec.c
   vpx/src/vpx_decoder.c
   vpx/src/vpx_decoder_compat.c
@@ -114,7 +114,7 @@ commonFiles=(
   vpx/src/vpx_image.c
   vpx_mem/vpx_mem.c
   vpx_scale/generic/gen_scalers.c
-  vpx_scale/generic/scalesystemdependant.c
+  vpx_scale/generic/scalesystemdependent.c
   vpx_scale/generic/vpxscale.c
   vpx_scale/generic/yv12config.c
   vpx_scale/generic/yv12extend.c
@@ -128,6 +128,7 @@ commonFiles=(
   vp8/common/entropymode.h
   vp8/common/entropymv.h
   vp8/common/extend.h
+  vp8/common/filter.h
   vp8/common/findnearmv.h
   vp8/common/g_common.h
   vp8/common/header.h
@@ -142,8 +143,6 @@ commonFiles=(
   vp8/common/postproc.h
   vp8/common/ppflags.h
   vp8/common/pragmas.h
-  vp8/common/predictdc.h
-  vp8/common/preproc.h
   vp8/common/quant_common.h
   vp8/common/recon.h
   vp8/common/reconinter.h
@@ -156,7 +155,7 @@ commonFiles=(
   vp8/common/threading.h
   vp8/common/treecoder.h
   vp8/common/type_aliases.h
-  vp8/common/vpxerrors.h
+  vp8/common/arm/bilinearfilter_arm.h
   vp8/common/arm/idct_arm.h
   vp8/common/arm/loopfilter_arm.h
   vp8/common/arm/recon_arm.h
@@ -174,9 +173,7 @@ commonFiles=(
   vp8/decoder/onyxd_int.h
   vp8/decoder/reconintra_mt.h
   vp8/decoder/treereader.h
-  vp8/decoder/arm/dboolhuff_arm.h
   vp8/decoder/arm/dequantize_arm.h
-  vp8/decoder/arm/detokenize_arm.h
   vp8/decoder/x86/dequantize_x86.h
   vpx/internal/vpx_codec_internal.h
   vpx/vp8cx.h
@@ -240,7 +237,6 @@ commonFiles=(
   vp8/common/arm/neon/recon16x16mb_neon.asm
   vp8/common/arm/neon/buildintrapredictorsmby_neon.asm
   vp8/common/arm/neon/save_neon_reg.asm
-  vp8/decoder/arm/detokenize.asm
   vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm
   vp8/decoder/arm/armv6/dequant_idct_v6.asm
   vp8/decoder/arm/armv6/dequantize_v6.asm
@@ -318,17 +314,5 @@ done
 # Patch to compile with Sun Studio on Solaris
 patch -p3 < solaris.patch
 
-# Patch to fix link with xcode4
-patch -p1 < xcode4.patch
-
-# Patch to fix data race on global function pointers
-patch -p3 < bug640935.patch
-
-# Patch to avoid text relocations on ARM
-patch -p3 < bug646815.patch
-
-# Patch to fix alignment problems with using ARM asm in Thumb mode.
-patch -p3 < bug666931.patch
-
-# Patch to make chroma planes 16-byte aligned.
-patch -p3 < bug671818.patch
+# Patch to fix errors including C headers in C++
+patch -p3 < compile_errors.patch

media/libvpx/vp8/common/alloccommon.c

@@ -16,18 +16,20 @@
 #include "findnearmv.h"
 #include "entropymode.h"
 #include "systemdependent.h"
-#include "vpxerrors.h"
 
 
 extern  void vp8_init_scan_order_mask();
 
-void vp8_update_mode_info_border(MODE_INFO *mi, int rows, int cols)
+static void update_mode_info_border(MODE_INFO *mi, int rows, int cols)
 {
     int i;
     vpx_memset(mi - cols - 2, 0, sizeof(MODE_INFO) * (cols + 1));
 
     for (i = 0; i < rows; i++)
     {
+        /* TODO(holmer): Bug? This updates the last element of each row
+         * rather than the border element!
+         */
         vpx_memset(&mi[i*cols-1], 0, sizeof(MODE_INFO));
     }
 }
@@ -44,9 +46,11 @@ void vp8_de_alloc_frame_buffers(VP8_COMMON *oci)
 
     vpx_free(oci->above_context);
     vpx_free(oci->mip);
+    vpx_free(oci->prev_mip);
 
     oci->above_context = 0;
     oci->mip = 0;
+    oci->prev_mip = 0;
 
 }
 
@@ -66,12 +70,12 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
 
     for (i = 0; i < NUM_YV12_BUFFERS; i++)
     {
-      oci->fb_idx_ref_cnt[0] = 0;
-
-      if (vp8_yv12_alloc_frame_buffer(&oci->yv12_fb[i],  width, height, VP8BORDERINPIXELS) < 0)
+        oci->fb_idx_ref_cnt[i] = 0;
+        oci->yv12_fb[i].flags = 0;
+        if (vp8_yv12_alloc_frame_buffer(&oci->yv12_fb[i], width, height, VP8BORDERINPIXELS) < 0)
         {
             vp8_de_alloc_frame_buffers(oci);
-            return ALLOC_FAILURE;
+            return 1;
         }
     }
 
@@ -88,13 +92,13 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
     if (vp8_yv12_alloc_frame_buffer(&oci->temp_scale_frame,   width, 16, VP8BORDERINPIXELS) < 0)
     {
         vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
+        return 1;
     }
 
     if (vp8_yv12_alloc_frame_buffer(&oci->post_proc_buffer, width, height, VP8BORDERINPIXELS) < 0)
     {
         vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
+        return 1;
     }
 
     oci->mb_rows = height >> 4;
@@ -106,21 +110,39 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
     if (!oci->mip)
     {
         vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
+        return 1;
     }
 
     oci->mi = oci->mip + oci->mode_info_stride + 1;
 
+    /* allocate memory for last frame MODE_INFO array */
+#if CONFIG_ERROR_CONCEALMENT
+    oci->prev_mip = vpx_calloc((oci->mb_cols + 1) * (oci->mb_rows + 1), sizeof(MODE_INFO));
+
+    if (!oci->prev_mip)
+    {
+        vp8_de_alloc_frame_buffers(oci);
+        return 1;
+    }
+
+    oci->prev_mi = oci->prev_mip + oci->mode_info_stride + 1;
+#else
+    oci->prev_mip = NULL;
+    oci->prev_mi = NULL;
+#endif
 
     oci->above_context = vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * oci->mb_cols, 1);
 
     if (!oci->above_context)
     {
         vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
+        return 1;
     }
 
-    vp8_update_mode_info_border(oci->mi, oci->mb_rows, oci->mb_cols);
+    update_mode_info_border(oci->mi, oci->mb_rows, oci->mb_cols);
+#if CONFIG_ERROR_CONCEALMENT
+    update_mode_info_border(oci->prev_mi, oci->mb_rows, oci->mb_cols);
+#endif
 
     return 0;
 }
@@ -130,32 +152,32 @@ void vp8_setup_version(VP8_COMMON *cm)
     {
     case 0:
         cm->no_lpf = 0;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 0;
         cm->full_pixel = 0;
         break;
     case 1:
         cm->no_lpf = 0;
-        cm->simpler_lpf = 1;
+        cm->filter_type = SIMPLE_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 0;
         break;
     case 2:
         cm->no_lpf = 1;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 0;
         break;
     case 3:
         cm->no_lpf = 1;
-        cm->simpler_lpf = 1;
+        cm->filter_type = SIMPLE_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 1;
         break;
     default:
         /*4,5,6,7 are reserved for future use*/
         cm->no_lpf = 0;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 0;
         cm->full_pixel = 0;
         break;
@@ -170,7 +192,7 @@ void vp8_create_common(VP8_COMMON *oci)
 
     oci->mb_no_coeff_skip = 1;
     oci->no_lpf = 0;
-    oci->simpler_lpf = 0;
+    oci->filter_type = NORMAL_LOOPFILTER;
     oci->use_bilinear_mc_filter = 0;
     oci->full_pixel = 0;
     oci->multi_token_partition = ONE_PARTITION;

media/libvpx/vp8/common/arm/arm_systemdependent.c

@@ -11,27 +11,30 @@
 
 #include "vpx_ports/config.h"
 #include "vpx_ports/arm.h"
-#include "g_common.h"
-#include "pragmas.h"
-#include "subpixel.h"
-#include "loopfilter.h"
-#include "recon.h"
-#include "idct.h"
-#include "onyxc_int.h"
+#include "vp8/common/g_common.h"
+#include "vp8/common/pragmas.h"
+#include "vp8/common/subpixel.h"
+#include "vp8/common/loopfilter.h"
+#include "vp8/common/recon.h"
+#include "vp8/common/idct.h"
+#include "vp8/common/onyxc_int.h"
 
 void vp8_arch_arm_common_init(VP8_COMMON *ctx)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
     int flags = arm_cpu_caps();
-    int has_edsp = flags & HAS_EDSP;
-    int has_media = flags & HAS_MEDIA;
-    int has_neon = flags & HAS_NEON;
     rtcd->flags = flags;
 
     /* Override default functions with fastest ones for this CPU. */
+#if HAVE_ARMV5TE
+    if (flags & HAS_EDSP)
+    {
+    }
+#endif
+
 #if HAVE_ARMV6
-    if (has_media)
+    if (flags & HAS_MEDIA)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_armv6;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_armv6;
@@ -51,9 +54,11 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_armv6;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_armv6;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_armv6;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_armv6;
+        rtcd->loopfilter.simple_mb_v =
+                vp8_loop_filter_simple_vertical_edge_armv6;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_armv6;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_armv6;
+        rtcd->loopfilter.simple_mb_h =
+                vp8_loop_filter_simple_horizontal_edge_armv6;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_armv6;
 
         rtcd->recon.copy16x16   = vp8_copy_mem16x16_v6;
@@ -66,7 +71,7 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
 #endif
 
 #if HAVE_ARMV7
-    if (has_neon)
+    if (flags & HAS_NEON)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_neon;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_neon;

media/libvpx/vp8/common/arm/armv6/bilinearfilter_v6.asm

@@ -15,33 +15,33 @@
     AREA    |.text|, CODE, READONLY  ; name this block of code
 
 ;-------------------------------------
-; r0    unsigned char *src_ptr,
-; r1    unsigned short *output_ptr,
-; r2    unsigned int src_pixels_per_line,
-; r3    unsigned int output_height,
-; stack    unsigned int output_width,
-; stack    const short *vp8_filter
+; r0    unsigned char  *src_ptr,
+; r1    unsigned short *dst_ptr,
+; r2    unsigned int    src_pitch,
+; r3    unsigned int    height,
+; stack unsigned int    width,
+; stack const short    *vp8_filter
 ;-------------------------------------
 ; The output is transposed stroed in output array to make it easy for second pass filtering.
 |vp8_filter_block2d_bil_first_pass_armv6| PROC
     stmdb   sp!, {r4 - r11, lr}
 
     ldr     r11, [sp, #40]                  ; vp8_filter address
-    ldr     r4, [sp, #36]                   ; output width
+    ldr     r4, [sp, #36]                   ; width
 
     mov     r12, r3                         ; outer-loop counter
-    sub     r2, r2, r4                      ; src increment for height loop
 
-    ;;IF ARCHITECTURE=6
-    pld     [r0]
-    ;;ENDIF
+    add     r7, r2, r4                      ; preload next row
+    pld     [r0, r7]
+
+    sub     r2, r2, r4                      ; src increment for height loop
 
     ldr     r5, [r11]                       ; load up filter coefficients
 
-    mov     r3, r3, lsl #1                  ; output_height*2
+    mov     r3, r3, lsl #1                  ; height*2
     add     r3, r3, #2                      ; plus 2 to make output buffer 4-bit aligned since height is actually (height+1)
 
-    mov     r11, r1                         ; save output_ptr for each row
+    mov     r11, r1                         ; save dst_ptr for each row
 
     cmp     r5, #128                        ; if filter coef = 128, then skip the filter
     beq     bil_null_1st_filter
@@ -96,9 +96,8 @@
     add     r0, r0, r2                      ; move to next input row
     subs    r12, r12, #1
 
-    ;;IF ARCHITECTURE=6
-    pld     [r0]
-    ;;ENDIF
+    add     r9, r2, r4, lsl #1              ; adding back block width
+    pld     [r0, r9]                        ; preload next row
 
     add     r11, r11, #2                    ; move over to next column
     mov     r1, r11
@@ -140,17 +139,17 @@
 
 ;---------------------------------
 ; r0    unsigned short *src_ptr,
-; r1    unsigned char *output_ptr,
-; r2    int output_pitch,
-; r3    unsigned int  output_height,
-; stack unsigned int  output_width,
-; stack const short *vp8_filter
+; r1    unsigned char  *dst_ptr,
+; r2    int             dst_pitch,
+; r3    unsigned int    height,
+; stack unsigned int    width,
+; stack const short    *vp8_filter
 ;---------------------------------
 |vp8_filter_block2d_bil_second_pass_armv6| PROC
     stmdb   sp!, {r4 - r11, lr}
 
     ldr     r11, [sp, #40]                  ; vp8_filter address
-    ldr     r4, [sp, #36]                   ; output width
+    ldr     r4, [sp, #36]                   ; width
 
     ldr     r5, [r11]                       ; load up filter coefficients
     mov     r12, r4                         ; outer-loop counter = width, since we work on transposed data matrix

media/libvpx/vp8/common/arm/armv6/copymem16x16_v6.asm

@@ -22,9 +22,7 @@
     ;push   {r4-r7}
 
     ;preload
-    pld     [r0]
-    pld     [r0, r1]
-    pld     [r0, r1, lsl #1]
+    pld     [r0, #31]                ; preload for next 16x16 block
 
     ands    r4, r0, #15
     beq     copy_mem16x16_fast
@@ -90,6 +88,8 @@ copy_mem16x16_1_loop
     ldrneb  r6, [r0, #2]
     ldrneb  r7, [r0, #3]
 
+    pld     [r0, #31]               ; preload for next 16x16 block
+
     bne     copy_mem16x16_1_loop
 
     ldmia       sp!, {r4 - r7}
@@ -121,6 +121,8 @@ copy_mem16x16_4_loop
     ldrne   r6, [r0, #8]
     ldrne   r7, [r0, #12]
 
+    pld     [r0, #31]               ; preload for next 16x16 block
+
     bne     copy_mem16x16_4_loop
 
     ldmia       sp!, {r4 - r7}
@@ -148,6 +150,7 @@ copy_mem16x16_8_loop
 
     add     r2, r2, r3
 
+    pld     [r0, #31]               ; preload for next 16x16 block
     bne     copy_mem16x16_8_loop
 
     ldmia       sp!, {r4 - r7}
@@ -171,6 +174,7 @@ copy_mem16x16_fast_loop
     ;stm        r2, {r4-r7}
     add     r2, r2, r3
 
+    pld     [r0, #31]               ; preload for next 16x16 block
     bne     copy_mem16x16_fast_loop
 
     ldmia       sp!, {r4 - r7}

media/libvpx/vp8/common/arm/armv6/filter_v6.asm

@@ -10,6 +10,8 @@
 
 
     EXPORT  |vp8_filter_block2d_first_pass_armv6|
+    EXPORT  |vp8_filter_block2d_first_pass_16x16_armv6|
+    EXPORT  |vp8_filter_block2d_first_pass_8x8_armv6|
     EXPORT  |vp8_filter_block2d_second_pass_armv6|
     EXPORT  |vp8_filter4_block2d_second_pass_armv6|
     EXPORT  |vp8_filter_block2d_first_pass_only_armv6|
@@ -40,11 +42,6 @@
     add     r12, r3, #16                    ; square off the output
     sub     sp, sp, #4
 
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, #-2]
-    ;;pld       [r0, #30]
-    ;;ENDIF
-
     ldr     r4, [r11]                       ; load up packed filter coefficients
     ldr     r5, [r11, #4]
     ldr     r6, [r11, #8]
@@ -101,15 +98,10 @@
 
     bne     width_loop_1st_6
 
-    ;;add       r9, r2, #30                 ; attempt to load 2 adjacent cache lines
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, r2]
-    ;;pld       [r0, r9]
-    ;;ENDIF
-
     ldr     r1, [sp]                        ; load and update dst address
     subs    r7, r7, #0x10000
     add     r0, r0, r2                      ; move to next input line
+
     add     r1, r1, #2                      ; move over to next column
     str     r1, [sp]
 
@@ -120,6 +112,192 @@
 
     ENDP
 
+; --------------------------
+; 16x16 version
+; -----------------------------
+|vp8_filter_block2d_first_pass_16x16_armv6| PROC
+    stmdb   sp!, {r4 - r11, lr}
+
+    ldr     r11, [sp, #40]                  ; vp8_filter address
+    ldr     r7, [sp, #36]                   ; output height
+
+    add     r4, r2, #18                     ; preload next low
+    pld     [r0, r4]
+
+    sub     r2, r2, r3                      ; inside loop increments input array,
+                                            ; so the height loop only needs to add
+                                            ; r2 - width to the input pointer
+
+    mov     r3, r3, lsl #1                  ; multiply width by 2 because using shorts
+    add     r12, r3, #16                    ; square off the output
+    sub     sp, sp, #4
+
+    ldr     r4, [r11]                       ; load up packed filter coefficients
+    ldr     r5, [r11, #4]
+    ldr     r6, [r11, #8]
+
+    str     r1, [sp]                        ; push destination to stack
+    mov     r7, r7, lsl #16                 ; height is top part of counter
+
+; six tap filter
+|height_loop_1st_16_6|
+    ldrb    r8, [r0, #-2]                   ; load source data
+    ldrb    r9, [r0, #-1]
+    ldrb    r10, [r0], #2
+    orr     r7, r7, r3, lsr #2              ; construct loop counter
+
+|width_loop_1st_16_6|
+    ldrb    r11, [r0, #-1]
+
+    pkhbt   lr, r8, r9, lsl #16             ; r9 | r8
+    pkhbt   r8, r9, r10, lsl #16            ; r10 | r9
+
+    ldrb    r9, [r0]
+
+    smuad   lr, lr, r4                      ; apply the filter
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+    smuad   r8, r8, r4
+    pkhbt   r11, r11, r9, lsl #16           ; r9 | r11
+
+    smlad   lr, r10, r5, lr
+    ldrb    r10, [r0, #1]
+    smlad   r8, r11, r5, r8
+    ldrb    r11, [r0, #2]
+
+    sub     r7, r7, #1
+
+    pkhbt   r9, r9, r10, lsl #16            ; r10 | r9
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+
+    smlad   lr, r9, r6, lr
+    smlad   r11, r10, r6, r8
+
+    ands    r10, r7, #0xff                  ; test loop counter
+
+    add     lr, lr, #0x40                   ; round_shift_and_clamp
+    ldrneb  r8, [r0, #-2]                   ; load data for next loop
+    usat    lr, #8, lr, asr #7
+    add     r11, r11, #0x40
+    ldrneb  r9, [r0, #-1]
+    usat    r11, #8, r11, asr #7
+
+    strh    lr, [r1], r12                   ; result is transposed and stored, which
+                                            ; will make second pass filtering easier.
+    ldrneb  r10, [r0], #2
+    strh    r11, [r1], r12
+
+    bne     width_loop_1st_16_6
+
+    ldr     r1, [sp]                        ; load and update dst address
+    subs    r7, r7, #0x10000
+    add     r0, r0, r2                      ; move to next input line
+
+    add     r11, r2, #34                    ; adding back block width(=16)
+    pld     [r0, r11]                       ; preload next low
+
+    add     r1, r1, #2                      ; move over to next column
+    str     r1, [sp]
+
+    bne     height_loop_1st_16_6
+
+    add     sp, sp, #4
+    ldmia   sp!, {r4 - r11, pc}
+
+    ENDP
+
+; --------------------------
+; 8x8 version
+; -----------------------------
+|vp8_filter_block2d_first_pass_8x8_armv6| PROC
+    stmdb   sp!, {r4 - r11, lr}
+
+    ldr     r11, [sp, #40]                  ; vp8_filter address
+    ldr     r7, [sp, #36]                   ; output height
+
+    add     r4, r2, #10                     ; preload next low
+    pld     [r0, r4]
+
+    sub     r2, r2, r3                      ; inside loop increments input array,
+                                            ; so the height loop only needs to add
+                                            ; r2 - width to the input pointer
+
+    mov     r3, r3, lsl #1                  ; multiply width by 2 because using shorts
+    add     r12, r3, #16                    ; square off the output
+    sub     sp, sp, #4
+
+    ldr     r4, [r11]                       ; load up packed filter coefficients
+    ldr     r5, [r11, #4]
+    ldr     r6, [r11, #8]
+
+    str     r1, [sp]                        ; push destination to stack
+    mov     r7, r7, lsl #16                 ; height is top part of counter
+
+; six tap filter
+|height_loop_1st_8_6|
+    ldrb    r8, [r0, #-2]                   ; load source data
+    ldrb    r9, [r0, #-1]
+    ldrb    r10, [r0], #2
+    orr     r7, r7, r3, lsr #2              ; construct loop counter
+
+|width_loop_1st_8_6|
+    ldrb    r11, [r0, #-1]
+
+    pkhbt   lr, r8, r9, lsl #16             ; r9 | r8
+    pkhbt   r8, r9, r10, lsl #16            ; r10 | r9
+
+    ldrb    r9, [r0]
+
+    smuad   lr, lr, r4                      ; apply the filter
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+    smuad   r8, r8, r4
+    pkhbt   r11, r11, r9, lsl #16           ; r9 | r11
+
+    smlad   lr, r10, r5, lr
+    ldrb    r10, [r0, #1]
+    smlad   r8, r11, r5, r8
+    ldrb    r11, [r0, #2]
+
+    sub     r7, r7, #1
+
+    pkhbt   r9, r9, r10, lsl #16            ; r10 | r9
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+
+    smlad   lr, r9, r6, lr
+    smlad   r11, r10, r6, r8
+
+    ands    r10, r7, #0xff                  ; test loop counter
+
+    add     lr, lr, #0x40                   ; round_shift_and_clamp
+    ldrneb  r8, [r0, #-2]                   ; load data for next loop
+    usat    lr, #8, lr, asr #7
+    add     r11, r11, #0x40
+    ldrneb  r9, [r0, #-1]
+    usat    r11, #8, r11, asr #7
+
+    strh    lr, [r1], r12                   ; result is transposed and stored, which
+                                            ; will make second pass filtering easier.
+    ldrneb  r10, [r0], #2
+    strh    r11, [r1], r12
+
+    bne     width_loop_1st_8_6
+
+    ldr     r1, [sp]                        ; load and update dst address
+    subs    r7, r7, #0x10000
+    add     r0, r0, r2                      ; move to next input line
+
+    add     r11, r2, #18                    ; adding back block width(=8)
+    pld     [r0, r11]                       ; preload next low
+
+    add     r1, r1, #2                      ; move over to next column
+    str     r1, [sp]
+
+    bne     height_loop_1st_8_6
+
+    add     sp, sp, #4
+    ldmia   sp!, {r4 - r11, pc}
+
+    ENDP
+
 ;---------------------------------
 ; r0    short         *src_ptr,
 ; r1    unsigned char *output_ptr,
@@ -262,6 +440,10 @@
 |vp8_filter_block2d_first_pass_only_armv6| PROC
     stmdb   sp!, {r4 - r11, lr}
 
+    add     r7, r2, r3                      ; preload next low
+    add     r7, r7, #2
+    pld     [r0, r7]
+
     ldr     r4, [sp, #36]                   ; output pitch
     ldr     r11, [sp, #40]                  ; HFilter address
     sub     sp, sp, #8
@@ -330,16 +512,15 @@
 
     bne     width_loop_1st_only_6
 
-    ;;add       r9, r2, #30                 ; attempt to load 2 adjacent cache lines
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, r2]
-    ;;pld       [r0, r9]
-    ;;ENDIF
-
     ldr     lr, [sp]                        ; load back output pitch
     ldr     r12, [sp, #4]                   ; load back output pitch
     subs    r7, r7, #1
     add     r0, r0, r12                     ; updata src for next loop
+
+    add     r11, r12, r3                    ; preload next low
+    add     r11, r11, #2
+    pld     [r0, r11]
+
     add     r1, r1, lr                      ; update dst for next loop
 
     bne     height_loop_1st_only_6

media/libvpx/vp8/common/arm/armv6/loopfilter_v6.asm

@@ -53,14 +53,11 @@ count       RN  r5
 
 ;r0     unsigned char *src_ptr,
 ;r1     int src_pixel_step,
-;r2     const char *flimit,
+;r2     const char *blimit,
 ;r3     const char *limit,
 ;stack  const char *thresh,
 ;stack  int  count
 
-;Note: All 16 elements in flimit are equal. So, in the code, only one load is needed
-;for flimit. Same way applies to limit and thresh.
-
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 |vp8_loop_filter_horizontal_edge_armv6| PROC
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
@@ -72,14 +69,18 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r9, [src], pstep            ; p3
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r10, [src], pstep           ; p2
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r11, [src], pstep           ; p1
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r6], #4                ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r6]                    ; thresh
+    orr         r2, r2, r2, lsl #8
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |Hnext8|
     ; vp8_filter_mask() function
@@ -253,12 +254,6 @@ count       RN  r5
 
     subs        count, count, #1
 
-    ;pld            [src]
-    ;pld            [src, pstep]
-    ;pld            [src, pstep, lsl #1]
-    ;pld            [src, pstep, lsl #2]
-    ;pld            [src, pstep, lsl #3]
-
     ldrne       r9, [src], pstep            ; p3
     ldrne       r10, [src], pstep           ; p2
     ldrne       r11, [src], pstep           ; p1
@@ -281,14 +276,18 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r9, [src], pstep            ; p3
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r10, [src], pstep           ; p2
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r11, [src], pstep           ; p1
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r6], #4                ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r6]                    ; thresh
+    orr         r2, r2, r2, lsl #8
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |MBHnext8|
 
@@ -590,15 +589,19 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r6, [src], pstep            ; load source data
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r7, [src], pstep
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r8, [src], pstep
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r12], #4               ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r12]                   ; thresh
+    orr         r2, r2, r2, lsl #8
     ldr         lr, [src], pstep
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |Vnext8|
 
@@ -857,18 +860,26 @@ count       RN  r5
     sub         src, src, #4                ; move src pointer down by 4
     ldr         count, [sp, #40]            ; count for 8-in-parallel
     ldr         r12, [sp, #36]              ; load thresh address
+    pld         [src, #23]                  ; preload for next block
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r6, [src], pstep            ; load source data
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
+    pld         [src, #23]
     ldr         r7, [src], pstep
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
+    pld         [src, #23]
     ldr         r8, [src], pstep
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r12], #4               ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r12]                   ; thresh
+    orr         r2, r2, r2, lsl #8
+    pld         [src, #23]
     ldr         lr, [src], pstep
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |MBVnext8|
     ; vp8_filter_mask() function
@@ -908,6 +919,7 @@ count       RN  r5
     str         lr, [sp, #8]
     ldr         lr, [src], pstep
 
+
     TRANSPOSE_MATRIX r6, r7, r8, lr, r9, r10, r11, r12
 
     ldr         lr, [sp, #8]                ; load back (f)limit accumulator
@@ -956,6 +968,7 @@ count       RN  r5
     beq         mbvskip_filter               ; skip filtering
 
 
+
     ;vp8_hevmask() function
     ;calculate high edge variance
 
@@ -1123,6 +1136,7 @@ count       RN  r5
     smlabb      r8, r6, lr, r7
     smlatb      r6, r6, lr, r7
     smlabb      r9, r10, lr, r7
+
     smlatb      r10, r10, lr, r7
     ssat        r8, #8, r8, asr #7
     ssat        r6, #8, r6, asr #7
@@ -1242,9 +1256,13 @@ count       RN  r5
     sub         src, src, #4
     subs        count, count, #1
 
+    pld         [src, #23]                  ; preload for next block
     ldrne       r6, [src], pstep            ; load source data
+    pld         [src, #23]
     ldrne       r7, [src], pstep
+    pld         [src, #23]
     ldrne       r8, [src], pstep
+    pld         [src, #23]
     ldrne       lr, [src], pstep
 
     bne         MBVnext8

media/libvpx/vp8/common/arm/armv6/simpleloopfilter_v6.asm

@@ -45,35 +45,28 @@
     MEND
 
 
+
 src         RN  r0
 pstep       RN  r1
 
 ;r0     unsigned char *src_ptr,
 ;r1     int src_pixel_step,
-;r2     const char *flimit,
-;r3     const char *limit,
-;stack  const char *thresh,
-;stack  int  count
-
-; All 16 elements in flimit are equal. So, in the code, only one load is needed
-; for flimit. Same applies to limit. thresh is not used in simple looopfilter
+;r2     const char *blimit
 
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 |vp8_loop_filter_simple_horizontal_edge_armv6| PROC
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    ldr         r12, [r3]                   ; limit
+    ldrb        r12, [r2]                   ; blimit
     ldr         r3, [src, -pstep, lsl #1]   ; p1
     ldr         r4, [src, -pstep]           ; p0
     ldr         r5, [src]                   ; q0
     ldr         r6, [src, pstep]            ; q1
-    ldr         r7, [r2]                    ; flimit
+    orr         r12, r12, r12, lsl #8       ; blimit
     ldr         r2, c0x80808080
-    ldr         r9, [sp, #40]               ; count for 8-in-parallel
-    uadd8       r7, r7, r7                  ; flimit * 2
-    mov         r9, r9, lsl #1              ; double the count. we're doing 4 at a time
-    uadd8       r12, r7, r12                ; flimit * 2 + limit
+    orr         r12, r12, r12, lsl #16      ; blimit
+    mov         r9, #4                      ; double the count. we're doing 4 at a time
     mov         lr, #0                      ; need 0 in a couple places
 
 |simple_hnext8|
@@ -148,30 +141,32 @@ pstep       RN  r1
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    ldr         r12, [r2]                   ; r12: flimit
+    ldrb        r12, [r2]                   ; r12: blimit
     ldr         r2, c0x80808080
-    ldr         r7, [r3]                    ; limit
+    orr         r12, r12, r12, lsl #8
 
     ; load soure data to r7, r8, r9, r10
     ldrh        r3, [src, #-2]
+    pld         [src, #23]                  ; preload for next block
     ldrh        r4, [src], pstep
-    uadd8       r12, r12, r12               ; flimit * 2
+    orr         r12, r12, r12, lsl #16
 
     ldrh        r5, [src, #-2]
+    pld         [src, #23]
     ldrh        r6, [src], pstep
-    uadd8       r12, r12, r7                ; flimit * 2 + limit
 
     pkhbt       r7, r3, r4, lsl #16
 
     ldrh        r3, [src, #-2]
+    pld         [src, #23]
     ldrh        r4, [src], pstep
-    ldr         r11, [sp, #40]              ; count (r11) for 8-in-parallel
 
     pkhbt       r8, r5, r6, lsl #16
 
     ldrh        r5, [src, #-2]
+    pld         [src, #23]
     ldrh        r6, [src], pstep
-    mov         r11, r11, lsl #1            ; 4-in-parallel
+    mov         r11, #4                     ; double the count. we're doing 4 at a time
 
 |simple_vnext8|
     ; vp8_simple_filter_mask() function
@@ -259,19 +254,23 @@ pstep       RN  r1
 
     ; load soure data to r7, r8, r9, r10
     ldrneh      r3, [src, #-2]
+    pld         [src, #23]                  ; preload for next block
     ldrneh      r4, [src], pstep
 
     ldrneh      r5, [src, #-2]
+    pld         [src, #23]
     ldrneh      r6, [src], pstep
 
     pkhbt       r7, r3, r4, lsl #16
 
     ldrneh      r3, [src, #-2]
+    pld         [src, #23]
     ldrneh      r4, [src], pstep
 
     pkhbt       r8, r5, r6, lsl #16
 
     ldrneh      r5, [src, #-2]
+    pld         [src, #23]
     ldrneh      r6, [src], pstep
 
     bne         simple_vnext8

media/libvpx/vp8/common/arm/armv6/sixtappredict8x4_v6.asm

@@ -35,6 +35,9 @@
     adr         r12, filter8_coeff
     sub         r0, r0, r1, lsl #1
 
+    add         r3, r1, #10                 ; preload next low
+    pld         [r0, r3]
+
     add         r2, r12, r2, lsl #4         ;calculate filter location
     add         r0, r0, #3                  ;adjust src only for loading convinience
 
@@ -110,6 +113,9 @@
 
     add         r0, r0, r1                  ; move to next input line
 
+    add         r11, r1, #18                ; preload next low. adding back block width(=8), which is subtracted earlier
+    pld         [r0, r11]
+
     bne         first_pass_hloop_v6
 
 ;second pass filter
@@ -243,8 +249,6 @@ skip_secondpass_hloop
     ENDP
 
 ;-----------------
-    AREA    subpelfilters8_dat, DATA, READWRITE         ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
 ;One word each is reserved. Label filter_coeff can be used to access the data.
 ;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
 filter8_coeff

media/libvpx/vp8/common/arm/bilinearfilter_arm.c

@@ -10,128 +10,29 @@
 
 
 #include <math.h>
-#include "subpixel.h"
-
-#define BLOCK_HEIGHT_WIDTH 4
-#define VP8_FILTER_WEIGHT 128
-#define VP8_FILTER_SHIFT  7
-
-static const short bilinear_filters[8][2] =
-{
-    { 128,   0 },
-    { 112,  16 },
-    {  96,  32 },
-    {  80,  48 },
-    {  64,  64 },
-    {  48,  80 },
-    {  32,  96 },
-    {  16, 112 }
-};
-
-
-extern void vp8_filter_block2d_bil_first_pass_armv6
-(
-    unsigned char *src_ptr,
-    unsigned short *output_ptr,
-    unsigned int src_pixels_per_line,
-    unsigned int output_height,
-    unsigned int output_width,
-    const short *vp8_filter
-);
-
-extern void vp8_filter_block2d_bil_second_pass_armv6
-(
-    unsigned short *src_ptr,
-    unsigned char  *output_ptr,
-    int output_pitch,
-    unsigned int  output_height,
-    unsigned int  output_width,
-    const short *vp8_filter
-);
-
-#if 0
-void vp8_filter_block2d_bil_first_pass_6
-(
-    unsigned char *src_ptr,
-    unsigned short *output_ptr,
-    unsigned int src_pixels_per_line,
-    unsigned int output_height,
-    unsigned int output_width,
-    const short *vp8_filter
-)
-{
-    unsigned int i, j;
-
-    for ( i=0; i<output_height; i++ )
-    {
-        for ( j=0; j<output_width; j++ )
-        {
-            /* Apply bilinear filter */
-            output_ptr[j] = ( ( (int)src_ptr[0]          * vp8_filter[0]) +
-                               ((int)src_ptr[1] * vp8_filter[1]) +
-                                (VP8_FILTER_WEIGHT/2) ) >> VP8_FILTER_SHIFT;
-            src_ptr++;
-        }
-
-        /* Next row... */
-        src_ptr    += src_pixels_per_line - output_width;
-        output_ptr += output_width;
-    }
-}
-
-void vp8_filter_block2d_bil_second_pass_6
-(
-    unsigned short *src_ptr,
-    unsigned char  *output_ptr,
-    int output_pitch,
-    unsigned int  output_height,
-    unsigned int  output_width,
-    const short *vp8_filter
-)
-{
-    unsigned int  i,j;
-    int  Temp;
-
-    for ( i=0; i<output_height; i++ )
-    {
-        for ( j=0; j<output_width; j++ )
-        {
-            /* Apply filter */
-            Temp =  ((int)src_ptr[0]         * vp8_filter[0]) +
-                    ((int)src_ptr[output_width] * vp8_filter[1]) +
-                    (VP8_FILTER_WEIGHT/2);
-            output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
-            src_ptr++;
-        }
-
-        /* Next row... */
-        /*src_ptr    += src_pixels_per_line - output_width;*/
-        output_ptr += output_pitch;
-    }
-}
-#endif
+#include "vp8/common/filter.h"
+#include "vp8/common/subpixel.h"
+#include "bilinearfilter_arm.h"
 
 void vp8_filter_block2d_bil_armv6
 (
     unsigned char *src_ptr,
-    unsigned char *output_ptr,
-    unsigned int   src_pixels_per_line,
+    unsigned char *dst_ptr,
+    unsigned int   src_pitch,
     unsigned int   dst_pitch,
-    const short      *HFilter,
-    const short      *VFilter,
+    const short   *HFilter,
+    const short   *VFilter,
     int            Width,
     int            Height
 )
 {
-
-    unsigned short FData[36*16]; /* Temp data bufffer used in filtering */
+    unsigned short FData[36*16]; /* Temp data buffer used in filtering */
 
     /* First filter 1-D horizontally... */
-    /* pixel_step = 1; */
-    vp8_filter_block2d_bil_first_pass_armv6(src_ptr, FData, src_pixels_per_line, Height + 1, Width, HFilter);
+    vp8_filter_block2d_bil_first_pass_armv6(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
 
     /* then 1-D vertically... */
-    vp8_filter_block2d_bil_second_pass_armv6(FData, output_ptr, dst_pitch, Height, Width, VFilter);
+    vp8_filter_block2d_bil_second_pass_armv6(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
 }
 
 
@@ -148,8 +49,8 @@ void vp8_bilinear_predict4x4_armv6
     const short  *HFilter;
     const short  *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
     vp8_filter_block2d_bil_armv6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
 }
@@ -167,8 +68,8 @@ void vp8_bilinear_predict8x8_armv6
     const short  *HFilter;
     const short  *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
     vp8_filter_block2d_bil_armv6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
 }
@@ -186,8 +87,8 @@ void vp8_bilinear_predict8x4_armv6
     const short  *HFilter;
     const short  *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
     vp8_filter_block2d_bil_armv6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
 }
@@ -205,8 +106,8 @@ void vp8_bilinear_predict16x16_armv6
     const short  *HFilter;
     const short  *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
     vp8_filter_block2d_bil_armv6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
 }

media/libvpx/vp8/common/arm/bilinearfilter_arm.h

@@ -0,0 +1,35 @@
+/*
+ *  Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+
+#ifndef BILINEARFILTER_ARM_H
+#define BILINEARFILTER_ARM_H
+
+extern void vp8_filter_block2d_bil_first_pass_armv6
+(
+    const unsigned char  *src_ptr,
+    unsigned short       *dst_ptr,
+    unsigned int          src_pitch,
+    unsigned int          height,
+    unsigned int          width,
+    const short          *vp8_filter
+);
+
+extern void vp8_filter_block2d_bil_second_pass_armv6
+(
+    const unsigned short *src_ptr,
+    unsigned char        *dst_ptr,
+    int                   dst_pitch,
+    unsigned int          height,
+    unsigned int          width,
+    const short         *vp8_filter
+);
+
+#endif /* BILINEARFILTER_ARM_H */

media/libvpx/vp8/common/arm/filter_arm.c

@@ -11,26 +11,10 @@
 
 #include "vpx_ports/config.h"
 #include <math.h>
-#include "subpixel.h"
+#include "vp8/common/filter.h"
+#include "vp8/common/subpixel.h"
 #include "vpx_ports/mem.h"
 
-#define BLOCK_HEIGHT_WIDTH 4
-#define VP8_FILTER_WEIGHT 128
-#define VP8_FILTER_SHIFT  7
-
-DECLARE_ALIGNED(16, static const short, sub_pel_filters[8][6]) =
-{
-    { 0,  0,  128,    0,   0,  0 },         /* note that 1/8 pel positions are just as per alpha -0.5 bicubic */
-    { 0, -6,  123,   12,  -1,  0 },
-    { 2, -11, 108,   36,  -8,  1 },         /* New 1/4 pel 6 tap filter */
-    { 0, -9,   93,   50,  -6,  0 },
-    { 3, -16,  77,   77, -16,  3 },         /* New 1/2 pel 6 tap filter */
-    { 0, -6,   50,   93,  -9,  0 },
-    { 1, -8,   36,  108, -11,  2 },         /* New 1/4 pel 6 tap filter */
-    { 0, -1,   12,  123,  -6,  0 },
-};
-
-
 extern void vp8_filter_block2d_first_pass_armv6
 (
     unsigned char *src_ptr,
@@ -41,6 +25,28 @@ extern void vp8_filter_block2d_first_pass_armv6
     const short *vp8_filter
 );
 
+// 8x8
+extern void vp8_filter_block2d_first_pass_8x8_armv6
+(
+    unsigned char *src_ptr,
+    short         *output_ptr,
+    unsigned int src_pixels_per_line,
+    unsigned int output_width,
+    unsigned int output_height,
+    const short *vp8_filter
+);
+
+// 16x16
+extern void vp8_filter_block2d_first_pass_16x16_armv6
+(
+    unsigned char *src_ptr,
+    short         *output_ptr,
+    unsigned int src_pixels_per_line,
+    unsigned int output_width,
+    unsigned int output_height,
+    const short *vp8_filter
+);
+
 extern void vp8_filter_block2d_second_pass_armv6
 (
     short         *src_ptr,
@@ -93,11 +99,11 @@ void vp8_sixtap_predict_armv6
 {
     const short  *HFilter;
     const short  *VFilter;
-    DECLARE_ALIGNED_ARRAY(4, short, FData, 12*4); /* Temp data bufffer used in filtering */
+    DECLARE_ALIGNED_ARRAY(4, short, FData, 12*4); /* Temp data buffer used in filtering */
 
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];       /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* Vfilter is null. First pass only */
     if (xoffset && !yoffset)
@@ -129,47 +135,6 @@ void vp8_sixtap_predict_armv6
     }
 }
 
-#if 0
-void vp8_sixtap_predict8x4_armv6
-(
-    unsigned char  *src_ptr,
-    int  src_pixels_per_line,
-    int  xoffset,
-    int  yoffset,
-    unsigned char *dst_ptr,
-    int  dst_pitch
-)
-{
-    const short  *HFilter;
-    const short  *VFilter;
-    DECLARE_ALIGNED_ARRAY(4, short, FData, 16*8); /* Temp data bufffer used in filtering */
-
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];       /* 6 tap */
-
-
-    /*if (xoffset && !yoffset)
-    {
-        vp8_filter_block2d_first_pass_only_armv6 (  src_ptr, dst_ptr, src_pixels_per_line, 8, dst_pitch, HFilter );
-    }*/
-    /* Hfilter is null. Second pass only */
-    /*else if (!xoffset && yoffset)
-    {
-        vp8_filter_block2d_second_pass_only_armv6 ( src_ptr, dst_ptr, src_pixels_per_line, 8, dst_pitch, VFilter );
-    }
-    else
-    {
-        if (yoffset & 0x1)
-            vp8_filter_block2d_first_pass_armv6 ( src_ptr-src_pixels_per_line, FData+1, src_pixels_per_line, 8, 7, HFilter );
-        else*/
-
-        vp8_filter_block2d_first_pass_armv6 ( src_ptr-(2*src_pixels_per_line), FData, src_pixels_per_line, 8, 9, HFilter );
-
-        vp8_filter_block2d_second_pass_armv6 ( FData+2, dst_ptr, dst_pitch, 4, 8, VFilter );
-    /*}*/
-}
-#endif
-
 void vp8_sixtap_predict8x8_armv6
 (
     unsigned char  *src_ptr,
@@ -182,10 +147,10 @@ void vp8_sixtap_predict8x8_armv6
 {
     const short  *HFilter;
     const short  *VFilter;
-    DECLARE_ALIGNED_ARRAY(4, short, FData, 16*8); /* Temp data bufffer used in filtering */
+    DECLARE_ALIGNED_ARRAY(4, short, FData, 16*8); /* Temp data buffer used in filtering */
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];       /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     if (xoffset && !yoffset)
     {
@@ -200,12 +165,12 @@ void vp8_sixtap_predict8x8_armv6
     {
         if (yoffset & 0x1)
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 8, 11, HFilter);
+            vp8_filter_block2d_first_pass_8x8_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 8, 11, HFilter);
             vp8_filter4_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 8, VFilter);
         }
         else
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 8, 13, HFilter);
+            vp8_filter_block2d_first_pass_8x8_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 8, 13, HFilter);
             vp8_filter_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 8, VFilter);
         }
     }
@@ -224,10 +189,10 @@ void vp8_sixtap_predict16x16_armv6
 {
     const short  *HFilter;
     const short  *VFilter;
-    DECLARE_ALIGNED_ARRAY(4, short, FData, 24*16);    /* Temp data bufffer used in filtering */
+    DECLARE_ALIGNED_ARRAY(4, short, FData, 24*16);    /* Temp data buffer used in filtering */
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];       /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     if (xoffset && !yoffset)
     {
@@ -242,12 +207,12 @@ void vp8_sixtap_predict16x16_armv6
     {
         if (yoffset & 0x1)
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 16, 19, HFilter);
+            vp8_filter_block2d_first_pass_16x16_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 16, 19, HFilter);
             vp8_filter4_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 16, VFilter);
         }
         else
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 16, 21, HFilter);
+            vp8_filter_block2d_first_pass_16x16_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 16, 21, HFilter);
             vp8_filter_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 16, VFilter);
         }
     }

media/libvpx/vp8/common/arm/loopfilter_arm.c

@@ -9,135 +9,107 @@
  */
 
 
-#include "vpx_ports/config.h"
-#include <math.h>
-#include "loopfilter.h"
-#include "onyxc_int.h"
+#include "vpx_config.h"
+#include "vp8/common/loopfilter.h"
+#include "vp8/common/onyxc_int.h"
 
+#if HAVE_ARMV6
 extern prototype_loopfilter(vp8_loop_filter_horizontal_edge_armv6);
 extern prototype_loopfilter(vp8_loop_filter_vertical_edge_armv6);
 extern prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_armv6);
 extern prototype_loopfilter(vp8_mbloop_filter_vertical_edge_armv6);
-extern prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_armv6);
-extern prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_armv6);
+#endif
 
-extern prototype_loopfilter(vp8_loop_filter_horizontal_edge_y_neon);
-extern prototype_loopfilter(vp8_loop_filter_vertical_edge_y_neon);
-extern prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_y_neon);
-extern prototype_loopfilter(vp8_mbloop_filter_vertical_edge_y_neon);
-extern prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_neon);
-extern prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_neon);
+#if HAVE_ARMV7
+typedef void loopfilter_y_neon(unsigned char *src, int pitch,
+        unsigned char blimit, unsigned char limit, unsigned char thresh);
+typedef void loopfilter_uv_neon(unsigned char *u, int pitch,
+        unsigned char blimit, unsigned char limit, unsigned char thresh,
+        unsigned char *v);
 
-extern loop_filter_uvfunction vp8_loop_filter_horizontal_edge_uv_neon;
-extern loop_filter_uvfunction vp8_loop_filter_vertical_edge_uv_neon;
-extern loop_filter_uvfunction vp8_mbloop_filter_horizontal_edge_uv_neon;
-extern loop_filter_uvfunction vp8_mbloop_filter_vertical_edge_uv_neon;
+extern loopfilter_y_neon vp8_loop_filter_horizontal_edge_y_neon;
+extern loopfilter_y_neon vp8_loop_filter_vertical_edge_y_neon;
+extern loopfilter_y_neon vp8_mbloop_filter_horizontal_edge_y_neon;
+extern loopfilter_y_neon vp8_mbloop_filter_vertical_edge_y_neon;
 
+extern loopfilter_uv_neon vp8_loop_filter_horizontal_edge_uv_neon;
+extern loopfilter_uv_neon vp8_loop_filter_vertical_edge_uv_neon;
+extern loopfilter_uv_neon vp8_mbloop_filter_horizontal_edge_uv_neon;
+extern loopfilter_uv_neon vp8_mbloop_filter_vertical_edge_uv_neon;
+#endif
 
 #if HAVE_ARMV6
 /*ARMV6 loopfilter functions*/
 /* Horizontal MB filtering */
 void vp8_loop_filter_mbh_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                               int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-void vp8_loop_filter_mbhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                                int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                               int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-void vp8_loop_filter_mbvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                                int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_armv6(unsigned char *y_ptr, int y_stride,
+                               const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, int y_stride,
+                               const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 12, y_stride, blimit);
 }
 #endif
 
@@ -145,93 +117,60 @@ void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsig
 /* NEON loopfilter functions */
 /* Horizontal MB filtering */
 void vp8_loop_filter_mbh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    unsigned char mblim = *lfi->mblim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+    vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, mblim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-}
-
-void vp8_loop_filter_mbhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, mblim, lim, hev_thr, v_ptr);
 }
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    unsigned char mblim = *lfi->mblim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, mblim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-}
-
-void vp8_loop_filter_mbvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, mblim, lim, hev_thr, v_ptr);
 }
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    unsigned char blim = *lfi->blim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 4 * y_stride, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 8 * y_stride, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 12 * y_stride, y_stride, blim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4 * uv_stride);
-}
-
-void vp8_loop_filter_bhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+        vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, blim, lim, hev_thr, v_ptr + 4 * uv_stride);
 }
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    unsigned char blim = *lfi->blim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 4, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 8, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 12, y_stride, blim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4);
-}
-
-void vp8_loop_filter_bvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+        vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, blim, lim, hev_thr, v_ptr + 4);
 }
 #endif

media/libvpx/vp8/common/arm/loopfilter_arm.h

@@ -12,15 +12,17 @@
 #ifndef LOOPFILTER_ARM_H
 #define LOOPFILTER_ARM_H
 
+#include "vpx_config.h"
+
 #if HAVE_ARMV6
 extern prototype_loopfilter_block(vp8_loop_filter_mbv_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_armv6);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_lf_normal_mb_v
@@ -36,28 +38,29 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_armv6
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_armv6
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_armv6
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_armv6
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_armv6
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_armv6
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_armv6
-#endif
-#endif
+#endif /* !CONFIG_RUNTIME_CPU_DETECT */
+
+#endif /* HAVE_ARMV6 */
 
 #if HAVE_ARMV7
 extern prototype_loopfilter_block(vp8_loop_filter_mbv_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_mbvs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_mbhs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_neon);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_lf_normal_mb_v
@@ -83,7 +86,8 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_neon
-#endif
-#endif
+#endif /* !CONFIG_RUNTIME_CPU_DETECT */
 
-#endif
+#endif /* HAVE_ARMV7 */
+
+#endif /* LOOPFILTER_ARM_H */

media/libvpx/vp8/common/arm/neon/bilinearpredict16x16_neon.asm

@@ -350,10 +350,7 @@ filt_blk2d_spo16x16_loop_neon
     ENDP
 
 ;-----------------
-    AREA    bifilters16_dat, DATA, READWRITE            ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
 bifilter16_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm

@@ -123,10 +123,7 @@ skip_secondpass_filter
     ENDP
 
 ;-----------------
-    AREA    bilinearfilters4_dat, DATA, READWRITE           ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
 bifilter4_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

media/libvpx/vp8/common/arm/neon/bilinearpredict8x4_neon.asm

@@ -128,10 +128,7 @@ skip_secondpass_filter
     ENDP
 
 ;-----------------
-    AREA    bifilters8x4_dat, DATA, READWRITE           ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
 bifilter8x4_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

media/libvpx/vp8/common/arm/neon/bilinearpredict8x8_neon.asm

@@ -176,10 +176,7 @@ skip_secondpass_filter
     ENDP
 
 ;-----------------
-    AREA    bifilters8_dat, DATA, READWRITE         ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
 bifilter8_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

media/libvpx/vp8/common/arm/neon/iwalsh_neon.asm

@@ -20,19 +20,16 @@
 |vp8_short_inv_walsh4x4_neon| PROC
 
     ; read in all four lines of values: d0->d3
-    vldm.64 r0, {q0, q1}
+    vld1.i16 {q0-q1}, [r0@128]
 
     ; first for loop
-
     vadd.s16 d4, d0, d3 ;a = [0] + [12]
-    vadd.s16 d5, d1, d2 ;b = [4] + [8]
-    vsub.s16 d6, d1, d2 ;c = [4] - [8]
-    vsub.s16 d7, d0, d3 ;d = [0] - [12]
+    vadd.s16 d6, d1, d2 ;b = [4] + [8]
+    vsub.s16 d5, d0, d3 ;d = [0] - [12]
+    vsub.s16 d7, d1, d2 ;c = [4] - [8]
 
-    vadd.s16 d0, d4, d5 ;a + b
-    vadd.s16 d1, d6, d7 ;c + d
-    vsub.s16 d2, d4, d5 ;a - b
-    vsub.s16 d3, d7, d6 ;d - c
+    vadd.s16 q0, q2, q3 ; a+b d+c
+    vsub.s16 q1, q2, q3 ; a-b d-c
 
     vtrn.32 d0, d2 ;d0:  0  1  8  9
                    ;d2:  2  3 10 11
@@ -47,29 +44,22 @@
     ; second for loop
 
     vadd.s16 d4, d0, d3 ;a = [0] + [3]
-    vadd.s16 d5, d1, d2 ;b = [1] + [2]
-    vsub.s16 d6, d1, d2 ;c = [1] - [2]
-    vsub.s16 d7, d0, d3 ;d = [0] - [3]
+    vadd.s16 d6, d1, d2 ;b = [1] + [2]
+    vsub.s16 d5, d0, d3 ;d = [0] - [3]
+    vsub.s16 d7, d1, d2 ;c = [1] - [2]
 
-    vadd.s16 d0, d4, d5 ;e = a + b
-    vadd.s16 d1, d6, d7 ;f = c + d
-    vsub.s16 d2, d4, d5 ;g = a - b
-    vsub.s16 d3, d7, d6 ;h = d - c
+    vmov.i16 q8, #3
 
-    vmov.i16 q2, #3
-    vadd.i16 q0, q0, q2 ;e/f += 3
-    vadd.i16 q1, q1, q2 ;g/h += 3
+    vadd.s16 q0, q2, q3 ; a+b d+c
+    vsub.s16 q1, q2, q3 ; a-b d-c
+
+    vadd.i16 q0, q0, q8 ;e/f += 3
+    vadd.i16 q1, q1, q8 ;g/h += 3
 
     vshr.s16 q0, q0, #3 ;e/f >> 3
     vshr.s16 q1, q1, #3 ;g/h >> 3
 
-    vtrn.32 d0, d2
-    vtrn.32 d1, d3
-    vtrn.16 d0, d1
-    vtrn.16 d2, d3
-
-    vstmia.16 r1!, {q0}
-    vstmia.16 r1!, {q1}
+    vst4.i16 {d0,d1,d2,d3}, [r1@128]
 
     bx lr
     ENDP    ; |vp8_short_inv_walsh4x4_neon|
@@ -77,19 +67,13 @@
 
 ;short vp8_short_inv_walsh4x4_1_neon(short *input, short *output)
 |vp8_short_inv_walsh4x4_1_neon| PROC
-    ; load a full line into a neon register
-    vld1.16  {q0}, [r0]
-    ; extract first element and replicate
-    vdup.16 q1, d0[0]
-    ; add 3 to all values
-    vmov.i16 q2, #3
-    vadd.i16 q3, q1, q2
-    ; right shift
-    vshr.s16 q3, q3, #3
-    ; write it back
-    vstmia.16 r1!, {q3}
-    vstmia.16 r1!, {q3}
-
+    ldrsh r2, [r0]          ; load input[0]
+    add r3, r2, #3          ; add 3
+    add r2, r1, #16         ; base for last 8 output
+    asr r0, r3, #3          ; right shift 3
+    vdup.16 q0, r0          ; load and duplicate
+    vst1.16 {q0}, [r1@128]  ; write back 8
+    vst1.16 {q0}, [r2@128]  ; write back last 8
     bx lr
     ENDP    ; |vp8_short_inv_walsh4x4_1_neon|
 

media/libvpx/vp8/common/arm/neon/loopfilter_neon.asm

@@ -14,109 +14,97 @@
     EXPORT  |vp8_loop_filter_vertical_edge_y_neon|
     EXPORT  |vp8_loop_filter_vertical_edge_uv_neon|
     ARM
-    REQUIRE8
-    PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
 
-; flimit, limit, and thresh should be positive numbers.
-; All 16 elements in these variables are equal.
-
-; void vp8_loop_filter_horizontal_edge_y_neon(unsigned char *src, int pitch,
-;                                             const signed char *flimit,
-;                                             const signed char *limit,
-;                                             const signed char *thresh,
-;                                             int count)
 ; r0    unsigned char *src
 ; r1    int pitch
-; r2    const signed char *flimit
-; r3    const signed char *limit
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_loop_filter_horizontal_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    sub         r2, r0, r1, lsl #2          ; move src pointer down by 4 lines
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    vdup.u8     q0, r2                     ; duplicate blimit
+    vdup.u8     q1, r3                     ; duplicate limit
+    sub         r2, r0, r1, lsl #2         ; move src pointer down by 4 lines
+    ldr         r3, [sp, #4]               ; load thresh
+    add         r12, r2, r1
+    add         r1, r1, r1
 
-    vld1.u8     {q3}, [r2], r1              ; p3
-    vld1.u8     {q4}, [r2], r1              ; p2
-    vld1.u8     {q5}, [r2], r1              ; p1
-    vld1.u8     {q6}, [r2], r1              ; p0
-    vld1.u8     {q7}, [r2], r1              ; q0
-    vld1.u8     {q8}, [r2], r1              ; q1
-    vld1.u8     {q9}, [r2], r1              ; q2
-    vld1.u8     {q10}, [r2]                 ; q3
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    sub         r0, r0, r1, lsl #1
+    vdup.u8     q2, r3                     ; duplicate thresh
+
+    vld1.u8     {q3}, [r2@128], r1              ; p3
+    vld1.u8     {q4}, [r12@128], r1             ; p2
+    vld1.u8     {q5}, [r2@128], r1              ; p1
+    vld1.u8     {q6}, [r12@128], r1             ; p0
+    vld1.u8     {q7}, [r2@128], r1              ; q0
+    vld1.u8     {q8}, [r12@128], r1             ; q1
+    vld1.u8     {q9}, [r2@128]                  ; q2
+    vld1.u8     {q10}, [r12@128]                ; q3
+
+    sub         r2, r2, r1, lsl #1
+    sub         r12, r12, r1, lsl #1
 
     bl          vp8_loop_filter_neon
 
-    vst1.u8     {q5}, [r0], r1              ; store op1
-    vst1.u8     {q6}, [r0], r1              ; store op0
-    vst1.u8     {q7}, [r0], r1              ; store oq0
-    vst1.u8     {q8}, [r0], r1              ; store oq1
+    vst1.u8     {q5}, [r2@128], r1              ; store op1
+    vst1.u8     {q6}, [r12@128], r1             ; store op0
+    vst1.u8     {q7}, [r2@128], r1              ; store oq0
+    vst1.u8     {q8}, [r12@128], r1             ; store oq1
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_horizontal_edge_y_neon|
 
-; void vp8_loop_filter_horizontal_edge_uv_neon(unsigned char *u, int pitch
-;                                              const signed char *flimit,
-;                                              const signed char *limit,
-;                                              const signed char *thresh,
-;                                              unsigned char *v)
+
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
 |vp8_loop_filter_horizontal_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
+    push        {lr}
+    vdup.u8     q0, r2                      ; duplicate blimit
+    vdup.u8     q1, r3                      ; duplicate limit
+    ldr         r12, [sp, #4]               ; load thresh
     ldr         r2, [sp, #8]                ; load v ptr
+    vdup.u8     q2, r12                     ; duplicate thresh
 
     sub         r3, r0, r1, lsl #2          ; move u pointer down by 4 lines
-    vld1.u8     {d6}, [r3], r1              ; p3
-    vld1.u8     {d8}, [r3], r1              ; p2
-    vld1.u8     {d10}, [r3], r1             ; p1
-    vld1.u8     {d12}, [r3], r1             ; p0
-    vld1.u8     {d14}, [r3], r1             ; q0
-    vld1.u8     {d16}, [r3], r1             ; q1
-    vld1.u8     {d18}, [r3], r1             ; q2
-    vld1.u8     {d20}, [r3]                 ; q3
-
-    ldr         r3, [sp, #4]                ; load thresh pointer
-
     sub         r12, r2, r1, lsl #2         ; move v pointer down by 4 lines
-    vld1.u8     {d7}, [r12], r1             ; p3
-    vld1.u8     {d9}, [r12], r1             ; p2
-    vld1.u8     {d11}, [r12], r1            ; p1
-    vld1.u8     {d13}, [r12], r1            ; p0
-    vld1.u8     {d15}, [r12], r1            ; q0
-    vld1.u8     {d17}, [r12], r1            ; q1
-    vld1.u8     {d19}, [r12], r1            ; q2
-    vld1.u8     {d21}, [r12]                ; q3
 
-    vld1.s8     {d4[], d5[]}, [r3]          ; thresh
+    vld1.u8     {d6}, [r3@64], r1              ; p3
+    vld1.u8     {d7}, [r12@64], r1             ; p3
+    vld1.u8     {d8}, [r3@64], r1              ; p2
+    vld1.u8     {d9}, [r12@64], r1             ; p2
+    vld1.u8     {d10}, [r3@64], r1             ; p1
+    vld1.u8     {d11}, [r12@64], r1            ; p1
+    vld1.u8     {d12}, [r3@64], r1             ; p0
+    vld1.u8     {d13}, [r12@64], r1            ; p0
+    vld1.u8     {d14}, [r3@64], r1             ; q0
+    vld1.u8     {d15}, [r12@64], r1            ; q0
+    vld1.u8     {d16}, [r3@64], r1             ; q1
+    vld1.u8     {d17}, [r12@64], r1            ; q1
+    vld1.u8     {d18}, [r3@64], r1             ; q2
+    vld1.u8     {d19}, [r12@64], r1            ; q2
+    vld1.u8     {d20}, [r3@64]                 ; q3
+    vld1.u8     {d21}, [r12@64]                ; q3
 
     bl          vp8_loop_filter_neon
 
     sub         r0, r0, r1, lsl #1
     sub         r2, r2, r1, lsl #1
 
-    vst1.u8     {d10}, [r0], r1             ; store u op1
-    vst1.u8     {d11}, [r2], r1             ; store v op1
-    vst1.u8     {d12}, [r0], r1             ; store u op0
-    vst1.u8     {d13}, [r2], r1             ; store v op0
-    vst1.u8     {d14}, [r0], r1             ; store u oq0
-    vst1.u8     {d15}, [r2], r1             ; store v oq0
-    vst1.u8     {d16}, [r0]                 ; store u oq1
-    vst1.u8     {d17}, [r2]                 ; store v oq1
+    vst1.u8     {d10}, [r0@64], r1             ; store u op1
+    vst1.u8     {d11}, [r2@64], r1             ; store v op1
+    vst1.u8     {d12}, [r0@64], r1             ; store u op0
+    vst1.u8     {d13}, [r2@64], r1             ; store v op0
+    vst1.u8     {d14}, [r0@64], r1             ; store u oq0
+    vst1.u8     {d15}, [r2@64], r1             ; store v oq0
+    vst1.u8     {d16}, [r0@64]                 ; store u oq1
+    vst1.u8     {d17}, [r2@64]                 ; store v oq1
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_horizontal_edge_uv_neon|
 
 ; void vp8_loop_filter_vertical_edge_y_neon(unsigned char *src, int pitch,
@@ -124,39 +112,38 @@
 ;                                           const signed char *limit,
 ;                                           const signed char *thresh,
 ;                                           int count)
-; r0    unsigned char *src,
-; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r0    unsigned char *src
+; r1    int pitch
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
+
 |vp8_loop_filter_vertical_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    sub         r2, r0, #4                  ; src ptr down by 4 columns
-    sub         r0, r0, #2                  ; dst ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    vdup.u8     q0, r2                     ; duplicate blimit
+    vdup.u8     q1, r3                     ; duplicate limit
+    sub         r2, r0, #4                 ; src ptr down by 4 columns
+    add         r1, r1, r1
+    ldr         r3, [sp, #4]               ; load thresh
+    add         r12, r2, r1, asr #1
 
-    vld1.u8     {d6}, [r2], r1              ; load first 8-line src data
-    vld1.u8     {d8}, [r2], r1
+    vld1.u8     {d6}, [r2], r1
+    vld1.u8     {d8}, [r12], r1
     vld1.u8     {d10}, [r2], r1
-    vld1.u8     {d12}, [r2], r1
+    vld1.u8     {d12}, [r12], r1
     vld1.u8     {d14}, [r2], r1
-    vld1.u8     {d16}, [r2], r1
+    vld1.u8     {d16}, [r12], r1
     vld1.u8     {d18}, [r2], r1
-    vld1.u8     {d20}, [r2], r1
-
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.u8     {d20}, [r12], r1
 
     vld1.u8     {d7}, [r2], r1              ; load second 8-line src data
-    vld1.u8     {d9}, [r2], r1
+    vld1.u8     {d9}, [r12], r1
     vld1.u8     {d11}, [r2], r1
-    vld1.u8     {d13}, [r2], r1
+    vld1.u8     {d13}, [r12], r1
     vld1.u8     {d15}, [r2], r1
-    vld1.u8     {d17}, [r2], r1
-    vld1.u8     {d19}, [r2], r1
-    vld1.u8     {d21}, [r2]
+    vld1.u8     {d17}, [r12], r1
+    vld1.u8     {d19}, [r2]
+    vld1.u8     {d21}, [r12]
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -164,6 +151,8 @@
     vtrn.32     q5, q9
     vtrn.32     q6, q10
 
+    vdup.u8     q2, r3                     ; duplicate thresh
+
     vtrn.16     q3, q5
     vtrn.16     q4, q6
     vtrn.16     q7, q9
@@ -178,28 +167,34 @@
 
     vswp        d12, d11
     vswp        d16, d13
+
+    sub         r0, r0, #2                 ; dst ptr
+
     vswp        d14, d12
     vswp        d16, d15
 
+    add         r12, r0, r1, asr #1
+
     ;store op1, op0, oq0, oq1
     vst4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
-    vst4.8      {d10[1], d11[1], d12[1], d13[1]}, [r0], r1
+    vst4.8      {d10[1], d11[1], d12[1], d13[1]}, [r12], r1
     vst4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r1
-    vst4.8      {d10[3], d11[3], d12[3], d13[3]}, [r0], r1
+    vst4.8      {d10[3], d11[3], d12[3], d13[3]}, [r12], r1
     vst4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r1
-    vst4.8      {d10[5], d11[5], d12[5], d13[5]}, [r0], r1
+    vst4.8      {d10[5], d11[5], d12[5], d13[5]}, [r12], r1
     vst4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r1
-    vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0], r1
-    vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r0], r1
-    vst4.8      {d14[1], d15[1], d16[1], d17[1]}, [r0], r1
-    vst4.8      {d14[2], d15[2], d16[2], d17[2]}, [r0], r1
-    vst4.8      {d14[3], d15[3], d16[3], d17[3]}, [r0], r1
-    vst4.8      {d14[4], d15[4], d16[4], d17[4]}, [r0], r1
-    vst4.8      {d14[5], d15[5], d16[5], d17[5]}, [r0], r1
-    vst4.8      {d14[6], d15[6], d16[6], d17[6]}, [r0], r1
-    vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r0]
+    vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r12], r1
 
-    ldmia       sp!, {pc}
+    vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r0], r1
+    vst4.8      {d14[1], d15[1], d16[1], d17[1]}, [r12], r1
+    vst4.8      {d14[2], d15[2], d16[2], d17[2]}, [r0], r1
+    vst4.8      {d14[3], d15[3], d16[3], d17[3]}, [r12], r1
+    vst4.8      {d14[4], d15[4], d16[4], d17[4]}, [r0], r1
+    vst4.8      {d14[5], d15[5], d16[5], d17[5]}, [r12], r1
+    vst4.8      {d14[6], d15[6], d16[6], d17[6]}, [r0]
+    vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r12]
+
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_vertical_edge_y_neon|
 
 ; void vp8_loop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch
@@ -209,38 +204,36 @@
 ;                                            unsigned char *v)
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
 |vp8_loop_filter_vertical_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r12, r0, #4                  ; move u pointer down by 4 columns
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-
+    push        {lr}
+    vdup.u8     q0, r2                      ; duplicate blimit
+    sub         r12, r0, #4                 ; move u pointer down by 4 columns
     ldr         r2, [sp, #8]                ; load v ptr
-
-    vld1.u8     {d6}, [r12], r1              ;load u data
-    vld1.u8     {d8}, [r12], r1
-    vld1.u8     {d10}, [r12], r1
-    vld1.u8     {d12}, [r12], r1
-    vld1.u8     {d14}, [r12], r1
-    vld1.u8     {d16}, [r12], r1
-    vld1.u8     {d18}, [r12], r1
-    vld1.u8     {d20}, [r12]
-
+    vdup.u8     q1, r3                      ; duplicate limit
     sub         r3, r2, #4                  ; move v pointer down by 4 columns
+
+    vld1.u8     {d6}, [r12], r1             ;load u data
     vld1.u8     {d7}, [r3], r1              ;load v data
+    vld1.u8     {d8}, [r12], r1
     vld1.u8     {d9}, [r3], r1
+    vld1.u8     {d10}, [r12], r1
     vld1.u8     {d11}, [r3], r1
+    vld1.u8     {d12}, [r12], r1
     vld1.u8     {d13}, [r3], r1
+    vld1.u8     {d14}, [r12], r1
     vld1.u8     {d15}, [r3], r1
+    vld1.u8     {d16}, [r12], r1
     vld1.u8     {d17}, [r3], r1
+    vld1.u8     {d18}, [r12], r1
     vld1.u8     {d19}, [r3], r1
+    vld1.u8     {d20}, [r12]
     vld1.u8     {d21}, [r3]
 
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    ldr        r12, [sp, #4]               ; load thresh
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -248,6 +241,8 @@
     vtrn.32     q5, q9
     vtrn.32     q6, q10
 
+    vdup.u8     q2, r12                     ; duplicate thresh
+
     vtrn.16     q3, q5
     vtrn.16     q4, q6
     vtrn.16     q7, q9
@@ -258,18 +253,16 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-
     bl          vp8_loop_filter_neon
 
-    sub         r0, r0, #2
-    sub         r2, r2, #2
-
     vswp        d12, d11
     vswp        d16, d13
     vswp        d14, d12
     vswp        d16, d15
 
+    sub         r0, r0, #2
+    sub         r2, r2, #2
+
     ;store op1, op0, oq0, oq1
     vst4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
     vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r2], r1
@@ -288,7 +281,7 @@
     vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0]
     vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r2]
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_vertical_edge_uv_neon|
 
 ; void vp8_loop_filter_neon();
@@ -308,7 +301,6 @@
 ; q9    q2
 ; q10   q3
 |vp8_loop_filter_neon| PROC
-    adr         r12, lf_coeff
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -317,42 +309,44 @@
     vabd.u8     q14, q8, q7                 ; abs(q1 - q0)
     vabd.u8     q3, q9, q8                  ; abs(q2 - q1)
     vabd.u8     q4, q10, q9                 ; abs(q3 - q2)
-    vabd.u8     q9, q6, q7                  ; abs(p0 - q0)
 
     vmax.u8     q11, q11, q12
     vmax.u8     q12, q13, q14
     vmax.u8     q3, q3, q4
     vmax.u8     q15, q11, q12
 
+    vabd.u8     q9, q6, q7                  ; abs(p0 - q0)
+
     ; vp8_hevmask
     vcgt.u8     q13, q13, q2                ; (abs(p1 - p0) > thresh)*-1
     vcgt.u8     q14, q14, q2                ; (abs(q1 - q0) > thresh)*-1
     vmax.u8     q15, q15, q3
 
-    vadd.u8     q0, q0, q0                  ; flimit * 2
-    vadd.u8     q0, q0, q1                  ; flimit * 2 + limit
-    vcge.u8     q15, q1, q15
+    vmov.u8     q10, #0x80                   ; 0x80
 
     vabd.u8     q2, q5, q8                  ; a = abs(p1 - q1)
     vqadd.u8    q9, q9, q9                  ; b = abs(p0 - q0) * 2
-    vshr.u8     q2, q2, #1                  ; a = a / 2
-    vqadd.u8    q9, q9, q2                  ; a = b + a
-    vcge.u8     q9, q0, q9                  ; (a > flimit * 2 + limit) * -1
 
-    vld1.u8     {q0}, [r12]!
+    vcge.u8     q15, q1, q15
 
     ; vp8_filter() function
     ; convert to signed
-    veor        q7, q7, q0                  ; qs0
-    veor        q6, q6, q0                  ; ps0
-    veor        q5, q5, q0                  ; ps1
-    veor        q8, q8, q0                  ; qs1
+    veor        q7, q7, q10                 ; qs0
+    vshr.u8     q2, q2, #1                  ; a = a / 2
+    veor        q6, q6, q10                 ; ps0
 
-    vld1.u8     {q10}, [r12]!
+    veor        q5, q5, q10                 ; ps1
+    vqadd.u8    q9, q9, q2                  ; a = b + a
+
+    veor        q8, q8, q10                 ; qs1
+
+    vmov.u8     q10, #3                     ; #3
 
     vsubl.s8    q2, d14, d12                ; ( qs0 - ps0)
     vsubl.s8    q11, d15, d13
 
+    vcge.u8     q9, q0, q9                  ; (a > flimit * 2 + limit) * -1
+
     vmovl.u8    q4, d20
 
     vqsub.s8    q1, q5, q8                  ; vp8_filter = clamp(ps1-qs1)
@@ -367,7 +361,7 @@
     vaddw.s8    q2, q2, d2
     vaddw.s8    q11, q11, d3
 
-    vld1.u8     {q9}, [r12]!
+    vmov.u8     q9, #4                      ; #4
 
     ; vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d2, q2
@@ -379,29 +373,25 @@
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
     vshr.s8     q1, q1, #3                  ; Filter1 >>= 3
 
+
     vqadd.s8    q11, q6, q2                 ; u = clamp(ps0 + Filter2)
     vqsub.s8    q10, q7, q1                 ; u = clamp(qs0 - Filter1)
 
     ; outer tap adjustments: ++vp8_filter >> 1
     vrshr.s8    q1, q1, #1
     vbic        q1, q1, q14                 ; vp8_filter &= ~hev
-
+    vmov.u8     q0, #0x80                   ; 0x80
     vqadd.s8    q13, q5, q1                 ; u = clamp(ps1 + vp8_filter)
     vqsub.s8    q12, q8, q1                 ; u = clamp(qs1 - vp8_filter)
 
-    veor        q5, q13, q0                 ; *op1 = u^0x80
     veor        q6, q11, q0                 ; *op0 = u^0x80
     veor        q7, q10, q0                 ; *oq0 = u^0x80
+    veor        q5, q13, q0                 ; *op1 = u^0x80
     veor        q8, q12, q0                 ; *oq1 = u^0x80
 
     bx          lr
     ENDP        ; |vp8_loop_filter_horizontal_edge_y_neon|
 
-    AREA    loopfilter_dat, DATA, READONLY
-lf_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
-    DCD     0x01010101, 0x01010101, 0x01010101, 0x01010101
+;-----------------
 
     END

media/libvpx/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm

@@ -9,108 +9,109 @@
 ;
 
 
-    EXPORT  |vp8_loop_filter_simple_horizontal_edge_neon|
+    ;EXPORT  |vp8_loop_filter_simple_horizontal_edge_neon|
+    EXPORT  |vp8_loop_filter_bhs_neon|
+    EXPORT  |vp8_loop_filter_mbhs_neon|
     ARM
-    REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;Note: flimit, limit, and thresh shpuld be positive numbers. All 16 elements in flimit
-;are equal. So, in the code, only one load is needed
-;for flimit. Same way applies to limit and thresh.
-; r0    unsigned char *s,
-; r1    int p, //pitch
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; stack(r4) const signed char *thresh,
-; //stack(r5)   int count --unused
+
+; r0    unsigned char *s, PRESERVE
+; r1    int p, PRESERVE
+; q1    limit, PRESERVE
 
 |vp8_loop_filter_simple_horizontal_edge_neon| PROC
-    sub         r0, r0, r1, lsl #1          ; move src pointer down by 2 lines
 
-    adr         r12, lfhy_coeff
-    vld1.u8     {q5}, [r0], r1              ; p1
-    vld1.s8     {d2[], d3[]}, [r2]          ; flimit
-    vld1.s8     {d26[], d27[]}, [r3]        ; limit -> q13
-    vld1.u8     {q6}, [r0], r1              ; p0
-    vld1.u8     {q0}, [r12]!                ; 0x80
-    vld1.u8     {q7}, [r0], r1              ; q0
-    vld1.u8     {q10}, [r12]!               ; 0x03
-    vld1.u8     {q8}, [r0]                  ; q1
+    sub         r3, r0, r1, lsl #1          ; move src pointer down by 2 lines
+
+    vld1.u8     {q7}, [r0@128], r1          ; q0
+    vld1.u8     {q5}, [r3@128], r1          ; p0
+    vld1.u8     {q8}, [r0@128]              ; q1
+    vld1.u8     {q6}, [r3@128]              ; p1
 
-    ;vp8_filter_mask() function
     vabd.u8     q15, q6, q7                 ; abs(p0 - q0)
     vabd.u8     q14, q5, q8                 ; abs(p1 - q1)
+
     vqadd.u8    q15, q15, q15               ; abs(p0 - q0) * 2
     vshr.u8     q14, q14, #1                ; abs(p1 - q1) / 2
+    vmov.u8     q0, #0x80                   ; 0x80
+    vmov.s16    q13, #3
     vqadd.u8    q15, q15, q14               ; abs(p0 - q0) * 2 + abs(p1 - q1) / 2
 
-    ;vp8_filter() function
     veor        q7, q7, q0                  ; qs0: q0 offset to convert to a signed value
     veor        q6, q6, q0                  ; ps0: p0 offset to convert to a signed value
     veor        q5, q5, q0                  ; ps1: p1 offset to convert to a signed value
     veor        q8, q8, q0                  ; qs1: q1 offset to convert to a signed value
 
-    vadd.u8     q1, q1, q1                  ; flimit * 2
-    vadd.u8     q1, q1, q13                 ; flimit * 2 + limit
-    vcge.u8     q15, q1, q15                ; (abs(p0 - q0)*2 + abs(p1-q1)/2 > flimit*2 + limit)*-1
+    vcge.u8     q15, q1, q15                ; (abs(p0 - q0)*2 + abs(p1-q1)/2 > limit)*-1
 
-;;;;;;;;;;
-    ;vqsub.s8   q2, q7, q6                  ; ( qs0 - ps0)
     vsubl.s8    q2, d14, d12                ; ( qs0 - ps0)
     vsubl.s8    q3, d15, d13
 
     vqsub.s8    q4, q5, q8                  ; q4: vp8_filter = vp8_signed_char_clamp(ps1-qs1)
 
-    ;vmul.i8    q2, q2, q10                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q11, q2, q2                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q12, q3, q3
+    vmul.s16    q2, q2, q13                 ;  3 * ( qs0 - ps0)
+    vmul.s16    q3, q3, q13
 
-    vld1.u8     {q9}, [r12]!                ; 0x04
-
-    vadd.s16    q2, q2, q11
-    vadd.s16    q3, q3, q12
+    vmov.u8     q10, #0x03                  ; 0x03
+    vmov.u8     q9, #0x04                   ; 0x04
 
     vaddw.s8    q2, q2, d8                  ; vp8_filter + 3 * ( qs0 - ps0)
     vaddw.s8    q3, q3, d9
 
-    ;vqadd.s8   q4, q4, q2                  ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d8, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d9, q3
-;;;;;;;;;;;;;
 
-    vand        q4, q4, q15                 ; vp8_filter &= mask
+    vand        q14, q4, q15                ; vp8_filter &= mask
 
-    vqadd.s8    q2, q4, q10                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
-    vqadd.s8    q4, q4, q9                  ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
+    vqadd.s8    q2, q14, q10                ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
+    vqadd.s8    q3, q14, q9                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
-    vshr.s8     q4, q4, #3                  ; Filter1 >>= 3
+    vshr.s8     q4, q3, #3                  ; Filter1 >>= 3
 
-    sub         r0, r0, r1, lsl #1
+    sub         r0, r0, r1
 
     ;calculate output
     vqadd.s8    q11, q6, q2                 ; u = vp8_signed_char_clamp(ps0 + Filter2)
     vqsub.s8    q10, q7, q4                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
 
-    add         r3, r0, r1
-
     veor        q6, q11, q0                 ; *op0 = u^0x80
     veor        q7, q10, q0                 ; *oq0 = u^0x80
 
-    vst1.u8     {q6}, [r0]                  ; store op0
-    vst1.u8     {q7}, [r3]                  ; store oq0
+    vst1.u8     {q6}, [r3@128]              ; store op0
+    vst1.u8     {q7}, [r0@128]              ; store oq0
 
     bx          lr
     ENDP        ; |vp8_loop_filter_simple_horizontal_edge_neon|
 
-;-----------------
-    AREA    hloopfiltery_dat, DATA, READWRITE           ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 16 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-lfhy_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
+
+|vp8_loop_filter_bhs_neon| PROC
+    push        {r4, lr}
+    ldrb        r3, [r2]                    ; load blim from mem
+    vdup.s8     q1, r3                      ; duplicate blim
+
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 4 * y_stride
+    bl          vp8_loop_filter_simple_horizontal_edge_neon
+    ; vp8_loop_filter_simple_horizontal_edge_neon preserves r0, r1 and q1
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 8* y_stride
+    bl          vp8_loop_filter_simple_horizontal_edge_neon
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 12 * y_stride
+    pop         {r4, lr}
+    b           vp8_loop_filter_simple_horizontal_edge_neon
+    ENDP        ;|vp8_loop_filter_bhs_neon|
+
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
+
+|vp8_loop_filter_mbhs_neon| PROC
+    ldrb        r3, [r2]                   ; load blim from mem
+    vdup.s8     q1, r3                     ; duplicate mblim
+    b           vp8_loop_filter_simple_horizontal_edge_neon
+    ENDP        ;|vp8_loop_filter_bhs_neon|
 
     END

media/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm

@@ -9,60 +9,54 @@
 ;
 
 
-    EXPORT  |vp8_loop_filter_simple_vertical_edge_neon|
+    ;EXPORT  |vp8_loop_filter_simple_vertical_edge_neon|
+    EXPORT |vp8_loop_filter_bvs_neon|
+    EXPORT |vp8_loop_filter_mbvs_neon|
     ARM
-    REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;Note: flimit, limit, and thresh should be positive numbers. All 16 elements in flimit
-;are equal. So, in the code, only one load is needed
-;for flimit. Same way applies to limit and thresh.
-; r0    unsigned char *s,
-; r1    int p, //pitch
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; stack(r4) const signed char *thresh,
-; //stack(r5)   int count --unused
+
+; r0    unsigned char *s, PRESERVE
+; r1    int p, PRESERVE
+; q1    limit, PRESERVE
 
 |vp8_loop_filter_simple_vertical_edge_neon| PROC
     sub         r0, r0, #2                  ; move src pointer down by 2 columns
+    add         r12, r1, r1
+    add         r3, r0, r1
 
-    vld4.8      {d6[0], d7[0], d8[0], d9[0]}, [r0], r1
-    vld1.s8     {d2[], d3[]}, [r2]          ; flimit
-    vld1.s8     {d26[], d27[]}, [r3]        ; limit -> q13
-    vld4.8      {d6[1], d7[1], d8[1], d9[1]}, [r0], r1
-    adr         r12, vlfy_coeff
-    vld4.8      {d6[2], d7[2], d8[2], d9[2]}, [r0], r1
-    vld4.8      {d6[3], d7[3], d8[3], d9[3]}, [r0], r1
-    vld4.8      {d6[4], d7[4], d8[4], d9[4]}, [r0], r1
-    vld4.8      {d6[5], d7[5], d8[5], d9[5]}, [r0], r1
-    vld4.8      {d6[6], d7[6], d8[6], d9[6]}, [r0], r1
-    vld4.8      {d6[7], d7[7], d8[7], d9[7]}, [r0], r1
+    vld4.8      {d6[0], d7[0], d8[0], d9[0]}, [r0], r12
+    vld4.8      {d6[1], d7[1], d8[1], d9[1]}, [r3], r12
+    vld4.8      {d6[2], d7[2], d8[2], d9[2]}, [r0], r12
+    vld4.8      {d6[3], d7[3], d8[3], d9[3]}, [r3], r12
+    vld4.8      {d6[4], d7[4], d8[4], d9[4]}, [r0], r12
+    vld4.8      {d6[5], d7[5], d8[5], d9[5]}, [r3], r12
+    vld4.8      {d6[6], d7[6], d8[6], d9[6]}, [r0], r12
+    vld4.8      {d6[7], d7[7], d8[7], d9[7]}, [r3], r12
 
-    vld4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
-    vld1.u8     {q0}, [r12]!                ; 0x80
-    vld4.8      {d10[1], d11[1], d12[1], d13[1]}, [r0], r1
-    vld1.u8     {q11}, [r12]!               ; 0x03
-    vld4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r1
-    vld1.u8     {q12}, [r12]!               ; 0x04
-    vld4.8      {d10[3], d11[3], d12[3], d13[3]}, [r0], r1
-    vld4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r1
-    vld4.8      {d10[5], d11[5], d12[5], d13[5]}, [r0], r1
-    vld4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r1
-    vld4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0], r1
+    vld4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r12
+    vld4.8      {d10[1], d11[1], d12[1], d13[1]}, [r3], r12
+    vld4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r12
+    vld4.8      {d10[3], d11[3], d12[3], d13[3]}, [r3], r12
+    vld4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r12
+    vld4.8      {d10[5], d11[5], d12[5], d13[5]}, [r3], r12
+    vld4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r12
+    vld4.8      {d10[7], d11[7], d12[7], d13[7]}, [r3]
 
     vswp        d7, d10
     vswp        d12, d9
-    ;vswp       q4, q5                      ; p1:q3, p0:q5, q0:q4, q1:q6
 
     ;vp8_filter_mask() function
     ;vp8_hevmask() function
     sub         r0, r0, r1, lsl #4
     vabd.u8     q15, q5, q4                 ; abs(p0 - q0)
     vabd.u8     q14, q3, q6                 ; abs(p1 - q1)
+
     vqadd.u8    q15, q15, q15               ; abs(p0 - q0) * 2
     vshr.u8     q14, q14, #1                ; abs(p1 - q1) / 2
+    vmov.u8     q0, #0x80                   ; 0x80
+    vmov.s16    q11, #3
     vqadd.u8    q15, q15, q14               ; abs(p0 - q0) * 2 + abs(p1 - q1) / 2
 
     veor        q4, q4, q0                  ; qs0: q0 offset to convert to a signed value
@@ -70,88 +64,91 @@
     veor        q3, q3, q0                  ; ps1: p1 offset to convert to a signed value
     veor        q6, q6, q0                  ; qs1: q1 offset to convert to a signed value
 
-    vadd.u8     q1, q1, q1                  ; flimit * 2
-    vadd.u8     q1, q1, q13                 ; flimit * 2 + limit
     vcge.u8     q15, q1, q15                ; abs(p0 - q0)*2 + abs(p1-q1)/2 > flimit*2 + limit)*-1
 
-    ;vp8_filter() function
-;;;;;;;;;;
-    ;vqsub.s8   q2, q5, q4                  ; ( qs0 - ps0)
     vsubl.s8    q2, d8, d10                 ; ( qs0 - ps0)
     vsubl.s8    q13, d9, d11
 
-    vqsub.s8    q1, q3, q6                  ; vp8_filter = vp8_signed_char_clamp(ps1-qs1)
+    vqsub.s8    q14, q3, q6                  ; vp8_filter = vp8_signed_char_clamp(ps1-qs1)
 
-    ;vmul.i8    q2, q2, q11                 ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
-    vadd.s16    q10, q2, q2                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q14, q13, q13
-    vadd.s16    q2, q2, q10
-    vadd.s16    q13, q13, q14
+    vmul.s16    q2, q2, q11                 ;  3 * ( qs0 - ps0)
+    vmul.s16    q13, q13, q11
 
-    ;vqadd.s8   q1, q1, q2
-    vaddw.s8    q2, q2, d2                  ; vp8_filter + 3 * ( qs0 - ps0)
-    vaddw.s8    q13, q13, d3
+    vmov.u8     q11, #0x03                  ; 0x03
+    vmov.u8     q12, #0x04                  ; 0x04
 
-    vqmovn.s16  d2, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
-    vqmovn.s16  d3, q13
+    vaddw.s8    q2, q2, d28                  ; vp8_filter + 3 * ( qs0 - ps0)
+    vaddw.s8    q13, q13, d29
+
+    vqmovn.s16  d28, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
+    vqmovn.s16  d29, q13
 
     add         r0, r0, #1
-    add         r2, r0, r1
-;;;;;;;;;;;
+    add         r3, r0, r1
 
-    vand        q1, q1, q15                 ; vp8_filter &= mask
+    vand        q14, q14, q15                 ; vp8_filter &= mask
 
-    vqadd.s8    q2, q1, q11                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
-    vqadd.s8    q1, q1, q12                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
+    vqadd.s8    q2, q14, q11                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
+    vqadd.s8    q3, q14, q12                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
-    vshr.s8     q1, q1, #3                  ; Filter1 >>= 3
+    vshr.s8     q14, q3, #3                  ; Filter1 >>= 3
 
     ;calculate output
-    vqsub.s8    q10, q4, q1                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
     vqadd.s8    q11, q5, q2                 ; u = vp8_signed_char_clamp(ps0 + Filter2)
+    vqsub.s8    q10, q4, q14                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
 
-    veor        q7, q10, q0                 ; *oq0 = u^0x80
     veor        q6, q11, q0                 ; *op0 = u^0x80
-
-    add         r3, r2, r1
+    veor        q7, q10, q0                 ; *oq0 = u^0x80
+    add         r12, r1, r1
     vswp        d13, d14
-    add         r12, r3, r1
 
     ;store op1, op0, oq0, oq1
-    vst2.8      {d12[0], d13[0]}, [r0]
-    vst2.8      {d12[1], d13[1]}, [r2]
-    vst2.8      {d12[2], d13[2]}, [r3]
-    vst2.8      {d12[3], d13[3]}, [r12], r1
-    add         r0, r12, r1
-    vst2.8      {d12[4], d13[4]}, [r12]
-    vst2.8      {d12[5], d13[5]}, [r0], r1
-    add         r2, r0, r1
-    vst2.8      {d12[6], d13[6]}, [r0]
-    vst2.8      {d12[7], d13[7]}, [r2], r1
-    add         r3, r2, r1
-    vst2.8      {d14[0], d15[0]}, [r2]
-    vst2.8      {d14[1], d15[1]}, [r3], r1
-    add         r12, r3, r1
-    vst2.8      {d14[2], d15[2]}, [r3]
-    vst2.8      {d14[3], d15[3]}, [r12], r1
-    add         r0, r12, r1
-    vst2.8      {d14[4], d15[4]}, [r12]
-    vst2.8      {d14[5], d15[5]}, [r0], r1
-    add         r2, r0, r1
-    vst2.8      {d14[6], d15[6]}, [r0]
-    vst2.8      {d14[7], d15[7]}, [r2]
+    vst2.8      {d12[0], d13[0]}, [r0], r12
+    vst2.8      {d12[1], d13[1]}, [r3], r12
+    vst2.8      {d12[2], d13[2]}, [r0], r12
+    vst2.8      {d12[3], d13[3]}, [r3], r12
+    vst2.8      {d12[4], d13[4]}, [r0], r12
+    vst2.8      {d12[5], d13[5]}, [r3], r12
+    vst2.8      {d12[6], d13[6]}, [r0], r12
+    vst2.8      {d12[7], d13[7]}, [r3], r12
+    vst2.8      {d14[0], d15[0]}, [r0], r12
+    vst2.8      {d14[1], d15[1]}, [r3], r12
+    vst2.8      {d14[2], d15[2]}, [r0], r12
+    vst2.8      {d14[3], d15[3]}, [r3], r12
+    vst2.8      {d14[4], d15[4]}, [r0], r12
+    vst2.8      {d14[5], d15[5]}, [r3], r12
+    vst2.8      {d14[6], d15[6]}, [r0], r12
+    vst2.8      {d14[7], d15[7]}, [r3]
 
     bx          lr
     ENDP        ; |vp8_loop_filter_simple_vertical_edge_neon|
 
-;-----------------
-    AREA    vloopfiltery_dat, DATA, READWRITE           ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 16 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-vlfy_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
 
+|vp8_loop_filter_bvs_neon| PROC
+    push        {r4, lr}
+    ldrb        r3, [r2]                   ; load blim from mem
+    mov         r4, r0
+    add         r0, r0, #4
+    vdup.s8     q1, r3                     ; duplicate blim
+    bl          vp8_loop_filter_simple_vertical_edge_neon
+    ; vp8_loop_filter_simple_vertical_edge_neon preserves  r1 and q1
+    add         r0, r4, #8
+    bl          vp8_loop_filter_simple_vertical_edge_neon
+    add         r0, r4, #12
+    pop         {r4, lr}
+    b           vp8_loop_filter_simple_vertical_edge_neon
+    ENDP        ;|vp8_loop_filter_bvs_neon|
+
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
+
+|vp8_loop_filter_mbvs_neon| PROC
+    ldrb        r3, [r2]                   ; load mblim from mem
+    vdup.s8     q1, r3                     ; duplicate mblim
+    b           vp8_loop_filter_simple_vertical_edge_neon
+    ENDP        ;|vp8_loop_filter_bvs_neon|
     END

media/libvpx/vp8/common/arm/neon/mbloopfilter_neon.asm

@@ -14,155 +14,143 @@
     EXPORT  |vp8_mbloop_filter_vertical_edge_y_neon|
     EXPORT  |vp8_mbloop_filter_vertical_edge_uv_neon|
     ARM
-    REQUIRE8
-    PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
 
-; flimit, limit, and thresh should be positive numbers.
-; All 16 elements in these variables are equal.
-
 ; void vp8_mbloop_filter_horizontal_edge_y_neon(unsigned char *src, int pitch,
-;                                               const signed char *flimit,
-;                                               const signed char *limit,
-;                                               const signed char *thresh,
-;                                               int count)
+;                                               const unsigned char *blimit,
+;                                               const unsigned char *limit,
+;                                               const unsigned char *thresh)
 ; r0    unsigned char *src,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_mbloop_filter_horizontal_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, r1, lsl #2          ; move src pointer down by 4 lines
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    add         r1, r1, r1                  ; double stride
+    ldr         r12, [sp, #4]               ; load thresh
+    sub         r0, r0, r1, lsl #1          ; move src pointer down by 4 lines
+    vdup.u8     q2, r12                     ; thresh
+    add         r12, r0, r1,  lsr #1        ; move src pointer up by 1 line
 
-    vld1.u8     {q3}, [r0], r1              ; p3
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    vld1.u8     {q4}, [r0], r1              ; p2
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vld1.u8     {q5}, [r0], r1              ; p1
-    vld1.u8     {q6}, [r0], r1              ; p0
-    vld1.u8     {q7}, [r0], r1              ; q0
-    vld1.u8     {q8}, [r0], r1              ; q1
-    vld1.u8     {q9}, [r0], r1              ; q2
-    vld1.u8     {q10}, [r0], r1             ; q3
+    vld1.u8     {q3}, [r0@128], r1              ; p3
+    vld1.u8     {q4}, [r12@128], r1             ; p2
+    vld1.u8     {q5}, [r0@128], r1              ; p1
+    vld1.u8     {q6}, [r12@128], r1             ; p0
+    vld1.u8     {q7}, [r0@128], r1              ; q0
+    vld1.u8     {q8}, [r12@128], r1             ; q1
+    vld1.u8     {q9}, [r0@128], r1              ; q2
+    vld1.u8     {q10}, [r12@128], r1            ; q3
 
     bl          vp8_mbloop_filter_neon
 
-    sub         r0, r0, r1, lsl #3
-    add         r0, r0, r1
-    add         r2, r0, r1
-    add         r3, r2, r1
+    sub         r12, r12, r1, lsl #2
+    add         r0, r12, r1, lsr #1
 
-    vst1.u8     {q4}, [r0]                  ; store op2
-    vst1.u8     {q5}, [r2]                  ; store op1
-    vst1.u8     {q6}, [r3], r1              ; store op0
-    add         r12, r3, r1
-    vst1.u8     {q7}, [r3]                  ; store oq0
-    vst1.u8     {q8}, [r12], r1             ; store oq1
-    vst1.u8     {q9}, [r12]             ; store oq2
+    vst1.u8     {q4}, [r12@128],r1         ; store op2
+    vst1.u8     {q5}, [r0@128],r1          ; store op1
+    vst1.u8     {q6}, [r12@128], r1        ; store op0
+    vst1.u8     {q7}, [r0@128],r1          ; store oq0
+    vst1.u8     {q8}, [r12@128]            ; store oq1
+    vst1.u8     {q9}, [r0@128]             ; store oq2
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_horizontal_edge_y_neon|
 
 ; void vp8_mbloop_filter_horizontal_edge_uv_neon(unsigned char *u, int pitch,
-;                                                const signed char *flimit,
-;                                                const signed char *limit,
-;                                                const signed char *thresh,
+;                                                const unsigned char *blimit,
+;                                                const unsigned char *limit,
+;                                                const unsigned char *thresh,
 ;                                                unsigned char *v)
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
+
 |vp8_mbloop_filter_horizontal_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, r1, lsl #2          ; move u pointer down by 4 lines
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    ldr         r3, [sp, #8]                ; load v ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
-    sub         r3, r3, r1, lsl #2          ; move v pointer down by 4 lines
+    push        {lr}
+    ldr         r12, [sp, #4]                 ; load thresh
+    sub         r0, r0, r1, lsl #2            ; move u pointer down by 4 lines
+    vdup.u8     q2, r12                       ; thresh
+    ldr         r12, [sp, #8]                 ; load v ptr
+    sub         r12, r12, r1, lsl #2          ; move v pointer down by 4 lines
 
-    vld1.u8     {d6}, [r0], r1              ; p3
-    vld1.u8     {d7}, [r3], r1              ; p3
-    vld1.u8     {d8}, [r0], r1              ; p2
-    vld1.u8     {d9}, [r3], r1              ; p2
-    vld1.u8     {d10}, [r0], r1             ; p1
-    vld1.u8     {d11}, [r3], r1             ; p1
-    vld1.u8     {d12}, [r0], r1             ; p0
-    vld1.u8     {d13}, [r3], r1             ; p0
-    vld1.u8     {d14}, [r0], r1             ; q0
-    vld1.u8     {d15}, [r3], r1             ; q0
-    vld1.u8     {d16}, [r0], r1             ; q1
-    vld1.u8     {d17}, [r3], r1             ; q1
-    vld1.u8     {d18}, [r0], r1             ; q2
-    vld1.u8     {d19}, [r3], r1             ; q2
-    vld1.u8     {d20}, [r0], r1             ; q3
-    vld1.u8     {d21}, [r3], r1             ; q3
-
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.u8     {d6}, [r0@64], r1              ; p3
+    vld1.u8     {d7}, [r12@64], r1              ; p3
+    vld1.u8     {d8}, [r0@64], r1              ; p2
+    vld1.u8     {d9}, [r12@64], r1              ; p2
+    vld1.u8     {d10}, [r0@64], r1             ; p1
+    vld1.u8     {d11}, [r12@64], r1             ; p1
+    vld1.u8     {d12}, [r0@64], r1             ; p0
+    vld1.u8     {d13}, [r12@64], r1             ; p0
+    vld1.u8     {d14}, [r0@64], r1             ; q0
+    vld1.u8     {d15}, [r12@64], r1             ; q0
+    vld1.u8     {d16}, [r0@64], r1             ; q1
+    vld1.u8     {d17}, [r12@64], r1             ; q1
+    vld1.u8     {d18}, [r0@64], r1             ; q2
+    vld1.u8     {d19}, [r12@64], r1             ; q2
+    vld1.u8     {d20}, [r0@64], r1             ; q3
+    vld1.u8     {d21}, [r12@64], r1             ; q3
 
     bl          vp8_mbloop_filter_neon
 
     sub         r0, r0, r1, lsl #3
-    sub         r3, r3, r1, lsl #3
+    sub         r12, r12, r1, lsl #3
 
     add         r0, r0, r1
-    add         r3, r3, r1
+    add         r12, r12, r1
 
-    vst1.u8     {d8}, [r0], r1              ; store u op2
-    vst1.u8     {d9}, [r3], r1              ; store v op2
-    vst1.u8     {d10}, [r0], r1             ; store u op1
-    vst1.u8     {d11}, [r3], r1             ; store v op1
-    vst1.u8     {d12}, [r0], r1             ; store u op0
-    vst1.u8     {d13}, [r3], r1             ; store v op0
-    vst1.u8     {d14}, [r0], r1             ; store u oq0
-    vst1.u8     {d15}, [r3], r1             ; store v oq0
-    vst1.u8     {d16}, [r0], r1             ; store u oq1
-    vst1.u8     {d17}, [r3], r1             ; store v oq1
-    vst1.u8     {d18}, [r0], r1             ; store u oq2
-    vst1.u8     {d19}, [r3], r1             ; store v oq2
+    vst1.u8     {d8}, [r0@64], r1              ; store u op2
+    vst1.u8     {d9}, [r12@64], r1              ; store v op2
+    vst1.u8     {d10}, [r0@64], r1             ; store u op1
+    vst1.u8     {d11}, [r12@64], r1             ; store v op1
+    vst1.u8     {d12}, [r0@64], r1             ; store u op0
+    vst1.u8     {d13}, [r12@64], r1             ; store v op0
+    vst1.u8     {d14}, [r0@64], r1             ; store u oq0
+    vst1.u8     {d15}, [r12@64], r1             ; store v oq0
+    vst1.u8     {d16}, [r0@64], r1             ; store u oq1
+    vst1.u8     {d17}, [r12@64], r1             ; store v oq1
+    vst1.u8     {d18}, [r0@64], r1             ; store u oq2
+    vst1.u8     {d19}, [r12@64], r1             ; store v oq2
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_horizontal_edge_uv_neon|
 
 ; void vp8_mbloop_filter_vertical_edge_y_neon(unsigned char *src, int pitch,
-;                                             const signed char *flimit,
-;                                             const signed char *limit,
-;                                             const signed char *thresh,
-;                                             int count)
+;                                             const unsigned char *blimit,
+;                                             const unsigned char *limit,
+;                                             const unsigned char *thresh)
 ; r0    unsigned char *src,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_mbloop_filter_vertical_edge_y_neon| PROC
-    stmdb       sp!, {lr}
+    push        {lr}
+    ldr         r12, [sp, #4]               ; load thresh
     sub         r0, r0, #4                  ; move src pointer down by 4 columns
+    vdup.s8     q2, r12                     ; thresh
+    add         r12, r0, r1, lsl #3         ; move src pointer down by 8 lines
 
     vld1.u8     {d6}, [r0], r1              ; load first 8-line src data
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    vld1.u8     {d7}, [r12], r1             ; load second 8-line src data
     vld1.u8     {d8}, [r0], r1
-    sub         sp, sp, #32
+    vld1.u8     {d9}, [r12], r1
     vld1.u8     {d10}, [r0], r1
+    vld1.u8     {d11}, [r12], r1
     vld1.u8     {d12}, [r0], r1
+    vld1.u8     {d13}, [r12], r1
     vld1.u8     {d14}, [r0], r1
+    vld1.u8     {d15}, [r12], r1
     vld1.u8     {d16}, [r0], r1
+    vld1.u8     {d17}, [r12], r1
     vld1.u8     {d18}, [r0], r1
+    vld1.u8     {d19}, [r12], r1
     vld1.u8     {d20}, [r0], r1
-
-    vld1.u8     {d7}, [r0], r1              ; load second 8-line src data
-    vld1.u8     {d9}, [r0], r1
-    vld1.u8     {d11}, [r0], r1
-    vld1.u8     {d13}, [r0], r1
-    vld1.u8     {d15}, [r0], r1
-    vld1.u8     {d17}, [r0], r1
-    vld1.u8     {d19}, [r0], r1
-    vld1.u8     {d21}, [r0], r1
+    vld1.u8     {d21}, [r12], r1
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -180,133 +168,11 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    mov         r12, sp
-    vst1.u8     {q3}, [r12]!
-    vst1.u8     {q10}, [r12]!
-
-    bl          vp8_mbloop_filter_neon
-
-    sub         r0, r0, r1, lsl #4
-
-    add         r2, r0, r1
-
-    add         r3, r2, r1
-
-    vld1.u8     {q3}, [sp]!
-    vld1.u8     {q10}, [sp]!
-
-    ;transpose to 16x8 matrix
-    vtrn.32     q3, q7
-    vtrn.32     q4, q8
-    vtrn.32     q5, q9
-    vtrn.32     q6, q10
-    add         r12, r3, r1
-
-    vtrn.16     q3, q5
-    vtrn.16     q4, q6
-    vtrn.16     q7, q9
-    vtrn.16     q8, q10
-
-    vtrn.8      q3, q4
-    vtrn.8      q5, q6
-    vtrn.8      q7, q8
-    vtrn.8      q9, q10
-
-    ;store op2, op1, op0, oq0, oq1, oq2
-    vst1.8      {d6}, [r0]
-    vst1.8      {d8}, [r2]
-    vst1.8      {d10}, [r3]
-    vst1.8      {d12}, [r12], r1
-    add         r0, r12, r1
-    vst1.8      {d14}, [r12]
-    vst1.8      {d16}, [r0], r1
-    add         r2, r0, r1
-    vst1.8      {d18}, [r0]
-    vst1.8      {d20}, [r2], r1
-    add         r3, r2, r1
-    vst1.8      {d7}, [r2]
-    vst1.8      {d9}, [r3], r1
-    add         r12, r3, r1
-    vst1.8      {d11}, [r3]
-    vst1.8      {d13}, [r12], r1
-    add         r0, r12, r1
-    vst1.8      {d15}, [r12]
-    vst1.8      {d17}, [r0], r1
-    add         r2, r0, r1
-    vst1.8      {d19}, [r0]
-    vst1.8      {d21}, [r2]
-
-    ldmia       sp!, {pc}
-    ENDP        ; |vp8_mbloop_filter_vertical_edge_y_neon|
-
-; void vp8_mbloop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch,
-;                                              const signed char *flimit,
-;                                              const signed char *limit,
-;                                              const signed char *thresh,
-;                                              unsigned char *v)
-; r0    unsigned char *u,
-; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  unsigned char *v
-|vp8_mbloop_filter_vertical_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, #4                  ; move src pointer down by 4 columns
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    ldr         r3, [sp, #8]                ; load v ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
-
-    sub         r3, r3, #4                  ; move v pointer down by 4 columns
-
-    vld1.u8     {d6}, [r0], r1              ;load u data
-    vld1.u8     {d7}, [r3], r1              ;load v data
-    vld1.u8     {d8}, [r0], r1
-    vld1.u8     {d9}, [r3], r1
-    vld1.u8     {d10}, [r0], r1
-    vld1.u8     {d11}, [r3], r1
-    vld1.u8     {d12}, [r0], r1
-    vld1.u8     {d13}, [r3], r1
-    vld1.u8     {d14}, [r0], r1
-    vld1.u8     {d15}, [r3], r1
-    vld1.u8     {d16}, [r0], r1
-    vld1.u8     {d17}, [r3], r1
-    vld1.u8     {d18}, [r0], r1
-    vld1.u8     {d19}, [r3], r1
-    vld1.u8     {d20}, [r0], r1
-    vld1.u8     {d21}, [r3], r1
-
-    ;transpose to 8x16 matrix
-    vtrn.32     q3, q7
-    vtrn.32     q4, q8
-    vtrn.32     q5, q9
-    vtrn.32     q6, q10
-
-    vtrn.16     q3, q5
-    vtrn.16     q4, q6
-    vtrn.16     q7, q9
-    vtrn.16     q8, q10
-
-    vtrn.8      q3, q4
-    vtrn.8      q5, q6
-    vtrn.8      q7, q8
-    vtrn.8      q9, q10
-
-    sub         sp, sp, #32
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    mov         r12, sp
-    vst1.u8     {q3}, [r12]!
-    vst1.u8     {q10}, [r12]!
-
-    bl          vp8_mbloop_filter_neon
-
     sub         r0, r0, r1, lsl #3
-    sub         r3, r3, r1, lsl #3
 
-    vld1.u8     {q3}, [sp]!
-    vld1.u8     {q10}, [sp]!
+    bl          vp8_mbloop_filter_neon
+
+    sub         r12, r12, r1, lsl #3
 
     ;transpose to 16x8 matrix
     vtrn.32     q3, q7
@@ -326,23 +192,118 @@
 
     ;store op2, op1, op0, oq0, oq1, oq2
     vst1.8      {d6}, [r0], r1
-    vst1.8      {d7}, [r3], r1
+    vst1.8      {d7}, [r12], r1
     vst1.8      {d8}, [r0], r1
-    vst1.8      {d9}, [r3], r1
+    vst1.8      {d9}, [r12], r1
     vst1.8      {d10}, [r0], r1
-    vst1.8      {d11}, [r3], r1
+    vst1.8      {d11}, [r12], r1
     vst1.8      {d12}, [r0], r1
-    vst1.8      {d13}, [r3], r1
+    vst1.8      {d13}, [r12], r1
     vst1.8      {d14}, [r0], r1
-    vst1.8      {d15}, [r3], r1
+    vst1.8      {d15}, [r12], r1
     vst1.8      {d16}, [r0], r1
-    vst1.8      {d17}, [r3], r1
+    vst1.8      {d17}, [r12], r1
     vst1.8      {d18}, [r0], r1
-    vst1.8      {d19}, [r3], r1
-    vst1.8      {d20}, [r0], r1
-    vst1.8      {d21}, [r3], r1
+    vst1.8      {d19}, [r12], r1
+    vst1.8      {d20}, [r0]
+    vst1.8      {d21}, [r12]
 
-    ldmia       sp!, {pc}
+    pop         {pc}
+    ENDP        ; |vp8_mbloop_filter_vertical_edge_y_neon|
+
+; void vp8_mbloop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch,
+;                                              const unsigned char *blimit,
+;                                              const unsigned char *limit,
+;                                              const unsigned char *thresh,
+;                                              unsigned char *v)
+; r0    unsigned char *u,
+; r1    int pitch,
+; r2    const signed char *flimit,
+; r3    const signed char *limit,
+; sp    const signed char *thresh,
+; sp+4  unsigned char *v
+|vp8_mbloop_filter_vertical_edge_uv_neon| PROC
+    push        {lr}
+    ldr         r12, [sp, #4]               ; load thresh
+    sub         r0, r0, #4                  ; move u pointer down by 4 columns
+    vdup.u8     q2, r12                     ; thresh
+    ldr         r12, [sp, #8]               ; load v ptr
+    sub         r12, r12, #4                ; move v pointer down by 4 columns
+
+    vld1.u8     {d6}, [r0], r1              ;load u data
+    vld1.u8     {d7}, [r12], r1             ;load v data
+    vld1.u8     {d8}, [r0], r1
+    vld1.u8     {d9}, [r12], r1
+    vld1.u8     {d10}, [r0], r1
+    vld1.u8     {d11}, [r12], r1
+    vld1.u8     {d12}, [r0], r1
+    vld1.u8     {d13}, [r12], r1
+    vld1.u8     {d14}, [r0], r1
+    vld1.u8     {d15}, [r12], r1
+    vld1.u8     {d16}, [r0], r1
+    vld1.u8     {d17}, [r12], r1
+    vld1.u8     {d18}, [r0], r1
+    vld1.u8     {d19}, [r12], r1
+    vld1.u8     {d20}, [r0], r1
+    vld1.u8     {d21}, [r12], r1
+
+    ;transpose to 8x16 matrix
+    vtrn.32     q3, q7
+    vtrn.32     q4, q8
+    vtrn.32     q5, q9
+    vtrn.32     q6, q10
+
+    vtrn.16     q3, q5
+    vtrn.16     q4, q6
+    vtrn.16     q7, q9
+    vtrn.16     q8, q10
+
+    vtrn.8      q3, q4
+    vtrn.8      q5, q6
+    vtrn.8      q7, q8
+    vtrn.8      q9, q10
+
+    sub         r0, r0, r1, lsl #3
+
+    bl          vp8_mbloop_filter_neon
+
+    sub         r12, r12, r1, lsl #3
+
+    ;transpose to 16x8 matrix
+    vtrn.32     q3, q7
+    vtrn.32     q4, q8
+    vtrn.32     q5, q9
+    vtrn.32     q6, q10
+
+    vtrn.16     q3, q5
+    vtrn.16     q4, q6
+    vtrn.16     q7, q9
+    vtrn.16     q8, q10
+
+    vtrn.8      q3, q4
+    vtrn.8      q5, q6
+    vtrn.8      q7, q8
+    vtrn.8      q9, q10
+
+    ;store op2, op1, op0, oq0, oq1, oq2
+    vst1.8      {d6}, [r0], r1
+    vst1.8      {d7}, [r12], r1
+    vst1.8      {d8}, [r0], r1
+    vst1.8      {d9}, [r12], r1
+    vst1.8      {d10}, [r0], r1
+    vst1.8      {d11}, [r12], r1
+    vst1.8      {d12}, [r0], r1
+    vst1.8      {d13}, [r12], r1
+    vst1.8      {d14}, [r0], r1
+    vst1.8      {d15}, [r12], r1
+    vst1.8      {d16}, [r0], r1
+    vst1.8      {d17}, [r12], r1
+    vst1.8      {d18}, [r0], r1
+    vst1.8      {d19}, [r12], r1
+    vst1.8      {d20}, [r0]
+    vst1.8      {d21}, [r12]
+
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_vertical_edge_uv_neon|
 
 ; void vp8_mbloop_filter_neon()
@@ -350,41 +311,33 @@
 ; functions do the necessary load, transpose (if necessary), preserve (if
 ; necessary) and store.
 
-; TODO:
-; The vertical filter writes p3/q3 back out because two 4 element writes are
-; much simpler than ordering and writing two 3 element sets (or three 2 elements
-; sets, or whichever other combinations are possible).
-; If we can preserve q3 and q10, the vertical filter will be able to avoid
-; storing those values on the stack and reading them back after the filter.
-
 ; r0,r1 PRESERVE
-; r2    flimit
-; r3    PRESERVE
-; q1    limit
+; r2    mblimit
+; r3    limit
+
 ; q2    thresh
-; q3    p3
+; q3    p3 PRESERVE
 ; q4    p2
 ; q5    p1
 ; q6    p0
 ; q7    q0
 ; q8    q1
 ; q9    q2
-; q10   q3
+; q10   q3 PRESERVE
 
 |vp8_mbloop_filter_neon| PROC
-    adr         r12, mblf_coeff
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
     vabd.u8     q12, q4, q5                 ; abs(p2 - p1)
     vabd.u8     q13, q5, q6                 ; abs(p1 - p0)
     vabd.u8     q14, q8, q7                 ; abs(q1 - q0)
-    vabd.u8     q3, q9, q8                  ; abs(q2 - q1)
+    vabd.u8     q1, q9, q8                  ; abs(q2 - q1)
     vabd.u8     q0, q10, q9                 ; abs(q3 - q2)
 
     vmax.u8     q11, q11, q12
     vmax.u8     q12, q13, q14
-    vmax.u8     q3, q3, q0
+    vmax.u8     q1, q1, q0
     vmax.u8     q15, q11, q12
 
     vabd.u8     q12, q6, q7                 ; abs(p0 - q0)
@@ -392,51 +345,53 @@
     ; vp8_hevmask
     vcgt.u8     q13, q13, q2                ; (abs(p1 - p0) > thresh) * -1
     vcgt.u8     q14, q14, q2                ; (abs(q1 - q0) > thresh) * -1
-    vmax.u8     q15, q15, q3
+    vmax.u8     q15, q15, q1
 
-    vld1.s8     {d4[], d5[]}, [r2]          ; flimit
+    vdup.u8     q1, r3                      ; limit
+    vdup.u8     q2, r2                      ; mblimit
 
-    vld1.u8     {q0}, [r12]!
+    vmov.u8     q0, #0x80                   ; 0x80
 
-    vadd.u8     q2, q2, q2                  ; flimit * 2
-    vadd.u8     q2, q2, q1                  ; flimit * 2 +  limit
     vcge.u8     q15, q1, q15
 
     vabd.u8     q1, q5, q8                  ; a = abs(p1 - q1)
     vqadd.u8    q12, q12, q12               ; b = abs(p0 - q0) * 2
-    vshr.u8     q1, q1, #1                  ; a = a / 2
-    vqadd.u8    q12, q12, q1                ; a = b + a
-    vcge.u8     q12, q2, q12                ; (a > flimit * 2 + limit) * -1
+    vmov.u16    q11, #3                     ; #3
 
     ; vp8_filter
     ; convert to signed
     veor        q7, q7, q0                  ; qs0
+    vshr.u8     q1, q1, #1                  ; a = a / 2
     veor        q6, q6, q0                  ; ps0
     veor        q5, q5, q0                  ; ps1
+
+    vqadd.u8    q12, q12, q1                ; a = b + a
+
     veor        q8, q8, q0                  ; qs1
     veor        q4, q4, q0                  ; ps2
     veor        q9, q9, q0                  ; qs2
 
     vorr        q14, q13, q14               ; vp8_hevmask
 
+    vcge.u8     q12, q2, q12                ; (a > flimit * 2 + limit) * -1
+
     vsubl.s8    q2, d14, d12                ; qs0 - ps0
     vsubl.s8    q13, d15, d13
 
     vqsub.s8    q1, q5, q8                  ; vp8_filter = clamp(ps1-qs1)
 
-    vadd.s16    q10, q2, q2                 ; 3 * (qs0 - ps0)
-    vadd.s16    q11, q13, q13
+    vmul.i16    q2, q2, q11                 ; 3 * ( qs0 - ps0)
+
     vand        q15, q15, q12               ; vp8_filter_mask
 
-    vadd.s16    q2, q2, q10
-    vadd.s16    q13, q13, q11
+    vmul.i16    q13, q13, q11
 
-    vld1.u8     {q12}, [r12]!               ; #3
+    vmov.u8     q12, #3                     ; #3
 
     vaddw.s8    q2, q2, d2                  ; vp8_filter + 3 * ( qs0 - ps0)
     vaddw.s8    q13, q13, d3
 
-    vld1.u8     {q11}, [r12]!               ; #4
+    vmov.u8     q11, #4                     ; #4
 
     ; vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d2, q2
@@ -444,27 +399,23 @@
 
     vand        q1, q1, q15                 ; vp8_filter &= mask
 
-    vld1.u8     {q15}, [r12]!               ; #63
-    ;
-    vand        q13, q1, q14                ; Filter2 &= hev
+    vmov.u16    q15, #63                    ; #63
 
-    vld1.u8     {d7}, [r12]!                ; #9
+    vand        q13, q1, q14                ; Filter2 &= hev
 
     vqadd.s8    q2, q13, q11                ; Filter1 = clamp(Filter2+4)
     vqadd.s8    q13, q13, q12               ; Filter2 = clamp(Filter2+3)
 
-    vld1.u8     {d6}, [r12]!                ; #18
+    vmov        q0, q15
 
     vshr.s8     q2, q2, #3                  ; Filter1 >>= 3
     vshr.s8     q13, q13, #3                ; Filter2 >>= 3
 
-    vmov        q10, q15
+    vmov        q11, q15
     vmov        q12, q15
 
     vqsub.s8    q7, q7, q2                  ; qs0 = clamp(qs0 - Filter1)
 
-    vld1.u8     {d5}, [r12]!                ; #27
-
     vqadd.s8    q6, q6, q13                 ; ps0 = clamp(ps0 + Filter2)
 
     vbic        q1, q1, q14                 ; vp8_filter &= ~hev
@@ -472,46 +423,47 @@
     ; roughly 1/7th difference across boundary
     ; roughly 2/7th difference across boundary
     ; roughly 3/7th difference across boundary
-    vmov        q11, q15
+
+    vmov.u8     d5, #9                      ; #9
+    vmov.u8     d4, #18                     ; #18
+
     vmov        q13, q15
     vmov        q14, q15
 
-    vmlal.s8    q10, d2, d7                 ; Filter2 * 9
-    vmlal.s8    q11, d3, d7
-    vmlal.s8    q12, d2, d6                 ; Filter2 * 18
-    vmlal.s8    q13, d3, d6
-    vmlal.s8    q14, d2, d5                 ; Filter2 * 27
+    vmlal.s8    q0, d2, d5                  ; 63 + Filter2 * 9
+    vmlal.s8    q11, d3, d5
+    vmov.u8     d5, #27                     ; #27
+    vmlal.s8    q12, d2, d4                 ; 63 + Filter2 * 18
+    vmlal.s8    q13, d3, d4
+    vmlal.s8    q14, d2, d5                 ; 63 + Filter2 * 27
     vmlal.s8    q15, d3, d5
-    vqshrn.s16  d20, q10, #7                ; u = clamp((63 + Filter2 * 9)>>7)
-    vqshrn.s16  d21, q11, #7
+
+    vqshrn.s16  d0, q0, #7                  ; u = clamp((63 + Filter2 * 9)>>7)
+    vqshrn.s16  d1, q11, #7
     vqshrn.s16  d24, q12, #7                ; u = clamp((63 + Filter2 * 18)>>7)
     vqshrn.s16  d25, q13, #7
     vqshrn.s16  d28, q14, #7                ; u = clamp((63 + Filter2 * 27)>>7)
     vqshrn.s16  d29, q15, #7
 
-    vqsub.s8    q11, q9, q10                ; s = clamp(qs2 - u)
-    vqadd.s8    q10, q4, q10                ; s = clamp(ps2 + u)
+    vmov.u8     q1, #0x80                   ; 0x80
+
+    vqsub.s8    q11, q9, q0                 ; s = clamp(qs2 - u)
+    vqadd.s8    q0, q4, q0                  ; s = clamp(ps2 + u)
     vqsub.s8    q13, q8, q12                ; s = clamp(qs1 - u)
     vqadd.s8    q12, q5, q12                ; s = clamp(ps1 + u)
     vqsub.s8    q15, q7, q14                ; s = clamp(qs0 - u)
     vqadd.s8    q14, q6, q14                ; s = clamp(ps0 + u)
-    veor        q9, q11, q0                 ; *oq2 = s^0x80
-    veor        q4, q10, q0                 ; *op2 = s^0x80
-    veor        q8, q13, q0                 ; *oq1 = s^0x80
-    veor        q5, q12, q0                 ; *op2 = s^0x80
-    veor        q7, q15, q0                 ; *oq0 = s^0x80
-    veor        q6, q14, q0                 ; *op0 = s^0x80
+
+    veor        q9, q11, q1                 ; *oq2 = s^0x80
+    veor        q4, q0, q1                  ; *op2 = s^0x80
+    veor        q8, q13, q1                 ; *oq1 = s^0x80
+    veor        q5, q12, q1                 ; *op2 = s^0x80
+    veor        q7, q15, q1                 ; *oq0 = s^0x80
+    veor        q6, q14, q1                 ; *op0 = s^0x80
 
     bx          lr
     ENDP        ; |vp8_mbloop_filter_neon|
 
-    AREA    mbloopfilter_dat, DATA, READONLY
-mblf_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
-    DCD     0x003f003f, 0x003f003f, 0x003f003f, 0x003f003f
-    DCD     0x09090909, 0x09090909, 0x12121212, 0x12121212
-    DCD     0x1b1b1b1b, 0x1b1b1b1b
+;-----------------
 
     END

media/libvpx/vp8/common/arm/neon/recon_neon.c

@@ -10,8 +10,8 @@
 
 
 #include "vpx_ports/config.h"
-#include "recon.h"
-#include "blockd.h"
+#include "vp8/common/recon.h"
+#include "vp8/common/blockd.h"
 
 extern void vp8_recon16x16mb_neon(unsigned char *pred_ptr, short *diff_ptr, unsigned char *dst_ptr, int ystride, unsigned char *udst_ptr, unsigned char *vdst_ptr);
 

media/libvpx/vp8/common/arm/neon/shortidct4x4llm_neon.asm

@@ -113,10 +113,7 @@
     ENDP
 
 ;-----------------
-    AREA    idct4x4_dat, DATA, READWRITE            ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
 idct_coeff
     DCD     0x4e7b4e7b, 0x8a8c8a8c
 

media/libvpx/vp8/common/arm/neon/sixtappredict16x16_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter16_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -33,7 +44,7 @@
 |vp8_sixtap_predict16x16_neon| PROC
     push            {r4-r5, lr}
 
-    adrl            r12, filter16_coeff
+    adr             r12, filter16_coeff
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
 
@@ -476,18 +487,4 @@ secondpass_only_inner_loop_neon
     ENDP
 
 ;-----------------
-    AREA    subpelfilters16_dat, DATA, READWRITE            ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-filter16_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
-
     END

media/libvpx/vp8/common/arm/neon/sixtappredict4x4_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter4_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict_neon| PROC
     push            {r4, lr}
 
-    adrl            r12, filter4_coeff
+    adr             r12, filter4_coeff
     ldr             r4, [sp, #8]            ;load parameters from stack
     ldr             lr, [sp, #12]           ;load parameters from stack
 
@@ -407,18 +418,5 @@ secondpass_filter4x4_only
     ENDP
 
 ;-----------------
-    AREA    subpelfilters4_dat, DATA, READWRITE         ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-filter4_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
 
     END

media/libvpx/vp8/common/arm/neon/sixtappredict8x4_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter8_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict8x4_neon| PROC
     push            {r4-r5, lr}
 
-    adrl            r12, filter8_coeff
+    adr             r12, filter8_coeff
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
 
@@ -458,18 +469,5 @@ secondpass_filter8x4_only
     ENDP
 
 ;-----------------
-    AREA    subpelfilters8_dat, DATA, READWRITE         ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-filter8_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
 
     END

media/libvpx/vp8/common/arm/neon/sixtappredict8x8_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter8_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict8x8_neon| PROC
     push            {r4-r5, lr}
 
-    adrl            r12, filter8_coeff
+    adr             r12, filter8_coeff
 
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
@@ -509,18 +520,5 @@ filt_blk2d_spo8x8_loop_neon
     ENDP
 
 ;-----------------
-    AREA    subpelfilters8_dat, DATA, READWRITE         ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-filter8_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
 
     END

media/libvpx/vp8/common/arm/reconintra_arm.c

@@ -10,10 +10,10 @@
 
 
 #include "vpx_ports/config.h"
-#include "blockd.h"
-#include "reconintra.h"
+#include "vp8/common/blockd.h"
+#include "vp8/common/reconintra.h"
 #include "vpx_mem/vpx_mem.h"
-#include "recon.h"
+#include "vp8/common/recon.h"
 
 #if HAVE_ARMV7
 extern void vp8_build_intra_predictors_mby_neon_func(

media/libvpx/vp8/common/arm/vpx_asm_offsets.c

@@ -1,87 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "vpx_ports/config.h"
-#include <stddef.h>
-
-#if CONFIG_VP8_ENCODER
-#include "vpx_scale/yv12config.h"
-#endif
-
-#if CONFIG_VP8_DECODER
-#include "onyxd_int.h"
-#endif
-
-#define DEFINE(sym, val) int sym = val;
-
-/*
-#define BLANK() asm volatile("\n->" : : )
-*/
-
-/*
- * int main(void)
- * {
- */
-
-#if CONFIG_VP8_DECODER || CONFIG_VP8_ENCODER
-DEFINE(yv12_buffer_config_y_width,              offsetof(YV12_BUFFER_CONFIG, y_width));
-DEFINE(yv12_buffer_config_y_height,             offsetof(YV12_BUFFER_CONFIG, y_height));
-DEFINE(yv12_buffer_config_y_stride,             offsetof(YV12_BUFFER_CONFIG, y_stride));
-DEFINE(yv12_buffer_config_uv_width,             offsetof(YV12_BUFFER_CONFIG, uv_width));
-DEFINE(yv12_buffer_config_uv_height,            offsetof(YV12_BUFFER_CONFIG, uv_height));
-DEFINE(yv12_buffer_config_uv_stride,            offsetof(YV12_BUFFER_CONFIG, uv_stride));
-DEFINE(yv12_buffer_config_y_buffer,             offsetof(YV12_BUFFER_CONFIG, y_buffer));
-DEFINE(yv12_buffer_config_u_buffer,             offsetof(YV12_BUFFER_CONFIG, u_buffer));
-DEFINE(yv12_buffer_config_v_buffer,             offsetof(YV12_BUFFER_CONFIG, v_buffer));
-DEFINE(yv12_buffer_config_border,               offsetof(YV12_BUFFER_CONFIG, border));
-#endif
-
-#if CONFIG_VP8_DECODER
-DEFINE(mb_diff,                                 offsetof(MACROBLOCKD, diff));
-DEFINE(mb_predictor,                            offsetof(MACROBLOCKD, predictor));
-DEFINE(mb_dst_y_stride,                         offsetof(MACROBLOCKD, dst.y_stride));
-DEFINE(mb_dst_y_buffer,                         offsetof(MACROBLOCKD, dst.y_buffer));
-DEFINE(mb_dst_u_buffer,                         offsetof(MACROBLOCKD, dst.u_buffer));
-DEFINE(mb_dst_v_buffer,                         offsetof(MACROBLOCKD, dst.v_buffer));
-DEFINE(mb_up_available,                         offsetof(MACROBLOCKD, up_available));
-DEFINE(mb_left_available,                       offsetof(MACROBLOCKD, left_available));
-
-DEFINE(detok_scan,                              offsetof(DETOK, scan));
-DEFINE(detok_ptr_block2leftabove,               offsetof(DETOK, ptr_block2leftabove));
-DEFINE(detok_coef_tree_ptr,                     offsetof(DETOK, vp8_coef_tree_ptr));
-DEFINE(detok_teb_base_ptr,                      offsetof(DETOK, teb_base_ptr));
-DEFINE(detok_norm_ptr,                          offsetof(DETOK, norm_ptr));
-DEFINE(detok_ptr_coef_bands_x,                  offsetof(DETOK, ptr_coef_bands_x));
-
-DEFINE(detok_A,                                 offsetof(DETOK, A));
-DEFINE(detok_L,                                 offsetof(DETOK, L));
-
-DEFINE(detok_qcoeff_start_ptr,                  offsetof(DETOK, qcoeff_start_ptr));
-DEFINE(detok_current_bc,                        offsetof(DETOK, current_bc));
-DEFINE(detok_coef_probs,                        offsetof(DETOK, coef_probs));
-DEFINE(detok_eob,                               offsetof(DETOK, eob));
-
-DEFINE(bool_decoder_user_buffer_end,            offsetof(BOOL_DECODER, user_buffer_end));
-DEFINE(bool_decoder_user_buffer,                offsetof(BOOL_DECODER, user_buffer));
-DEFINE(bool_decoder_value,                      offsetof(BOOL_DECODER, value));
-DEFINE(bool_decoder_count,                      offsetof(BOOL_DECODER, count));
-DEFINE(bool_decoder_range,                      offsetof(BOOL_DECODER, range));
-
-DEFINE(tokenextrabits_min_val,                  offsetof(TOKENEXTRABITS, min_val));
-DEFINE(tokenextrabits_length,                   offsetof(TOKENEXTRABITS, Length));
-#endif
-
-//add asserts for any offset that is not supported by assembly code
-//add asserts for any size that is not supported by assembly code
-/*
- * return 0;
- * }
- */

media/libvpx/vp8/common/blockd.c

@@ -12,8 +12,6 @@
 #include "blockd.h"
 #include "vpx_mem/vpx_mem.h"
 
-const int vp8_block2type[25] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 1};
-
 const unsigned char vp8_block2left[25] =
 {
     0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8

media/libvpx/vp8/common/blockd.h

@@ -28,11 +28,6 @@ void vpx_log(const char *format, ...);
 #define DCPREDSIMTHRESH 0
 #define DCPREDCNTTHRESH 3
 
-#define Y1CONTEXT 0
-#define UCONTEXT 1
-#define VCONTEXT 2
-#define Y2CONTEXT 3
-
 #define MB_FEATURE_TREE_PROBS   3
 #define MAX_MB_SEGMENTS         4
 
@@ -48,6 +43,11 @@ typedef struct
     int r, c;
 } POS;
 
+#define PLANE_TYPE_Y_NO_DC    0
+#define PLANE_TYPE_Y2         1
+#define PLANE_TYPE_UV         2
+#define PLANE_TYPE_Y_WITH_DC  3
+
 
 typedef char ENTROPY_CONTEXT;
 typedef struct
@@ -58,8 +58,6 @@ typedef struct
     ENTROPY_CONTEXT y2;
 } ENTROPY_CONTEXT_PLANES;
 
-extern const int vp8_block2type[25];
-
 extern const unsigned char vp8_block2left[25];
 extern const unsigned char vp8_block2above[25];
 
@@ -139,16 +137,11 @@ typedef enum
    modes for the Y blocks to the left and above us; for interframes, there
    is a single probability table. */
 
-typedef struct
+union b_mode_info
 {
-    B_PREDICTION_MODE mode;
-    union
-    {
-        int as_int;
-        MV  as_mv;
-    } mv;
-} B_MODE_INFO;
-
+    B_PREDICTION_MODE as_mode;
+    int_mv mv;
+};
 
 typedef enum
 {
@@ -163,38 +156,26 @@ typedef struct
 {
     MB_PREDICTION_MODE mode, uv_mode;
     MV_REFERENCE_FRAME ref_frame;
-    union
-    {
-        int as_int;
-        MV  as_mv;
-    } mv;
+    int_mv mv;
 
     unsigned char partitioning;
     unsigned char mb_skip_coeff;                                /* does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens */
-    unsigned char dc_diff;
     unsigned char need_to_clamp_mvs;
-
     unsigned char segment_id;                  /* Which set of segmentation parameters should be used for this MB */
-
-    unsigned char force_no_skip; /* encoder only */
 } MB_MODE_INFO;
 
-
 typedef struct
 {
     MB_MODE_INFO mbmi;
-    B_MODE_INFO bmi[16];
+    union b_mode_info bmi[16];
 } MODE_INFO;
 
-
 typedef struct
 {
     short *qcoeff;
     short *dqcoeff;
     unsigned char  *predictor;
     short *diff;
-    short *reference;
-
     short *dequant;
 
     /* 16 Y blocks, 4 U blocks, 4 V blocks each with 16 entries */
@@ -208,15 +189,13 @@ typedef struct
 
     int eob;
 
-    B_MODE_INFO bmi;
-
+    union b_mode_info bmi;
 } BLOCKD;
 
-typedef struct
+typedef struct MacroBlockD
 {
     DECLARE_ALIGNED(16, short, diff[400]);      /* from idct diff */
     DECLARE_ALIGNED(16, unsigned char,  predictor[384]);
-/* not used    DECLARE_ALIGNED(16, short, reference[384]); */
     DECLARE_ALIGNED(16, short, qcoeff[400]);
     DECLARE_ALIGNED(16, short, dqcoeff[400]);
     DECLARE_ALIGNED(16, char,  eobs[25]);
@@ -273,6 +252,9 @@ typedef struct
     int mb_to_top_edge;
     int mb_to_bottom_edge;
 
+    int ref_frame_cost[MAX_REF_FRAMES];
+
+
     unsigned int frames_since_golden;
     unsigned int frames_till_alt_ref_frame;
     vp8_subpix_fn_t  subpixel_predict;
@@ -282,6 +264,16 @@ typedef struct
 
     void *current_bc;
 
+    int corrupted;
+
+#if ARCH_X86 || ARCH_X86_64
+    /* This is an intermediate buffer currently used in sub-pixel motion search
+     * to keep a copy of the reference area. This buffer can be used for other
+     * purpose.
+     */
+    DECLARE_ALIGNED(32, unsigned char, y_buf[22*32]);
+#endif
+
 #if CONFIG_RUNTIME_CPU_DETECT
     struct VP8_COMMON_RTCD  *rtcd;
 #endif
@@ -291,4 +283,20 @@ typedef struct
 extern void vp8_build_block_doffsets(MACROBLOCKD *x);
 extern void vp8_setup_block_dptrs(MACROBLOCKD *x);
 
+static void update_blockd_bmi(MACROBLOCKD *xd)
+{
+    int i;
+    int is_4x4;
+    is_4x4 = (xd->mode_info_context->mbmi.mode == SPLITMV) ||
+              (xd->mode_info_context->mbmi.mode == B_PRED);
+
+    if (is_4x4)
+    {
+        for (i = 0; i < 16; i++)
+        {
+            xd->block[i].bmi = xd->mode_info_context->bmi[i];
+        }
+    }
+}
+
 #endif  /* __INC_BLOCKD_H */

media/libvpx/vp8/common/coefupdateprobs.h

@@ -12,7 +12,7 @@
 /* Update probabilities for the nodes in the token entropy tree.
    Generated file included by entropy.c */
 
-const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1] =
+const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] =
 {
     {
         {

media/libvpx/vp8/common/debugmodes.c

@@ -97,7 +97,7 @@ void vp8_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols, int f
                 bindex = (b_row & 3) * 4 + (b_col & 3);
 
                 if (mi[mb_index].mbmi.mode == B_PRED)
-                    fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].mode);
+                    fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].as_mode);
                 else
                     fprintf(mvs, "xx ");
 

media/libvpx/vp8/common/defaultcoefcounts.c

@@ -0,0 +1,225 @@
+/*
+ *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "defaultcoefcounts.h"
+
+/* Generated file, included by entropy.c */
+
+const unsigned int vp8_default_coef_counts[BLOCK_TYPES]
+                                          [COEF_BANDS]
+                                          [PREV_COEF_CONTEXTS]
+                                          [MAX_ENTROPY_TOKENS] =
+{
+
+    {
+        /* Block Type ( 0 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {30190, 26544, 225,  24,   4,   0,   0,   0,   0,   0,   0, 4171593,},
+            {26846, 25157, 1241, 130,  26,   6,   1,   0,   0,   0,   0, 149987,},
+            {10484, 9538, 1006, 160,  36,  18,   0,   0,   0,   0,   0, 15104,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {25842, 40456, 1126,  83,  11,   2,   0,   0,   0,   0,   0,   0,},
+            {9338, 8010, 512,  73,   7,   3,   2,   0,   0,   0,   0, 43294,},
+            {1047, 751, 149,  31,  13,   6,   1,   0,   0,   0,   0, 879,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {26136, 9826, 252,  13,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {8134, 5574, 191,  14,   2,   0,   0,   0,   0,   0,   0, 35302,},
+            { 605, 677, 116,   9,   1,   0,   0,   0,   0,   0,   0, 611,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {10263, 15463, 283,  17,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {2773, 2191, 128,   9,   2,   2,   0,   0,   0,   0,   0, 10073,},
+            { 134, 125,  32,   4,   0,   2,   0,   0,   0,   0,   0,  50,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {10483, 2663,  23,   1,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {2137, 1251,  27,   1,   1,   0,   0,   0,   0,   0,   0, 14362,},
+            { 116, 156,  14,   2,   1,   0,   0,   0,   0,   0,   0, 190,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {40977, 27614, 412,  28,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {6113, 5213, 261,  22,   3,   0,   0,   0,   0,   0,   0, 26164,},
+            { 382, 312,  50,  14,   2,   0,   0,   0,   0,   0,   0, 345,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,  26,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,  13,   0,   0,   0,   0,   0,   0,   0,   0,   0, 319,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   8,},
+        },
+    },
+    {
+        /* Block Type ( 1 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {3268, 19382, 1043, 250,  93,  82,  49,  26,  17,   8,  25, 82289,},
+            {8758, 32110, 5436, 1832, 827, 668, 420, 153,  24,   0,   3, 52914,},
+            {9337, 23725, 8487, 3954, 2107, 1836, 1069, 399,  59,   0,   0, 18620,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {12419, 8420, 452,  62,   9,   1,   0,   0,   0,   0,   0,   0,},
+            {11715, 8705, 693,  92,  15,   7,   2,   0,   0,   0,   0, 53988,},
+            {7603, 8585, 2306, 778, 270, 145,  39,   5,   0,   0,   0, 9136,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {15938, 14335, 1207, 184,  55,  13,   4,   1,   0,   0,   0,   0,},
+            {7415, 6829, 1138, 244,  71,  26,   7,   0,   0,   0,   0, 9980,},
+            {1580, 1824, 655, 241,  89,  46,  10,   2,   0,   0,   0, 429,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {19453, 5260, 201,  19,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {9173, 3758, 213,  22,   1,   1,   0,   0,   0,   0,   0, 9820,},
+            {1689, 1277, 276,  51,  17,   4,   0,   0,   0,   0,   0, 679,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {12076, 10667, 620,  85,  19,   9,   5,   0,   0,   0,   0,   0,},
+            {4665, 3625, 423,  55,  19,   9,   0,   0,   0,   0,   0, 5127,},
+            { 415, 440, 143,  34,  20,   7,   2,   0,   0,   0,   0, 101,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {12183, 4846, 115,  11,   1,   0,   0,   0,   0,   0,   0,   0,},
+            {4226, 3149, 177,  21,   2,   0,   0,   0,   0,   0,   0, 7157,},
+            { 375, 621, 189,  51,  11,   4,   1,   0,   0,   0,   0, 198,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {61658, 37743, 1203,  94,  10,   3,   0,   0,   0,   0,   0,   0,},
+            {15514, 11563, 903, 111,  14,   5,   0,   0,   0,   0,   0, 25195,},
+            { 929, 1077, 291,  78,  14,   7,   1,   0,   0,   0,   0, 507,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0, 990,  15,   3,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0, 412,  13,   0,   0,   0,   0,   0,   0,   0,   0, 1641,},
+            {   0,  18,   7,   1,   0,   0,   0,   0,   0,   0,   0,  30,},
+        },
+    },
+    {
+        /* Block Type ( 2 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            { 953, 24519, 628, 120,  28,  12,   4,   0,   0,   0,   0, 2248798,},
+            {1525, 25654, 2647, 617, 239, 143,  42,   5,   0,   0,   0, 66837,},
+            {1180, 11011, 3001, 1237, 532, 448, 239,  54,   5,   0,   0, 7122,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {1356, 2220,  67,  10,   4,   1,   0,   0,   0,   0,   0,   0,},
+            {1450, 2544, 102,  18,   4,   3,   0,   0,   0,   0,   0, 57063,},
+            {1182, 2110, 470, 130,  41,  21,   0,   0,   0,   0,   0, 6047,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            { 370, 3378, 200,  30,   5,   4,   1,   0,   0,   0,   0,   0,},
+            { 293, 1006, 131,  29,  11,   0,   0,   0,   0,   0,   0, 5404,},
+            { 114, 387,  98,  23,   4,   8,   1,   0,   0,   0,   0, 236,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            { 579, 194,   4,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            { 395, 213,   5,   1,   0,   0,   0,   0,   0,   0,   0, 4157,},
+            { 119, 122,   4,   0,   0,   0,   0,   0,   0,   0,   0, 300,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {  38, 557,  19,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  21, 114,  12,   1,   0,   0,   0,   0,   0,   0,   0, 427,},
+            {   0,   5,   0,   0,   0,   0,   0,   0,   0,   0,   0,   7,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {  52,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  18,   6,   0,   0,   0,   0,   0,   0,   0,   0,   0, 652,},
+            {   1,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  30,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            { 640, 569,  10,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  25,  77,   2,   0,   0,   0,   0,   0,   0,   0,   0, 517,},
+            {   4,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   3,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+        },
+    },
+    {
+        /* Block Type ( 3 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {2506, 20161, 2707, 767, 261, 178, 107,  30,  14,   3,   0, 100694,},
+            {8806, 36478, 8817, 3268, 1280, 850, 401, 114,  42,   0,   0, 58572,},
+            {11003, 27214, 11798, 5716, 2482, 2072, 1048, 175,  32,   0,   0, 19284,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {9738, 11313, 959, 205,  70,  18,  11,   1,   0,   0,   0,   0,},
+            {12628, 15085, 1507, 273,  52,  19,   9,   0,   0,   0,   0, 54280,},
+            {10701, 15846, 5561, 1926, 813, 570, 249,  36,   0,   0,   0, 6460,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {6781, 22539, 2784, 634, 182, 123,  20,   4,   0,   0,   0,   0,},
+            {6263, 11544, 2649, 790, 259, 168,  27,   5,   0,   0,   0, 20539,},
+            {3109, 4075, 2031, 896, 457, 386, 158,  29,   0,   0,   0, 1138,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {11515, 4079, 465,  73,   5,  14,   2,   0,   0,   0,   0,   0,},
+            {9361, 5834, 650,  96,  24,   8,   4,   0,   0,   0,   0, 22181,},
+            {4343, 3974, 1360, 415, 132,  96,  14,   1,   0,   0,   0, 1267,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {4787, 9297, 823, 168,  44,  12,   4,   0,   0,   0,   0,   0,},
+            {3619, 4472, 719, 198,  60,  31,   3,   0,   0,   0,   0, 8401,},
+            {1157, 1175, 483, 182,  88,  31,   8,   0,   0,   0,   0, 268,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {8299, 1226,  32,   5,   1,   0,   0,   0,   0,   0,   0,   0,},
+            {3502, 1568,  57,   4,   1,   1,   0,   0,   0,   0,   0, 9811,},
+            {1055, 1070, 166,  29,   6,   1,   0,   0,   0,   0,   0, 527,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {27414, 27927, 1989, 347,  69,  26,   0,   0,   0,   0,   0,   0,},
+            {5876, 10074, 1574, 341,  91,  24,   4,   0,   0,   0,   0, 21954,},
+            {1571, 2171, 778, 324, 124,  65,  16,   0,   0,   0,   0, 979,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,  29,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,  23,   0,   0,   0,   0,   0,   0,   0,   0,   0, 459,},
+            {   0,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  13,},
+        },
+    },
+};

media/libvpx/vp8/common/defaultcoefcounts.h

@@ -8,214 +8,14 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#ifndef __DEFAULTCOEFCOUNTS_H
+#define __DEFAULTCOEFCOUNTS_H
 
-/* Generated file, included by entropy.c */
+#include "entropy.h"
 
-static const unsigned int default_coef_counts [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens] =
-{
+extern const unsigned int vp8_default_coef_counts[BLOCK_TYPES]
+                                                 [COEF_BANDS]
+                                                 [PREV_COEF_CONTEXTS]
+                                                 [MAX_ENTROPY_TOKENS];
 
-    {
-        /* Block Type ( 0 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {30190, 26544, 225,  24,   4,   0,   0,   0,   0,   0,   0, 4171593,},
-            {26846, 25157, 1241, 130,  26,   6,   1,   0,   0,   0,   0, 149987,},
-            {10484, 9538, 1006, 160,  36,  18,   0,   0,   0,   0,   0, 15104,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {25842, 40456, 1126,  83,  11,   2,   0,   0,   0,   0,   0,   0,},
-            {9338, 8010, 512,  73,   7,   3,   2,   0,   0,   0,   0, 43294,},
-            {1047, 751, 149,  31,  13,   6,   1,   0,   0,   0,   0, 879,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {26136, 9826, 252,  13,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {8134, 5574, 191,  14,   2,   0,   0,   0,   0,   0,   0, 35302,},
-            { 605, 677, 116,   9,   1,   0,   0,   0,   0,   0,   0, 611,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {10263, 15463, 283,  17,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {2773, 2191, 128,   9,   2,   2,   0,   0,   0,   0,   0, 10073,},
-            { 134, 125,  32,   4,   0,   2,   0,   0,   0,   0,   0,  50,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {10483, 2663,  23,   1,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {2137, 1251,  27,   1,   1,   0,   0,   0,   0,   0,   0, 14362,},
-            { 116, 156,  14,   2,   1,   0,   0,   0,   0,   0,   0, 190,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {40977, 27614, 412,  28,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {6113, 5213, 261,  22,   3,   0,   0,   0,   0,   0,   0, 26164,},
-            { 382, 312,  50,  14,   2,   0,   0,   0,   0,   0,   0, 345,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,  26,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,  13,   0,   0,   0,   0,   0,   0,   0,   0,   0, 319,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   8,},
-        },
-    },
-    {
-        /* Block Type ( 1 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {3268, 19382, 1043, 250,  93,  82,  49,  26,  17,   8,  25, 82289,},
-            {8758, 32110, 5436, 1832, 827, 668, 420, 153,  24,   0,   3, 52914,},
-            {9337, 23725, 8487, 3954, 2107, 1836, 1069, 399,  59,   0,   0, 18620,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {12419, 8420, 452,  62,   9,   1,   0,   0,   0,   0,   0,   0,},
-            {11715, 8705, 693,  92,  15,   7,   2,   0,   0,   0,   0, 53988,},
-            {7603, 8585, 2306, 778, 270, 145,  39,   5,   0,   0,   0, 9136,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {15938, 14335, 1207, 184,  55,  13,   4,   1,   0,   0,   0,   0,},
-            {7415, 6829, 1138, 244,  71,  26,   7,   0,   0,   0,   0, 9980,},
-            {1580, 1824, 655, 241,  89,  46,  10,   2,   0,   0,   0, 429,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {19453, 5260, 201,  19,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {9173, 3758, 213,  22,   1,   1,   0,   0,   0,   0,   0, 9820,},
-            {1689, 1277, 276,  51,  17,   4,   0,   0,   0,   0,   0, 679,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {12076, 10667, 620,  85,  19,   9,   5,   0,   0,   0,   0,   0,},
-            {4665, 3625, 423,  55,  19,   9,   0,   0,   0,   0,   0, 5127,},
-            { 415, 440, 143,  34,  20,   7,   2,   0,   0,   0,   0, 101,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {12183, 4846, 115,  11,   1,   0,   0,   0,   0,   0,   0,   0,},
-            {4226, 3149, 177,  21,   2,   0,   0,   0,   0,   0,   0, 7157,},
-            { 375, 621, 189,  51,  11,   4,   1,   0,   0,   0,   0, 198,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {61658, 37743, 1203,  94,  10,   3,   0,   0,   0,   0,   0,   0,},
-            {15514, 11563, 903, 111,  14,   5,   0,   0,   0,   0,   0, 25195,},
-            { 929, 1077, 291,  78,  14,   7,   1,   0,   0,   0,   0, 507,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0, 990,  15,   3,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0, 412,  13,   0,   0,   0,   0,   0,   0,   0,   0, 1641,},
-            {   0,  18,   7,   1,   0,   0,   0,   0,   0,   0,   0,  30,},
-        },
-    },
-    {
-        /* Block Type ( 2 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            { 953, 24519, 628, 120,  28,  12,   4,   0,   0,   0,   0, 2248798,},
-            {1525, 25654, 2647, 617, 239, 143,  42,   5,   0,   0,   0, 66837,},
-            {1180, 11011, 3001, 1237, 532, 448, 239,  54,   5,   0,   0, 7122,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {1356, 2220,  67,  10,   4,   1,   0,   0,   0,   0,   0,   0,},
-            {1450, 2544, 102,  18,   4,   3,   0,   0,   0,   0,   0, 57063,},
-            {1182, 2110, 470, 130,  41,  21,   0,   0,   0,   0,   0, 6047,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            { 370, 3378, 200,  30,   5,   4,   1,   0,   0,   0,   0,   0,},
-            { 293, 1006, 131,  29,  11,   0,   0,   0,   0,   0,   0, 5404,},
-            { 114, 387,  98,  23,   4,   8,   1,   0,   0,   0,   0, 236,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            { 579, 194,   4,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            { 395, 213,   5,   1,   0,   0,   0,   0,   0,   0,   0, 4157,},
-            { 119, 122,   4,   0,   0,   0,   0,   0,   0,   0,   0, 300,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {  38, 557,  19,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  21, 114,  12,   1,   0,   0,   0,   0,   0,   0,   0, 427,},
-            {   0,   5,   0,   0,   0,   0,   0,   0,   0,   0,   0,   7,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {  52,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  18,   6,   0,   0,   0,   0,   0,   0,   0,   0,   0, 652,},
-            {   1,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  30,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            { 640, 569,  10,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  25,  77,   2,   0,   0,   0,   0,   0,   0,   0,   0, 517,},
-            {   4,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   3,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-        },
-    },
-    {
-        /* Block Type ( 3 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {2506, 20161, 2707, 767, 261, 178, 107,  30,  14,   3,   0, 100694,},
-            {8806, 36478, 8817, 3268, 1280, 850, 401, 114,  42,   0,   0, 58572,},
-            {11003, 27214, 11798, 5716, 2482, 2072, 1048, 175,  32,   0,   0, 19284,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {9738, 11313, 959, 205,  70,  18,  11,   1,   0,   0,   0,   0,},
-            {12628, 15085, 1507, 273,  52,  19,   9,   0,   0,   0,   0, 54280,},
-            {10701, 15846, 5561, 1926, 813, 570, 249,  36,   0,   0,   0, 6460,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {6781, 22539, 2784, 634, 182, 123,  20,   4,   0,   0,   0,   0,},
-            {6263, 11544, 2649, 790, 259, 168,  27,   5,   0,   0,   0, 20539,},
-            {3109, 4075, 2031, 896, 457, 386, 158,  29,   0,   0,   0, 1138,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {11515, 4079, 465,  73,   5,  14,   2,   0,   0,   0,   0,   0,},
-            {9361, 5834, 650,  96,  24,   8,   4,   0,   0,   0,   0, 22181,},
-            {4343, 3974, 1360, 415, 132,  96,  14,   1,   0,   0,   0, 1267,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {4787, 9297, 823, 168,  44,  12,   4,   0,   0,   0,   0,   0,},
-            {3619, 4472, 719, 198,  60,  31,   3,   0,   0,   0,   0, 8401,},
-            {1157, 1175, 483, 182,  88,  31,   8,   0,   0,   0,   0, 268,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {8299, 1226,  32,   5,   1,   0,   0,   0,   0,   0,   0,   0,},
-            {3502, 1568,  57,   4,   1,   1,   0,   0,   0,   0,   0, 9811,},
-            {1055, 1070, 166,  29,   6,   1,   0,   0,   0,   0,   0, 527,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {27414, 27927, 1989, 347,  69,  26,   0,   0,   0,   0,   0,   0,},
-            {5876, 10074, 1574, 341,  91,  24,   4,   0,   0,   0,   0, 21954,},
-            {1571, 2171, 778, 324, 124,  65,  16,   0,   0,   0,   0, 979,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,  29,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,  23,   0,   0,   0,   0,   0,   0,   0,   0,   0, 459,},
-            {   0,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  13,},
-        },
-    },
-};
+#endif //__DEFAULTCOEFCOUNTS_H

media/libvpx/vp8/common/entropy.c

@@ -26,8 +26,32 @@ typedef vp8_prob Prob;
 
 #include "coefupdateprobs.h"
 
-DECLARE_ALIGNED(16, cuchar, vp8_coef_bands[16]) = { 0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7};
-DECLARE_ALIGNED(16, cuchar, vp8_prev_token_class[MAX_ENTROPY_TOKENS]) = { 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0};
+DECLARE_ALIGNED(16, const unsigned char, vp8_norm[256]) =
+{
+    0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4,
+    3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+};
+
+DECLARE_ALIGNED(16, cuchar, vp8_coef_bands[16]) =
+{ 0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7};
+
+DECLARE_ALIGNED(16, cuchar, vp8_prev_token_class[MAX_ENTROPY_TOKENS]) =
+{ 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0};
+
 DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d[16]) =
 {
     0,  1,  4,  8,
@@ -36,6 +60,14 @@ DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d[16]) =
     7, 11, 14, 15,
 };
 
+DECLARE_ALIGNED(16, const short, vp8_default_inv_zig_zag[16]) =
+{
+    1,  2,  6,  7,
+    3,  5,  8, 13,
+    4,  9, 12, 14,
+   10, 11, 15, 16
+};
+
 DECLARE_ALIGNED(16, short, vp8_default_zig_zag_mask[16]);
 
 const int vp8_mb_feature_data_bits[MB_LVL_MAX] = {7, 6};
@@ -57,7 +89,7 @@ const vp8_tree_index vp8_coef_tree[ 22] =     /* corresponding _CONTEXT_NODEs */
     -DCT_VAL_CATEGORY5, -DCT_VAL_CATEGORY6   /* 10 = CAT_FIVE */
 };
 
-struct vp8_token_struct vp8_coef_encodings[vp8_coef_tokens];
+struct vp8_token_struct vp8_coef_encodings[MAX_ENTROPY_TOKENS];
 
 /* Trees for extra bits.  Probabilities are constant and
    do not depend on previously encoded bits */
@@ -106,23 +138,20 @@ static void init_bit_trees()
     init_bit_tree(cat6, 11);
 }
 
-
-static vp8bc_index_t bcc1[1], bcc2[2], bcc3[3], bcc4[4], bcc5[5], bcc6[11];
-
 vp8_extra_bit_struct vp8_extra_bits[12] =
 {
-    { 0, 0, 0, 0, 0},
-    { 0, 0, 0, 0, 1},
-    { 0, 0, 0, 0, 2},
-    { 0, 0, 0, 0, 3},
-    { 0, 0, 0, 0, 4},
-    { cat1, Pcat1, bcc1, 1, 5},
-    { cat2, Pcat2, bcc2, 2, 7},
-    { cat3, Pcat3, bcc3, 3, 11},
-    { cat4, Pcat4, bcc4, 4, 19},
-    { cat5, Pcat5, bcc5, 5, 35},
-    { cat6, Pcat6, bcc6, 11, 67},
-    { 0, 0, 0, 0, 0}
+    { 0, 0, 0, 0},
+    { 0, 0, 0, 1},
+    { 0, 0, 0, 2},
+    { 0, 0, 0, 3},
+    { 0, 0, 0, 4},
+    { cat1, Pcat1, 1, 5},
+    { cat2, Pcat2, 2, 7},
+    { cat3, Pcat3, 3, 11},
+    { cat4, Pcat4, 4, 19},
+    { cat5, Pcat5, 5, 35},
+    { cat6, Pcat6, 11, 67},
+    { 0, 0, 0, 0}
 };
 #include "defaultcoefcounts.h"
 
@@ -140,10 +169,12 @@ void vp8_default_coef_probs(VP8_COMMON *pc)
 
             do
             {
-                unsigned int branch_ct [vp8_coef_tokens-1] [2];
+                unsigned int branch_ct [ENTROPY_NODES] [2];
                 vp8_tree_probs_from_distribution(
-                    vp8_coef_tokens, vp8_coef_encodings, vp8_coef_tree,
-                    pc->fc.coef_probs [h][i][k], branch_ct, default_coef_counts [h][i][k],
+                    MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
+                    pc->fc.coef_probs[h][i][k],
+                    branch_ct,
+                    vp8_default_coef_counts[h][i][k],
                     256, 1);
 
             }

media/libvpx/vp8/common/entropy.h

@@ -24,25 +24,23 @@
 #define FOUR_TOKEN              4       /* 4         Extra Bits 0+1 */
 #define DCT_VAL_CATEGORY1       5       /* 5-6       Extra Bits 1+1 */
 #define DCT_VAL_CATEGORY2       6       /* 7-10      Extra Bits 2+1 */
-#define DCT_VAL_CATEGORY3       7       /* 11-26     Extra Bits 4+1 */
-#define DCT_VAL_CATEGORY4       8       /* 11-26     Extra Bits 5+1 */
-#define DCT_VAL_CATEGORY5       9       /* 27-58     Extra Bits 5+1 */
-#define DCT_VAL_CATEGORY6       10      /* 59+       Extra Bits 11+1 */
+#define DCT_VAL_CATEGORY3       7       /* 11-18     Extra Bits 3+1 */
+#define DCT_VAL_CATEGORY4       8       /* 19-34     Extra Bits 4+1 */
+#define DCT_VAL_CATEGORY5       9       /* 35-66     Extra Bits 5+1 */
+#define DCT_VAL_CATEGORY6       10      /* 67+       Extra Bits 11+1 */
 #define DCT_EOB_TOKEN           11      /* EOB       Extra Bits 0+0 */
 
-#define vp8_coef_tokens 12
-#define MAX_ENTROPY_TOKENS vp8_coef_tokens
+#define MAX_ENTROPY_TOKENS 12
 #define ENTROPY_NODES 11
 
 extern const vp8_tree_index vp8_coef_tree[];
 
-extern struct vp8_token_struct vp8_coef_encodings[vp8_coef_tokens];
+extern struct vp8_token_struct vp8_coef_encodings[MAX_ENTROPY_TOKENS];
 
 typedef struct
 {
     vp8_tree_p tree;
     const vp8_prob *prob;
-    vp8bc_index_t *prob_bc;
     int Len;
     int base_val;
 } vp8_extra_bit_struct;
@@ -86,15 +84,16 @@ extern DECLARE_ALIGNED(16, const unsigned char, vp8_coef_bands[16]);
 /*# define DC_TOKEN_CONTEXTS        3*/ /* 00, 0!0, !0!0 */
 #   define PREV_COEF_CONTEXTS       3
 
-extern DECLARE_ALIGNED(16, const unsigned char, vp8_prev_token_class[vp8_coef_tokens]);
+extern DECLARE_ALIGNED(16, const unsigned char, vp8_prev_token_class[MAX_ENTROPY_TOKENS]);
 
-extern const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1];
+extern const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
 
 
 struct VP8Common;
 void vp8_default_coef_probs(struct VP8Common *);
 
 extern DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d[16]);
+extern DECLARE_ALIGNED(16, const short, vp8_default_inv_zig_zag[16]);
 extern short vp8_default_zig_zag_mask[16];
 extern const int vp8_mb_feature_data_bits[MB_LVL_MAX];
 

media/libvpx/vp8/common/entropymode.c

@@ -33,11 +33,11 @@ typedef enum
     SUBMVREF_LEFT_ABOVE_ZED
 } sumvfref_t;
 
-int vp8_mv_cont(const MV *l, const MV *a)
+int vp8_mv_cont(const int_mv *l, const int_mv *a)
 {
-    int lez = (l->row == 0 && l->col == 0);
-    int aez = (a->row == 0 && a->col == 0);
-    int lea = (l->row == a->row && l->col == a->col);
+    int lez = (l->as_int == 0);
+    int aez = (a->as_int == 0);
+    int lea = (l->as_int == a->as_int);
 
     if (lea && lez)
         return SUBMVREF_LEFT_ABOVE_ZED;

media/libvpx/vp8/common/entropymode.h

@@ -25,7 +25,7 @@ extern const int vp8_mbsplit_count [VP8_NUMMBSPLITS];    /* # of subsets */
 
 extern const vp8_prob vp8_mbsplit_probs [VP8_NUMMBSPLITS-1];
 
-extern int vp8_mv_cont(const MV *l, const MV *a);
+extern int vp8_mv_cont(const int_mv *l, const int_mv *a);
 #define SUBMVREF_COUNT 5
 extern const vp8_prob vp8_sub_mv_ref_prob2 [SUBMVREF_COUNT][VP8_SUBMVREFS-1];
 

media/libvpx/vp8/common/entropymv.h

@@ -18,6 +18,8 @@ enum
 {
     mv_max  = 1023,              /* max absolute value of a MV component */
     MVvals = (2 * mv_max) + 1,   /* # possible values "" */
+    mvfp_max  = 255,              /* max absolute value of a full pixel MV component */
+    MVfpvals = (2 * mvfp_max) +1, /* # possible full pixel MV values */
 
     mvlong_width = 10,       /* Large MVs have 9 bit magnitudes */
     mvnum_short = 8,         /* magnitudes 0 through 7 */

media/libvpx/vp8/common/extend.c

@@ -13,10 +13,12 @@
 #include "vpx_mem/vpx_mem.h"
 
 
-static void extend_plane_borders
+static void copy_and_extend_plane
 (
     unsigned char *s, /* source */
-    int sp,           /* pitch */
+    int sp,           /* source pitch */
+    unsigned char *d, /* destination */
+    int dp,           /* destination pitch */
     int h,            /* height */
     int w,            /* width */
     int et,           /* extend top border */
@@ -25,7 +27,6 @@ static void extend_plane_borders
     int er            /* extend right border */
 )
 {
-
     int i;
     unsigned char *src_ptr1, *src_ptr2;
     unsigned char *dest_ptr1, *dest_ptr2;
@@ -34,68 +35,73 @@ static void extend_plane_borders
     /* copy the left and right most columns out */
     src_ptr1 = s;
     src_ptr2 = s + w - 1;
-    dest_ptr1 = s - el;
-    dest_ptr2 = s + w;
+    dest_ptr1 = d - el;
+    dest_ptr2 = d + w;
 
-    for (i = 0; i < h - 0 + 1; i++)
+    for (i = 0; i < h; i++)
     {
-        /* Some linkers will complain if we call vpx_memset with el set to a
-         * constant 0.
-         */
-        if (el)
-            vpx_memset(dest_ptr1, src_ptr1[0], el);
+        vpx_memset(dest_ptr1, src_ptr1[0], el);
+        vpx_memcpy(dest_ptr1 + el, src_ptr1, w);
         vpx_memset(dest_ptr2, src_ptr2[0], er);
         src_ptr1  += sp;
         src_ptr2  += sp;
-        dest_ptr1 += sp;
-        dest_ptr2 += sp;
+        dest_ptr1 += dp;
+        dest_ptr2 += dp;
     }
 
-    /* Now copy the top and bottom source lines into each line of the respective borders */
-    src_ptr1 = s - el;
-    src_ptr2 = s + sp * (h - 1) - el;
-    dest_ptr1 = s + sp * (-et) - el;
-    dest_ptr2 = s + sp * (h) - el;
-    linesize = el + er + w + 1;
+    /* Now copy the top and bottom lines into each line of the respective
+     * borders
+     */
+    src_ptr1 = d - el;
+    src_ptr2 = d + dp * (h - 1) - el;
+    dest_ptr1 = d + dp * (-et) - el;
+    dest_ptr2 = d + dp * (h) - el;
+    linesize = el + er + w;
 
-    for (i = 0; i < (int)et; i++)
+    for (i = 0; i < et; i++)
     {
         vpx_memcpy(dest_ptr1, src_ptr1, linesize);
-        dest_ptr1 += sp;
+        dest_ptr1 += dp;
     }
 
-    for (i = 0; i < (int)eb; i++)
+    for (i = 0; i < eb; i++)
     {
         vpx_memcpy(dest_ptr2, src_ptr2, linesize);
-        dest_ptr2 += sp;
+        dest_ptr2 += dp;
     }
 }
 
 
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height)
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst)
 {
-    int er = 0xf & (16 - (width & 0xf));
-    int eb = 0xf & (16 - (height & 0xf));
+    int et = dst->border;
+    int el = dst->border;
+    int eb = dst->border + dst->y_height - src->y_height;
+    int er = dst->border + dst->y_width - src->y_width;
 
-    /* check for non multiples of 16 */
-    if (er != 0 || eb != 0)
-    {
-        extend_plane_borders(ybf->y_buffer, ybf->y_stride, height, width, 0, 0, eb, er);
+    copy_and_extend_plane(src->y_buffer, src->y_stride,
+                          dst->y_buffer, dst->y_stride,
+                          src->y_height, src->y_width,
+                          et, el, eb, er);
 
-        /* adjust for uv */
-        height = (height + 1) >> 1;
-        width  = (width  + 1) >> 1;
-        er = 0x7 & (8 - (width  & 0x7));
-        eb = 0x7 & (8 - (height & 0x7));
+    et = dst->border >> 1;
+    el = dst->border >> 1;
+    eb = (dst->border >> 1) + dst->uv_height - src->uv_height;
+    er = (dst->border >> 1) + dst->uv_width - src->uv_width;
 
-        if (er || eb)
-        {
-            extend_plane_borders(ybf->u_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-            extend_plane_borders(ybf->v_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-        }
-    }
+    copy_and_extend_plane(src->u_buffer, src->uv_stride,
+                          dst->u_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
+
+    copy_and_extend_plane(src->v_buffer, src->uv_stride,
+                          dst->v_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
 }
 
+
 /* note the extension is only for the last row, for intra prediction purpose */
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr)
 {

media/libvpx/vp8/common/extend.h

@@ -14,8 +14,8 @@
 
 #include "vpx_scale/yv12config.h"
 
-void Extend(YV12_BUFFER_CONFIG *ybf);
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr);
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height);
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst);
 
 #endif

media/libvpx/vp8/common/filter.c

@@ -10,13 +10,10 @@
 
 
 #include <stdlib.h>
+#include "filter.h"
+#include "vpx_ports/mem.h"
 
-#define BLOCK_HEIGHT_WIDTH 4
-#define VP8_FILTER_WEIGHT 128
-#define VP8_FILTER_SHIFT  7
-
-
-static const int bilinear_filters[8][2] =
+DECLARE_ALIGNED(16, const short, vp8_bilinear_filters[8][2]) =
 {
     { 128,   0 },
     { 112,  16 },
@@ -28,8 +25,7 @@ static const int bilinear_filters[8][2] =
     {  16, 112 }
 };
 
-
-static const short sub_pel_filters[8][6] =
+DECLARE_ALIGNED(16, const short, vp8_sub_pel_filters[8][6]) =
 {
 
     { 0,  0,  128,    0,   0,  0 },         /* note that 1/8 pel positions are just as per alpha -0.5 bicubic */
@@ -40,12 +36,9 @@ static const short sub_pel_filters[8][6] =
     { 0, -6,   50,   93,  -9,  0 },
     { 1, -8,   36,  108, -11,  2 },         /* New 1/4 pel 6 tap filter */
     { 0, -1,   12,  123,  -6,  0 },
-
-
-
 };
 
-void vp8_filter_block2d_first_pass
+static void filter_block2d_first_pass
 (
     unsigned char *src_ptr,
     int *output_ptr,
@@ -89,7 +82,7 @@ void vp8_filter_block2d_first_pass
     }
 }
 
-void vp8_filter_block2d_second_pass
+static void filter_block2d_second_pass
 (
     int *src_ptr,
     unsigned char *output_ptr,
@@ -136,7 +129,7 @@ void vp8_filter_block2d_second_pass
 }
 
 
-void vp8_filter_block2d
+static void filter_block2d
 (
     unsigned char  *src_ptr,
     unsigned char  *output_ptr,
@@ -146,42 +139,16 @@ void vp8_filter_block2d
     const short  *VFilter
 )
 {
-    int FData[9*4]; /* Temp data bufffer used in filtering */
+    int FData[9*4]; /* Temp data buffer used in filtering */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 4, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 4, HFilter);
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 8, output_ptr, output_pitch, 4, 4, 4, 4, VFilter);
+    filter_block2d_second_pass(FData + 8, output_ptr, output_pitch, 4, 4, 4, 4, VFilter);
 }
 
 
-void vp8_block_variation_c
-(
-    unsigned char  *src_ptr,
-    int   src_pixels_per_line,
-    int *HVar,
-    int *VVar
-)
-{
-    int i, j;
-    unsigned char *Ptr = src_ptr;
-
-    for (i = 0; i < 4; i++)
-    {
-        for (j = 0; j < 4; j++)
-        {
-            *HVar += abs((int)Ptr[j] - (int)Ptr[j+1]);
-            *VVar += abs((int)Ptr[j] - (int)Ptr[j+src_pixels_per_line]);
-        }
-
-        Ptr += src_pixels_per_line;
-    }
-}
-
-
-
-
 void vp8_sixtap_predict_c
 (
     unsigned char  *src_ptr,
@@ -195,10 +162,10 @@ void vp8_sixtap_predict_c
     const short  *HFilter;
     const short  *VFilter;
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];   /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
-    vp8_filter_block2d(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter);
+    filter_block2d(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter);
 }
 void vp8_sixtap_predict8x8_c
 (
@@ -212,17 +179,17 @@ void vp8_sixtap_predict8x8_c
 {
     const short  *HFilter;
     const short  *VFilter;
-    int FData[13*16];   /* Temp data bufffer used in filtering */
+    int FData[13*16];   /* Temp data buffer used in filtering */
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];   /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 13, 8, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 13, 8, HFilter);
 
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 8, 8, VFilter);
+    filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 8, 8, VFilter);
 
 }
 
@@ -238,17 +205,17 @@ void vp8_sixtap_predict8x4_c
 {
     const short  *HFilter;
     const short  *VFilter;
-    int FData[13*16];   /* Temp data bufffer used in filtering */
+    int FData[13*16];   /* Temp data buffer used in filtering */
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];   /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 8, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 8, HFilter);
 
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 4, 8, VFilter);
+    filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 4, 8, VFilter);
 
 }
 
@@ -264,17 +231,17 @@ void vp8_sixtap_predict16x16_c
 {
     const short  *HFilter;
     const short  *VFilter;
-    int FData[21*24];   /* Temp data bufffer used in filtering */
+    int FData[21*24];   /* Temp data buffer used in filtering */
 
 
-    HFilter = sub_pel_filters[xoffset];   /* 6 tap */
-    VFilter = sub_pel_filters[yoffset];   /* 6 tap */
+    HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
+    VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 21, 16, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 21, 16, HFilter);
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 32, dst_ptr, dst_pitch, 16, 16, 16, 16, VFilter);
+    filter_block2d_second_pass(FData + 32, dst_ptr, dst_pitch, 16, 16, 16, 16, VFilter);
 
 }
 
@@ -283,57 +250,50 @@ void vp8_sixtap_predict16x16_c
  *
  *  ROUTINE       : filter_block2d_bil_first_pass
  *
- *  INPUTS        : UINT8  *src_ptr          : Pointer to source block.
- *                  UINT32 src_pixels_per_line : Stride of input block.
- *                  UINT32 pixel_step        : Offset between filter input samples (see notes).
- *                  UINT32 output_height     : Input block height.
- *                  UINT32 output_width      : Input block width.
- *                  INT32  *vp8_filter          : Array of 2 bi-linear filter taps.
+ *  INPUTS        : UINT8  *src_ptr    : Pointer to source block.
+ *                  UINT32  src_stride : Stride of source block.
+ *                  UINT32  height     : Block height.
+ *                  UINT32  width      : Block width.
+ *                  INT32  *vp8_filter : Array of 2 bi-linear filter taps.
  *
- *  OUTPUTS       : INT32 *output_ptr        : Pointer to filtered block.
+ *  OUTPUTS       : INT32  *dst_ptr    : Pointer to filtered block.
  *
  *  RETURNS       : void
  *
- *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block in
- *                  either horizontal or vertical direction to produce the
- *                  filtered output block. Used to implement first-pass
- *                  of 2-D separable filter.
+ *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block
+ *                  in the horizontal direction to produce the filtered output
+ *                  block. Used to implement first-pass of 2-D separable filter.
  *
  *  SPECIAL NOTES : Produces INT32 output to retain precision for next pass.
  *                  Two filter taps should sum to VP8_FILTER_WEIGHT.
- *                  pixel_step defines whether the filter is applied
- *                  horizontally (pixel_step=1) or vertically (pixel_step=stride).
- *                  It defines the offset required to move from one input
- *                  to the next.
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil_first_pass
+static void filter_block2d_bil_first_pass
 (
-    unsigned char *src_ptr,
-    unsigned short *output_ptr,
-    unsigned int src_pixels_per_line,
-    int pixel_step,
-    unsigned int output_height,
-    unsigned int output_width,
-    const int *vp8_filter
+    unsigned char  *src_ptr,
+    unsigned short *dst_ptr,
+    unsigned int    src_stride,
+    unsigned int    height,
+    unsigned int    width,
+    const short    *vp8_filter
 )
 {
     unsigned int i, j;
 
-    for (i = 0; i < output_height; i++)
+    for (i = 0; i < height; i++)
     {
-        for (j = 0; j < output_width; j++)
+        for (j = 0; j < width; j++)
         {
             /* Apply bilinear filter */
-            output_ptr[j] = (((int)src_ptr[0]          * vp8_filter[0]) +
-                             ((int)src_ptr[pixel_step] * vp8_filter[1]) +
-                             (VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
+            dst_ptr[j] = (((int)src_ptr[0] * vp8_filter[0]) +
+                          ((int)src_ptr[1] * vp8_filter[1]) +
+                          (VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
             src_ptr++;
         }
 
         /* Next row... */
-        src_ptr    += src_pixels_per_line - output_width;
-        output_ptr += output_width;
+        src_ptr += src_stride - width;
+        dst_ptr += width;
     }
 }
 
@@ -341,60 +301,51 @@ void vp8_filter_block2d_bil_first_pass
  *
  *  ROUTINE       : filter_block2d_bil_second_pass
  *
- *  INPUTS        : INT32  *src_ptr          : Pointer to source block.
- *                  UINT32 src_pixels_per_line : Stride of input block.
- *                  UINT32 pixel_step        : Offset between filter input samples (see notes).
- *                  UINT32 output_height     : Input block height.
- *                  UINT32 output_width      : Input block width.
- *                  INT32  *vp8_filter          : Array of 2 bi-linear filter taps.
+ *  INPUTS        : INT32  *src_ptr    : Pointer to source block.
+ *                  UINT32  dst_pitch  : Destination block pitch.
+ *                  UINT32  height     : Block height.
+ *                  UINT32  width      : Block width.
+ *                  INT32  *vp8_filter : Array of 2 bi-linear filter taps.
  *
- *  OUTPUTS       : UINT16 *output_ptr       : Pointer to filtered block.
+ *  OUTPUTS       : UINT16 *dst_ptr    : Pointer to filtered block.
  *
  *  RETURNS       : void
  *
- *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block in
- *                  either horizontal or vertical direction to produce the
- *                  filtered output block. Used to implement second-pass
- *                  of 2-D separable filter.
+ *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block
+ *                  in the vertical direction to produce the filtered output
+ *                  block. Used to implement second-pass of 2-D separable filter.
  *
  *  SPECIAL NOTES : Requires 32-bit input as produced by filter_block2d_bil_first_pass.
  *                  Two filter taps should sum to VP8_FILTER_WEIGHT.
- *                  pixel_step defines whether the filter is applied
- *                  horizontally (pixel_step=1) or vertically (pixel_step=stride).
- *                  It defines the offset required to move from one input
- *                  to the next.
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil_second_pass
+static void filter_block2d_bil_second_pass
 (
     unsigned short *src_ptr,
-    unsigned char  *output_ptr,
-    int output_pitch,
-    unsigned int  src_pixels_per_line,
-    unsigned int  pixel_step,
-    unsigned int  output_height,
-    unsigned int  output_width,
-    const int *vp8_filter
+    unsigned char  *dst_ptr,
+    int             dst_pitch,
+    unsigned int    height,
+    unsigned int    width,
+    const short    *vp8_filter
 )
 {
     unsigned int  i, j;
     int  Temp;
 
-    for (i = 0; i < output_height; i++)
+    for (i = 0; i < height; i++)
     {
-        for (j = 0; j < output_width; j++)
+        for (j = 0; j < width; j++)
         {
             /* Apply filter */
-            Temp = ((int)src_ptr[0]         * vp8_filter[0]) +
-                   ((int)src_ptr[pixel_step] * vp8_filter[1]) +
+            Temp = ((int)src_ptr[0]     * vp8_filter[0]) +
+                   ((int)src_ptr[width] * vp8_filter[1]) +
                    (VP8_FILTER_WEIGHT / 2);
-            output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
+            dst_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
             src_ptr++;
         }
 
         /* Next row... */
-        src_ptr    += src_pixels_per_line - output_width;
-        output_ptr += output_pitch;
+        dst_ptr += dst_pitch;
     }
 }
 
@@ -404,11 +355,14 @@ void vp8_filter_block2d_bil_second_pass
  *  ROUTINE       : filter_block2d_bil
  *
  *  INPUTS        : UINT8  *src_ptr          : Pointer to source block.
- *                  UINT32 src_pixels_per_line : Stride of input block.
- *                  INT32  *HFilter         : Array of 2 horizontal filter taps.
- *                  INT32  *VFilter         : Array of 2 vertical filter taps.
+ *                  UINT32  src_pitch        : Stride of source block.
+ *                  UINT32  dst_pitch        : Stride of destination block.
+ *                  INT32  *HFilter          : Array of 2 horizontal filter taps.
+ *                  INT32  *VFilter          : Array of 2 vertical filter taps.
+ *                  INT32  Width             : Block width
+ *                  INT32  Height            : Block height
  *
- *  OUTPUTS       : UINT16 *output_ptr       : Pointer to filtered block.
+ *  OUTPUTS       : UINT16 *dst_ptr       : Pointer to filtered block.
  *
  *  RETURNS       : void
  *
@@ -419,26 +373,26 @@ void vp8_filter_block2d_bil_second_pass
  *  SPECIAL NOTES : The largest block size can be handled here is 16x16
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil
+static void filter_block2d_bil
 (
     unsigned char *src_ptr,
-    unsigned char *output_ptr,
-    unsigned int   src_pixels_per_line,
+    unsigned char *dst_ptr,
+    unsigned int   src_pitch,
     unsigned int   dst_pitch,
-    const int      *HFilter,
-    const int      *VFilter,
+    const short   *HFilter,
+    const short   *VFilter,
     int            Width,
     int            Height
 )
 {
 
-    unsigned short FData[17*16];    /* Temp data bufffer used in filtering */
+    unsigned short FData[17*16];    /* Temp data buffer used in filtering */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_bil_first_pass(src_ptr, FData, src_pixels_per_line, 1, Height + 1, Width, HFilter);
+    filter_block2d_bil_first_pass(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
 
     /* then 1-D vertically... */
-    vp8_filter_block2d_bil_second_pass(FData, output_ptr, dst_pitch, Width, Width, Height, Width, VFilter);
+    filter_block2d_bil_second_pass(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
 }
 
 
@@ -452,11 +406,11 @@ void vp8_bilinear_predict4x4_c
     int dst_pitch
 )
 {
-    const int  *HFilter;
-    const int  *VFilter;
+    const short *HFilter;
+    const short *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 #if 0
     {
         int i;
@@ -464,19 +418,19 @@ void vp8_bilinear_predict4x4_c
         unsigned char temp2[16];
 
         bilinear_predict4x4_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, temp1, 4);
-        vp8_filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
+        filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
 
         for (i = 0; i < 16; i++)
         {
             if (temp1[i] != temp2[i])
             {
                 bilinear_predict4x4_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, temp1, 4);
-                vp8_filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
+                filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
             }
         }
     }
 #endif
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
 
 }
 
@@ -490,13 +444,13 @@ void vp8_bilinear_predict8x8_c
     int  dst_pitch
 )
 {
-    const int  *HFilter;
-    const int  *VFilter;
+    const short *HFilter;
+    const short *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
 
 }
 
@@ -510,13 +464,13 @@ void vp8_bilinear_predict8x4_c
     int  dst_pitch
 )
 {
-    const int  *HFilter;
-    const int  *VFilter;
+    const short *HFilter;
+    const short *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
 
 }
 
@@ -530,11 +484,11 @@ void vp8_bilinear_predict16x16_c
     int  dst_pitch
 )
 {
-    const int  *HFilter;
-    const int  *VFilter;
+    const short *HFilter;
+    const short *VFilter;
 
-    HFilter = bilinear_filters[xoffset];
-    VFilter = bilinear_filters[yoffset];
+    HFilter = vp8_bilinear_filters[xoffset];
+    VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
 }

media/libvpx/vp8/common/filter.h

@@ -1,5 +1,5 @@
 /*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *  Copyright (c) 2011 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
@@ -9,14 +9,14 @@
  */
 
 
-#ifndef DETOKENIZE_ARM_H
-#define DETOKENIZE_ARM_H
+#ifndef FILTER_H
+#define FILTER_H
 
-#if HAVE_ARMV6
-#if CONFIG_ARM_ASM_DETOK
-void vp8_init_detokenizer(VP8D_COMP *dx);
-void vp8_decode_mb_tokens_v6(DETOK *detoken, int type);
-#endif
-#endif
+#define BLOCK_HEIGHT_WIDTH 4
+#define VP8_FILTER_WEIGHT 128
+#define VP8_FILTER_SHIFT  7
 
-#endif
+extern const short vp8_bilinear_filters[8][2];
+extern const short vp8_sub_pel_filters[8][6];
+
+#endif //FILTER_H

media/libvpx/vp8/common/findnearmv.c

@@ -11,54 +11,23 @@
 
 #include "findnearmv.h"
 
-#define FINDNEAR_SEARCH_SITES   3
+const unsigned char vp8_mbsplit_offset[4][16] = {
+    { 0,  8,  0,  0,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  2,  0,  0,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  2,  8, 10,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9,  10, 11, 12, 13, 14, 15}
+};
 
 /* Predict motion vectors using those from already-decoded nearby blocks.
    Note that we only consider one 4x4 subblock from each candidate 16x16
    macroblock.   */
-
-typedef union
-{
-    unsigned int as_int;
-    MV           as_mv;
-} int_mv;        /* facilitates rapid equality tests */
-
-static void mv_bias(const MODE_INFO *x, int refframe, int_mv *mvp, const int *ref_frame_sign_bias)
-{
-    MV xmv;
-    xmv = x->mbmi.mv.as_mv;
-
-    if (ref_frame_sign_bias[x->mbmi.ref_frame] != ref_frame_sign_bias[refframe])
-    {
-        xmv.row *= -1;
-        xmv.col *= -1;
-    }
-
-    mvp->as_mv = xmv;
-}
-
-
-void vp8_clamp_mv(MV *mv, const MACROBLOCKD *xd)
-{
-    if (mv->col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN))
-        mv->col = xd->mb_to_left_edge - LEFT_TOP_MARGIN;
-    else if (mv->col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN)
-        mv->col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN;
-
-    if (mv->row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN))
-        mv->row = xd->mb_to_top_edge - LEFT_TOP_MARGIN;
-    else if (mv->row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN)
-        mv->row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN;
-}
-
-
 void vp8_find_near_mvs
 (
     MACROBLOCKD *xd,
     const MODE_INFO *here,
-    MV *nearest,
-    MV *nearby,
-    MV *best_mv,
+    int_mv *nearest,
+    int_mv *nearby,
+    int_mv *best_mv,
     int cnt[4],
     int refframe,
     int *ref_frame_sign_bias
@@ -82,7 +51,7 @@ void vp8_find_near_mvs
         if (above->mbmi.mv.as_int)
         {
             (++mv)->as_int = above->mbmi.mv.as_int;
-            mv_bias(above, refframe, mv, ref_frame_sign_bias);
+            mv_bias(ref_frame_sign_bias[above->mbmi.ref_frame], refframe, mv, ref_frame_sign_bias);
             ++cntx;
         }
 
@@ -97,7 +66,7 @@ void vp8_find_near_mvs
             int_mv this_mv;
 
             this_mv.as_int = left->mbmi.mv.as_int;
-            mv_bias(left, refframe, &this_mv, ref_frame_sign_bias);
+            mv_bias(ref_frame_sign_bias[left->mbmi.ref_frame], refframe, &this_mv, ref_frame_sign_bias);
 
             if (this_mv.as_int != mv->as_int)
             {
@@ -119,7 +88,7 @@ void vp8_find_near_mvs
             int_mv this_mv;
 
             this_mv.as_int = aboveleft->mbmi.mv.as_int;
-            mv_bias(aboveleft, refframe, &this_mv, ref_frame_sign_bias);
+            mv_bias(ref_frame_sign_bias[aboveleft->mbmi.ref_frame], refframe, &this_mv, ref_frame_sign_bias);
 
             if (this_mv.as_int != mv->as_int)
             {
@@ -162,13 +131,14 @@ void vp8_find_near_mvs
         near_mvs[CNT_INTRA] = near_mvs[CNT_NEAREST];
 
     /* Set up return values */
-    *best_mv = near_mvs[0].as_mv;
-    *nearest = near_mvs[CNT_NEAREST].as_mv;
-    *nearby = near_mvs[CNT_NEAR].as_mv;
+    best_mv->as_int = near_mvs[0].as_int;
+    nearest->as_int = near_mvs[CNT_NEAREST].as_int;
+    nearby->as_int = near_mvs[CNT_NEAR].as_int;
 
-    vp8_clamp_mv(nearest, xd);
-    vp8_clamp_mv(nearby, xd);
-    vp8_clamp_mv(best_mv, xd); /*TODO: move this up before the copy*/
+    //TODO: move clamp outside findnearmv
+    vp8_clamp_mv2(nearest, xd);
+    vp8_clamp_mv2(nearby, xd);
+    vp8_clamp_mv2(best_mv, xd);
 }
 
 vp8_prob *vp8_mv_ref_probs(
@@ -183,26 +153,3 @@ vp8_prob *vp8_mv_ref_probs(
     return p;
 }
 
-const B_MODE_INFO *vp8_left_bmi(const MODE_INFO *cur_mb, int b)
-{
-    if (!(b & 3))
-    {
-        /* On L edge, get from MB to left of us */
-        --cur_mb;
-        b += 4;
-    }
-
-    return cur_mb->bmi + b - 1;
-}
-
-const B_MODE_INFO *vp8_above_bmi(const MODE_INFO *cur_mb, int b, int mi_stride)
-{
-    if (!(b >> 2))
-    {
-        /* On top edge, get from MB above us */
-        cur_mb -= mi_stride;
-        b += 16;
-    }
-
-    return cur_mb->bmi + b - 4;
-}

media/libvpx/vp8/common/findnearmv.h

@@ -17,11 +17,65 @@
 #include "modecont.h"
 #include "treecoder.h"
 
+
+static void mv_bias(int refmb_ref_frame_sign_bias, int refframe, int_mv *mvp, const int *ref_frame_sign_bias)
+{
+    MV xmv;
+    xmv = mvp->as_mv;
+
+    if (refmb_ref_frame_sign_bias != ref_frame_sign_bias[refframe])
+    {
+        xmv.row *= -1;
+        xmv.col *= -1;
+    }
+
+    mvp->as_mv = xmv;
+}
+
+#define LEFT_TOP_MARGIN (16 << 3)
+#define RIGHT_BOTTOM_MARGIN (16 << 3)
+static void vp8_clamp_mv2(int_mv *mv, const MACROBLOCKD *xd)
+{
+    if (mv->as_mv.col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN))
+        mv->as_mv.col = xd->mb_to_left_edge - LEFT_TOP_MARGIN;
+    else if (mv->as_mv.col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN)
+        mv->as_mv.col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN;
+
+    if (mv->as_mv.row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN))
+        mv->as_mv.row = xd->mb_to_top_edge - LEFT_TOP_MARGIN;
+    else if (mv->as_mv.row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN)
+        mv->as_mv.row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN;
+}
+
+static void vp8_clamp_mv(int_mv *mv, int mb_to_left_edge, int mb_to_right_edge,
+                         int mb_to_top_edge, int mb_to_bottom_edge)
+{
+    mv->as_mv.col = (mv->as_mv.col < mb_to_left_edge) ?
+        mb_to_left_edge : mv->as_mv.col;
+    mv->as_mv.col = (mv->as_mv.col > mb_to_right_edge) ?
+        mb_to_right_edge : mv->as_mv.col;
+    mv->as_mv.row = (mv->as_mv.row < mb_to_top_edge) ?
+        mb_to_top_edge : mv->as_mv.row;
+    mv->as_mv.row = (mv->as_mv.row > mb_to_bottom_edge) ?
+        mb_to_bottom_edge : mv->as_mv.row;
+}
+static unsigned int vp8_check_mv_bounds(int_mv *mv, int mb_to_left_edge,
+                                int mb_to_right_edge, int mb_to_top_edge,
+                                int mb_to_bottom_edge)
+{
+    unsigned int need_to_clamp;
+    need_to_clamp = (mv->as_mv.col < mb_to_left_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.col > mb_to_right_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.row < mb_to_top_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.row > mb_to_bottom_edge) ? 1 : 0;
+    return need_to_clamp;
+}
+
 void vp8_find_near_mvs
 (
     MACROBLOCKD *xd,
     const MODE_INFO *here,
-    MV *nearest, MV *nearby, MV *best,
+    int_mv *nearest, int_mv *nearby, int_mv *best,
     int near_mv_ref_cts[4],
     int refframe,
     int *ref_frame_sign_bias
@@ -31,12 +85,89 @@ vp8_prob *vp8_mv_ref_probs(
     vp8_prob p[VP8_MVREFS-1], const int near_mv_ref_ct[4]
 );
 
-const B_MODE_INFO *vp8_left_bmi(const MODE_INFO *cur_mb, int b);
+extern const unsigned char vp8_mbsplit_offset[4][16];
 
-const B_MODE_INFO *vp8_above_bmi(const MODE_INFO *cur_mb, int b, int mi_stride);
 
-#define LEFT_TOP_MARGIN (16 << 3)
-#define RIGHT_BOTTOM_MARGIN (16 << 3)
+static int left_block_mv(const MODE_INFO *cur_mb, int b)
+{
+    if (!(b & 3))
+    {
+        /* On L edge, get from MB to left of us */
+        --cur_mb;
 
+        if(cur_mb->mbmi.mode != SPLITMV)
+            return cur_mb->mbmi.mv.as_int;
+        b += 4;
+    }
+
+    return (cur_mb->bmi + b - 1)->mv.as_int;
+}
+
+static int above_block_mv(const MODE_INFO *cur_mb, int b, int mi_stride)
+{
+    if (!(b >> 2))
+    {
+        /* On top edge, get from MB above us */
+        cur_mb -= mi_stride;
+
+        if(cur_mb->mbmi.mode != SPLITMV)
+            return cur_mb->mbmi.mv.as_int;
+        b += 16;
+    }
+
+    return (cur_mb->bmi + b - 4)->mv.as_int;
+}
+static B_PREDICTION_MODE left_block_mode(const MODE_INFO *cur_mb, int b)
+{
+    if (!(b & 3))
+    {
+        /* On L edge, get from MB to left of us */
+        --cur_mb;
+        switch (cur_mb->mbmi.mode)
+        {
+            case B_PRED:
+              return (cur_mb->bmi + b + 3)->as_mode;
+            case DC_PRED:
+                return B_DC_PRED;
+            case V_PRED:
+                return B_VE_PRED;
+            case H_PRED:
+                return B_HE_PRED;
+            case TM_PRED:
+                return B_TM_PRED;
+            default:
+                return B_DC_PRED;
+        }
+    }
+
+    return (cur_mb->bmi + b - 1)->as_mode;
+}
+
+static B_PREDICTION_MODE above_block_mode(const MODE_INFO *cur_mb, int b, int mi_stride)
+{
+    if (!(b >> 2))
+    {
+        /* On top edge, get from MB above us */
+        cur_mb -= mi_stride;
+
+        switch (cur_mb->mbmi.mode)
+        {
+            case B_PRED:
+              return (cur_mb->bmi + b + 12)->as_mode;
+            case DC_PRED:
+                return B_DC_PRED;
+            case V_PRED:
+                return B_VE_PRED;
+            case H_PRED:
+                return B_HE_PRED;
+            case TM_PRED:
+                return B_TM_PRED;
+            default:
+                return B_DC_PRED;
+        }
+    }
+
+    return (cur_mb->bmi + b - 4)->as_mode;
+}
 
 #endif

media/libvpx/vp8/common/generic/systemdependent.c

@@ -10,16 +10,61 @@
 
 
 #include "vpx_ports/config.h"
-#include "g_common.h"
-#include "subpixel.h"
-#include "loopfilter.h"
-#include "recon.h"
-#include "idct.h"
-#include "onyxc_int.h"
+#include "vp8/common/g_common.h"
+#include "vp8/common/subpixel.h"
+#include "vp8/common/loopfilter.h"
+#include "vp8/common/recon.h"
+#include "vp8/common/idct.h"
+#include "vp8/common/onyxc_int.h"
+
+#if CONFIG_MULTITHREAD
+#if HAVE_UNISTD_H
+#include <unistd.h>
+#elif defined(_WIN32)
+#include <windows.h>
+typedef void (WINAPI *PGNSI)(LPSYSTEM_INFO);
+#endif
+#endif
 
 extern void vp8_arch_x86_common_init(VP8_COMMON *ctx);
 extern void vp8_arch_arm_common_init(VP8_COMMON *ctx);
 
+#if CONFIG_MULTITHREAD
+static int get_cpu_count()
+{
+    int core_count = 16;
+
+#if HAVE_UNISTD_H
+#if defined(_SC_NPROCESSORS_ONLN)
+    core_count = sysconf(_SC_NPROCESSORS_ONLN);
+#elif defined(_SC_NPROC_ONLN)
+    core_count = sysconf(_SC_NPROC_ONLN);
+#endif
+#elif defined(_WIN32)
+    {
+        PGNSI pGNSI;
+        SYSTEM_INFO sysinfo;
+
+        /* Call GetNativeSystemInfo if supported or
+         * GetSystemInfo otherwise. */
+
+        pGNSI = (PGNSI) GetProcAddress(
+                GetModuleHandle(TEXT("kernel32.dll")), "GetNativeSystemInfo");
+        if (pGNSI != NULL)
+            pGNSI(&sysinfo);
+        else
+            GetSystemInfo(&sysinfo);
+
+        core_count = sysinfo.dwNumberOfProcessors;
+    }
+#else
+    /* other platforms */
+#endif
+
+    return core_count > 0 ? core_count : 1;
+}
+#endif
+
 void vp8_machine_specific_config(VP8_COMMON *ctx)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -43,6 +88,12 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
         vp8_build_intra_predictors_mby;
     rtcd->recon.build_intra_predictors_mby_s =
         vp8_build_intra_predictors_mby_s;
+    rtcd->recon.build_intra_predictors_mbuv =
+        vp8_build_intra_predictors_mbuv;
+    rtcd->recon.build_intra_predictors_mbuv_s =
+        vp8_build_intra_predictors_mbuv_s;
+    rtcd->recon.intra4x4_predict =
+        vp8_intra4x4_predict;
 
     rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_c;
     rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_c;
@@ -57,17 +108,19 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_c;
     rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_c;
     rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_c;
-    rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_c;
+    rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_c;
     rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_c;
-    rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_c;
+    rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_c;
     rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_c;
 
-#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_PSNR)
-    rtcd->postproc.down        = vp8_mbpost_proc_down_c;
-    rtcd->postproc.across      = vp8_mbpost_proc_across_ip_c;
-    rtcd->postproc.downacross  = vp8_post_proc_down_and_across_c;
-    rtcd->postproc.addnoise    = vp8_plane_add_noise_c;
-    rtcd->postproc.blend_mb    = vp8_blend_mb_c;
+#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_INTERNAL_STATS)
+    rtcd->postproc.down             = vp8_mbpost_proc_down_c;
+    rtcd->postproc.across           = vp8_mbpost_proc_across_ip_c;
+    rtcd->postproc.downacross       = vp8_post_proc_down_and_across_c;
+    rtcd->postproc.addnoise         = vp8_plane_add_noise_c;
+    rtcd->postproc.blend_mb_inner   = vp8_blend_mb_inner_c;
+    rtcd->postproc.blend_mb_outer   = vp8_blend_mb_outer_c;
+    rtcd->postproc.blend_b          = vp8_blend_b_c;
 #endif
 
 #endif
@@ -80,4 +133,7 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     vp8_arch_arm_common_init(ctx);
 #endif
 
+#if CONFIG_MULTITHREAD
+    ctx->processor_core_count = get_cpu_count();
+#endif /* CONFIG_MULTITHREAD */
 }

media/libvpx/vp8/common/loopfilter.c

@@ -9,162 +9,149 @@
  */
 
 
-#include "vpx_ports/config.h"
+#include "vpx_config.h"
 #include "loopfilter.h"
 #include "onyxc_int.h"
+#include "vpx_mem/vpx_mem.h"
 
 typedef unsigned char uc;
 
-
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_c);
 prototype_loopfilter(vp8_loop_filter_vertical_edge_c);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_c);
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
+
+prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
+prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
 
 /* Horizontal MB filtering */
-void vp8_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                           unsigned char *v_ptr, int y_stride, int uv_stride,
+                           loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-void vp8_loop_filter_mbhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Vertical MB Filtering */
-void vp8_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                           unsigned char *v_ptr, int y_stride, int uv_stride,
+                           loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-void vp8_loop_filter_mbvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Horizontal B Filtering */
-void vp8_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                          unsigned char *v_ptr, int y_stride, int uv_stride,
+                          loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_c(unsigned char *y_ptr, int y_stride,
+                           const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 /* Vertical B Filtering */
-void vp8_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                          unsigned char *v_ptr, int y_stride, int uv_stride,
+                          loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_c(unsigned char *y_ptr, int y_stride,
+                           const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit);
 }
 
-void vp8_init_loop_filter(VP8_COMMON *cm)
+static void lf_init_lut(loop_filter_info_n *lfi)
 {
-    loop_filter_info *lfi = cm->lf_info;
-    LOOPFILTERTYPE lft = cm->filter_type;
-    int sharpness_lvl = cm->sharpness_level;
-    int frame_type = cm->frame_type;
-    int i, j;
+    int filt_lvl;
 
-    int block_inside_limit = 0;
-    int HEVThresh;
-    const int yhedge_boost  = 2;
-    const int uvhedge_boost = 2;
-
-    /* For each possible value for the loop filter fill out a "loop_filter_info" entry. */
-    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    for (filt_lvl = 0; filt_lvl <= MAX_LOOP_FILTER; filt_lvl++)
     {
-        int filt_lvl = i;
-
-        if (frame_type == KEY_FRAME)
+        if (filt_lvl >= 40)
         {
-            if (filt_lvl >= 40)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 2;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 3;
+        }
+        else if (filt_lvl >= 20)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 2;
+        }
+        else if (filt_lvl >= 15)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 1;
         }
         else
         {
-            if (filt_lvl >= 40)
-                HEVThresh = 3;
-            else if (filt_lvl >= 20)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 0;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 0;
         }
+    }
+
+    lfi->mode_lf_lut[DC_PRED] = 1;
+    lfi->mode_lf_lut[V_PRED] = 1;
+    lfi->mode_lf_lut[H_PRED] = 1;
+    lfi->mode_lf_lut[TM_PRED] = 1;
+    lfi->mode_lf_lut[B_PRED]  = 0;
+
+    lfi->mode_lf_lut[ZEROMV]  = 1;
+    lfi->mode_lf_lut[NEARESTMV] = 2;
+    lfi->mode_lf_lut[NEARMV] = 2;
+    lfi->mode_lf_lut[NEWMV] = 2;
+    lfi->mode_lf_lut[SPLITMV] = 3;
+
+}
+
+void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
+                                      int sharpness_lvl)
+{
+    int i;
+
+    /* For each possible value for the loop filter fill out limits */
+    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    {
+        int filt_lvl = i;
+        int block_inside_limit = 0;
 
         /* Set loop filter paramaeters that control sharpness. */
         block_inside_limit = filt_lvl >> (sharpness_lvl > 0);
@@ -179,181 +166,143 @@ void vp8_init_loop_filter(VP8_COMMON *cm)
         if (block_inside_limit < 1)
             block_inside_limit = 1;
 
-        for (j = 0; j < 16; j++)
-        {
-            lfi[i].lim[j] = block_inside_limit;
-            lfi[i].mbflim[j] = filt_lvl + yhedge_boost;
-            lfi[i].mbthr[j] = HEVThresh;
-            lfi[i].flim[j] = filt_lvl;
-            lfi[i].thr[j] = HEVThresh;
-            lfi[i].uvlim[j] = block_inside_limit;
-            lfi[i].uvmbflim[j] = filt_lvl + uvhedge_boost;
-            lfi[i].uvmbthr[j] = HEVThresh;
-            lfi[i].uvflim[j] = filt_lvl;
-            lfi[i].uvthr[j] = HEVThresh;
-        }
-
-    }
-
-    /* Set up the function pointers depending on the type of loop filtering selected */
-    if (lft == NORMAL_LOOPFILTER)
-    {
-        cm->lf_mbv = LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v);
-        cm->lf_bv  = LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v);
-        cm->lf_mbh = LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h);
-        cm->lf_bh  = LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h);
-    }
-    else
-    {
-        cm->lf_mbv = LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v);
-        cm->lf_bv  = LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v);
-        cm->lf_mbh = LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h);
-        cm->lf_bh  = LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h);
+        vpx_memset(lfi->lim[i], block_inside_limit, SIMD_WIDTH);
+        vpx_memset(lfi->blim[i], (2 * filt_lvl + block_inside_limit),
+                SIMD_WIDTH);
+        vpx_memset(lfi->mblim[i], (2 * (filt_lvl + 2) + block_inside_limit),
+                SIMD_WIDTH);
     }
 }
 
-/* Put vp8_init_loop_filter() in vp8dx_create_decompressor(). Only call vp8_frame_init_loop_filter() while decoding
- * each frame. Check last_frame_type to skip the function most of times.
- */
-void vp8_frame_init_loop_filter(loop_filter_info *lfi, int frame_type)
+void vp8_loop_filter_init(VP8_COMMON *cm)
 {
-    int HEVThresh;
-    int i, j;
+    loop_filter_info_n *lfi = &cm->lf_info;
+    int i;
 
-    /* For each possible value for the loop filter fill out a "loop_filter_info" entry. */
-    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    /* init limits for given sharpness*/
+    vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+    cm->last_sharpness_level = cm->sharpness_level;
+
+    /* init LUT for lvl  and hev thr picking */
+    lf_init_lut(lfi);
+
+    /* init hev threshold const vectors */
+    for(i = 0; i < 4 ; i++)
     {
-        int filt_lvl = i;
-
-        if (frame_type == KEY_FRAME)
-        {
-            if (filt_lvl >= 40)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
-        }
-        else
-        {
-            if (filt_lvl >= 40)
-                HEVThresh = 3;
-            else if (filt_lvl >= 20)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
-        }
-
-        for (j = 0; j < 16; j++)
-        {
-            /*lfi[i].lim[j] = block_inside_limit;
-            lfi[i].mbflim[j] = filt_lvl+yhedge_boost;*/
-            lfi[i].mbthr[j] = HEVThresh;
-            /*lfi[i].flim[j] = filt_lvl;*/
-            lfi[i].thr[j] = HEVThresh;
-            /*lfi[i].uvlim[j] = block_inside_limit;
-            lfi[i].uvmbflim[j] = filt_lvl+uvhedge_boost;*/
-            lfi[i].uvmbthr[j] = HEVThresh;
-            /*lfi[i].uvflim[j] = filt_lvl;*/
-            lfi[i].uvthr[j] = HEVThresh;
-        }
+        vpx_memset(lfi->hev_thr[i], i, SIMD_WIDTH);
     }
 }
 
-
-void vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int *filter_level)
+void vp8_loop_filter_frame_init(VP8_COMMON *cm,
+                                MACROBLOCKD *mbd,
+                                int default_filt_lvl)
 {
-    MB_MODE_INFO *mbmi = &mbd->mode_info_context->mbmi;
+    int seg,  /* segment number */
+        ref,  /* index in ref_lf_deltas */
+        mode; /* index in mode_lf_deltas */
 
-    if (mbd->mode_ref_lf_delta_enabled)
+    loop_filter_info_n *lfi = &cm->lf_info;
+
+    /* update limits if sharpness has changed */
+    if(cm->last_sharpness_level != cm->sharpness_level)
     {
+        vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+        cm->last_sharpness_level = cm->sharpness_level;
+    }
+
+    for(seg = 0; seg < MAX_MB_SEGMENTS; seg++)
+    {
+        int lvl_seg = default_filt_lvl;
+        int lvl_ref, lvl_mode;
+
+        /* Note the baseline filter values for each segment */
+        if (mbd->segmentation_enabled)
+        {
+            /* Abs value */
+            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
+            {
+                lvl_seg = mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+            }
+            else  /* Delta Value */
+            {
+                lvl_seg += mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+                lvl_seg = (lvl_seg > 0) ? ((lvl_seg > 63) ? 63: lvl_seg) : 0;
+            }
+        }
+
+        if (!mbd->mode_ref_lf_delta_enabled)
+        {
+            /* we could get rid of this if we assume that deltas are set to
+             * zero when not in use; encoder always uses deltas
+             */
+            vpx_memset(lfi->lvl[seg][0], lvl_seg, 4 * 4 );
+            continue;
+        }
+
+        lvl_ref = lvl_seg;
+
+        /* INTRA_FRAME */
+        ref = INTRA_FRAME;
+
         /* Apply delta for reference frame */
-        *filter_level += mbd->ref_lf_deltas[mbmi->ref_frame];
+        lvl_ref += mbd->ref_lf_deltas[ref];
 
-        /* Apply delta for mode */
-        if (mbmi->ref_frame == INTRA_FRAME)
+        /* Apply delta for Intra modes */
+        mode = 0; /* B_PRED */
+        /* Only the split mode BPRED has a further special case */
+        lvl_mode = lvl_ref +  mbd->mode_lf_deltas[mode];
+        lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
+
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        mode = 1; /* all the rest of Intra modes */
+        lvl_mode = (lvl_ref > 0) ? (lvl_ref > 63 ? 63 : lvl_ref)  : 0; /* clamp */
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        /* LAST, GOLDEN, ALT */
+        for(ref = 1; ref < MAX_REF_FRAMES; ref++)
         {
-            /* Only the split mode BPRED has a further special case */
-            if (mbmi->mode == B_PRED)
-                *filter_level +=  mbd->mode_lf_deltas[0];
+            int lvl_ref = lvl_seg;
+
+            /* Apply delta for reference frame */
+            lvl_ref += mbd->ref_lf_deltas[ref];
+
+            /* Apply delta for Inter modes */
+            for (mode = 1; mode < 4; mode++)
+            {
+                lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
+                lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
+
+                lfi->lvl[seg][ref][mode] = lvl_mode;
+            }
         }
-        else
-        {
-            /* Zero motion mode */
-            if (mbmi->mode == ZEROMV)
-                *filter_level +=  mbd->mode_lf_deltas[1];
-
-            /* Split MB motion mode */
-            else if (mbmi->mode == SPLITMV)
-                *filter_level +=  mbd->mode_lf_deltas[3];
-
-            /* All other inter motion modes (Nearest, Near, New) */
-            else
-                *filter_level +=  mbd->mode_lf_deltas[2];
-        }
-
-        /* Range check */
-        if (*filter_level > MAX_LOOP_FILTER)
-            *filter_level = MAX_LOOP_FILTER;
-        else if (*filter_level < 0)
-            *filter_level = 0;
     }
 }
 
-
 void vp8_loop_filter_frame
 (
     VP8_COMMON *cm,
-    MACROBLOCKD *mbd,
-    int default_filt_lvl
+    MACROBLOCKD *mbd
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
-    loop_filter_info *lfi = cm->lf_info;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
     FRAME_TYPE frame_type = cm->frame_type;
 
     int mb_row;
     int mb_col;
 
-
-    int baseline_filter_level[MAX_MB_SEGMENTS];
     int filter_level;
-    int alt_flt_enabled = mbd->segmentation_enabled;
 
-    int i;
     unsigned char *y_ptr, *u_ptr, *v_ptr;
 
-    mbd->mode_info_context = cm->mi;          /* Point at base of Mb MODE_INFO list */
-
-    /* Note the baseline filter values for each segment */
-    if (alt_flt_enabled)
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-        {
-            /* Abs value */
-            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-            /* Delta Value */
-            else
-            {
-                baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-                baseline_filter_level[i] = (baseline_filter_level[i] >= 0) ? ((baseline_filter_level[i] <= MAX_LOOP_FILTER) ? baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  /* Clamp to valid range */
-            }
-        }
-    }
-    else
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-            baseline_filter_level[i] = default_filt_lvl;
-    }
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
 
     /* Initialize the loop filter for this frame. */
-    if ((cm->last_filter_type != cm->filter_type) || (cm->last_sharpness_level != cm->sharpness_level))
-        vp8_init_loop_filter(cm);
-    else if (frame_type != cm->last_frame_type)
-        vp8_frame_init_loop_filter(lfi, frame_type);
+    vp8_loop_filter_frame_init(cm, mbd, cm->filter_level);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer;
@@ -365,101 +314,108 @@ void vp8_loop_filter_frame
     {
         for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
         {
-            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                            mode_info_context->mbmi.mode != SPLITMV &&
+                            mode_info_context->mbmi.mb_skip_coeff);
 
-            filter_level = baseline_filter_level[Segment];
+            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
 
-            /* Distance of Mb to the various image edges.
-             * These specified to 8th pel as they are always compared to values that are in 1/8th pel units
-             * Apply any context driven MB level adjustment
-             */
-            vp8_adjust_mb_lf_value(mbd, &filter_level);
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
 
             if (filter_level)
             {
-                if (mb_col > 0)
-                    cm->lf_mbv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
 
-                /* don't apply across umv border */
-                if (mb_row > 0)
-                    cm->lf_mbh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
             u_ptr += 8;
             v_ptr += 8;
 
-            mbd->mode_info_context++;     /* step to next MB */
+            mode_info_context++;     /* step to next MB */
         }
 
         y_ptr += post->y_stride  * 16 - post->y_width;
         u_ptr += post->uv_stride *  8 - post->uv_width;
         v_ptr += post->uv_stride *  8 - post->uv_width;
 
-        mbd->mode_info_context++;         /* Skip border mb */
+        mode_info_context++;         /* Skip border mb */
     }
 }
 
-
 void vp8_loop_filter_frame_yonly
 (
     VP8_COMMON *cm,
     MACROBLOCKD *mbd,
-    int default_filt_lvl,
-    int sharpness_lvl
+    int default_filt_lvl
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
 
-    int i;
     unsigned char *y_ptr;
     int mb_row;
     int mb_col;
 
-    loop_filter_info *lfi = cm->lf_info;
-    int baseline_filter_level[MAX_MB_SEGMENTS];
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
     int filter_level;
-    int alt_flt_enabled = mbd->segmentation_enabled;
     FRAME_TYPE frame_type = cm->frame_type;
 
-    (void) sharpness_lvl;
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
 
-    /*MODE_INFO * this_mb_mode_info = cm->mi;*/ /* Point at base of Mb MODE_INFO list */
-    mbd->mode_info_context = cm->mi;          /* Point at base of Mb MODE_INFO list */
-
-    /* Note the baseline filter values for each segment */
-    if (alt_flt_enabled)
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-        {
-            /* Abs value */
-            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-            /* Delta Value */
-            else
-            {
-                baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-                baseline_filter_level[i] = (baseline_filter_level[i] >= 0) ? ((baseline_filter_level[i] <= MAX_LOOP_FILTER) ? baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  /* Clamp to valid range */
-            }
-        }
-    }
-    else
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-            baseline_filter_level[i] = default_filt_lvl;
-    }
+#if 0
+    if(default_filt_lvl == 0) /* no filter applied */
+        return;
+#endif
 
     /* Initialize the loop filter for this frame. */
-    if ((cm->last_filter_type != cm->filter_type) || (cm->last_sharpness_level != cm->sharpness_level))
-        vp8_init_loop_filter(cm);
-    else if (frame_type != cm->last_frame_type)
-        vp8_frame_init_loop_filter(lfi, frame_type);
+    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer;
@@ -469,72 +425,106 @@ void vp8_loop_filter_frame_yonly
     {
         for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
         {
-            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
-            filter_level = baseline_filter_level[Segment];
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                            mode_info_context->mbmi.mode != SPLITMV &&
+                            mode_info_context->mbmi.mb_skip_coeff);
 
-            /* Apply any context driven MB level adjustment */
-            vp8_adjust_mb_lf_value(mbd, &filter_level);
+            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
 
             if (filter_level)
             {
-                if (mb_col > 0)
-                    cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                /* don't apply across umv border */
-                if (mb_row > 0)
-                    cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
-            mbd->mode_info_context ++;        /* step to next MB */
+            mode_info_context ++;        /* step to next MB */
 
         }
 
         y_ptr += post->y_stride  * 16 - post->y_width;
-        mbd->mode_info_context ++;            /* Skip border mb */
+        mode_info_context ++;            /* Skip border mb */
     }
 
 }
 
-
 void vp8_loop_filter_partial_frame
 (
     VP8_COMMON *cm,
     MACROBLOCKD *mbd,
-    int default_filt_lvl,
-    int sharpness_lvl,
-    int Fraction
+    int default_filt_lvl
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
 
-    int i;
     unsigned char *y_ptr;
     int mb_row;
     int mb_col;
-    /*int mb_rows = post->y_height >> 4;*/
     int mb_cols = post->y_width  >> 4;
 
-    int linestocopy;
+    int linestocopy, i;
+
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
 
-    loop_filter_info *lfi = cm->lf_info;
-    int baseline_filter_level[MAX_MB_SEGMENTS];
     int filter_level;
     int alt_flt_enabled = mbd->segmentation_enabled;
     FRAME_TYPE frame_type = cm->frame_type;
 
-    (void) sharpness_lvl;
+    const MODE_INFO *mode_info_context;
 
-    /*MODE_INFO * this_mb_mode_info = cm->mi + (post->y_height>>5) * (mb_cols + 1);*/ /* Point at base of Mb MODE_INFO list */
-    mbd->mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);        /* Point at base of Mb MODE_INFO list */
+    int lvl_seg[MAX_MB_SEGMENTS];
 
-    linestocopy = (post->y_height >> (4 + Fraction));
+    mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);
+
+    /* 3 is a magic number. 4 is probably magic too */
+    linestocopy = (post->y_height >> (4 + 3));
 
     if (linestocopy < 1)
         linestocopy = 1;
@@ -542,32 +532,27 @@ void vp8_loop_filter_partial_frame
     linestocopy <<= 4;
 
     /* Note the baseline filter values for each segment */
+    /* See vp8_loop_filter_frame_init. Rather than call that for each change
+     * to default_filt_lvl, copy the relevant calculation here.
+     */
     if (alt_flt_enabled)
     {
         for (i = 0; i < MAX_MB_SEGMENTS; i++)
-        {
-            /* Abs value */
+        {    /* Abs value */
             if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+            {
+                lvl_seg[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+            }
             /* Delta Value */
             else
             {
-                baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-                baseline_filter_level[i] = (baseline_filter_level[i] >= 0) ? ((baseline_filter_level[i] <= MAX_LOOP_FILTER) ? baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  /* Clamp to valid range */
+                lvl_seg[i] = default_filt_lvl
+                        + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+                lvl_seg[i] = (lvl_seg[i] > 0) ?
+                        ((lvl_seg[i] > 63) ? 63: lvl_seg[i]) : 0;
             }
         }
     }
-    else
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-            baseline_filter_level[i] = default_filt_lvl;
-    }
-
-    /* Initialize the loop filter for this frame. */
-    if ((cm->last_filter_type != cm->filter_type) || (cm->last_sharpness_level != cm->sharpness_level))
-        vp8_init_loop_filter(cm);
-    else if (frame_type != cm->last_frame_type)
-        vp8_frame_init_loop_filter(lfi, frame_type);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer + (post->y_height >> 5) * 16 * post->y_stride;
@@ -577,28 +562,64 @@ void vp8_loop_filter_partial_frame
     {
         for (mb_col = 0; mb_col < mb_cols; mb_col++)
         {
-            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
-            filter_level = baseline_filter_level[Segment];
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                           mode_info_context->mbmi.mode != SPLITMV &&
+                           mode_info_context->mbmi.mb_skip_coeff);
+
+            if (alt_flt_enabled)
+                filter_level = lvl_seg[mode_info_context->mbmi.segment_id];
+            else
+                filter_level = default_filt_lvl;
 
             if (filter_level)
             {
-                if (mb_col > 0)
-                    cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
-            mbd->mode_info_context += 1;      /* step to next MB */
+            mode_info_context += 1;      /* step to next MB */
         }
 
         y_ptr += post->y_stride  * 16 - post->y_width;
-        mbd->mode_info_context += 1;          /* Skip border mb */
+        mode_info_context += 1;          /* Skip border mb */
     }
 }

media/libvpx/vp8/common/loopfilter.h

@@ -13,6 +13,7 @@
 #define loopfilter_h
 
 #include "vpx_ports/mem.h"
+#include "vpx_config.h"
 
 #define MAX_LOOP_FILTER 63
 
@@ -22,32 +23,45 @@ typedef enum
     SIMPLE_LOOPFILTER = 1
 } LOOPFILTERTYPE;
 
-/* FRK
- * Need to align this structure so when it is declared and
+#if ARCH_ARM
+#define SIMD_WIDTH 1
+#else
+#define SIMD_WIDTH 16
+#endif
+
+/* Need to align this structure so when it is declared and
  * passed it can be loaded into vector registers.
  */
 typedef struct
 {
-    DECLARE_ALIGNED(16, signed char, lim[16]);
-    DECLARE_ALIGNED(16, signed char, flim[16]);
-    DECLARE_ALIGNED(16, signed char, thr[16]);
-    DECLARE_ALIGNED(16, signed char, mbflim[16]);
-    DECLARE_ALIGNED(16, signed char, mbthr[16]);
-    DECLARE_ALIGNED(16, signed char, uvlim[16]);
-    DECLARE_ALIGNED(16, signed char, uvflim[16]);
-    DECLARE_ALIGNED(16, signed char, uvthr[16]);
-    DECLARE_ALIGNED(16, signed char, uvmbflim[16]);
-    DECLARE_ALIGNED(16, signed char, uvmbthr[16]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, mblim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, blim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, lim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, hev_thr[4][SIMD_WIDTH]);
+    unsigned char lvl[4][4][4];
+    unsigned char hev_thr_lut[2][MAX_LOOP_FILTER + 1];
+    unsigned char mode_lf_lut[10];
+} loop_filter_info_n;
+
+typedef struct
+{
+    const unsigned char * mblim;
+    const unsigned char * blim;
+    const unsigned char * lim;
+    const unsigned char * hev_thr;
 } loop_filter_info;
 
 
 #define prototype_loopfilter(sym) \
-    void sym(unsigned char *src, int pitch, const signed char *flimit,\
-             const signed char *limit, const signed char *thresh, int count)
+    void sym(unsigned char *src, int pitch, const unsigned char *blimit,\
+             const unsigned char *limit, const unsigned char *thresh, int count)
 
 #define prototype_loopfilter_block(sym) \
-    void sym(unsigned char *y, unsigned char *u, unsigned char *v,\
-             int ystride, int uv_stride, loop_filter_info *lfi, int simpler)
+    void sym(unsigned char *y, unsigned char *u, unsigned char *v, \
+             int ystride, int uv_stride, loop_filter_info *lfi)
+
+#define prototype_simple_loopfilter(sym) \
+    void sym(unsigned char *y, int ystride, const unsigned char *blimit)
 
 #if ARCH_X86 || ARCH_X86_64
 #include "x86/loopfilter_x86.h"
@@ -77,38 +91,39 @@ extern prototype_loopfilter_block(vp8_lf_normal_mb_h);
 #endif
 extern prototype_loopfilter_block(vp8_lf_normal_b_h);
 
-
 #ifndef vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_c
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_mb_v);
+extern prototype_simple_loopfilter(vp8_lf_simple_mb_v);
 
 #ifndef vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_b_v);
+extern prototype_simple_loopfilter(vp8_lf_simple_b_v);
 
 #ifndef vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_c
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_mb_h);
+extern prototype_simple_loopfilter(vp8_lf_simple_mb_h);
 
 #ifndef vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_b_h);
+extern prototype_simple_loopfilter(vp8_lf_simple_b_h);
 
 typedef prototype_loopfilter_block((*vp8_lf_block_fn_t));
+typedef prototype_simple_loopfilter((*vp8_slf_block_fn_t));
+
 typedef struct
 {
     vp8_lf_block_fn_t  normal_mb_v;
     vp8_lf_block_fn_t  normal_b_v;
     vp8_lf_block_fn_t  normal_mb_h;
     vp8_lf_block_fn_t  normal_b_h;
-    vp8_lf_block_fn_t  simple_mb_v;
-    vp8_lf_block_fn_t  simple_b_v;
-    vp8_lf_block_fn_t  simple_mb_h;
-    vp8_lf_block_fn_t  simple_b_h;
+    vp8_slf_block_fn_t  simple_mb_v;
+    vp8_slf_block_fn_t  simple_b_v;
+    vp8_slf_block_fn_t  simple_mb_h;
+    vp8_slf_block_fn_t  simple_b_h;
 } vp8_loopfilter_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -121,10 +136,33 @@ typedef void loop_filter_uvfunction
 (
     unsigned char *u,   /* source pointer */
     int p,              /* pitch */
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     unsigned char *v
 );
 
+/* assorted loopfilter functions which get used elsewhere */
+struct VP8Common;
+struct MacroBlockD;
+
+void vp8_loop_filter_init(struct VP8Common *cm);
+
+void vp8_loop_filter_frame_init(struct VP8Common *cm,
+                                struct MacroBlockD *mbd,
+                                int default_filt_lvl);
+
+void vp8_loop_filter_frame(struct VP8Common *cm, struct MacroBlockD *mbd);
+
+void vp8_loop_filter_partial_frame(struct VP8Common *cm,
+                                   struct MacroBlockD *mbd,
+                                   int default_filt_lvl);
+
+void vp8_loop_filter_frame_yonly(struct VP8Common *cm,
+                                 struct MacroBlockD *mbd,
+                                 int default_filt_lvl);
+
+void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
+                                      int sharpness_lvl);
+
 #endif

media/libvpx/vp8/common/loopfilter_filters.c

@@ -28,8 +28,9 @@ static __inline signed char vp8_signed_char_clamp(int t)
 
 
 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */
-static __inline signed char vp8_filter_mask(signed char limit, signed char flimit,
-                                     uc p3, uc p2, uc p1, uc p0, uc q0, uc q1, uc q2, uc q3)
+static __inline signed char vp8_filter_mask(uc limit, uc blimit,
+                                     uc p3, uc p2, uc p1, uc p0,
+                                     uc q0, uc q1, uc q2, uc q3)
 {
     signed char mask = 0;
     mask |= (abs(p3 - p2) > limit) * -1;
@@ -38,13 +39,13 @@ static __inline signed char vp8_filter_mask(signed char limit, signed char flimi
     mask |= (abs(q1 - q0) > limit) * -1;
     mask |= (abs(q2 - q1) > limit) * -1;
     mask |= (abs(q3 - q2) > limit) * -1;
-    mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  > flimit * 2 + limit) * -1;
+    mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  > blimit) * -1;
     mask = ~mask;
     return mask;
 }
 
 /* is there high variance internal edge ( 11111111 yes, 00000000 no) */
-static __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0, uc q1)
+static __inline signed char vp8_hevmask(uc thresh, uc p1, uc p0, uc q0, uc q1)
 {
     signed char hev = 0;
     hev  |= (abs(p1 - p0) > thresh) * -1;
@@ -52,7 +53,8 @@ static __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0,
     return hev;
 }
 
-static __inline void vp8_filter(signed char mask, signed char hev, uc *op1, uc *op0, uc *oq0, uc *oq1)
+static __inline void vp8_filter(signed char mask, uc hev, uc *op1,
+        uc *op0, uc *oq0, uc *oq1)
 
 {
     signed char ps0, qs0;
@@ -102,9 +104,9 @@ void vp8_loop_filter_horizontal_edge_c
 (
     unsigned char *s,
     int p, /* pitch */
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -117,11 +119,11 @@ void vp8_loop_filter_horizontal_edge_c
      */
     do
     {
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4*p], s[-3*p], s[-2*p], s[-1*p],
                                s[0*p], s[1*p], s[2*p], s[3*p]);
 
-        hev = vp8_hevmask(thresh[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        hev = vp8_hevmask(thresh[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
 
         vp8_filter(mask, hev, s - 2 * p, s - 1 * p, s, s + 1 * p);
 
@@ -134,9 +136,9 @@ void vp8_loop_filter_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -149,10 +151,10 @@ void vp8_loop_filter_vertical_edge_c
      */
     do
     {
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4], s[-3], s[-2], s[-1], s[0], s[1], s[2], s[3]);
 
-        hev = vp8_hevmask(thresh[i], s[-2], s[-1], s[0], s[1]);
+        hev = vp8_hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
 
         vp8_filter(mask, hev, s - 2, s - 1, s, s + 1);
 
@@ -161,7 +163,7 @@ void vp8_loop_filter_vertical_edge_c
     while (++i < count * 8);
 }
 
-static __inline void vp8_mbfilter(signed char mask, signed char hev,
+static __inline void vp8_mbfilter(signed char mask, uc hev,
                            uc *op2, uc *op1, uc *op0, uc *oq0, uc *oq1, uc *oq2)
 {
     signed char s, u;
@@ -220,9 +222,9 @@ void vp8_mbloop_filter_horizontal_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -236,11 +238,11 @@ void vp8_mbloop_filter_horizontal_edge_c
     do
     {
 
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4*p], s[-3*p], s[-2*p], s[-1*p],
                                s[0*p], s[1*p], s[2*p], s[3*p]);
 
-        hev = vp8_hevmask(thresh[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        hev = vp8_hevmask(thresh[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
 
         vp8_mbfilter(mask, hev, s - 3 * p, s - 2 * p, s - 1 * p, s, s + 1 * p, s + 2 * p);
 
@@ -255,9 +257,9 @@ void vp8_mbloop_filter_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -268,10 +270,10 @@ void vp8_mbloop_filter_vertical_edge_c
     do
     {
 
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4], s[-3], s[-2], s[-1], s[0], s[1], s[2], s[3]);
 
-        hev = vp8_hevmask(thresh[i], s[-2], s[-1], s[0], s[1]);
+        hev = vp8_hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
 
         vp8_mbfilter(mask, hev, s - 3, s - 2, s - 1, s, s + 1, s + 2);
 
@@ -282,13 +284,13 @@ void vp8_mbloop_filter_vertical_edge_c
 }
 
 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */
-static __inline signed char vp8_simple_filter_mask(signed char limit, signed char flimit, uc p1, uc p0, uc q0, uc q1)
+static __inline signed char vp8_simple_filter_mask(uc blimit, uc p1, uc p0, uc q0, uc q1)
 {
 /* Why does this cause problems for win32?
  * error C2143: syntax error : missing ';' before 'type'
  *  (void) limit;
  */
-    signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  <= flimit * 2 + limit) * -1;
+    signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  <= blimit) * -1;
     return mask;
 }
 
@@ -321,47 +323,37 @@ void vp8_loop_filter_simple_horizontal_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
-    int count
+    const unsigned char *blimit
 )
 {
     signed char mask = 0;
     int i = 0;
-    (void) thresh;
 
     do
     {
-        /*mask = vp8_simple_filter_mask( limit[i], flimit[i],s[-1*p],s[0*p]);*/
-        mask = vp8_simple_filter_mask(limit[i], flimit[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        mask = vp8_simple_filter_mask(blimit[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
         vp8_simple_filter(mask, s - 2 * p, s - 1 * p, s, s + 1 * p);
         ++s;
     }
-    while (++i < count * 8);
+    while (++i < 16);
 }
 
 void vp8_loop_filter_simple_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
-    int count
+    const unsigned char *blimit
 )
 {
     signed char mask = 0;
     int i = 0;
-    (void) thresh;
 
     do
     {
-        /*mask = vp8_simple_filter_mask( limit[i], flimit[i],s[-1],s[0]);*/
-        mask = vp8_simple_filter_mask(limit[i], flimit[i], s[-2], s[-1], s[0], s[1]);
+        mask = vp8_simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]);
         vp8_simple_filter(mask, s - 2, s - 1, s, s + 1);
         s += p;
     }
-    while (++i < count * 8);
+    while (++i < 16);
 
 }

media/libvpx/vp8/common/mbpitch.c

@@ -17,7 +17,7 @@ typedef enum
     DEST = 1
 } BLOCKSET;
 
-void vp8_setup_block
+static void setup_block
 (
     BLOCKD *b,
     int mv_stride,
@@ -43,7 +43,8 @@ void vp8_setup_block
 
 }
 
-void vp8_setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
+
+static void setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
 {
     int block;
 
@@ -64,16 +65,16 @@ void vp8_setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
 
     for (block = 0; block < 16; block++) /* y blocks */
     {
-        vp8_setup_block(&x->block[block], x->dst.y_stride, y, x->dst.y_stride,
+        setup_block(&x->block[block], x->dst.y_stride, y, x->dst.y_stride,
                         (block >> 2) * 4 * x->dst.y_stride + (block & 3) * 4, bs);
     }
 
     for (block = 16; block < 20; block++) /* U and V blocks */
     {
-        vp8_setup_block(&x->block[block], x->dst.uv_stride, u, x->dst.uv_stride,
+        setup_block(&x->block[block], x->dst.uv_stride, u, x->dst.uv_stride,
                         ((block - 16) >> 1) * 4 * x->dst.uv_stride + (block & 1) * 4, bs);
 
-        vp8_setup_block(&x->block[block+4], x->dst.uv_stride, v, x->dst.uv_stride,
+        setup_block(&x->block[block+4], x->dst.uv_stride, v, x->dst.uv_stride,
                         ((block - 16) >> 1) * 4 * x->dst.uv_stride + (block & 1) * 4, bs);
     }
 }
@@ -124,6 +125,6 @@ void vp8_build_block_doffsets(MACROBLOCKD *x)
 {
 
     /* handle the destination pitch features */
-    vp8_setup_macroblock(x, DEST);
-    vp8_setup_macroblock(x, PRED);
+    setup_macroblock(x, DEST);
+    setup_macroblock(x, PRED);
 }

media/libvpx/vp8/common/mv.h

@@ -11,6 +11,7 @@
 
 #ifndef __INC_MV_H
 #define __INC_MV_H
+#include "vpx/vpx_integer.h"
 
 typedef struct
 {
@@ -18,4 +19,10 @@ typedef struct
     short col;
 } MV;
 
+typedef union
+{
+    uint32_t  as_int;
+    MV        as_mv;
+} int_mv;        /* facilitates faster equality tests and copies */
+
 #endif

media/libvpx/vp8/common/onyx.h

@@ -18,6 +18,7 @@ extern "C"
 #endif
 
 #include "vpx/internal/vpx_codec_internal.h"
+#include "vpx/vp8cx.h"
 #include "vpx_scale/yv12config.h"
 #include "type_aliases.h"
 #include "ppflags.h"
@@ -45,7 +46,8 @@ extern "C"
     typedef enum
     {
         USAGE_STREAM_FROM_SERVER    = 0x0,
-        USAGE_LOCAL_FILE_PLAYBACK   = 0x1
+        USAGE_LOCAL_FILE_PLAYBACK   = 0x1,
+        USAGE_CONSTRAINED_QUALITY   = 0x2
     } END_USAGE;
 
 
@@ -107,6 +109,7 @@ extern "C"
         int noise_sensitivity;   // parameter used for applying pre processing blur: recommendation 0
         int Sharpness;          // parameter used for sharpening output: recommendation 0:
         int cpu_used;
+        unsigned int rc_max_intra_bitrate_pct;
 
         // mode ->
         //(0)=Realtime/Live Encoding. This mode is optimized for realtim encoding (for example, capturing
@@ -137,8 +140,9 @@ extern "C"
 
         int end_usage; // vbr or cbr
 
-        // shoot to keep buffer full at all times by undershooting a bit 95 recommended
+        // buffer targeting aggressiveness
         int under_shoot_pct;
+        int over_shoot_pct;
 
         // buffering parameters
         int starting_buffer_level;  // in seconds
@@ -149,6 +153,7 @@ extern "C"
         int fixed_q;
         int worst_allowed_q;
         int best_allowed_q;
+        int cq_level;
 
         // allow internal resizing ( currently disabled in the build !!!!!)
         int allow_spatial_resampling;
@@ -179,16 +184,20 @@ extern "C"
         int token_partitions; // how many token partitions to create for multi core decoding
         int encode_breakout;  // early breakout encode threshold : for video conf recommend 800
 
-        int error_resilient_mode;  // if running over udp networks provides decodable frames after a
-        // dropped packet
+        unsigned int error_resilient_mode; // Bitfield defining the error
+                                   // resiliency features to enable. Can provide
+                                   // decodable frames after losses in previous
+                                   // frames and decodable partitions after
+                                   // losses in the same frame.
 
         int arnr_max_frames;
         int arnr_strength ;
         int arnr_type     ;
 
-
         struct vpx_fixed_buf         two_pass_stats_in;
         struct vpx_codec_pkt_list  *output_pkt_list;
+
+        vp8e_tuning tuning;
     } VP8_CONFIG;
 
 
@@ -202,9 +211,9 @@ extern "C"
 
 // receive a frames worth of data caller can assume that a copy of this frame is made
 // and not just a copy of the pointer..
-    int vp8_receive_raw_frame(VP8_PTR comp, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, INT64 time_stamp, INT64 end_time_stamp);
-    int vp8_get_compressed_data(VP8_PTR comp, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, INT64 *time_stamp, INT64 *time_end, int flush);
-    int vp8_get_preview_raw_frame(VP8_PTR comp, YV12_BUFFER_CONFIG *dest, int deblock_level, int noise_level, int flags);
+    int vp8_receive_raw_frame(VP8_PTR comp, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, int64_t time_stamp, int64_t end_time_stamp);
+    int vp8_get_compressed_data(VP8_PTR comp, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, int64_t *time_stamp, int64_t *time_end, int flush);
+    int vp8_get_preview_raw_frame(VP8_PTR comp, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t *flags);
 
     int vp8_use_as_reference(VP8_PTR comp, int ref_frame_flags);
     int vp8_update_reference(VP8_PTR comp, int ref_frame_flags);

media/libvpx/vp8/common/onyxc_int.h

@@ -19,7 +19,9 @@
 #include "entropy.h"
 #include "idct.h"
 #include "recon.h"
+#if CONFIG_POSTPROC
 #include "postproc.h"
+#endif
 
 /*#ifdef PACKET_TESTING*/
 #include "header.h"
@@ -35,13 +37,15 @@ void vp8_initialize_common(void);
 
 #define NUM_YV12_BUFFERS 4
 
+#define MAX_PARTITIONS 9
+
 typedef struct frame_contexts
 {
     vp8_prob bmode_prob [VP8_BINTRAMODES-1];
     vp8_prob ymode_prob [VP8_YMODES-1];   /* interframe intra mode probs */
     vp8_prob uv_mode_prob [VP8_UV_MODES-1];
     vp8_prob sub_mv_ref_prob [VP8_SUBMVREFS-1];
-    vp8_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1];
+    vp8_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
     MV_CONTEXT mvc[2];
     MV_CONTEXT pre_mvc[2];  /* not to caculate the mvcost for the frame if mvc doesn't change. */
 } FRAME_CONTEXT;
@@ -73,7 +77,9 @@ typedef struct VP8_COMMON_RTCD
     vp8_recon_rtcd_vtable_t       recon;
     vp8_subpix_rtcd_vtable_t      subpix;
     vp8_loopfilter_rtcd_vtable_t  loopfilter;
+#if CONFIG_POSTPROC
     vp8_postproc_rtcd_vtable_t    postproc;
+#endif
     int                           flags;
 #else
     int unused;
@@ -81,6 +87,7 @@ typedef struct VP8_COMMON_RTCD
 } VP8_COMMON_RTCD;
 
 typedef struct VP8Common
+
 {
     struct vpx_internal_error_info  error;
 
@@ -105,7 +112,8 @@ typedef struct VP8Common
     YV12_BUFFER_CONFIG post_proc_buffer;
     YV12_BUFFER_CONFIG temp_scale_frame;
 
-    FRAME_TYPE last_frame_type;  /* Add to check if vp8_frame_init_loop_filter() can be skipped. */
+
+    FRAME_TYPE last_frame_type;  /* Save last frame's frame type for motion search. */
     FRAME_TYPE frame_type;
 
     int show_frame;
@@ -119,7 +127,6 @@ typedef struct VP8Common
     /* profile settings */
     int mb_no_coeff_skip;
     int no_lpf;
-    int simpler_lpf;
     int use_bilinear_mc_filter;
     int full_pixel;
 
@@ -140,16 +147,15 @@ typedef struct VP8Common
 
     MODE_INFO *mip; /* Base of allocated array */
     MODE_INFO *mi;  /* Corresponds to upper left visible macroblock */
+    MODE_INFO *prev_mip; /* MODE_INFO array 'mip' from last decoded frame */
+    MODE_INFO *prev_mi;  /* 'mi' from last frame (points into prev_mip) */
 
 
     INTERPOLATIONFILTERTYPE mcomp_filter_type;
-    LOOPFILTERTYPE last_filter_type;
     LOOPFILTERTYPE filter_type;
-    loop_filter_info lf_info[MAX_LOOP_FILTER+1];
-    prototype_loopfilter_block((*lf_mbv));
-    prototype_loopfilter_block((*lf_mbh));
-    prototype_loopfilter_block((*lf_bv));
-    prototype_loopfilter_block((*lf_bh));
+
+    loop_filter_info_n lf_info;
+
     int filter_level;
     int last_sharpness_level;
     int sharpness_level;
@@ -196,13 +202,12 @@ typedef struct VP8Common
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD rtcd;
 #endif
+#if CONFIG_MULTITHREAD
+    int processor_core_count;
+#endif
+#if CONFIG_POSTPROC
     struct postproc_state  postproc_state;
+#endif
 } VP8_COMMON;
 
-
-void vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int *filter_level);
-void vp8_init_loop_filter(VP8_COMMON *cm);
-void vp8_frame_init_loop_filter(loop_filter_info *lfi, int frame_type);
-extern void vp8_loop_filter_frame(VP8_COMMON *cm,    MACROBLOCKD *mbd,  int filt_val);
-
 #endif

media/libvpx/vp8/common/onyxd.h

@@ -22,6 +22,7 @@ extern "C"
 #include "vpx_scale/yv12config.h"
 #include "ppflags.h"
 #include "vpx_ports/mem.h"
+#include "vpx/vpx_codec.h"
 
     typedef void   *VP8D_PTR;
     typedef struct
@@ -31,6 +32,8 @@ extern "C"
         int     Version;
         int     postprocess;
         int     max_threads;
+        int     error_concealment;
+        int     input_partition;
     } VP8D_CONFIG;
     typedef enum
     {
@@ -50,11 +53,11 @@ extern "C"
 
     int vp8dx_get_setting(VP8D_PTR comp, VP8D_SETTING oxst);
 
-    int vp8dx_receive_compressed_data(VP8D_PTR comp, unsigned long size, const unsigned char *dest, INT64 time_stamp);
-    int vp8dx_get_raw_frame(VP8D_PTR comp, YV12_BUFFER_CONFIG *sd, INT64 *time_stamp, INT64 *time_end_stamp, int deblock_level,  int noise_level, int flags);
+    int vp8dx_receive_compressed_data(VP8D_PTR comp, unsigned long size, const unsigned char *dest, int64_t time_stamp);
+    int vp8dx_get_raw_frame(VP8D_PTR comp, YV12_BUFFER_CONFIG *sd, int64_t *time_stamp, int64_t *time_end_stamp, vp8_ppflags_t *flags);
 
-    int vp8dx_get_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
-    int vp8dx_set_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_get_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_set_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
 
     VP8D_PTR vp8dx_create_decompressor(VP8D_CONFIG *oxcf);
 

media/libvpx/vp8/common/postproc.c

@@ -26,7 +26,7 @@
     ( (0.439*(float)(t>>16)) - (0.368*(float)(t>>8&0xff)) - (0.071*(float)(t&0xff)) + 128)
 
 /* global constants */
-
+#if CONFIG_POSTPROC_VISUALIZER
 static const unsigned char MB_PREDICTION_MODE_colors[MB_MODE_COUNT][3] =
 {
     { RGB_TO_YUV(0x98FB98) },   /* PaleGreen */
@@ -41,13 +41,32 @@ static const unsigned char MB_PREDICTION_MODE_colors[MB_MODE_COUNT][3] =
     { RGB_TO_YUV(0xFF0000) }    /* Red */
 };
 
-static const unsigned char MV_REFERENCE_FRAME_colors[MB_MODE_COUNT][3] =
+static const unsigned char B_PREDICTION_MODE_colors[B_MODE_COUNT][3] =
+{
+    { RGB_TO_YUV(0x6633ff) },   /* Purple */
+    { RGB_TO_YUV(0xcc33ff) },   /* Magenta */
+    { RGB_TO_YUV(0xff33cc) },   /* Pink */
+    { RGB_TO_YUV(0xff3366) },   /* Coral */
+    { RGB_TO_YUV(0x3366ff) },   /* Blue */
+    { RGB_TO_YUV(0xed00f5) },   /* Dark Blue */
+    { RGB_TO_YUV(0x2e00b8) },   /* Dark Purple */
+    { RGB_TO_YUV(0xff6633) },   /* Orange */
+    { RGB_TO_YUV(0x33ccff) },   /* Light Blue */
+    { RGB_TO_YUV(0x8ab800) },   /* Green */
+    { RGB_TO_YUV(0xffcc33) },   /* Light Orange */
+    { RGB_TO_YUV(0x33ffcc) },   /* Aqua */
+    { RGB_TO_YUV(0x66ff33) },   /* Light Green */
+    { RGB_TO_YUV(0xccff33) },   /* Yellow */
+};
+
+static const unsigned char MV_REFERENCE_FRAME_colors[MAX_REF_FRAMES][3] =
 {
     { RGB_TO_YUV(0x00ff00) },   /* Blue */
     { RGB_TO_YUV(0x0000ff) },   /* Green */
     { RGB_TO_YUV(0xffff00) },   /* Yellow */
     { RGB_TO_YUV(0xff0000) },   /* Red */
 };
+#endif
 
 static const short kernel5[] =
 {
@@ -192,7 +211,7 @@ void vp8_post_proc_down_and_across_c
     }
 }
 
-int vp8_q2mbl(int x)
+static int q2mbl(int x)
 {
     if (x < 20) x = 20;
 
@@ -295,8 +314,8 @@ static void vp8_deblock_and_de_macro_block(YV12_BUFFER_CONFIG         *source,
     (void) flag;
 
     POSTPROC_INVOKE(rtcd, downacross)(source->y_buffer, post->y_buffer, source->y_stride,  post->y_stride, source->y_height, source->y_width,  ppl);
-    POSTPROC_INVOKE(rtcd, across)(post->y_buffer, post->y_stride, post->y_height, post->y_width, vp8_q2mbl(q));
-    POSTPROC_INVOKE(rtcd, down)(post->y_buffer, post->y_stride, post->y_height, post->y_width, vp8_q2mbl(q));
+    POSTPROC_INVOKE(rtcd, across)(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
+    POSTPROC_INVOKE(rtcd, down)(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
 
     POSTPROC_INVOKE(rtcd, downacross)(source->u_buffer, post->u_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
     POSTPROC_INVOKE(rtcd, downacross)(source->v_buffer, post->v_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
@@ -476,7 +495,7 @@ void vp8_plane_add_noise_c(unsigned char *Start, char *noise,
  * edges unblended to give distinction to macro blocks in areas
  * filled with the same color block.
  */
-void vp8_blend_mb_c (unsigned char *y, unsigned char *u, unsigned char *v,
+void vp8_blend_mb_inner_c (unsigned char *y, unsigned char *u, unsigned char *v,
                         int y1, int u1, int v1, int alpha, int stride)
 {
     int i, j;
@@ -484,10 +503,10 @@ void vp8_blend_mb_c (unsigned char *y, unsigned char *u, unsigned char *v,
     int u1_const = u1*((1<<16)-alpha);
     int v1_const = v1*((1<<16)-alpha);
 
-    y += stride + 2;
-    for (i = 0; i < 14; i++)
+    y += 2*stride + 2;
+    for (i = 0; i < 12; i++)
     {
-        for (j = 0; j < 14; j++)
+        for (j = 0; j < 12; j++)
         {
             y[j] = (y[j]*alpha + y1_const)>>16;
         }
@@ -511,6 +530,104 @@ void vp8_blend_mb_c (unsigned char *y, unsigned char *u, unsigned char *v,
     }
 }
 
+/* Blend only the edge of the macro block.  Leave center
+ * unblended to allow for other visualizations to be layered.
+ */
+void vp8_blend_mb_outer_c (unsigned char *y, unsigned char *u, unsigned char *v,
+                        int y1, int u1, int v1, int alpha, int stride)
+{
+    int i, j;
+    int y1_const = y1*((1<<16)-alpha);
+    int u1_const = u1*((1<<16)-alpha);
+    int v1_const = v1*((1<<16)-alpha);
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 16; j++)
+        {
+            y[j] = (y[j]*alpha + y1_const)>>16;
+        }
+        y += stride;
+    }
+
+    for (i = 0; i < 12; i++)
+    {
+        y[0]  = (y[0]*alpha  + y1_const)>>16;
+        y[1]  = (y[1]*alpha  + y1_const)>>16;
+        y[14] = (y[14]*alpha + y1_const)>>16;
+        y[15] = (y[15]*alpha + y1_const)>>16;
+        y += stride;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 16; j++)
+        {
+            y[j] = (y[j]*alpha + y1_const)>>16;
+        }
+        y += stride;
+    }
+
+    stride >>= 1;
+
+    for (j = 0; j < 8; j++)
+    {
+        u[j] = (u[j]*alpha + u1_const)>>16;
+        v[j] = (v[j]*alpha + v1_const)>>16;
+    }
+    u += stride;
+    v += stride;
+
+    for (i = 0; i < 6; i++)
+    {
+        u[0] = (u[0]*alpha + u1_const)>>16;
+        v[0] = (v[0]*alpha + v1_const)>>16;
+
+        u[7] = (u[7]*alpha + u1_const)>>16;
+        v[7] = (v[7]*alpha + v1_const)>>16;
+
+        u += stride;
+        v += stride;
+    }
+
+    for (j = 0; j < 8; j++)
+    {
+        u[j] = (u[j]*alpha + u1_const)>>16;
+        v[j] = (v[j]*alpha + v1_const)>>16;
+    }
+}
+
+void vp8_blend_b_c (unsigned char *y, unsigned char *u, unsigned char *v,
+                        int y1, int u1, int v1, int alpha, int stride)
+{
+    int i, j;
+    int y1_const = y1*((1<<16)-alpha);
+    int u1_const = u1*((1<<16)-alpha);
+    int v1_const = v1*((1<<16)-alpha);
+
+    for (i = 0; i < 4; i++)
+    {
+        for (j = 0; j < 4; j++)
+        {
+            y[j] = (y[j]*alpha + y1_const)>>16;
+        }
+        y += stride;
+    }
+
+    stride >>= 1;
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 2; j++)
+        {
+            u[j] = (u[j]*alpha + u1_const)>>16;
+            v[j] = (v[j]*alpha + v1_const)>>16;
+        }
+        u += stride;
+        v += stride;
+    }
+}
+
 static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int height)
 {
     int dx;
@@ -522,7 +639,7 @@ static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int hei
         dy = *y1 - y0;
 
         *x1 = width;
-        if (dy)
+        if (dx)
             *y1 = ((width-x0)*dy)/dx + y0;
     }
     if (*x1 < 0)
@@ -531,7 +648,7 @@ static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int hei
         dy = *y1 - y0;
 
         *x1 = 0;
-        if (dy)
+        if (dx)
             *y1 = ((0-x0)*dy)/dx + y0;
     }
     if (*y1 > height)
@@ -540,7 +657,7 @@ static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int hei
         dy = *y1 - y0;
 
         *y1 = height;
-        if (dx)
+        if (dy)
             *x1 = ((height-y0)*dx)/dy + x0;
     }
     if (*y1 < 0)
@@ -549,7 +666,7 @@ static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int hei
         dy = *y1 - y0;
 
         *y1 = 0;
-        if (dx)
+        if (dy)
             *x1 = ((0-y0)*dx)/dy + x0;
     }
 }
@@ -561,10 +678,12 @@ static void constrain_line (int x0, int *x1, int y0, int *y1, int width, int hei
 #define RTCD_VTABLE(oci) NULL
 #endif
 
-int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_level, int noise_level, int flags)
+int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t *ppflags)
 {
-    char message[512];
     int q = oci->filter_level * 10 / 6;
+    int flags = ppflags->post_proc_flag;
+    int deblock_level = ppflags->deblocking_level;
+    int noise_level = ppflags->noise_level;
 
     if (!oci->frame_to_show)
         return -1;
@@ -621,8 +740,10 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
          oci->post_proc_buffer.y_stride);
     }
 
-    if (flags & VP8D_DEBUG_LEVEL1)
+#if CONFIG_POSTPROC_VISUALIZER
+    if (flags & VP8D_DEBUG_TXT_FRAME_INFO)
     {
+        char message[512];
         sprintf(message, "F%1dG%1dQ%3dF%3dP%d_s%dx%d",
                 (oci->frame_type == KEY_FRAME),
                 oci->refresh_golden_frame,
@@ -633,7 +754,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
         vp8_blit_text(message, oci->post_proc_buffer.y_buffer, oci->post_proc_buffer.y_stride);
     }
 
-    if (flags & VP8D_DEBUG_LEVEL2)
+    if (flags & VP8D_DEBUG_TXT_MBLK_MODES)
     {
         int i, j;
         unsigned char *y_ptr;
@@ -665,7 +786,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
         }
     }
 
-    if (flags & VP8D_DEBUG_LEVEL3)
+    if (flags & VP8D_DEBUG_TXT_DC_DIFF)
     {
         int i, j;
         unsigned char *y_ptr;
@@ -683,11 +804,14 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
             for (j = 0; j < mb_cols; j++)
             {
                 char zz[4];
+                int dc_diff = !(mi[mb_index].mbmi.mode != B_PRED &&
+                              mi[mb_index].mbmi.mode != SPLITMV &&
+                              mi[mb_index].mbmi.mb_skip_coeff);
 
                 if (oci->frame_type == KEY_FRAME)
                     sprintf(zz, "a");
                 else
-                    sprintf(zz, "%c", mi[mb_index].mbmi.dc_diff + '0');
+                    sprintf(zz, "%c", dc_diff + '0');
 
                 vp8_blit_text(zz, y_ptr, post->y_stride);
                 mb_index ++;
@@ -700,78 +824,162 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
         }
     }
 
-    if (flags & VP8D_DEBUG_LEVEL4)
+    if (flags & VP8D_DEBUG_TXT_RATE_INFO)
     {
+        char message[512];
         sprintf(message, "Bitrate: %10.2f frame_rate: %10.2f ", oci->bitrate, oci->framerate);
         vp8_blit_text(message, oci->post_proc_buffer.y_buffer, oci->post_proc_buffer.y_stride);
-#if 0
-        int i, j;
-        unsigned char *y_ptr;
-        YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
-        int mb_rows = post->y_height >> 4;
-        int mb_cols = post->y_width  >> 4;
-        int mb_index = 0;
-        MODE_INFO *mi = oci->mi;
-
-        y_ptr = post->y_buffer + 4 * post->y_stride + 4;
-
-        /* vp8_filter each macro block */
-        for (i = 0; i < mb_rows; i++)
-        {
-            for (j = 0; j < mb_cols; j++)
-            {
-                char zz[4];
-
-                sprintf(zz, "%c", mi[mb_index].mbmi.dc_diff + '0');
-                vp8_blit_text(zz, y_ptr, post->y_stride);
-                mb_index ++;
-                y_ptr += 16;
-            }
-
-            mb_index ++; /* border */
-            y_ptr += post->y_stride  * 16 - post->y_width;
-
-        }
-
-#endif
-
     }
 
     /* Draw motion vectors */
-    if (flags & VP8D_DEBUG_LEVEL5)
+    if ((flags & VP8D_DEBUG_DRAW_MV) && ppflags->display_mv_flag)
     {
         YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
         int width  = post->y_width;
         int height = post->y_height;
-        int mb_cols = width  >> 4;
         unsigned char *y_buffer = oci->post_proc_buffer.y_buffer;
         int y_stride = oci->post_proc_buffer.y_stride;
         MODE_INFO *mi = oci->mi;
         int x0, y0;
 
-        for (y0 = 8; y0 < (height + 8); y0 += 16)
+        for (y0 = 0; y0 < height; y0 += 16)
         {
-            for (x0 = 8; x0 < (width + 8); x0 += 16)
+            for (x0 = 0; x0 < width; x0 += 16)
             {
-               int x1, y1;
-               if (mi->mbmi.mode >= NEARESTMV)
+                int x1, y1;
+
+                if (!(ppflags->display_mv_flag & (1<<mi->mbmi.mode)))
+                {
+                    mi++;
+                    continue;
+                }
+
+                if (mi->mbmi.mode == SPLITMV)
+                {
+                    switch (mi->mbmi.partitioning)
+                    {
+                        case 0 :    /* mv_top_bottom */
+                        {
+                            union b_mode_info *bmi = &mi->bmi[0];
+                            MV *mv = &bmi->mv.as_mv;
+
+                            x1 = x0 + 8 + (mv->col >> 3);
+                            y1 = y0 + 4 + (mv->row >> 3);
+
+                            constrain_line (x0+8, &x1, y0+4, &y1, width, height);
+                            vp8_blit_line  (x0+8,  x1, y0+4,  y1, y_buffer, y_stride);
+
+                            bmi = &mi->bmi[8];
+
+                            x1 = x0 + 8 + (mv->col >> 3);
+                            y1 = y0 +12 + (mv->row >> 3);
+
+                            constrain_line (x0+8, &x1, y0+12, &y1, width, height);
+                            vp8_blit_line  (x0+8,  x1, y0+12,  y1, y_buffer, y_stride);
+
+                            break;
+                        }
+                        case 1 :    /* mv_left_right */
+                        {
+                            union b_mode_info *bmi = &mi->bmi[0];
+                            MV *mv = &bmi->mv.as_mv;
+
+                            x1 = x0 + 4 + (mv->col >> 3);
+                            y1 = y0 + 8 + (mv->row >> 3);
+
+                            constrain_line (x0+4, &x1, y0+8, &y1, width, height);
+                            vp8_blit_line  (x0+4,  x1, y0+8,  y1, y_buffer, y_stride);
+
+                            bmi = &mi->bmi[2];
+
+                            x1 = x0 +12 + (mv->col >> 3);
+                            y1 = y0 + 8 + (mv->row >> 3);
+
+                            constrain_line (x0+12, &x1, y0+8, &y1, width, height);
+                            vp8_blit_line  (x0+12,  x1, y0+8,  y1, y_buffer, y_stride);
+
+                            break;
+                        }
+                        case 2 :    /* mv_quarters   */
+                        {
+                            union b_mode_info *bmi = &mi->bmi[0];
+                            MV *mv = &bmi->mv.as_mv;
+
+                            x1 = x0 + 4 + (mv->col >> 3);
+                            y1 = y0 + 4 + (mv->row >> 3);
+
+                            constrain_line (x0+4, &x1, y0+4, &y1, width, height);
+                            vp8_blit_line  (x0+4,  x1, y0+4,  y1, y_buffer, y_stride);
+
+                            bmi = &mi->bmi[2];
+
+                            x1 = x0 +12 + (mv->col >> 3);
+                            y1 = y0 + 4 + (mv->row >> 3);
+
+                            constrain_line (x0+12, &x1, y0+4, &y1, width, height);
+                            vp8_blit_line  (x0+12,  x1, y0+4,  y1, y_buffer, y_stride);
+
+                            bmi = &mi->bmi[8];
+
+                            x1 = x0 + 4 + (mv->col >> 3);
+                            y1 = y0 +12 + (mv->row >> 3);
+
+                            constrain_line (x0+4, &x1, y0+12, &y1, width, height);
+                            vp8_blit_line  (x0+4,  x1, y0+12,  y1, y_buffer, y_stride);
+
+                            bmi = &mi->bmi[10];
+
+                            x1 = x0 +12 + (mv->col >> 3);
+                            y1 = y0 +12 + (mv->row >> 3);
+
+                            constrain_line (x0+12, &x1, y0+12, &y1, width, height);
+                            vp8_blit_line  (x0+12,  x1, y0+12,  y1, y_buffer, y_stride);
+                            break;
+                        }
+                        default :
+                        {
+                            union b_mode_info *bmi = mi->bmi;
+                            int bx0, by0;
+
+                            for (by0 = y0; by0 < (y0+16); by0 += 4)
+                            {
+                                for (bx0 = x0; bx0 < (x0+16); bx0 += 4)
+                                {
+                                    MV *mv = &bmi->mv.as_mv;
+
+                                    x1 = bx0 + 2 + (mv->col >> 3);
+                                    y1 = by0 + 2 + (mv->row >> 3);
+
+                                    constrain_line (bx0+2, &x1, by0+2, &y1, width, height);
+                                    vp8_blit_line  (bx0+2,  x1, by0+2,  y1, y_buffer, y_stride);
+
+                                    bmi++;
+                                }
+                            }
+                        }
+                    }
+                }
+                else if (mi->mbmi.mode >= NEARESTMV)
                 {
                     MV *mv = &mi->mbmi.mv.as_mv;
+                    const int lx0 = x0 + 8;
+                    const int ly0 = y0 + 8;
 
-                    x1 = x0 + (mv->col >> 3);
-                    y1 = y0 + (mv->row >> 3);
+                    x1 = lx0 + (mv->col >> 3);
+                    y1 = ly0 + (mv->row >> 3);
 
-                    if (x1 != x0 && y1 != y0)
+                    if (x1 != lx0 && y1 != ly0)
                     {
-                        constrain_line (x0, &x1, y0-1, &y1, width, height);
-                        vp8_blit_line  (x0,  x1, y0-1,  y1, y_buffer, y_stride);
+                        constrain_line (lx0, &x1, ly0-1, &y1, width, height);
+                        vp8_blit_line  (lx0,  x1, ly0-1,  y1, y_buffer, y_stride);
 
-                        constrain_line (x0, &x1, y0+1, &y1, width, height);
-                        vp8_blit_line  (x0,  x1, y0+1,  y1, y_buffer, y_stride);
+                        constrain_line (lx0, &x1, ly0+1, &y1, width, height);
+                        vp8_blit_line  (lx0,  x1, ly0+1,  y1, y_buffer, y_stride);
                     }
                     else
-                        vp8_blit_line  (x0,  x1, y0,  y1, y_buffer, y_stride);
+                        vp8_blit_line  (lx0,  x1, ly0,  y1, y_buffer, y_stride);
                 }
+
                 mi++;
             }
             mi++;
@@ -779,9 +987,10 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
     }
 
     /* Color in block modes */
-    if (flags & VP8D_DEBUG_LEVEL6)
+    if ((flags & VP8D_DEBUG_CLR_BLK_MODES)
+        && (ppflags->display_mb_modes_flag || ppflags->display_b_modes_flag))
     {
-        int i, j;
+        int y, x;
         YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
         int width  = post->y_width;
         int height = post->y_height;
@@ -791,18 +1000,54 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
         int y_stride = oci->post_proc_buffer.y_stride;
         MODE_INFO *mi = oci->mi;
 
-        for (i = 0; i < height; i += 16)
+        for (y = 0; y < height; y += 16)
         {
-            for (j = 0; j < width; j += 16)
+            for (x = 0; x < width; x += 16)
             {
                 int Y = 0, U = 0, V = 0;
 
-                Y = MB_PREDICTION_MODE_colors[mi->mbmi.mode][0];
-                U = MB_PREDICTION_MODE_colors[mi->mbmi.mode][1];
-                V = MB_PREDICTION_MODE_colors[mi->mbmi.mode][2];
+                if (mi->mbmi.mode == B_PRED &&
+                    ((ppflags->display_mb_modes_flag & B_PRED) || ppflags->display_b_modes_flag))
+                {
+                    int by, bx;
+                    unsigned char *yl, *ul, *vl;
+                    union b_mode_info *bmi = mi->bmi;
 
-                POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_mb)
-                    (&y_ptr[j], &u_ptr[j>>1], &v_ptr[j>>1], Y, U, V, 0xc000, y_stride);
+                    yl = y_ptr + x;
+                    ul = u_ptr + (x>>1);
+                    vl = v_ptr + (x>>1);
+
+                    for (by = 0; by < 16; by += 4)
+                    {
+                        for (bx = 0; bx < 16; bx += 4)
+                        {
+                            if ((ppflags->display_b_modes_flag & (1<<mi->mbmi.mode))
+                                || (ppflags->display_mb_modes_flag & B_PRED))
+                            {
+                                Y = B_PREDICTION_MODE_colors[bmi->as_mode][0];
+                                U = B_PREDICTION_MODE_colors[bmi->as_mode][1];
+                                V = B_PREDICTION_MODE_colors[bmi->as_mode][2];
+
+                                POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_b)
+                                    (yl+bx, ul+(bx>>1), vl+(bx>>1), Y, U, V, 0xc000, y_stride);
+                            }
+                            bmi++;
+                        }
+
+                        yl += y_stride*4;
+                        ul += y_stride*1;
+                        vl += y_stride*1;
+                    }
+                }
+                else if (ppflags->display_mb_modes_flag & (1<<mi->mbmi.mode))
+                {
+                    Y = MB_PREDICTION_MODE_colors[mi->mbmi.mode][0];
+                    U = MB_PREDICTION_MODE_colors[mi->mbmi.mode][1];
+                    V = MB_PREDICTION_MODE_colors[mi->mbmi.mode][2];
+
+                    POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_mb_inner)
+                        (y_ptr+x, u_ptr+(x>>1), v_ptr+(x>>1), Y, U, V, 0xc000, y_stride);
+                }
 
                 mi++;
             }
@@ -815,9 +1060,9 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
     }
 
     /* Color in frame reference blocks */
-    if (flags & VP8D_DEBUG_LEVEL7)
+    if ((flags & VP8D_DEBUG_CLR_FRM_REF_BLKS) && ppflags->display_ref_frame_flag)
     {
-        int i, j;
+        int y, x;
         YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
         int width  = post->y_width;
         int height = post->y_height;
@@ -827,18 +1072,21 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
         int y_stride = oci->post_proc_buffer.y_stride;
         MODE_INFO *mi = oci->mi;
 
-        for (i = 0; i < height; i += 16)
+        for (y = 0; y < height; y += 16)
         {
-            for (j = 0; j < width; j +=16)
+            for (x = 0; x < width; x +=16)
             {
                 int Y = 0, U = 0, V = 0;
 
-                Y = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][0];
-                U = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][1];
-                V = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][2];
+                if (ppflags->display_ref_frame_flag & (1<<mi->mbmi.ref_frame))
+                {
+                    Y = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][0];
+                    U = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][1];
+                    V = MV_REFERENCE_FRAME_colors[mi->mbmi.ref_frame][2];
 
-                POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_mb)
-                    (&y_ptr[j], &u_ptr[j>>1], &v_ptr[j>>1], Y, U, V, 0xc000, y_stride);
+                    POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_mb_outer)
+                        (y_ptr+x, u_ptr+(x>>1), v_ptr+(x>>1), Y, U, V, 0xc000, y_stride);
+                }
 
                 mi++;
             }
@@ -849,6 +1097,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, int deblock_l
             mi++;
         }
     }
+#endif
 
     *dest = oci->post_proc_buffer;
 

media/libvpx/vp8/common/postproc.h

@@ -24,7 +24,15 @@
               char whiteclamp[16], char bothclamp[16],\
               unsigned int w, unsigned int h, int pitch)
 
-#define prototype_postproc_blend_mb(sym)\
+#define prototype_postproc_blend_mb_inner(sym)\
+    void sym (unsigned char *y, unsigned char *u, unsigned char *v,\
+              int y1, int u1, int v1, int alpha, int stride)
+
+#define prototype_postproc_blend_mb_outer(sym)\
+    void sym (unsigned char *y, unsigned char *u, unsigned char *v,\
+              int y1, int u1, int v1, int alpha, int stride)
+
+#define prototype_postproc_blend_b(sym)\
     void sym (unsigned char *y, unsigned char *u, unsigned char *v,\
               int y1, int u1, int v1, int alpha, int stride)
 
@@ -52,22 +60,36 @@ extern prototype_postproc(vp8_postproc_downacross);
 #endif
 extern prototype_postproc_addnoise(vp8_postproc_addnoise);
 
-#ifndef vp8_postproc_blend_mb
-#define vp8_postproc_blend_mb vp8_blend_mb_c
+#ifndef vp8_postproc_blend_mb_inner
+#define vp8_postproc_blend_mb_inner vp8_blend_mb_inner_c
 #endif
-extern prototype_postproc_blend_mb(vp8_postproc_blend_mb);
+extern prototype_postproc_blend_mb_inner(vp8_postproc_blend_mb_inner);
+
+#ifndef vp8_postproc_blend_mb_outer
+#define vp8_postproc_blend_mb_outer vp8_blend_mb_outer_c
+#endif
+extern prototype_postproc_blend_mb_outer(vp8_postproc_blend_mb_outer);
+
+#ifndef vp8_postproc_blend_b
+#define vp8_postproc_blend_b vp8_blend_b_c
+#endif
+extern prototype_postproc_blend_b(vp8_postproc_blend_b);
 
 typedef prototype_postproc((*vp8_postproc_fn_t));
 typedef prototype_postproc_inplace((*vp8_postproc_inplace_fn_t));
 typedef prototype_postproc_addnoise((*vp8_postproc_addnoise_fn_t));
-typedef prototype_postproc_blend_mb((*vp8_postproc_blend_mb_fn_t));
+typedef prototype_postproc_blend_mb_inner((*vp8_postproc_blend_mb_inner_fn_t));
+typedef prototype_postproc_blend_mb_outer((*vp8_postproc_blend_mb_outer_fn_t));
+typedef prototype_postproc_blend_b((*vp8_postproc_blend_b_fn_t));
 typedef struct
 {
-    vp8_postproc_inplace_fn_t   down;
-    vp8_postproc_inplace_fn_t   across;
-    vp8_postproc_fn_t           downacross;
-    vp8_postproc_addnoise_fn_t  addnoise;
-    vp8_postproc_blend_mb_fn_t  blend_mb;
+    vp8_postproc_inplace_fn_t           down;
+    vp8_postproc_inplace_fn_t           across;
+    vp8_postproc_fn_t                   downacross;
+    vp8_postproc_addnoise_fn_t          addnoise;
+    vp8_postproc_blend_mb_inner_fn_t    blend_mb_inner;
+    vp8_postproc_blend_mb_outer_fn_t    blend_mb_outer;
+    vp8_postproc_blend_b_fn_t           blend_b;
 } vp8_postproc_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -89,7 +111,7 @@ struct postproc_state
 #include "onyxc_int.h"
 #include "ppflags.h"
 int vp8_post_proc_frame(struct VP8Common *oci, YV12_BUFFER_CONFIG *dest,
-                        int deblock_level, int noise_level, int flags);
+                        vp8_ppflags_t *flags);
 
 
 void vp8_de_noise(YV12_BUFFER_CONFIG         *source,

media/libvpx/vp8/common/ppflags.h

@@ -13,17 +13,28 @@
 #define __INC_PPFLAGS_H
 enum
 {
-    VP8D_NOFILTERING    = 0,
-    VP8D_DEBLOCK        = 1<<0,
-    VP8D_DEMACROBLOCK   = 1<<1,
-    VP8D_ADDNOISE       = 1<<2,
-    VP8D_DEBUG_LEVEL1   = 1<<3,
-    VP8D_DEBUG_LEVEL2   = 1<<4,
-    VP8D_DEBUG_LEVEL3   = 1<<5,
-    VP8D_DEBUG_LEVEL4   = 1<<6,
-    VP8D_DEBUG_LEVEL5   = 1<<7,
-    VP8D_DEBUG_LEVEL6   = 1<<8,
-    VP8D_DEBUG_LEVEL7   = 1<<9
+    VP8D_NOFILTERING            = 0,
+    VP8D_DEBLOCK                = 1<<0,
+    VP8D_DEMACROBLOCK           = 1<<1,
+    VP8D_ADDNOISE               = 1<<2,
+    VP8D_DEBUG_TXT_FRAME_INFO   = 1<<3,
+    VP8D_DEBUG_TXT_MBLK_MODES   = 1<<4,
+    VP8D_DEBUG_TXT_DC_DIFF      = 1<<5,
+    VP8D_DEBUG_TXT_RATE_INFO    = 1<<6,
+    VP8D_DEBUG_DRAW_MV          = 1<<7,
+    VP8D_DEBUG_CLR_BLK_MODES    = 1<<8,
+    VP8D_DEBUG_CLR_FRM_REF_BLKS = 1<<9
 };
 
+typedef struct
+{
+    int post_proc_flag;
+    int deblocking_level;
+    int noise_level;
+    int display_ref_frame_flag;
+    int display_mb_modes_flag;
+    int display_b_modes_flag;
+    int display_mv_flag;
+} vp8_ppflags_t;
+
 #endif

media/libvpx/vp8/common/predictdc.c

@@ -1,44 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include <stdlib.h>
-#include "blockd.h"
-
-
-void vp8_predict_dc(short *lastdc, short *thisdc, short quant, short *cons)
-{
-    int diff;
-    int sign;
-    int last_dc = *lastdc;
-    int this_dc = *thisdc;
-
-    if (*cons  > DCPREDCNTTHRESH)
-    {
-        this_dc += last_dc;
-    }
-
-    diff = abs(last_dc - this_dc);
-    sign  = (last_dc >> 31) ^(this_dc >> 31);
-    sign |= (!last_dc | !this_dc);
-
-    if (sign)
-    {
-        *cons = 0;
-    }
-    else
-    {
-        if (diff <= DCPREDSIMTHRESH * quant)
-            (*cons)++ ;
-    }
-
-    *thisdc = this_dc;
-    *lastdc = this_dc;
-}

media/libvpx/vp8/common/predictdc.h

@@ -1,18 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#ifndef __PREDICTDC_H
-#define __PREDICTDC_H
-
-void uvvp8_predict_dc(short *lastdc, short *thisdc, short quant, short *cons);
-void vp8_predict_dc(short *lastdc, short *thisdc, short quant, short *cons);
-
-#endif

media/libvpx/vp8/common/preproc.h

@@ -1,46 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-/****************************************************************************
-*
-*   Module Title :     preproc.h
-*
-*   Description  :     simple preprocessor
-*
-****************************************************************************/
-
-#ifndef __INC_PREPROC_H
-#define __INC_PREPROC_H
-
-/****************************************************************************
-*  Types
-****************************************************************************/
-
-typedef struct
-{
-    unsigned char *frame_buffer;
-    int frame;
-    unsigned int *fixed_divide;
-
-    unsigned char *frame_buffer_alloc;
-    unsigned int *fixed_divide_alloc;
-} pre_proc_instance;
-
-/****************************************************************************
-*  Functions.
-****************************************************************************/
-void pre_proc_machine_specific_config(void);
-void delete_pre_proc(pre_proc_instance *ppi);
-int init_pre_proc(pre_proc_instance *ppi, int frame_size);
-extern void spatial_filter_c(pre_proc_instance *ppi, unsigned char *s, unsigned char *d, int width, int height, int pitch, int strength);
-extern void (*temp_filter)(pre_proc_instance *ppi, unsigned char *s, unsigned char *d, int bytes, int strength);
-
-#endif

media/libvpx/vp8/common/recon.h

@@ -26,6 +26,9 @@
 #define prototype_build_intra_predictors(sym) \
     void sym(MACROBLOCKD *x)
 
+#define prototype_intra4x4_predict(sym) \
+    void sym(BLOCKD *x, int b_mode, unsigned char *predictor)
+
 struct vp8_recon_rtcd_vtable;
 
 #if ARCH_X86 || ARCH_X86_64
@@ -88,11 +91,30 @@ extern prototype_build_intra_predictors\
 extern prototype_build_intra_predictors\
     (vp8_recon_build_intra_predictors_mby_s);
 
+#ifndef vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mbuv);
+
+#ifndef vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mbuv_s);
+
+#ifndef vp8_recon_intra4x4_predict
+#define vp8_recon_intra4x4_predict vp8_intra4x4_predict
+#endif
+extern prototype_intra4x4_predict\
+    (vp8_recon_intra4x4_predict);
+
 
 typedef prototype_copy_block((*vp8_copy_block_fn_t));
 typedef prototype_recon_block((*vp8_recon_fn_t));
 typedef prototype_recon_macroblock((*vp8_recon_mb_fn_t));
 typedef prototype_build_intra_predictors((*vp8_build_intra_pred_fn_t));
+typedef prototype_intra4x4_predict((*vp8_intra4x4_pred_fn_t));
 typedef struct vp8_recon_rtcd_vtable
 {
     vp8_copy_block_fn_t  copy16x16;
@@ -105,6 +127,9 @@ typedef struct vp8_recon_rtcd_vtable
     vp8_recon_mb_fn_t    recon_mby;
     vp8_build_intra_pred_fn_t  build_intra_predictors_mby_s;
     vp8_build_intra_pred_fn_t  build_intra_predictors_mby;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mbuv_s;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mbuv;
+    vp8_intra4x4_pred_fn_t intra4x4_predict;
 } vp8_recon_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -113,6 +138,5 @@ typedef struct vp8_recon_rtcd_vtable
 #define RECON_INVOKE(ctx,fn) vp8_recon_##fn
 #endif
 
-void vp8_recon_intra4x4mb(const vp8_recon_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
 void vp8_recon_intra_mbuv(const vp8_recon_rtcd_vtable_t *rtcd, MACROBLOCKD *x);
 #endif

media/libvpx/vp8/common/reconinter.c

@@ -10,6 +10,7 @@
 
 
 #include "vpx_ports/config.h"
+#include "vpx/vpx_integer.h"
 #include "recon.h"
 #include "subpixel.h"
 #include "blockd.h"
@@ -18,12 +19,6 @@
 #include "onyxc_int.h"
 #endif
 
-/* use this define on systems where unaligned int reads and writes are
- * not allowed, i.e. ARM architectures
- */
-/*#define MUST_BE_ALIGNED*/
-
-
 static const int bbb[4] = {0, 2, 8, 10};
 
 
@@ -39,7 +34,7 @@ void vp8_copy_mem16x16_c(
 
     for (r = 0; r < 16; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -58,10 +53,10 @@ void vp8_copy_mem16x16_c(
         dst[15] = src[15];
 
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
-        ((int *)dst)[2] = ((int *)src)[2] ;
-        ((int *)dst)[3] = ((int *)src)[3] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
+        ((uint32_t *)dst)[2] = ((uint32_t *)src)[2] ;
+        ((uint32_t *)dst)[3] = ((uint32_t *)src)[3] ;
 
 #endif
         src += src_stride;
@@ -81,7 +76,7 @@ void vp8_copy_mem8x8_c(
 
     for (r = 0; r < 8; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -91,8 +86,8 @@ void vp8_copy_mem8x8_c(
         dst[6] = src[6];
         dst[7] = src[7];
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
 #endif
         src += src_stride;
         dst += dst_stride;
@@ -111,7 +106,7 @@ void vp8_copy_mem8x4_c(
 
     for (r = 0; r < 4; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -121,8 +116,8 @@ void vp8_copy_mem8x4_c(
         dst[6] = src[6];
         dst[7] = src[7];
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
 #endif
         src += src_stride;
         dst += dst_stride;
@@ -154,13 +149,13 @@ void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
 
         for (r = 0; r < 4; r++)
         {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
             pred_ptr[0]  = ptr[0];
             pred_ptr[1]  = ptr[1];
             pred_ptr[2]  = ptr[2];
             pred_ptr[3]  = ptr[3];
 #else
-            *(int *)pred_ptr = *(int *)ptr ;
+            *(uint32_t *)pred_ptr = *(uint32_t *)ptr ;
 #endif
             pred_ptr     += pitch;
             ptr         += d->pre_stride;
@@ -168,7 +163,7 @@ void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
     }
 }
 
-void vp8_build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
+static void build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 {
     unsigned char *ptr_base;
     unsigned char *ptr;
@@ -187,7 +182,7 @@ void vp8_build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
     }
 }
 
-void vp8_build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
+static void build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 {
     unsigned char *ptr_base;
     unsigned char *ptr;
@@ -207,12 +202,12 @@ void vp8_build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 }
 
 
+/*encoder only*/
 void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 {
     int i;
 
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
         unsigned char *uptr, *vptr;
         unsigned char *upred_ptr = &x->predictor[256];
@@ -246,7 +241,7 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
             BLOCKD *d1 = &x->block[i+1];
 
             if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                vp8_build_inter_predictors2b(x, d0, 8);
+                build_inter_predictors2b(x, d0, 8);
             else
             {
                 vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
@@ -257,158 +252,132 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 }
 
 /*encoder only*/
-void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
+void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x)
 {
+    unsigned char *ptr_base;
+    unsigned char *ptr;
+    unsigned char *pred_ptr = x->predictor;
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+    int pre_stride = x->block[0].pre_stride;
 
-  if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-      x->mode_info_context->mbmi.mode != SPLITMV)
+    ptr_base = x->pre.y_buffer;
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
     {
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *pred_ptr = x->predictor;
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->block[0].pre_stride;
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
+    }
+}
 
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+void vp8_build_inter16x16_predictors_mb(MACROBLOCKD *x,
+                                        unsigned char *dst_y,
+                                        unsigned char *dst_u,
+                                        unsigned char *dst_v,
+                                        int dst_ystride,
+                                        int dst_uvstride)
+{
+    int offset;
+    unsigned char *ptr;
+    unsigned char *uptr, *vptr;
 
-        if ((mv_row | mv_col) & 7)
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+
+    unsigned char *ptr_base = x->pre.y_buffer;
+    int pre_stride = x->block[0].pre_stride;
+
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_y, dst_ystride);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_y, dst_ystride);
+    }
+
+    mv_row = x->block[16].bmi.mv.as_mv.row;
+    mv_col = x->block[16].bmi.mv.as_mv.col;
+    pre_stride >>= 1;
+    offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
+    uptr = x->pre.u_buffer + offset;
+    vptr = x->pre.v_buffer + offset;
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, dst_u, dst_uvstride);
+        x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, dst_v, dst_uvstride);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, dst_u, dst_uvstride);
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, dst_v, dst_uvstride);
+    }
+
+}
+
+void vp8_build_inter4x4_predictors_mb(MACROBLOCKD *x)
+{
+    int i;
+
+    if (x->mode_info_context->mbmi.partitioning < 3)
+    {
+        for (i = 0; i < 4; i++)
         {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
+            BLOCKD *d = &x->block[bbb[i]];
+            build_inter_predictors4b(x, d, 16);
         }
     }
     else
     {
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
+        for (i = 0; i < 16; i += 2)
         {
-            for (i = 0; i < 4; i++)
+            BLOCKD *d0 = &x->block[i];
+            BLOCKD *d1 = &x->block[i+1];
+
+            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
+                build_inter_predictors2b(x, d0, 16);
+            else
             {
-                BLOCKD *d = &x->block[bbb[i]];
-                vp8_build_inter_predictors4b(x, d, 16);
+                vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
+                vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
             }
 
         }
+
+    }
+
+    for (i = 16; i < 24; i += 2)
+    {
+        BLOCKD *d0 = &x->block[i];
+        BLOCKD *d1 = &x->block[i+1];
+
+        if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
+            build_inter_predictors2b(x, d0, 8);
         else
         {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                    vp8_build_inter_predictors2b(x, d0, 16);
-                else
-                {
-                    vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
-                    vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
-                }
-
-            }
+            vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
+            vp8_build_inter_predictors_b(d1, 8, x->subpixel_predict);
         }
     }
 }
 
 void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
 {
-
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
-        int offset;
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *uptr, *vptr;
-        unsigned char *pred_ptr = x->predictor;
-        unsigned char *upred_ptr = &x->predictor[256];
-        unsigned char *vpred_ptr = &x->predictor[320];
-
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->block[0].pre_stride;
-
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
-        }
-
-        mv_row = x->block[16].bmi.mv.as_mv.row;
-        mv_col = x->block[16].bmi.mv.as_mv.col;
-        pre_stride >>= 1;
-        offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
-        uptr = x->pre.u_buffer + offset;
-        vptr = x->pre.v_buffer + offset;
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, upred_ptr, 8);
-            x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vpred_ptr, 8);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, upred_ptr, 8);
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vpred_ptr, 8);
-        }
+        vp8_build_inter16x16_predictors_mb(x, x->predictor, &x->predictor[256],
+                                           &x->predictor[320], 16, 8);
     }
     else
     {
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
-        {
-            for (i = 0; i < 4; i++)
-            {
-                BLOCKD *d = &x->block[bbb[i]];
-                vp8_build_inter_predictors4b(x, d, 16);
-            }
-        }
-        else
-        {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                    vp8_build_inter_predictors2b(x, d0, 16);
-                else
-                {
-                    vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
-                    vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
-                }
-
-            }
-
-        }
-
-        for (i = 16; i < 24; i += 2)
-        {
-            BLOCKD *d0 = &x->block[i];
-            BLOCKD *d1 = &x->block[i+1];
-
-            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                vp8_build_inter_predictors2b(x, d0, 8);
-            else
-            {
-                vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
-                vp8_build_inter_predictors_b(d1, 8, x->subpixel_predict);
-            }
-
-        }
-
+        vp8_build_inter4x4_predictors_mb(x);
     }
 }
 
@@ -492,202 +461,5 @@ void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel)
 }
 
 
-/* The following functions are wriiten for skip_recon_mb() to call. Since there is no recon in this
- * situation, we can write the result directly to dst buffer instead of writing it to predictor
- * buffer and then copying it to dst buffer.
- */
-static void vp8_build_inter_predictors_b_s(BLOCKD *d, unsigned char *dst_ptr, vp8_subpix_fn_t sppf)
-{
-    int r;
-    unsigned char *ptr_base;
-    unsigned char *ptr;
-    /*unsigned char *pred_ptr = d->predictor;*/
-    int dst_stride = d->dst_stride;
-    int pre_stride = d->pre_stride;
-
-    ptr_base = *(d->base_pre);
-
-    if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
-    {
-        ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-        sppf(ptr, pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, dst_stride);
-    }
-    else
-    {
-        ptr_base += d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-        ptr = ptr_base;
-
-        for (r = 0; r < 4; r++)
-        {
-#ifdef MUST_BE_ALIGNED
-            dst_ptr[0]   = ptr[0];
-            dst_ptr[1]   = ptr[1];
-            dst_ptr[2]   = ptr[2];
-            dst_ptr[3]   = ptr[3];
-#else
-            *(int *)dst_ptr = *(int *)ptr ;
-#endif
-            dst_ptr      += dst_stride;
-            ptr         += pre_stride;
-        }
-    }
-}
 
 
-
-void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
-{
-    /*unsigned char *pred_ptr = x->block[0].predictor;
-    unsigned char *dst_ptr = *(x->block[0].base_dst) + x->block[0].dst;*/
-    unsigned char *pred_ptr = x->predictor;
-    unsigned char *dst_ptr = x->dst.y_buffer;
-
-    if (x->mode_info_context->mbmi.mode != SPLITMV)
-    {
-        int offset;
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *uptr, *vptr;
-        /*unsigned char *pred_ptr = x->predictor;
-        unsigned char *upred_ptr = &x->predictor[256];
-        unsigned char *vpred_ptr = &x->predictor[320];*/
-        unsigned char *udst_ptr = x->dst.u_buffer;
-        unsigned char *vdst_ptr = x->dst.v_buffer;
-
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->dst.y_stride; /*x->block[0].pre_stride;*/
-
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-
-        mv_row = x->block[16].bmi.mv.as_mv.row;
-        mv_col = x->block[16].bmi.mv.as_mv.col;
-        pre_stride >>= 1;
-        offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
-        uptr = x->pre.u_buffer + offset;
-        vptr = x->pre.v_buffer + offset;
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, udst_ptr, x->dst.uv_stride);
-            x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vdst_ptr, x->dst.uv_stride);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, udst_ptr, x->dst.uv_stride);
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vdst_ptr, x->dst.uv_stride);
-        }
-    }
-    else
-    {
-        /* note: this whole ELSE part is not executed at all. So, no way to test the correctness of my modification. Later,
-         * if sth is wrong, go back to what it is in build_inter_predictors_mb.
-         */
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
-        {
-            for (i = 0; i < 4; i++)
-            {
-                BLOCKD *d = &x->block[bbb[i]];
-                /*vp8_build_inter_predictors4b(x, d, 16);*/
-
-                {
-                    unsigned char *ptr_base;
-                    unsigned char *ptr;
-                    unsigned char *pred_ptr = d->predictor;
-
-                    ptr_base = *(d->base_pre);
-                    ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-
-                    if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
-                    {
-                        x->subpixel_predict8x8(ptr, d->pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
-                    else
-                    {
-                        RECON_INVOKE(&x->rtcd->recon, copy8x8)(ptr, d->pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
-                }
-            }
-        }
-        else
-        {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                {
-                    /*vp8_build_inter_predictors2b(x, d0, 16);*/
-                    unsigned char *ptr_base;
-                    unsigned char *ptr;
-                    unsigned char *pred_ptr = d0->predictor;
-
-                    ptr_base = *(d0->base_pre);
-                    ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
-
-                    if (d0->bmi.mv.as_mv.row & 7 || d0->bmi.mv.as_mv.col & 7)
-                    {
-                        x->subpixel_predict8x4(ptr, d0->pre_stride, d0->bmi.mv.as_mv.col & 7, d0->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride);
-                    }
-                    else
-                    {
-                        RECON_INVOKE(&x->rtcd->recon, copy8x4)(ptr, d0->pre_stride, dst_ptr, x->dst.y_stride);
-                    }
-                }
-                else
-                {
-                    vp8_build_inter_predictors_b_s(d0, dst_ptr, x->subpixel_predict);
-                    vp8_build_inter_predictors_b_s(d1, dst_ptr, x->subpixel_predict);
-                }
-            }
-        }
-
-        for (i = 16; i < 24; i += 2)
-        {
-            BLOCKD *d0 = &x->block[i];
-            BLOCKD *d1 = &x->block[i+1];
-
-            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-            {
-                /*vp8_build_inter_predictors2b(x, d0, 8);*/
-                unsigned char *ptr_base;
-                unsigned char *ptr;
-                unsigned char *pred_ptr = d0->predictor;
-
-                ptr_base = *(d0->base_pre);
-                ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
-
-                if (d0->bmi.mv.as_mv.row & 7 || d0->bmi.mv.as_mv.col & 7)
-                {
-                    x->subpixel_predict8x4(ptr, d0->pre_stride,
-                        d0->bmi.mv.as_mv.col & 7,
-                        d0->bmi.mv.as_mv.row & 7,
-                        dst_ptr, x->dst.uv_stride);
-                }
-                else
-                {
-                    RECON_INVOKE(&x->rtcd->recon, copy8x4)(ptr,
-                        d0->pre_stride, dst_ptr, x->dst.uv_stride);
-                }
-            }
-            else
-            {
-                vp8_build_inter_predictors_b_s(d0, dst_ptr, x->subpixel_predict);
-                vp8_build_inter_predictors_b_s(d1, dst_ptr, x->subpixel_predict);
-            }
-        }
-    }
-}

media/libvpx/vp8/common/reconinter.h

@@ -13,9 +13,15 @@
 #define __INC_RECONINTER_H
 
 extern void vp8_build_inter_predictors_mb(MACROBLOCKD *x);
-extern void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x);
+extern void vp8_build_inter16x16_predictors_mb(MACROBLOCKD *x,
+                                               unsigned char *dst_y,
+                                               unsigned char *dst_u,
+                                               unsigned char *dst_v,
+                                               int dst_ystride,
+                                               int dst_uvstride);
 
-extern void vp8_build_inter_predictors_mby(MACROBLOCKD *x);
+
+extern void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x);
 extern void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel);
 extern void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf);
 extern void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x);

media/libvpx/vp8/common/reconintra.h

@@ -14,9 +14,4 @@
 
 extern void init_intra_left_above_pixels(MACROBLOCKD *x);
 
-extern void vp8_build_intra_predictors_mbuv(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mbuv_s(MACROBLOCKD *x);
-
-extern void vp8_predict_intra4x4(BLOCKD *x, int b_mode, unsigned char *Predictor);
-
 #endif

media/libvpx/vp8/common/reconintra4x4.c

@@ -14,7 +14,7 @@
 #include "vpx_mem/vpx_mem.h"
 #include "reconintra.h"
 
-void vp8_predict_intra4x4(BLOCKD *x,
+void vp8_intra4x4_predict(BLOCKD *x,
                           int b_mode,
                           unsigned char *predictor)
 {
@@ -313,89 +313,3 @@ void vp8_intra_prediction_down_copy(MACROBLOCKD *x)
 }
 
 
-void vp8_recon_intra4x4mb(const vp8_recon_rtcd_vtable_t *rtcd, MACROBLOCKD *x)
-{
-    int i;
-
-    vp8_intra_prediction_down_copy(x);
-
-#if ARCH_ARM
-    {
-        BLOCKD *b = &x->block[0];
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-        b += 1;
-
-        vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-    }
-#else
-    for (i = 0; i < 16; i++)
-    {
-        BLOCKD *b = &x->block[i];
-
-        vp8_predict_intra4x4(b, x->block[i].bmi.mode, x->block[i].predictor);
-        RECON_INVOKE(rtcd, recon)(b->predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride);
-    }
-#endif
-
-    vp8_recon_intra_mbuv(rtcd, x);
-
-}

media/libvpx/vp8/common/threading.h

@@ -12,7 +12,7 @@
 #ifndef _PTHREAD_EMULATION
 #define _PTHREAD_EMULATION
 
-#define VPXINFINITE 10000       /* 10second. */
+#if CONFIG_OS_SUPPORT && CONFIG_MULTITHREAD
 
 /* Thread management macros */
 #ifdef _WIN32
@@ -26,7 +26,7 @@
 #define pthread_t HANDLE
 #define pthread_attr_t DWORD
 #define pthread_create(thhandle,attr,thfunc,tharg) (int)((*thhandle=(HANDLE)_beginthreadex(NULL,0,(unsigned int (__stdcall *)(void *))thfunc,tharg,0,NULL))==NULL)
-#define pthread_join(thread, result) ((WaitForSingleObject((thread),VPXINFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
+#define pthread_join(thread, result) ((WaitForSingleObject((thread),INFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
 #define pthread_detach(thread) if(thread!=NULL)CloseHandle(thread)
 #define thread_sleep(nms) Sleep(nms)
 #define pthread_cancel(thread) terminate_thread(thread,0)
@@ -36,6 +36,7 @@
 #define pthread_self() GetCurrentThreadId()
 #else
 #ifdef __APPLE__
+#include <mach/mach_init.h>
 #include <mach/semaphore.h>
 #include <mach/task.h>
 #include <time.h>
@@ -58,9 +59,9 @@
 #ifdef _WIN32
 #define sem_t HANDLE
 #define pause(voidpara) __asm PAUSE
-#define sem_init(sem, sem_attr1, sem_init_value) (int)((*sem = CreateEvent(NULL,FALSE,FALSE,NULL))==NULL)
-#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,VPXINFINITE))
-#define sem_post(sem) SetEvent(*sem)
+#define sem_init(sem, sem_attr1, sem_init_value) (int)((*sem = CreateSemaphore(NULL,0,32768,NULL))==NULL)
+#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,INFINITE))
+#define sem_post(sem) ReleaseSemaphore(*sem,1,NULL)
 #define sem_destroy(sem) if(*sem)((int)(CloseHandle(*sem))==TRUE)
 #define thread_sleep(nms) Sleep(nms)
 
@@ -88,4 +89,6 @@
 #define x86_pause_hint()
 #endif
 
+#endif /* CONFIG_OS_SUPPORT && CONFIG_MULTITHREAD */
+
 #endif

media/libvpx/vp8/common/vpxerrors.h

@@ -1,13 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-
-#define ALLOC_FAILURE -2

media/libvpx/vp8/common/x86/idctllm_mmx.asm

@@ -14,18 +14,18 @@
 ; /****************************************************************************
 ; * Notes:
 ; *
-; * This implementation makes use of 16 bit fixed point verio of two multiply
+; * This implementation makes use of 16 bit fixed point version of two multiply
 ; * constants:
 ; *        1.   sqrt(2) * cos (pi/8)
-; *         2.   sqrt(2) * sin (pi/8)
-; * Becuase the first constant is bigger than 1, to maintain the same 16 bit
-; * fixed point prrcision as the second one, we use a trick of
+; *        2.   sqrt(2) * sin (pi/8)
+; * Because the first constant is bigger than 1, to maintain the same 16 bit
+; * fixed point precision as the second one, we use a trick of
 ; *        x * a = x + x*(a-1)
 ; * so
 ; *        x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1).
 ; *
-; * For     the second constant, becuase of the 16bit version is 35468, which
-; * is bigger than 32768, in signed 16 bit multiply, it become a negative
+; * For the second constant, because of the 16bit version is 35468, which
+; * is bigger than 32768, in signed 16 bit multiply, it becomes a negative
 ; * number.
 ; *        (x * (unsigned)35468 >> 16) = x * (signed)35468 >> 16 + x
 ; *

media/libvpx/vp8/common/x86/idctllm_sse2.asm

@@ -32,9 +32,6 @@ sym(idct_dequant_0_2x_sse2):
         mov         rdx,            arg(1) ; dequant
         mov         rax,            arg(0) ; qcoeff
 
-    ; Zero out xmm7, for use unpacking
-        pxor        xmm7,           xmm7
-
         movd        xmm4,           [rax]
         movd        xmm5,           [rdx]
 
@@ -43,9 +40,12 @@ sym(idct_dequant_0_2x_sse2):
 
         pmullw      xmm4,           xmm5
 
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
+
     ; clear coeffs
-        movd        [rax],          xmm7
-        movd        [rax+32],       xmm7
+        movd        [rax],          xmm5
+        movd        [rax+32],       xmm5
 ;pshufb
         pshuflw     xmm4,           xmm4,       00000000b
         pshufhw     xmm4,           xmm4,       00000000b
@@ -62,10 +62,10 @@ sym(idct_dequant_0_2x_sse2):
         lea         rcx,            [3*rcx]
         movq        xmm3,           [rax+rcx]
 
-        punpcklbw   xmm0,           xmm7
-        punpcklbw   xmm1,           xmm7
-        punpcklbw   xmm2,           xmm7
-        punpcklbw   xmm3,           xmm7
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
         mov         rax,            arg(3) ; dst
         movsxd      rdx,            dword ptr arg(4) ; dst_stride
@@ -77,10 +77,10 @@ sym(idct_dequant_0_2x_sse2):
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; store blocks back out
         movq        [rax],          xmm0
@@ -102,6 +102,7 @@ sym(idct_dequant_full_2x_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -347,6 +348,7 @@ sym(idct_dequant_full_2x_sse2):
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -377,8 +379,8 @@ sym(idct_dequant_dc_0_2x_sse2):
         mov         rdi,            arg(3) ; dst
         mov         rdx,            arg(5) ; dc
 
-    ; Zero out xmm7, for use unpacking
-        pxor        xmm7,           xmm7
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
 
     ; load up 2 dc words here == 2*16 = doubleword
         movd        xmm4,           [rdx]
@@ -398,10 +400,10 @@ sym(idct_dequant_dc_0_2x_sse2):
         psraw       xmm4,           3
 
     ; Predict buffer needs to be expanded from bytes to words
-        punpcklbw   xmm0,           xmm7
-        punpcklbw   xmm1,           xmm7
-        punpcklbw   xmm2,           xmm7
-        punpcklbw   xmm3,           xmm7
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
     ; Add to predict buffer
         paddw       xmm0,           xmm4
@@ -410,10 +412,10 @@ sym(idct_dequant_dc_0_2x_sse2):
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; Load destination stride before writing out,
     ;   doesn't need to persist
@@ -441,6 +443,7 @@ sym(idct_dequant_dc_full_2x_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -692,6 +695,7 @@ sym(idct_dequant_dc_full_2x_sse2):
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret

media/libvpx/vp8/common/x86/iwalsh_sse2.asm

@@ -17,7 +17,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 2
-    SAVE_XMM
+    SAVE_XMM 6
     push        rsi
     push        rdi
     ; end prolog
@@ -41,7 +41,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     movdqa    xmm4, xmm0
     punpcklqdq  xmm0, xmm3          ;d1 a1
     punpckhqdq  xmm4, xmm3          ;c1 b1
-    movd    xmm7, eax
+    movd    xmm6, eax
 
     movdqa    xmm1, xmm4          ;c1 b1
     paddw   xmm4, xmm0          ;dl+cl a1+b1 aka op[4] op[0]
@@ -66,7 +66,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     pshufd    xmm2, xmm1, 4eh       ;ip[8] ip[12]
     movdqa    xmm3, xmm4          ;ip[4] ip[0]
 
-    pshufd    xmm7, xmm7, 0       ;03 03 03 03 03 03 03 03
+    pshufd    xmm6, xmm6, 0       ;03 03 03 03 03 03 03 03
 
     paddw   xmm4, xmm2          ;ip[4]+ip[8] ip[0]+ip[12] aka b1 a1
     psubw   xmm3, xmm2          ;ip[4]-ip[8] ip[0]-ip[12] aka c1 d1
@@ -90,8 +90,8 @@ sym(vp8_short_inv_walsh4x4_sse2):
     punpcklwd xmm5, xmm0          ; 31 21 11 01 30 20 10 00
     punpckhwd xmm1, xmm0          ; 33 23 13 03 32 22 12 02
 ;~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-    paddw   xmm5, xmm7
-    paddw   xmm1, xmm7
+    paddw   xmm5, xmm6
+    paddw   xmm1, xmm6
 
     psraw   xmm5, 3
     psraw   xmm1, 3

media/libvpx/vp8/common/x86/loopfilter_mmx.asm

@@ -16,7 +16,7 @@
 ;(
 ;    unsigned char *src_ptr,
 ;    int src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int  count
@@ -122,12 +122,10 @@ next8_h:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm2              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx, arg(2) ;flimit           ; get flimit
-        movq        mm2, [rdx]            ; flimit mm2
-        paddb       mm2, mm2              ; flimit*2 (less than 255)
-        paddb       mm7, mm2              ; flimit * 2 + limit (less than 255)
+        mov         rdx, arg(2) ;blimit           ; get blimit
+        movq        mm7, [rdx]            ; blimit
 
-        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,    mm5
         pxor        mm5,    mm5
         pcmpeqb     mm1,    mm5           ; mask mm1
@@ -230,7 +228,7 @@ next8_h:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -406,9 +404,9 @@ next8_v:
         pand        mm5,        [GLOBAL(tfe)]               ; set lsb of each byte to zero
         psrlw       mm5,        1                           ; abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                      ;
+        mov         rdx,        arg(2) ;blimit                      ;
 
-        movq        mm2,        [rdx]                       ;flimit  mm2
+        movq        mm4,        [rdx]                       ;blimit
         movq        mm1,        mm3                         ; mm1=mm3=p0
 
         movq        mm7,        mm6                         ; mm7=mm6=q0
@@ -419,10 +417,7 @@ next8_v:
         paddusb     mm1,        mm1                         ; abs(q0-p0)*2
         paddusb     mm1,        mm5                         ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        paddb       mm2,        mm2                         ; flimit*2 (less than 255)
-        paddb       mm4,        mm2                         ; flimit * 2 + limit (less than 255)
-
-        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,        mm0;                        ; mask
 
         pxor        mm0,        mm0
@@ -603,7 +598,7 @@ next8_v:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -719,17 +714,15 @@ next8_mbh:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm2              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx, arg(2) ;flimit           ; get flimit
-        movq        mm2, [rdx]            ; flimit mm2
-        paddb       mm2, mm2              ; flimit*2 (less than 255)
-        paddb       mm7, mm2              ; flimit * 2 + limit (less than 255)
+        mov         rdx, arg(2) ;blimit           ; get blimit
+        movq        mm7, [rdx]            ; blimit
 
-        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,    mm5
         pxor        mm5,    mm5
         pcmpeqb     mm1,    mm5           ; mask mm1
 
-        ; mm1 = mask, mm0=q0,  mm7 = flimit, t0 = abs(q0-q1) t1 = abs(p1-p0)
+        ; mm1 = mask, mm0=q0,  mm7 = blimit, t0 = abs(q0-q1) t1 = abs(p1-p0)
         ; mm6 = p0,
 
         ; calculate high edge variance
@@ -922,7 +915,7 @@ next8_mbh:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -1108,9 +1101,9 @@ next8_mbv:
         pand        mm5,        [GLOBAL(tfe)]               ; set lsb of each byte to zero
         psrlw       mm5,        1                           ; abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                      ;
+        mov         rdx,        arg(2) ;blimit                      ;
 
-        movq        mm2,        [rdx]                       ;flimit  mm2
+        movq        mm4,        [rdx]                       ;blimit
         movq        mm1,        mm3                         ; mm1=mm3=p0
 
         movq        mm7,        mm6                         ; mm7=mm6=q0
@@ -1121,10 +1114,7 @@ next8_mbv:
         paddusb     mm1,        mm1                         ; abs(q0-p0)*2
         paddusb     mm1,        mm5                         ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        paddb       mm2,        mm2                         ; flimit*2 (less than 255)
-        paddb       mm4,        mm2                         ; flimit * 2 + limit (less than 255)
-
-        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,        mm0;                        ; mask
 
         pxor        mm0,        mm0
@@ -1392,16 +1382,13 @@ next8_mbv:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit
 ;)
 global sym(vp8_loop_filter_simple_horizontal_edge_mmx)
 sym(vp8_loop_filter_simple_horizontal_edge_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
+    SHADOW_ARGS_TO_STACK 3
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1410,14 +1397,10 @@ sym(vp8_loop_filter_simple_horizontal_edge_mmx):
         mov         rsi, arg(0) ;src_ptr
         movsxd      rax, dword ptr arg(1) ;src_pixel_step     ; destination pitch?
 
-        movsxd      rcx, dword ptr arg(5) ;count
+        mov         rcx, 2                ; count
 nexts8_h:
-        mov         rdx, arg(3) ;limit
-        movq        mm7, [rdx]
-        mov         rdx, arg(2) ;flimit           ; get flimit
+        mov         rdx, arg(2) ;blimit           ; get blimit
         movq        mm3, [rdx]            ;
-        paddb       mm3, mm3              ; flimit*2 (less than 255)
-        paddb       mm3, mm7              ; flimit * 2 + limit (less than 255)
 
         mov         rdi, rsi              ; rdi points to row +1 for indirect addressing
         add         rdi, rax
@@ -1445,7 +1428,7 @@ nexts8_h:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm1              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        psubusb     mm5, mm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5, mm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        mm3, mm3
         pcmpeqb     mm5, mm3
 
@@ -1515,16 +1498,13 @@ nexts8_h:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit
 ;)
 global sym(vp8_loop_filter_simple_vertical_edge_mmx)
 sym(vp8_loop_filter_simple_vertical_edge_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
+    SHADOW_ARGS_TO_STACK 3
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1539,7 +1519,7 @@ sym(vp8_loop_filter_simple_vertical_edge_mmx):
         movsxd      rax, dword ptr arg(1) ;src_pixel_step     ; destination pitch?
 
         lea         rsi, [rsi + rax*4- 2];  ;
-        movsxd      rcx, dword ptr arg(5) ;count
+        mov         rcx, 2                                      ; count
 nexts8_v:
 
         lea         rdi,        [rsi + rax];
@@ -1602,14 +1582,10 @@ nexts8_v:
         paddusb     mm5,        mm5                             ; abs(p0-q0)*2
         paddusb     mm5,        mm6                             ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                          ; get flimit
+        mov         rdx,        arg(2) ;blimit                          ; get blimit
         movq        mm7,        [rdx]
-        mov         rdx,        arg(3)                          ; get limit
-        movq        mm6,        [rdx]
-        paddb       mm7,        mm7                             ; flimit*2 (less than 255)
-        paddb       mm7,        mm6                             ; flimit * 2 + limit (less than 255)
 
-        psubusb     mm5,        mm7                             ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,        mm7                             ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        mm7,        mm7
         pcmpeqb     mm5,        mm7                             ; mm5 = mask
 

media/libvpx/vp8/common/x86/loopfilter_sse2.asm

@@ -110,7 +110,7 @@
         psubusb     xmm6,                   xmm5              ; p1-=p0
 
         por         xmm6,                   xmm4              ; abs(p1 - p0)
-        mov         rdx,                    arg(2)            ; get flimit
+        mov         rdx,                    arg(2)            ; get blimit
 
         movdqa        t1,                   xmm6              ; save to t1
 
@@ -123,7 +123,7 @@
         psubusb     xmm1,                   xmm7
         por         xmm2,                   xmm3              ; abs(p1-q1)
 
-        movdqa      xmm4,                   XMMWORD PTR [rdx] ; flimit
+        movdqa      xmm7,                   XMMWORD PTR [rdx] ; blimit
 
         movdqa      xmm3,                   xmm0              ; q0
         pand        xmm2,                   [GLOBAL(tfe)]     ; set lsb of each byte to zero
@@ -134,13 +134,11 @@
         psrlw       xmm2,                   1                 ; abs(p1-q1)/2
 
         psubusb     xmm5,                   xmm3              ; p0-=q0
-        paddb       xmm4,                   xmm4              ; flimit*2 (less than 255)
 
         psubusb     xmm3,                   xmm6              ; q0-=p0
         por         xmm5,                   xmm3              ; abs(p0 - q0)
 
         paddusb     xmm5,                   xmm5              ; abs(p0-q0)*2
-        paddb       xmm7,                   xmm4              ; flimit * 2 + limit (less than 255)
 
         movdqa      xmm4,                   t0                ; hev get abs (q1 - q0)
 
@@ -150,7 +148,7 @@
 
         movdqa      xmm2,                   XMMWORD PTR [rdx] ; hev
 
-        psubusb     xmm5,                   xmm7              ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5,                   xmm7              ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         psubusb     xmm4,                   xmm2              ; hev
 
         psubusb     xmm3,                   xmm2              ; hev
@@ -278,7 +276,7 @@
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -288,7 +286,7 @@ sym(vp8_loop_filter_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -328,7 +326,7 @@ sym(vp8_loop_filter_horizontal_edge_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -338,7 +336,7 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -574,7 +572,7 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -584,7 +582,7 @@ sym(vp8_mbloop_filter_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -624,7 +622,7 @@ sym(vp8_mbloop_filter_horizontal_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -634,7 +632,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -904,7 +902,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         movdqa      xmm4,               XMMWORD PTR [rdx]; limit
 
         pmaxub      xmm0,               xmm7
-        mov         rdx,                arg(2)          ; flimit
+        mov         rdx,                arg(2)          ; blimit
 
         psubusb     xmm0,               xmm4
         movdqa      xmm5,               xmm2            ; q1
@@ -921,12 +919,11 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         psrlw       xmm5,               1               ; abs(p1-q1)/2
         psubusb     xmm6,               xmm3            ; q0-p0
 
-        movdqa      xmm2,               XMMWORD PTR [rdx]; flimit
+        movdqa      xmm4,               XMMWORD PTR [rdx]; blimit
 
         mov         rdx,                arg(4)          ; get thresh
 
         por         xmm1,               xmm6            ; abs(q0-p0)
-        paddb       xmm2,               xmm2            ; flimit*2 (less than 255)
 
         movdqa      xmm6,               t0              ; get abs (q1 - q0)
 
@@ -939,10 +936,9 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         paddusb     xmm1,               xmm5            ; abs (p0 - q0) *2 + abs(p1-q1)/2
         psubusb     xmm6,               xmm7            ; abs(q1 - q0) > thresh
 
-        paddb       xmm4,               xmm2            ; flimit * 2 + limit (less than 255)
         psubusb     xmm3,               xmm7            ; abs(p1 - p0)> thresh
 
-        psubusb     xmm1,               xmm4            ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm1,               xmm4            ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         xmm6,               xmm3            ; abs(q1 - q0) > thresh || abs(p1 - p0) > thresh
 
         por         xmm1,               xmm0            ; mask
@@ -1014,7 +1010,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -1024,7 +1020,7 @@ sym(vp8_loop_filter_vertical_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1081,7 +1077,7 @@ sym(vp8_loop_filter_vertical_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -1091,7 +1087,7 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1239,7 +1235,7 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -1249,7 +1245,7 @@ sym(vp8_mbloop_filter_vertical_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1308,7 +1304,7 @@ sym(vp8_mbloop_filter_vertical_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -1318,7 +1314,7 @@ sym(vp8_mbloop_filter_vertical_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1376,17 +1372,14 @@ sym(vp8_mbloop_filter_vertical_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit,
 ;)
 global sym(vp8_loop_filter_simple_horizontal_edge_sse2)
 sym(vp8_loop_filter_simple_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1394,13 +1387,8 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
 
         mov         rsi, arg(0)             ;src_ptr
         movsxd      rax, dword ptr arg(1)   ;src_pixel_step     ; destination pitch?
-        mov         rdx, arg(2) ;flimit     ; get flimit
+        mov         rdx, arg(2)             ;blimit
         movdqa      xmm3, XMMWORD PTR [rdx]
-        mov         rdx, arg(3) ;limit
-        movdqa      xmm7, XMMWORD PTR [rdx]
-
-        paddb       xmm3, xmm3              ; flimit*2 (less than 255)
-        paddb       xmm3, xmm7              ; flimit * 2 + limit (less than 255)
 
         mov         rdi, rsi                ; rdi points to row +1 for indirect addressing
         add         rdi, rax
@@ -1428,7 +1416,7 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
         paddusb     xmm5, xmm5              ; abs(p0-q0)*2
         paddusb     xmm5, xmm1              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        psubusb     xmm5, xmm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5, xmm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        xmm3, xmm3
         pcmpeqb     xmm5, xmm3
 
@@ -1493,17 +1481,14 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit,
 ;)
 global sym(vp8_loop_filter_simple_vertical_edge_sse2)
 sym(vp8_loop_filter_simple_vertical_edge_sse2):
     push        rbp         ; save old base pointer value.
     mov         rbp, rsp    ; set new base pointer value.
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM 7
     GET_GOT     rbx         ; save callee-saved reg
     push        rsi
     push        rdi
@@ -1607,14 +1592,10 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
         paddusb     xmm5,       xmm5                            ; abs(p0-q0)*2
         paddusb     xmm5,       xmm6                            ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx,        arg(2)                          ;flimit
+        mov         rdx,        arg(2)                          ;blimit
         movdqa      xmm7, XMMWORD PTR [rdx]
-        mov         rdx,        arg(3)                          ; get limit
-        movdqa      xmm6, XMMWORD PTR [rdx]
-        paddb       xmm7,        xmm7                           ; flimit*2 (less than 255)
-        paddb       xmm7,        xmm6                           ; flimit * 2 + limit (less than 255)
 
-        psubusb     xmm5,        xmm7                           ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5,        xmm7                           ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        xmm7,        xmm7
         pcmpeqb     xmm5,        xmm7                           ; mm5 = mask
 

media/libvpx/vp8/common/x86/loopfilter_x86.c

@@ -9,30 +9,18 @@
  */
 
 
-#include "vpx_ports/config.h"
-#include "loopfilter.h"
-
-prototype_loopfilter(vp8_loop_filter_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_c);
-prototype_loopfilter(vp8_mbloop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
+#include "vpx_config.h"
+#include "vp8/common/loopfilter.h"
 
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_mmx);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_mmx);
 prototype_loopfilter(vp8_loop_filter_vertical_edge_mmx);
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_mmx);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_mmx);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_mmx);
 
 prototype_loopfilter(vp8_loop_filter_vertical_edge_sse2);
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_sse2);
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_sse2);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_sse2);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_sse2);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_sse2);
-prototype_loopfilter(vp8_fast_loop_filter_vertical_edges_sse2);
 
 extern loop_filter_uvfunction vp8_loop_filter_horizontal_edge_uv_sse2;
 extern loop_filter_uvfunction vp8_loop_filter_vertical_edge_uv_sse2;
@@ -42,113 +30,77 @@ extern loop_filter_uvfunction vp8_mbloop_filter_vertical_edge_uv_sse2;
 #if HAVE_MMX
 /* Horizontal MB filtering */
 void vp8_loop_filter_mbh_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-
-void vp8_loop_filter_mbhs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
-}
-
-
-void vp8_loop_filter_mbvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                            int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
-void vp8_loop_filter_bhs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_mmx(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                            int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
-void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 12, y_stride, blimit);
 }
 #endif
 
@@ -156,113 +108,65 @@ void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigne
 /* Horizontal MB filtering */
 #if HAVE_SSE2
 void vp8_loop_filter_mbh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-}
-
-
-void vp8_loop_filter_mbhs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, v_ptr);
 }
 
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+    vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-}
-
-
-void vp8_loop_filter_mbvs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+        vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, v_ptr);
 }
 
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4 * uv_stride);
+        vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, v_ptr + 4 * uv_stride);
 }
 
 
-void vp8_loop_filter_bhs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_sse2(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4);
+        vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, v_ptr + 4);
 }
 
 
-void vp8_loop_filter_bvs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_sse2(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, blimit);
 }
 
 #endif
-
-#if 0
-void vp8_fast_loop_filter_vertical_edges_sse(unsigned char *y_ptr,
-        int y_stride,
-        loop_filter_info *lfi)
-{
-
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-}
-#endif

media/libvpx/vp8/common/x86/loopfilter_x86.h

@@ -24,10 +24,10 @@ extern prototype_loopfilter_block(vp8_loop_filter_mbv_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_mmx);
 
 
 #if !CONFIG_RUNTIME_CPU_DETECT
@@ -44,13 +44,13 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_mmx);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_mmx
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_mmx
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_mmx
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_mmx
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_mmx
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_mmx
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_mmx
@@ -63,10 +63,10 @@ extern prototype_loopfilter_block(vp8_loop_filter_mbv_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_sse2);
 
 
 #if !CONFIG_RUNTIME_CPU_DETECT
@@ -83,13 +83,13 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_sse2);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_sse2
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_sse2
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_sse2
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_sse2
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_sse2
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_sse2
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_sse2

media/libvpx/vp8/common/x86/postproc_sse2.asm

@@ -26,7 +26,7 @@ sym(vp8_post_proc_down_and_across_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -256,7 +256,7 @@ sym(vp8_mbpost_proc_down_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -456,7 +456,7 @@ sym(vp8_mbpost_proc_across_ip_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

media/libvpx/vp8/common/x86/recon_sse2.asm

@@ -67,7 +67,7 @@ sym(vp8_recon4b_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
-    SAVE_XMM
+    SAVE_XMM 7
     push        rsi
     push        rdi
     ; end prolog
@@ -229,3 +229,460 @@ sym(vp8_copy_mem16x16_sse2):
     UNSHADOW_ARGS
     pop         rbp
     ret
+
+
+;void vp8_intra_pred_uv_dc_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dc_mmx2)
+sym(vp8_intra_pred_uv_dc_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from top
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        mm0,        mm0
+    movq        mm1,        [rsi]
+    psadbw      mm1,        mm0
+
+    ; from left
+    dec         rsi
+    lea         rdi,        [rax*3]
+    movzx       ecx,        byte [rsi+rax]
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    lea         rsi,        [rsi+rax*4]
+    movzx       edx,        byte [rsi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*4]
+    add         ecx,        edx
+
+    ; add up
+    pextrw      edx,        mm1, 0x0
+    lea         edx,        [edx+ecx+8]
+    sar         edx,        4
+    movd        mm1,        edx
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dctop_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dctop_mmx2)
+sym(vp8_intra_pred_uv_dctop_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from top
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        mm0,        mm0
+    movq        mm1,        [rsi]
+    psadbw      mm1,        mm0
+
+    ; add up
+    paddw       mm1,        [GLOBAL(dc_4)]
+    psraw       mm1,        3
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dcleft_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dcleft_mmx2)
+sym(vp8_intra_pred_uv_dcleft_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from left
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    dec         rsi
+    lea         rdi,        [rax*3]
+    movzx       ecx,        byte [rsi]
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    lea         rsi,        [rsi+rax*4]
+    movzx       edx,        byte [rsi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    lea         edx,        [ecx+edx+4]
+
+    ; add up
+    shr         edx,        3
+    movd        mm1,        edx
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dc128_mmx(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dc128_mmx)
+sym(vp8_intra_pred_uv_dc128_mmx):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    ; end prolog
+
+    ; write out
+    movq        mm1,        [GLOBAL(dc_128)]
+    mov         rax,        arg(0) ;dst;
+    movsxd      rdx,        dword ptr arg(1) ;dst_stride
+    lea         rcx,        [rdx*3]
+
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+    lea         rax,        [rax+rdx*4]
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+
+    ; begin epilog
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_tm_sse2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+%macro vp8_intra_pred_uv_tm 1
+global sym(vp8_intra_pred_uv_tm_%1)
+sym(vp8_intra_pred_uv_tm_%1):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; read top row
+    mov         edx,        4
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        xmm0,       xmm0
+%ifidn %1, ssse3
+    movdqa      xmm2,       [GLOBAL(dc_1024)]
+%endif
+    movq        xmm1,       [rsi]
+    punpcklbw   xmm1,       xmm0
+
+    ; set up left ptrs ans subtract topleft
+    movd        xmm3,       [rsi-1]
+    lea         rsi,        [rsi+rax-1]
+%ifidn %1, sse2
+    punpcklbw   xmm3,       xmm0
+    pshuflw     xmm3,       xmm3, 0x0
+    punpcklqdq  xmm3,       xmm3
+%else
+    pshufb      xmm3,       xmm2
+%endif
+    psubw       xmm1,       xmm3
+
+    ; set up dest ptrs
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+
+vp8_intra_pred_uv_tm_%1_loop:
+    movd        xmm3,       [rsi]
+    movd        xmm5,       [rsi+rax]
+%ifidn %1, sse2
+    punpcklbw   xmm3,       xmm0
+    punpcklbw   xmm5,       xmm0
+    pshuflw     xmm3,       xmm3, 0x0
+    pshuflw     xmm5,       xmm5, 0x0
+    punpcklqdq  xmm3,       xmm3
+    punpcklqdq  xmm5,       xmm5
+%else
+    pshufb      xmm3,       xmm2
+    pshufb      xmm5,       xmm2
+%endif
+    paddw       xmm3,       xmm1
+    paddw       xmm5,       xmm1
+    packuswb    xmm3,       xmm5
+    movq  [rdi    ],        xmm3
+    movhps[rdi+rcx],        xmm3
+    lea         rsi,        [rsi+rax*2]
+    lea         rdi,        [rdi+rcx*2]
+    dec         edx
+    jnz vp8_intra_pred_uv_tm_%1_loop
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+%endmacro
+
+vp8_intra_pred_uv_tm sse2
+vp8_intra_pred_uv_tm ssse3
+
+;void vp8_intra_pred_uv_ve_mmx(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_ve_mmx)
+sym(vp8_intra_pred_uv_ve_mmx):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    ; end prolog
+
+    ; read from top
+    mov         rax,        arg(2) ;src;
+    movsxd      rdx,        dword ptr arg(3) ;src_stride;
+    sub         rax,        rdx
+    movq        mm1,        [rax]
+
+    ; write out
+    mov         rax,        arg(0) ;dst;
+    movsxd      rdx,        dword ptr arg(1) ;dst_stride
+    lea         rcx,        [rdx*3]
+
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+    lea         rax,        [rax+rdx*4]
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+
+    ; begin epilog
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_ho_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+%macro vp8_intra_pred_uv_ho 1
+global sym(vp8_intra_pred_uv_ho_%1)
+sym(vp8_intra_pred_uv_ho_%1):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+%ifidn %1, ssse3
+%ifndef GET_GOT_SAVE_ARG
+    push        rbx
+%endif
+    GET_GOT     rbx
+%endif
+    ; end prolog
+
+    ; read from left and write out
+%ifidn %1, mmx2
+    mov         edx,        4
+%endif
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+%ifidn %1, ssse3
+    lea         rdx,        [rcx*3]
+    movdqa      xmm2,       [GLOBAL(dc_00001111)]
+    lea         rbx,        [rax*3]
+%endif
+    dec         rsi
+%ifidn %1, mmx2
+vp8_intra_pred_uv_ho_%1_loop:
+    movd        mm0,        [rsi]
+    movd        mm1,        [rsi+rax]
+    punpcklbw   mm0,        mm0
+    punpcklbw   mm1,        mm1
+    pshufw      mm0,        mm0, 0x0
+    pshufw      mm1,        mm1, 0x0
+    movq  [rdi    ],        mm0
+    movq  [rdi+rcx],        mm1
+    lea         rsi,        [rsi+rax*2]
+    lea         rdi,        [rdi+rcx*2]
+    dec         edx
+    jnz vp8_intra_pred_uv_ho_%1_loop
+%else
+    movd        xmm0,       [rsi]
+    movd        xmm3,       [rsi+rax]
+    movd        xmm1,       [rsi+rax*2]
+    movd        xmm4,       [rsi+rbx]
+    punpcklbw   xmm0,       xmm3
+    punpcklbw   xmm1,       xmm4
+    pshufb      xmm0,       xmm2
+    pshufb      xmm1,       xmm2
+    movq   [rdi    ],       xmm0
+    movhps [rdi+rcx],       xmm0
+    movq [rdi+rcx*2],       xmm1
+    movhps [rdi+rdx],       xmm1
+    lea         rsi,        [rsi+rax*4]
+    lea         rdi,        [rdi+rcx*4]
+    movd        xmm0,       [rsi]
+    movd        xmm3,       [rsi+rax]
+    movd        xmm1,       [rsi+rax*2]
+    movd        xmm4,       [rsi+rbx]
+    punpcklbw   xmm0,       xmm3
+    punpcklbw   xmm1,       xmm4
+    pshufb      xmm0,       xmm2
+    pshufb      xmm1,       xmm2
+    movq   [rdi    ],       xmm0
+    movhps [rdi+rcx],       xmm0
+    movq [rdi+rcx*2],       xmm1
+    movhps [rdi+rdx],       xmm1
+%endif
+
+    ; begin epilog
+%ifidn %1, ssse3
+    RESTORE_GOT
+%ifndef GET_GOT_SAVE_ARG
+    pop         rbx
+%endif
+%endif
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+%endmacro
+
+vp8_intra_pred_uv_ho mmx2
+vp8_intra_pred_uv_ho ssse3
+
+SECTION_RODATA
+dc_128:
+    times 8 db 128
+dc_4:
+    times 4 dw 4
+align 16
+dc_1024:
+    times 8 dw 0x400
+align 16
+dc_00001111:
+    times 8 db 0
+    times 8 db 1

media/libvpx/vp8/common/x86/recon_wrapper_sse2.c

@@ -0,0 +1,96 @@
+/*
+ *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "vp8/common/recon.h"
+#include "recon_x86.h"
+#include "vpx_mem/vpx_mem.h"
+
+#define build_intra_predictors_mbuv_prototype(sym) \
+    void sym(unsigned char *dst, int dst_stride, \
+             const unsigned char *src, int src_stride)
+typedef build_intra_predictors_mbuv_prototype((*build_intra_predictors_mbuv_fn_t));
+
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dc_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dctop_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dcleft_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dc128_mmx);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ho_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ho_ssse3);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ve_mmx);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_tm_sse2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_tm_ssse3);
+
+static void vp8_build_intra_predictors_mbuv_x86(MACROBLOCKD *x,
+                                                unsigned char *dst_u,
+                                                unsigned char *dst_v,
+                                                int dst_stride,
+                                                build_intra_predictors_mbuv_fn_t tm_func,
+                                                build_intra_predictors_mbuv_fn_t ho_func)
+{
+    int mode = x->mode_info_context->mbmi.uv_mode;
+    build_intra_predictors_mbuv_fn_t fn;
+    int src_stride = x->dst.uv_stride;
+
+    switch (mode) {
+        case  V_PRED: fn = vp8_intra_pred_uv_ve_mmx; break;
+        case  H_PRED: fn = ho_func; break;
+        case TM_PRED: fn = tm_func; break;
+        case DC_PRED:
+            if (x->up_available) {
+                if (x->left_available) {
+                    fn = vp8_intra_pred_uv_dc_mmx2; break;
+                } else {
+                    fn = vp8_intra_pred_uv_dctop_mmx2; break;
+                }
+            } else if (x->left_available) {
+                fn = vp8_intra_pred_uv_dcleft_mmx2; break;
+            } else {
+                fn = vp8_intra_pred_uv_dc128_mmx; break;
+            }
+            break;
+        default: return;
+    }
+
+    fn(dst_u, dst_stride, x->dst.u_buffer, src_stride);
+    fn(dst_v, dst_stride, x->dst.v_buffer, src_stride);
+}
+
+void vp8_build_intra_predictors_mbuv_sse2(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, &x->predictor[256],
+                                        &x->predictor[320], 8,
+                                        vp8_intra_pred_uv_tm_sse2,
+                                        vp8_intra_pred_uv_ho_mmx2);
+}
+
+void vp8_build_intra_predictors_mbuv_ssse3(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, &x->predictor[256],
+                                        &x->predictor[320], 8,
+                                        vp8_intra_pred_uv_tm_ssse3,
+                                        vp8_intra_pred_uv_ho_ssse3);
+}
+
+void vp8_build_intra_predictors_mbuv_s_sse2(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, x->dst.u_buffer,
+                                        x->dst.v_buffer, x->dst.uv_stride,
+                                        vp8_intra_pred_uv_tm_sse2,
+                                        vp8_intra_pred_uv_ho_mmx2);
+}
+
+void vp8_build_intra_predictors_mbuv_s_ssse3(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, x->dst.u_buffer,
+                                        x->dst.v_buffer, x->dst.uv_stride,
+                                        vp8_intra_pred_uv_tm_ssse3,
+                                        vp8_intra_pred_uv_ho_ssse3);
+}

media/libvpx/vp8/common/x86/recon_x86.h

@@ -46,6 +46,8 @@ extern prototype_copy_block(vp8_copy_mem16x16_mmx);
 extern prototype_recon_block(vp8_recon2b_sse2);
 extern prototype_recon_block(vp8_recon4b_sse2);
 extern prototype_copy_block(vp8_copy_mem16x16_sse2);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_sse2);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_s_sse2);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_recon_recon2
@@ -57,6 +59,26 @@ extern prototype_copy_block(vp8_copy_mem16x16_sse2);
 #undef  vp8_recon_copy16x16
 #define vp8_recon_copy16x16 vp8_copy_mem16x16_sse2
 
+#undef  vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv_sse2
+
+#undef  vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s_sse2
+
+#endif
+#endif
+
+#if HAVE_SSSE3
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_ssse3);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_s_ssse3);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+#undef  vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv_ssse3
+
+#undef  vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s_ssse3
+
 #endif
 #endif
 #endif

media/libvpx/vp8/common/x86/subpixel_mmx.asm

@@ -113,97 +113,6 @@ nextrow:
     ret
 
 
-;
-; THIS FUNCTION APPEARS TO BE UNUSED
-;
-;void vp8_filter_block1d_v6_mmx
-;(
-;   short *src_ptr,
-;   unsigned char *output_ptr,
-;   unsigned int pixels_per_line,
-;   unsigned int pixel_step,
-;   unsigned int output_height,
-;   unsigned int output_width,
-;   short * vp8_filter
-;)
-global sym(vp8_filter_block1d_v6_mmx)
-sym(vp8_filter_block1d_v6_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        movq      mm5, [GLOBAL(rd)]
-        push        rbx
-        mov         rbx, arg(6) ;vp8_filter
-        movq      mm1, [rbx + 16]             ; do both the negative taps first!!!
-        movq      mm2, [rbx + 32]         ;
-        movq      mm6, [rbx + 48]        ;
-        movq      mm7, [rbx + 64]        ;
-
-        movsxd      rdx, dword ptr arg(2) ;pixels_per_line
-        mov         rdi, arg(1) ;output_ptr
-        mov         rsi, arg(0) ;src_ptr
-        sub         rsi, rdx
-        sub         rsi, rdx
-        movsxd      rcx, DWORD PTR arg(4) ;output_height
-        movsxd      rax, DWORD PTR arg(5) ;output_width      ; destination pitch?
-        pxor        mm0, mm0              ; mm0 = 00000000
-
-
-nextrow_v:
-        movq        mm3, [rsi+rdx]        ; mm3 = p0..p8  = row -1
-        pmullw      mm3, mm1              ; mm3 *= kernel 1 modifiers.
-
-
-        movq        mm4, [rsi + 4*rdx]      ; mm4 = p0..p3  = row 2
-        pmullw      mm4, mm7              ; mm4 *= kernel 4 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 2*rdx]           ; mm4 = p0..p3  = row 0
-        pmullw      mm4, mm2              ; mm4 *= kernel 2 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi]            ; mm4 = p0..p3  = row -2
-        pmullw      mm4, [rbx]            ; mm4 *= kernel 0 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        add         rsi, rdx              ; move source forward 1 line to avoid 3 * pitch
-        movq        mm4, [rsi + 2*rdx]     ; mm4 = p0..p3  = row 1
-        pmullw      mm4, mm6              ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 4*rdx]    ; mm4 = p0..p3  = row 3
-        pmullw      mm4, [rbx +80]        ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        paddsw      mm3, mm5               ; mm3 += round value
-        psraw       mm3, VP8_FILTER_SHIFT     ; mm3 /= 128
-        packuswb    mm3, mm0              ; pack and saturate
-
-        movd        [rdi],mm3             ; store the results in the destination
-
-        add         rdi,rax;
-
-        dec         rcx                   ; decrement count
-        jnz         nextrow_v             ; next row
-
-        pop         rbx
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
 ;void vp8_filter_block1dc_v6_mmx
 ;(
 ;   short *src_ptr,

media/libvpx/vp8/common/x86/subpixel_sse2.asm

@@ -37,7 +37,7 @@ sym(vp8_filter_block1d8_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -157,7 +157,7 @@ sym(vp8_filter_block1d16_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -333,7 +333,7 @@ sym(vp8_filter_block1d8_v6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -428,7 +428,7 @@ sym(vp8_filter_block1d16_v6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -538,7 +538,7 @@ sym(vp8_filter_block1d8_h6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -651,7 +651,7 @@ sym(vp8_filter_block1d16_h6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -816,7 +816,7 @@ sym(vp8_filter_block1d8_v6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -908,7 +908,6 @@ sym(vp8_unpack_block1d16_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    ;SAVE_XMM                          ;xmm6, xmm7 are not used here.
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -948,7 +947,6 @@ unpack_block1d16_h6_sse2_rowloop:
     pop rdi
     pop rsi
     RESTORE_GOT
-    ;RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -969,7 +967,7 @@ sym(vp8_bilinear_predict16x16_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1238,7 +1236,7 @@ sym(vp8_bilinear_predict8x8_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

media/libvpx/vp8/common/x86/subpixel_ssse3.asm

@@ -39,6 +39,7 @@ sym(vp8_filter_block1d8_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -107,6 +108,7 @@ filter_block1d8_h6_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -162,6 +164,7 @@ filter_block1d8_h4_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -179,7 +182,7 @@ sym(vp8_filter_block1d16_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -194,10 +197,6 @@ sym(vp8_filter_block1d16_h6_ssse3):
 
     mov         rdi, arg(2)                     ;output_ptr
 
-;;
-;;    cmp         esi, DWORD PTR [rax]
-;;    je          vp8_filter_block1d16_h4_ssse3
-
     mov         rsi, arg(0)                     ;src_ptr
 
     movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
@@ -271,61 +270,7 @@ filter_block1d16_h6_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-vp8_filter_block1d16_h4_ssse3:
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-filter_block1d16_h4_rowloop_ssse3:
-    movdqu      xmm1,   XMMWORD PTR [rsi - 2]
-
-    movdqa      xmm2, xmm1
-    pshufb      xmm1, [GLOBAL(shuf2b)]
-    pshufb      xmm2, [GLOBAL(shuf3b)]
-    pmaddubsw   xmm1, xmm5
-
-    movdqu      xmm3,   XMMWORD PTR [rsi + 6]
-
-    pmaddubsw   xmm2, xmm6
-    movdqa      xmm0, xmm3
-    pshufb      xmm3, [GLOBAL(shuf3b)]
-    pshufb      xmm0, [GLOBAL(shuf2b)]
-
-    paddsw      xmm1, [GLOBAL(rd)]
-    paddsw      xmm1, xmm2
-
-    pmaddubsw   xmm0, xmm5
-    pmaddubsw   xmm3, xmm6
-
-    psraw       xmm1, 7
-    packuswb    xmm1, xmm1
-    lea         rsi,    [rsi + rax]
-    paddsw      xmm3, xmm0
-    paddsw      xmm3, [GLOBAL(rd)]
-    psraw       xmm3, 7
-    packuswb    xmm3, xmm3
-
-    punpcklqdq  xmm1, xmm3
-
-    movdqa      XMMWORD Ptr [rdi], xmm1
-
-    add         rdi, rdx
-    dec         rcx
-    jnz         filter_block1d16_h4_rowloop_ssse3
-
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -344,6 +289,7 @@ sym(vp8_filter_block1d4_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -451,6 +397,7 @@ filter_block1d4_h4_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -471,6 +418,7 @@ sym(vp8_filter_block1d16_v6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -566,6 +514,7 @@ vp8_filter_block1d16_v6_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -638,6 +587,7 @@ vp8_filter_block1d16_v4_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -656,6 +606,7 @@ sym(vp8_filter_block1d8_v6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -728,6 +679,7 @@ vp8_filter_block1d8_v6_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -776,6 +728,7 @@ vp8_filter_block1d8_v4_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -932,7 +885,7 @@ sym(vp8_bilinear_predict16x16_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1195,7 +1148,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

media/libvpx/vp8/common/x86/vp8_asm_stubs.c

@@ -11,7 +11,7 @@
 
 #include "vpx_ports/config.h"
 #include "vpx_ports/mem.h"
-#include "subpixel.h"
+#include "vp8/common/subpixel.h"
 
 extern const short vp8_six_tap_mmx[8][6*8];
 extern const short vp8_bilinear_filters_mmx[8][2*8];

media/libvpx/vp8/common/x86/x86_systemdependent.c

@@ -9,25 +9,21 @@
  */
 
 
-#include "vpx_ports/config.h"
+#include "vpx_config.h"
 #include "vpx_ports/x86.h"
-#include "g_common.h"
-#include "subpixel.h"
-#include "loopfilter.h"
-#include "recon.h"
-#include "idct.h"
-#include "pragmas.h"
-#include "onyxc_int.h"
+#include "vp8/common/g_common.h"
+#include "vp8/common/subpixel.h"
+#include "vp8/common/loopfilter.h"
+#include "vp8/common/recon.h"
+#include "vp8/common/idct.h"
+#include "vp8/common/pragmas.h"
+#include "vp8/common/onyxc_int.h"
 
 void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
     int flags = x86_simd_caps();
-    int mmx_enabled = flags & HAS_MMX;
-    int xmm_enabled = flags & HAS_SSE;
-    int wmt_enabled = flags & HAS_SSE2;
-    int SSSE3Enabled = flags & HAS_SSSE3;
 
     /* Note:
      *
@@ -39,7 +35,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
     /* Override default functions with fastest ones for this CPU. */
 #if HAVE_MMX
 
-    if (mmx_enabled)
+    if (flags & HAS_MMX)
     {
         rtcd->idct.idct1        = vp8_short_idct4x4llm_1_mmx;
         rtcd->idct.idct16       = vp8_short_idct4x4llm_mmx;
@@ -67,9 +63,9 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_mmx;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_mmx;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_mmx;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_mmx;
+        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_mmx;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_mmx;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_mmx;
+        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_mmx;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_mmx;
 
 #if CONFIG_POSTPROC
@@ -83,11 +79,15 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 #endif
 #if HAVE_SSE2
 
-    if (wmt_enabled)
+    if (flags & HAS_SSE2)
     {
         rtcd->recon.recon2      = vp8_recon2b_sse2;
         rtcd->recon.recon4      = vp8_recon4b_sse2;
         rtcd->recon.copy16x16   = vp8_copy_mem16x16_sse2;
+        rtcd->recon.build_intra_predictors_mbuv =
+            vp8_build_intra_predictors_mbuv_sse2;
+        rtcd->recon.build_intra_predictors_mbuv_s =
+            vp8_build_intra_predictors_mbuv_s_sse2;
 
         rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_sse2;
 
@@ -101,9 +101,9 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_sse2;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_sse2;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_sse2;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_sse2;
+        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_sse2;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_sse2;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_sse2;
+        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_sse2;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_sse2;
 
 #if CONFIG_POSTPROC
@@ -118,7 +118,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 
 #if HAVE_SSSE3
 
-    if (SSSE3Enabled)
+    if (flags & HAS_SSSE3)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_ssse3;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_ssse3;
@@ -126,6 +126,11 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->subpix.sixtap4x4     = vp8_sixtap_predict4x4_ssse3;
         rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_ssse3;
         rtcd->subpix.bilinear8x8   = vp8_bilinear_predict8x8_ssse3;
+
+        rtcd->recon.build_intra_predictors_mbuv =
+            vp8_build_intra_predictors_mbuv_ssse3;
+        rtcd->recon.build_intra_predictors_mbuv_s =
+            vp8_build_intra_predictors_mbuv_s_ssse3;
     }
 #endif
 

media/libvpx/vp8/decoder/arm/arm_dsystemdependent.c

@@ -11,23 +11,24 @@
 
 #include "vpx_ports/config.h"
 #include "vpx_ports/arm.h"
-#include "blockd.h"
-#include "pragmas.h"
-#include "postproc.h"
-#include "dboolhuff.h"
-#include "dequantize.h"
-#include "onyxd_int.h"
+#include "vp8/common/blockd.h"
+#include "vp8/common/pragmas.h"
+#include "vp8/decoder/dequantize.h"
+#include "vp8/decoder/onyxd_int.h"
 
 void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
     int flags = pbi->common.rtcd.flags;
-    int has_edsp = flags & HAS_EDSP;
-    int has_media = flags & HAS_MEDIA;
-    int has_neon = flags & HAS_NEON;
+
+#if HAVE_ARMV5TE
+    if (flags & HAS_EDSP)
+    {
+    }
+#endif
 
 #if HAVE_ARMV6
-    if (has_media)
+    if (flags & HAS_MEDIA)
     {
         pbi->dequant.block               = vp8_dequantize_b_v6;
         pbi->dequant.idct_add            = vp8_dequant_idct_add_v6;
@@ -35,17 +36,11 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
         pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_v6;
         pbi->dequant.idct_add_y_block    = vp8_dequant_idct_add_y_block_v6;
         pbi->dequant.idct_add_uv_block   = vp8_dequant_idct_add_uv_block_v6;
-#if 0 /*For use with RTCD, when implemented*/
-        pbi->dboolhuff.start             = vp8dx_start_decode_c;
-        pbi->dboolhuff.fill              = vp8dx_bool_decoder_fill_c;
-        pbi->dboolhuff.debool            = vp8dx_decode_bool_c;
-        pbi->dboolhuff.devalue           = vp8dx_decode_value_c;
-#endif
     }
 #endif
 
 #if HAVE_ARMV7
-    if (has_neon)
+    if (flags & HAS_NEON)
     {
         pbi->dequant.block               = vp8_dequantize_b_neon;
         pbi->dequant.idct_add            = vp8_dequant_idct_add_neon;
@@ -54,12 +49,6 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
         pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_neon;
         pbi->dequant.idct_add_y_block    = vp8_dequant_idct_add_y_block_neon;
         pbi->dequant.idct_add_uv_block   = vp8_dequant_idct_add_uv_block_neon;
-#if 0 /*For use with RTCD, when implemented*/
-        pbi->dboolhuff.start             = vp8dx_start_decode_c;
-        pbi->dboolhuff.fill              = vp8dx_bool_decoder_fill_c;
-        pbi->dboolhuff.debool            = vp8dx_decode_bool_c;
-        pbi->dboolhuff.devalue           = vp8dx_decode_value_c;
-#endif
     }
 #endif
 #endif

media/libvpx/vp8/decoder/arm/armv6/idct_blk_v6.c

@@ -9,8 +9,8 @@
  */
 
 #include "vpx_ports/config.h"
-#include "idct.h"
-#include "dequantize.h"
+#include "vp8/common/idct.h"
+#include "vp8/decoder/dequantize.h"
 
 void vp8_dequant_dc_idct_add_y_block_v6
             (short *q, short *dq, unsigned char *pre,

media/libvpx/vp8/decoder/arm/dboolhuff_arm.h

@@ -1,43 +0,0 @@
-#ifndef DBOOLHUFF_ARM_H
-#define DBOOLHUFF_ARM_H
-
-/* JLK
- * There are currently no arm-optimized versions of
- * these functions. As they are implemented, they
- * can be uncommented below and added to
- * arm/dsystemdependent.c
- *
- * The existing asm code is likely so different as
- * to be useless. However, its been left (for now)
- * for reference.
- */
-#if 0
-#if HAVE_ARMV6
-#undef vp8_dbool_start
-#define vp8_dbool_start vp8dx_start_decode_v6
-
-#undef vp8_dbool_fill
-#define vp8_dbool_fill vp8_bool_decoder_fill_v6
-
-#undef vp8_dbool_debool
-#define vp8_dbool_debool vp8_decode_bool_v6
-
-#undef vp8_dbool_devalue
-#define vp8_dbool_devalue vp8_decode_value_v6
-#endif /* HAVE_ARMV6 */
-
-#if HAVE_ARMV7
-#undef vp8_dbool_start
-#define vp8_dbool_start vp8dx_start_decode_neon
-
-#undef vp8_dbool_fill
-#define vp8_dbool_fill vp8_bool_decoder_fill_neon
-
-#undef vp8_dbool_debool
-#define vp8_dbool_debool vp8_decode_bool_neon
-
-#undef vp8_dbool_devalue
-#define vp8_dbool_devalue vp8_decode_value_neon
-#endif /* HAVE_ARMV7 */
-#endif
-#endif /* DBOOLHUFF_ARM_H */

media/libvpx/vp8/decoder/arm/dequantize_arm.c

@@ -10,9 +10,8 @@
 
 
 #include "vpx_ports/config.h"
-#include "dequantize.h"
-#include "predictdc.h"
-#include "idct.h"
+#include "vp8/decoder/dequantize.h"
+#include "vp8/common/idct.h"
 #include "vpx_mem/vpx_mem.h"
 
 #if HAVE_ARMV7
@@ -27,7 +26,6 @@ extern void vp8_dequantize_b_loop_v6(short *Q, short *DQC, short *DQ);
 
 void vp8_dequantize_b_neon(BLOCKD *d)
 {
-    int i;
     short *DQ  = d->dqcoeff;
     short *Q   = d->qcoeff;
     short *DQC = d->dequant;
@@ -39,7 +37,6 @@ void vp8_dequantize_b_neon(BLOCKD *d)
 #if HAVE_ARMV6
 void vp8_dequantize_b_v6(BLOCKD *d)
 {
-    int i;
     short *DQ  = d->dqcoeff;
     short *Q   = d->qcoeff;
     short *DQC = d->dequant;

media/libvpx/vp8/decoder/arm/detokenize.asm

@@ -1,320 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_decode_mb_tokens_v6|
-
-    AREA    |.text|, CODE, READONLY  ; name this block of code
-
-    INCLUDE vpx_asm_offsets.asm
-
-l_qcoeff    EQU     0
-l_i         EQU     4
-l_type      EQU     8
-l_stop      EQU     12
-l_c         EQU     16
-l_l_ptr     EQU     20
-l_a_ptr     EQU     24
-l_bc        EQU     28
-l_coef_ptr  EQU     32
-l_stacksize EQU     64
-
-
-;; constant offsets -- these should be created at build time
-c_block2above_offset         EQU 25
-c_entropy_nodes              EQU 11
-c_dct_eob_token              EQU 11
-
-|vp8_decode_mb_tokens_v6| PROC
-    stmdb       sp!, {r4 - r11, lr}
-    sub         sp, sp, #l_stacksize
-    mov         r7, r1                      ; type
-    mov         r9, r0                      ; detoken
-
-    ldr         r1, [r9, #detok_current_bc]
-    ldr         r0, [r9, #detok_qcoeff_start_ptr]
-    mov         r11, #0                     ; i
-    mov         r3, #16                     ; stop
-
-    cmp         r7, #1                      ; type ?= 1
-    addeq       r11, r11, #24               ; i = 24
-    addeq       r3, r3, #8                  ; stop = 24
-    addeq       r0, r0, #3, 24              ; qcoefptr += 24*16
-
-    str         r0, [sp, #l_qcoeff]
-    str         r11, [sp, #l_i]
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-    str         r1, [sp, #l_bc]
-
-    add         lr, r9, r7, lsl #2          ; detoken + type*4
-
-    ldr         r8, [r1, #bool_decoder_user_buffer]
-
-    ldr         r10, [lr, #detok_coef_probs]
-    ldr         r5, [r1, #bool_decoder_count]
-    ldr         r6, [r1, #bool_decoder_range]
-    ldr         r4, [r1, #bool_decoder_value]
-
-    str         r10, [sp, #l_coef_ptr]
-
-BLOCK_LOOP
-    ldr         r3, [r9, #detok_ptr_block2leftabove]
-    ldr         r1, [r9, #detok_L]
-    ldr         r2, [r9, #detok_A]
-    ldrb        r12, [r3, r11]!             ; block2left[i]
-    ldrb        r3, [r3, #c_block2above_offset]; block2above[i]
-
-    cmp         r7, #0                      ; c = !type
-    moveq       r7, #1
-    movne       r7, #0
-
-    ldrb        r0, [r1, r12]!              ; *(L += block2left[i])
-    ldrb        r3, [r2, r3]!               ; *(A += block2above[i])
-    mov         lr, #c_entropy_nodes        ; ENTROPY_NODES = 11
-
-; VP8_COMBINEENTROPYCONTETEXTS(t, *a, *l) => t = ((*a) != 0) + ((*l) !=0)
-    cmp         r0, #0                      ; *l ?= 0
-    movne       r0, #1
-    cmp         r3, #0                      ; *a ?= 0
-    addne       r0, r0, #1                  ; t
-
-    str         r1, [sp, #l_l_ptr]          ; save &l
-    str         r2, [sp, #l_a_ptr]          ; save &a
-    smlabb      r0, r0, lr, r10             ; Prob = coef_probs + (t * ENTROPY_NODES)
-    mov         r1, #0                      ; t = 0
-    str         r7, [sp, #l_c]
-
-    ;align 4
-COEFF_LOOP
-    ldr         r3, [r9, #detok_ptr_coef_bands_x]
-    ldr         lr, [r9, #detok_coef_tree_ptr]
-    ;STALL
-    ldrb        r3, [r3, r7]                ; coef_bands_x[c]
-    ;STALL
-    ;STALL
-    add         r0, r0, r3                  ; Prob += coef_bands_x[c]
-
-get_token_loop
-    ldrb        r2, [r0, +r1, asr #1]       ; Prob[t >> 1]
-    mov         r3, r6, lsl #8              ; range << 8
-    sub         r3, r3, #256                ; (range << 8) - (1 << 8)
-    mov         r10, #1                     ; 1
-
-    smlawb      r2, r3, r2, r10             ; split = 1 + (((range-1) * probability) >> 8)
-
-    ldrb        r12, [r8]                   ; load cx data byte in stall slot : r8 = bufptr
-    ;++
-
-    subs        r3, r4, r2, lsl #24         ; value-(split<<24): used later to calculate shift for NORMALIZE
-    addhs       r1, r1, #1                  ; t += 1
-    movhs       r4, r3                      ; value -= bigsplit (split << 24)
-    subhs       r2, r6, r2                  ; range -= split
- ;   movlo       r6, r2                      ; range = split
-
-    ldrsb     r1, [lr, r1]                  ; t = onyx_coef_tree_ptr[t]
-
-; NORMALIZE
-    clz         r3, r2                      ; vp8dx_bitreader_norm[range] + 24
-    sub         r3, r3, #24                 ; vp8dx_bitreader_norm[range]
-    subs        r5, r5, r3                  ; count -= shift
-    mov         r6, r2, lsl r3              ; range <<= shift
-    mov         r4, r4, lsl r3              ; value <<= shift
-
-; if count <= 0, += BR_COUNT; value |= *bufptr++ << (BR_COUNT-count); BR_COUNT = 8, but need to upshift values by +16
-    addle         r5, r5, #8                ; count += 8
-    rsble         r3, r5, #24               ; 24 - count
-    addle         r8, r8, #1                ; bufptr++
-    orrle         r4, r4, r12, lsl r3       ; value |= *bufptr << shift + 16
-
-    cmp         r1, #0                      ; t ?= 0
-    bgt         get_token_loop              ; while (t > 0)
-
-    cmn         r1, #c_dct_eob_token        ; if(t == -DCT_EOB_TOKEN)
-    beq         END_OF_BLOCK                ; break
-
-    rsb         lr, r1, #0                  ; v = -t;
-
-    cmp         lr, #4                      ; if(v > FOUR_TOKEN)
-    ble         SKIP_EXTRABITS
-
-    ldr         r3, [r9, #detok_teb_base_ptr]
-    mov         r11, #1                     ; 1 in split = 1 + ... nope, v+= 1 << bits_count
-    add         r7, r3, lr, lsl #4          ; detok_teb_base_ptr + (v << 4)
-
-    ldrsh       lr, [r7, #tokenextrabits_min_val] ; v = teb_ptr->min_val
-    ldrsh       r0, [r7, #tokenextrabits_length] ; bits_count = teb_ptr->Length
-
-extrabits_loop
-    add         r3, r0, r7                  ; &teb_ptr->Probs[bits_count]
-
-    ldrb        r2, [r3, #4]                ; probability. why +4?
-    mov         r3, r6, lsl #8              ; range << 8
-    sub         r3, r3, #256                ; range << 8 + 1 << 8
-
-    smlawb      r2, r3, r2, r11             ; split = 1 +  (((range-1) * probability) >> 8)
-
-    ldrb        r12, [r8]                   ; *bufptr
-    ;++
-
-    subs        r10, r4, r2, lsl #24        ; value - (split<<24)
-    movhs       r4, r10                     ; value = value - (split << 24)
-    subhs       r2, r6, r2                  ; range = range - split
-    addhs       lr, lr, r11, lsl r0         ; v += ((UINT16)1<<bits_count)
-
-; NORMALIZE
-    clz         r3, r2                      ; shift - leading zeros in split
-    sub         r3, r3, #24                 ; don't count first 3 bytes
-    subs        r5, r5, r3                  ; count -= shift
-    mov         r6, r2, lsl r3              ; range = range << shift
-    mov         r4, r4, lsl r3              ; value <<= shift
-
-    addle       r5, r5, #8                  ; count += BR_COUNT
-    addle       r8, r8, #1                  ; bufptr++
-    rsble       r3, r5, #24                 ; BR_COUNT - count
-    orrle       r4, r4, r12, lsl r3         ; value |= *bufptr << (BR_COUNT - count)
-
-    subs        r0, r0, #1                  ; bits_count --
-    bpl         extrabits_loop
-
-
-SKIP_EXTRABITS
-    ldr         r11, [sp, #l_qcoeff]
-    ldr         r0, [sp, #l_coef_ptr]       ; Prob = coef_probs
-
-    cmp         r1, #0                      ; check for nonzero token - if (t)
-    beq         SKIP_EOB_CHECK              ; if t is zero, we will skip the eob table chec
-
-    add         r3, r6, #1                  ; range + 1
-    mov         r2, r3, lsr #1              ; split = (range + 1) >> 1
-
-    subs        r3, r4, r2, lsl #24         ; value - (split<<24)
-    movhs       r4, r3                      ; value -= (split << 24)
-    subhs       r2, r6, r2                  ; range -= split
-    mvnhs       r3, lr                      ; -v
-    addhs       lr, r3, #1                  ; v = (v ^ -1) + 1
-
-; NORMALIZE
-    clz         r3, r2                      ; leading 0s in split
-    sub         r3, r3, #24                 ; shift
-    subs        r5, r5, r3                  ; count -= shift
-    mov         r6, r2, lsl r3              ; range <<= shift
-    mov         r4, r4, lsl r3              ; value <<= shift
-    ldrleb      r2, [r8], #1                ; *(bufptr++)
-    addle       r5, r5, #8                  ; count += 8
-    rsble       r3, r5, #24                 ; BR_COUNT - count
-    orrle       r4, r4, r2, lsl r3          ; value |= *bufptr << (BR_COUNT - count)
-
-    add         r0, r0, #11                 ; Prob += ENTROPY_NODES (11)
-
-    cmn         r1, #1                      ; t < -ONE_TOKEN
-
-    addlt       r0, r0, #11                 ; Prob += ENTROPY_NODES (11)
-
-    mvn         r1, #1                      ; t = -1 ???? C is -2
-
-SKIP_EOB_CHECK
-    ldr         r7, [sp, #l_c]              ; c
-    ldr         r3, [r9, #detok_scan]
-    add         r1, r1, #2                  ; t+= 2
-    cmp         r7, #15                     ; c should will be one higher
-
-    ldr         r3, [r3, +r7, lsl #2]       ; scan[c] this needs pre-inc c value
-    add         r7, r7, #1                  ; c++
-    add         r3, r11, r3, lsl #1         ; qcoeff + scan[c]
-
-    str         r7, [sp, #l_c]              ; store c
-    strh        lr, [r3]                    ; qcoef_ptr[scan[c]] = v
-
-    blt         COEFF_LOOP
-
-    sub         r7, r7, #1                  ; if(t != -DCT_EOB_TOKEN) --c
-
-END_OF_BLOCK
-    ldr         r3, [sp, #l_type]           ; type
-    ldr         r10, [sp, #l_coef_ptr]      ; coef_ptr
-    ldr         r0, [sp, #l_qcoeff]         ; qcoeff
-    ldr         r11, [sp, #l_i]             ; i
-    ldr         r12, [sp, #l_stop]          ; stop
-
-    cmp         r3, #0                      ; type ?= 0
-    moveq       r1, #1
-    movne       r1, #0
-    add         r3, r11, r9                 ; detok + i
-
-    cmp         r7, r1                      ; c ?= !type
-    strb        r7, [r3, #detok_eob]        ; eob[i] = c
-
-    ldr         r7, [sp, #l_l_ptr]          ; l
-    ldr         r2, [sp, #l_a_ptr]          ; a
-    movne       r3, #1                      ; t
-    moveq       r3, #0
-
-    add         r0, r0, #32                 ; qcoeff += 32 (16 * 2?)
-    add         r11, r11, #1                ; i++
-    strb        r3, [r7]                    ; *l = t
-    strb        r3, [r2]                    ; *a = t
-    str         r0, [sp, #l_qcoeff]         ; qcoeff
-    str         r11, [sp, #l_i]             ; i
-
-    cmp         r11, r12                    ; i < stop
-    ldr         r7, [sp, #l_type]           ; type
-
-    blt         BLOCK_LOOP
-
-    cmp         r11, #25                    ; i ?= 25
-    bne         ln2_decode_mb_to
-
-    ldr         r12, [r9, #detok_qcoeff_start_ptr]
-    ldr         r10, [r9, #detok_coef_probs]
-    mov         r7, #0                      ; type/i = 0
-    mov         r3, #16                     ; stop = 16
-    str         r12, [sp, #l_qcoeff]        ; qcoeff_ptr = qcoeff_start_ptr
-    str         r7, [sp, #l_i]
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-
-    str         r10, [sp, #l_coef_ptr]      ; coef_probs = coef_probs[type=0]
-
-    b           BLOCK_LOOP
-
-ln2_decode_mb_to
-    cmp         r11, #16                    ; i ?= 16
-    bne         ln1_decode_mb_to
-
-    mov         r10, #detok_coef_probs
-    add         r10, r10, #2*4              ; coef_probs[type]
-    ldr         r10, [r9, r10]              ; detok + detok_coef_probs[type]
-
-    mov         r7, #2                      ; type = 2
-    mov         r3, #24                     ; stop = 24
-
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-
-    str         r10, [sp, #l_coef_ptr]      ; coef_probs = coef_probs[type]
-    b           BLOCK_LOOP
-
-ln1_decode_mb_to
-    ldr         r2, [sp, #l_bc]
-    mov         r0, #0
-    nop
-
-    str         r8, [r2, #bool_decoder_user_buffer]
-    str         r5, [r2, #bool_decoder_count]
-    str         r4, [r2, #bool_decoder_value]
-    str         r6, [r2, #bool_decoder_range]
-
-    add         sp, sp, #l_stacksize
-    ldmia       sp!, {r4 - r11, pc}
-
-    ENDP  ; |vp8_decode_mb_tokens_v6|
-
-    END

media/libvpx/vp8/decoder/arm/neon/dequant_idct_neon.asm

@@ -35,7 +35,7 @@
 
     ldr             r1, [sp, #4]            ; stride
 
-    adr             r12, _CONSTANTS_
+    adr             r12, cospi8sqrt2minus1  ; pointer to the first constant
 
     vmul.i16        q1, q3, q5              ;input for short_idct4x4llm_neon
     vmul.i16        q2, q4, q6
@@ -123,7 +123,6 @@
     ENDP           ; |vp8_dequant_idct_add_neon|
 
 ; Constant Pool
-_CONSTANTS_       EQU cospi8sqrt2minus1
 cospi8sqrt2minus1 DCD 0x4e7b4e7b
 sinpi8sqrt2       DCD 0x8a8c8a8c
 

media/libvpx/vp8/decoder/arm/neon/idct_blk_neon.c

@@ -9,8 +9,8 @@
  */
 
 #include "vpx_ports/config.h"
-#include "idct.h"
-#include "dequantize.h"
+#include "vp8/common/idct.h"
+#include "vp8/decoder/dequantize.h"
 
 /* place these declarations here because we don't want to maintain them
  * outside of this scope

media/libvpx/vp8/decoder/arm/neon/idct_dequant_dc_full_2x_neon.asm

@@ -41,7 +41,7 @@
     ldr             r1, [sp, #4]
     vld1.32         {d31[1]}, [r12]
 
-    adr             r2, _CONSTANTS_
+    adr             r2, cospi8sqrt2minus1   ; pointer to the first constant
 
     ldrh            r12, [r1], #2           ; lo *dc
     ldrh            r1, [r1]                ; hi *dc
@@ -198,7 +198,6 @@
     ENDP           ; |idct_dequant_dc_full_2x_neon|
 
 ; Constant Pool
-_CONSTANTS_       EQU cospi8sqrt2minus1
 cospi8sqrt2minus1 DCD 0x4e7b
 ; because the lowest bit in 0x8a8c is 0, we can pre-shift this
 sinpi8sqrt2       DCD 0x4546