зеркало из https://github.com/mozilla/pjs.git
684 строки
27 KiB
Diff
684 строки
27 KiB
Diff
diff --git a/media/libvpx/vp8/common/arm/armv6/sixtappredict8x4_v6.asm b/media/libvpx/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
|
|
--- a/media/libvpx/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
|
|
+++ b/media/libvpx/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
|
|
@@ -27,17 +27,17 @@
|
|
stmdb sp!, {r4 - r11, lr}
|
|
str r3, [sp, #-184]! ;reserve space on stack for temporary storage, store yoffset
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
add lr, sp, #4 ;point to temporary buffer
|
|
beq skip_firstpass_filter
|
|
|
|
;first-pass filter
|
|
- ldr r12, _filter8_coeff_
|
|
+ adr r12, filter8_coeff
|
|
sub r0, r0, r1, lsl #1
|
|
|
|
add r2, r12, r2, lsl #4 ;calculate filter location
|
|
add r0, r0, #3 ;adjust src only for loading convinience
|
|
|
|
ldr r3, [r2] ; load up packed filter coefficients
|
|
ldr r4, [r2, #4]
|
|
ldr r5, [r2, #8]
|
|
@@ -116,17 +116,17 @@
|
|
secondpass_filter
|
|
ldr r3, [sp], #4 ; load back yoffset
|
|
ldr r0, [sp, #216] ; load dst address from stack 180+36
|
|
ldr r1, [sp, #220] ; load dst stride from stack 180+40
|
|
|
|
cmp r3, #0
|
|
beq skip_secondpass_filter
|
|
|
|
- ldr r12, _filter8_coeff_
|
|
+ adr r12, filter8_coeff
|
|
add lr, r12, r3, lsl #4 ;calculate filter location
|
|
|
|
mov r2, #0x00080000
|
|
|
|
ldr r3, [lr] ; load up packed filter coefficients
|
|
ldr r4, [lr, #4]
|
|
ldr r5, [lr, #8]
|
|
|
|
@@ -242,18 +242,16 @@ 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_
|
|
- DCD filter8_coeff
|
|
filter8_coeff
|
|
DCD 0x00000000, 0x00000080, 0x00000000, 0x00000000
|
|
DCD 0xfffa0000, 0x000c007b, 0x0000ffff, 0x00000000
|
|
DCD 0xfff50002, 0x0024006c, 0x0001fff8, 0x00000000
|
|
DCD 0xfff70000, 0x0032005d, 0x0000fffa, 0x00000000
|
|
DCD 0xfff00003, 0x004d004d, 0x0003fff0, 0x00000000
|
|
DCD 0xfffa0000, 0x005d0032, 0x0000fff7, 0x00000000
|
|
DCD 0xfff80001, 0x006c0024, 0x0002fff5, 0x00000000
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/bilinearpredict16x16_neon.asm b/media/libvpx/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; r4 unsigned char *dst_ptr,
|
|
; stack(r5) int dst_pitch
|
|
|
|
|vp8_bilinear_predict16x16_neon| PROC
|
|
push {r4-r5, lr}
|
|
|
|
- ldr r12, _bifilter16_coeff_
|
|
+ adr r12, bifilter16_coeff
|
|
ldr r4, [sp, #12] ;load parameters from stack
|
|
ldr r5, [sp, #16] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq secondpass_bfilter16x16_only
|
|
|
|
add r2, r12, r2, lsl #3 ;calculate filter location
|
|
|
|
@@ -349,14 +349,12 @@ 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 bifilter16_coeff
|
|
bifilter16_coeff
|
|
DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm b/media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; r4 unsigned char *dst_ptr,
|
|
; stack(lr) int dst_pitch
|
|
|
|
|vp8_bilinear_predict4x4_neon| PROC
|
|
push {r4, lr}
|
|
|
|
- ldr r12, _bifilter4_coeff_
|
|
+ adr r12, bifilter4_coeff
|
|
ldr r4, [sp, #8] ;load parameters from stack
|
|
ldr lr, [sp, #12] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq skip_firstpass_filter
|
|
|
|
;First pass: output_height lines x output_width columns (5x4)
|
|
vld1.u8 {d2}, [r0], r1 ;load src data
|
|
@@ -122,14 +122,12 @@ 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 bifilter4_coeff
|
|
bifilter4_coeff
|
|
DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/bilinearpredict8x4_neon.asm b/media/libvpx/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; r4 unsigned char *dst_ptr,
|
|
; stack(lr) int dst_pitch
|
|
|
|
|vp8_bilinear_predict8x4_neon| PROC
|
|
push {r4, lr}
|
|
|
|
- ldr r12, _bifilter8x4_coeff_
|
|
+ adr r12, bifilter8x4_coeff
|
|
ldr r4, [sp, #8] ;load parameters from stack
|
|
ldr lr, [sp, #12] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq skip_firstpass_filter
|
|
|
|
;First pass: output_height lines x output_width columns (5x8)
|
|
add r2, r12, r2, lsl #3 ;calculate filter location
|
|
@@ -127,14 +127,12 @@ 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 bifilter8x4_coeff
|
|
bifilter8x4_coeff
|
|
DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/bilinearpredict8x8_neon.asm b/media/libvpx/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; r4 unsigned char *dst_ptr,
|
|
; stack(lr) int dst_pitch
|
|
|
|
|vp8_bilinear_predict8x8_neon| PROC
|
|
push {r4, lr}
|
|
|
|
- ldr r12, _bifilter8_coeff_
|
|
+ adr r12, bifilter8_coeff
|
|
ldr r4, [sp, #8] ;load parameters from stack
|
|
ldr lr, [sp, #12] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq skip_firstpass_filter
|
|
|
|
;First pass: output_height lines x output_width columns (9x8)
|
|
add r2, r12, r2, lsl #3 ;calculate filter location
|
|
@@ -175,14 +175,12 @@ 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 bifilter8_coeff
|
|
bifilter8_coeff
|
|
DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/loopfilter_neon.asm b/media/libvpx/vp8/common/arm/neon/loopfilter_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/loopfilter_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/loopfilter_neon.asm
|
|
@@ -303,17 +303,17 @@
|
|
; q4 p2
|
|
; q5 p1
|
|
; q6 p0
|
|
; q7 q0
|
|
; q8 q1
|
|
; q9 q2
|
|
; q10 q3
|
|
|vp8_loop_filter_neon| PROC
|
|
- ldr r12, _lf_coeff_
|
|
+ adr r12, lf_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 q4, q10, q9 ; abs(q3 - q2)
|
|
@@ -393,17 +393,15 @@
|
|
veor q6, q11, q0 ; *op0 = u^0x80
|
|
veor q7, q10, q0 ; *oq0 = 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 lf_coeff
|
|
lf_coeff
|
|
DCD 0x80808080, 0x80808080, 0x80808080, 0x80808080
|
|
DCD 0x03030303, 0x03030303, 0x03030303, 0x03030303
|
|
DCD 0x04040404, 0x04040404, 0x04040404, 0x04040404
|
|
DCD 0x01010101, 0x01010101, 0x01010101, 0x01010101
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm b/media/libvpx/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
|
|
@@ -23,17 +23,17 @@
|
|
; r2 const signed char *flimit,
|
|
; r3 const signed char *limit,
|
|
; stack(r4) const signed char *thresh,
|
|
; //stack(r5) int count --unused
|
|
|
|
|vp8_loop_filter_simple_horizontal_edge_neon| PROC
|
|
sub r0, r0, r1, lsl #1 ; move src pointer down by 2 lines
|
|
|
|
- ldr r12, _lfhy_coeff_
|
|
+ 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
|
|
@@ -103,16 +103,14 @@
|
|
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 lfhy_coeff
|
|
lfhy_coeff
|
|
DCD 0x80808080, 0x80808080, 0x80808080, 0x80808080
|
|
DCD 0x03030303, 0x03030303, 0x03030303, 0x03030303
|
|
DCD 0x04040404, 0x04040404, 0x04040404, 0x04040404
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm b/media/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
|
|
@@ -27,17 +27,17 @@
|
|
|
|
|vp8_loop_filter_simple_vertical_edge_neon| PROC
|
|
sub r0, r0, #2 ; move src pointer down by 2 columns
|
|
|
|
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
|
|
- ldr r12, _vlfy_coeff_
|
|
+ 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 {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
|
|
@@ -144,16 +144,14 @@
|
|
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 vlfy_coeff
|
|
vlfy_coeff
|
|
DCD 0x80808080, 0x80808080, 0x80808080, 0x80808080
|
|
DCD 0x03030303, 0x03030303, 0x03030303, 0x03030303
|
|
DCD 0x04040404, 0x04040404, 0x04040404, 0x04040404
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/mbloopfilter_neon.asm b/media/libvpx/vp8/common/arm/neon/mbloopfilter_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/mbloopfilter_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/mbloopfilter_neon.asm
|
|
@@ -367,17 +367,17 @@
|
|
; q5 p1
|
|
; q6 p0
|
|
; q7 q0
|
|
; q8 q1
|
|
; q9 q2
|
|
; q10 q3
|
|
|
|
|vp8_mbloop_filter_neon| PROC
|
|
- ldr r12, _mblf_coeff_
|
|
+ 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 q0, q10, q9 ; abs(q3 - q2)
|
|
@@ -501,18 +501,16 @@
|
|
veor q5, q12, q0 ; *op2 = s^0x80
|
|
veor q7, q15, q0 ; *oq0 = s^0x80
|
|
veor q6, q14, q0 ; *op0 = s^0x80
|
|
|
|
bx lr
|
|
ENDP ; |vp8_mbloop_filter_neon|
|
|
|
|
AREA mbloopfilter_dat, DATA, READONLY
|
|
-_mblf_coeff_
|
|
- DCD mblf_coeff
|
|
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
|
|
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/shortidct4x4llm_neon.asm b/media/libvpx/vp8/common/arm/neon/shortidct4x4llm_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/shortidct4x4llm_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/shortidct4x4llm_neon.asm
|
|
@@ -26,17 +26,17 @@
|
|
;static const int sinpi8sqrt2 =35468;
|
|
;static const int rounding = 0;
|
|
;Optimization note: The resulted data from dequantization are signed 13-bit data that is
|
|
;in the range of [-4096, 4095]. This allows to use "vqdmulh"(neon) instruction since
|
|
;it won't go out of range (13+16+1=30bits<32bits). This instruction gives the high half
|
|
;result of the multiplication that is needed in IDCT.
|
|
|
|
|vp8_short_idct4x4llm_neon| PROC
|
|
- ldr r12, _idct_coeff_
|
|
+ adr r12, idct_coeff
|
|
vld1.16 {q1, q2}, [r0]
|
|
vld1.16 {d0}, [r12]
|
|
|
|
vswp d3, d4 ;q2(vp[4] vp[12])
|
|
|
|
vqdmulh.s16 q3, q2, d0[2]
|
|
vqdmulh.s16 q4, q2, d0[0]
|
|
|
|
@@ -112,16 +112,14 @@
|
|
|
|
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 idct_coeff
|
|
idct_coeff
|
|
DCD 0x4e7b4e7b, 0x8a8c8a8c
|
|
|
|
;20091, 20091, 35468, 35468
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/sixtappredict16x16_neon.asm b/media/libvpx/vp8/common/arm/neon/sixtappredict16x16_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/sixtappredict16x16_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/sixtappredict16x16_neon.asm
|
|
@@ -28,17 +28,17 @@
|
|
; that the result can be a large positive number (> 2^15-1), which could be confused as a
|
|
; negtive number. To avoid that error, apply filter coeffs in the order of 0, 1, 4 ,5 ,2,
|
|
; which ensures that the result stays in s16 range. Finally, saturated add the result by
|
|
; applying 3rd filter coeff. Same applys to other filter functions.
|
|
|
|
|vp8_sixtap_predict16x16_neon| PROC
|
|
push {r4-r5, lr}
|
|
|
|
- ldr r12, _filter16_coeff_
|
|
+ adrl r12, filter16_coeff
|
|
ldr r4, [sp, #12] ;load parameters from stack
|
|
ldr r5, [sp, #16] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq secondpass_filter16x16_only
|
|
|
|
add r2, r12, r2, lsl #5 ;calculate filter location
|
|
|
|
@@ -475,18 +475,16 @@ 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 filter16_coeff
|
|
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
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/sixtappredict4x4_neon.asm b/media/libvpx/vp8/common/arm/neon/sixtappredict4x4_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/sixtappredict4x4_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/sixtappredict4x4_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; stack(r4) unsigned char *dst_ptr,
|
|
; stack(lr) int dst_pitch
|
|
|
|
|vp8_sixtap_predict_neon| PROC
|
|
push {r4, lr}
|
|
|
|
- ldr r12, _filter4_coeff_
|
|
+ adrl r12, filter4_coeff
|
|
ldr r4, [sp, #8] ;load parameters from stack
|
|
ldr lr, [sp, #12] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq secondpass_filter4x4_only
|
|
|
|
add r2, r12, r2, lsl #5 ;calculate filter location
|
|
|
|
@@ -406,18 +406,16 @@ 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 filter4_coeff
|
|
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
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/sixtappredict8x4_neon.asm b/media/libvpx/vp8/common/arm/neon/sixtappredict8x4_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/sixtappredict8x4_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/sixtappredict8x4_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; r4 unsigned char *dst_ptr,
|
|
; stack(r5) int dst_pitch
|
|
|
|
|vp8_sixtap_predict8x4_neon| PROC
|
|
push {r4-r5, lr}
|
|
|
|
- ldr r12, _filter8_coeff_
|
|
+ adrl r12, filter8_coeff
|
|
ldr r4, [sp, #12] ;load parameters from stack
|
|
ldr r5, [sp, #16] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq secondpass_filter8x4_only
|
|
|
|
add r2, r12, r2, lsl #5 ;calculate filter location
|
|
|
|
@@ -457,18 +457,16 @@ 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 filter8_coeff
|
|
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
|
|
diff --git a/media/libvpx/vp8/common/arm/neon/sixtappredict8x8_neon.asm b/media/libvpx/vp8/common/arm/neon/sixtappredict8x8_neon.asm
|
|
--- a/media/libvpx/vp8/common/arm/neon/sixtappredict8x8_neon.asm
|
|
+++ b/media/libvpx/vp8/common/arm/neon/sixtappredict8x8_neon.asm
|
|
@@ -20,17 +20,17 @@
|
|
; r2 int xoffset,
|
|
; r3 int yoffset,
|
|
; stack(r4) unsigned char *dst_ptr,
|
|
; stack(r5) int dst_pitch
|
|
|
|
|vp8_sixtap_predict8x8_neon| PROC
|
|
push {r4-r5, lr}
|
|
|
|
- ldr r12, _filter8_coeff_
|
|
+ adrl r12, filter8_coeff
|
|
|
|
ldr r4, [sp, #12] ;load parameters from stack
|
|
ldr r5, [sp, #16] ;load parameters from stack
|
|
|
|
cmp r2, #0 ;skip first_pass filter if xoffset=0
|
|
beq secondpass_filter8x8_only
|
|
|
|
add r2, r12, r2, lsl #5 ;calculate filter location
|
|
@@ -508,18 +508,16 @@ 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 filter8_coeff
|
|
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
|
|
diff --git a/media/libvpx/vp8/decoder/arm/neon/dequant_idct_neon.asm b/media/libvpx/vp8/decoder/arm/neon/dequant_idct_neon.asm
|
|
--- a/media/libvpx/vp8/decoder/arm/neon/dequant_idct_neon.asm
|
|
+++ b/media/libvpx/vp8/decoder/arm/neon/dequant_idct_neon.asm
|
|
@@ -30,17 +30,17 @@
|
|
ldr r1, [sp] ; pitch
|
|
vld1.32 {d14[0]}, [r2], r1
|
|
vld1.32 {d14[1]}, [r2], r1
|
|
vld1.32 {d15[0]}, [r2], r1
|
|
vld1.32 {d15[1]}, [r2]
|
|
|
|
ldr r1, [sp, #4] ; stride
|
|
|
|
- ldr r12, _CONSTANTS_
|
|
+ adr r12, _CONSTANTS_
|
|
|
|
vmul.i16 q1, q3, q5 ;input for short_idct4x4llm_neon
|
|
vmul.i16 q2, q4, q6
|
|
|
|
;|short_idct4x4llm_neon| PROC
|
|
vld1.16 {d0}, [r12]
|
|
vswp d3, d4 ;q2(vp[4] vp[12])
|
|
|
|
@@ -118,13 +118,13 @@
|
|
vst1.32 {d1[0]}, [r3], r1
|
|
vst1.32 {d1[1]}, [r3]
|
|
|
|
bx lr
|
|
|
|
ENDP ; |vp8_dequant_idct_add_neon|
|
|
|
|
; Constant Pool
|
|
-_CONSTANTS_ DCD cospi8sqrt2minus1
|
|
+_CONSTANTS_ EQU cospi8sqrt2minus1
|
|
cospi8sqrt2minus1 DCD 0x4e7b4e7b
|
|
sinpi8sqrt2 DCD 0x8a8c8a8c
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/decoder/arm/neon/idct_dequant_dc_full_2x_neon.asm b/media/libvpx/vp8/decoder/arm/neon/idct_dequant_dc_full_2x_neon.asm
|
|
--- a/media/libvpx/vp8/decoder/arm/neon/idct_dequant_dc_full_2x_neon.asm
|
|
+++ b/media/libvpx/vp8/decoder/arm/neon/idct_dequant_dc_full_2x_neon.asm
|
|
@@ -36,17 +36,17 @@
|
|
vld1.32 {d29[0]}, [r2], r1
|
|
vld1.32 {d29[1]}, [r12], r1
|
|
vld1.32 {d30[0]}, [r2], r1
|
|
vld1.32 {d30[1]}, [r12], r1
|
|
vld1.32 {d31[0]}, [r2]
|
|
ldr r1, [sp, #4]
|
|
vld1.32 {d31[1]}, [r12]
|
|
|
|
- ldr r2, _CONSTANTS_
|
|
+ adr r2, _CONSTANTS_
|
|
|
|
ldrh r12, [r1], #2 ; lo *dc
|
|
ldrh r1, [r1] ; hi *dc
|
|
|
|
; dequant: q[i] = q[i] * dq[i]
|
|
vmul.i16 q2, q2, q0
|
|
vmul.i16 q3, q3, q1
|
|
vmul.i16 q4, q4, q0
|
|
@@ -193,14 +193,14 @@
|
|
vst1.32 {d3[0]}, [r3]
|
|
vst1.32 {d3[1]}, [r2]
|
|
|
|
bx lr
|
|
|
|
ENDP ; |idct_dequant_dc_full_2x_neon|
|
|
|
|
; Constant Pool
|
|
-_CONSTANTS_ DCD cospi8sqrt2minus1
|
|
+_CONSTANTS_ EQU cospi8sqrt2minus1
|
|
cospi8sqrt2minus1 DCD 0x4e7b
|
|
; because the lowest bit in 0x8a8c is 0, we can pre-shift this
|
|
sinpi8sqrt2 DCD 0x4546
|
|
|
|
END
|
|
diff --git a/media/libvpx/vp8/decoder/arm/neon/idct_dequant_full_2x_neon.asm b/media/libvpx/vp8/decoder/arm/neon/idct_dequant_full_2x_neon.asm
|
|
--- a/media/libvpx/vp8/decoder/arm/neon/idct_dequant_full_2x_neon.asm
|
|
+++ b/media/libvpx/vp8/decoder/arm/neon/idct_dequant_full_2x_neon.asm
|
|
@@ -35,17 +35,17 @@
|
|
vld1.32 {d28[1]}, [r12], r1 ; r pre
|
|
vld1.32 {d29[0]}, [r2], r1
|
|
vld1.32 {d29[1]}, [r12], r1
|
|
vld1.32 {d30[0]}, [r2], r1
|
|
vld1.32 {d30[1]}, [r12], r1
|
|
vld1.32 {d31[0]}, [r2]
|
|
vld1.32 {d31[1]}, [r12]
|
|
|
|
- ldr r2, _CONSTANTS_
|
|
+ adr r2, _CONSTANTS_
|
|
|
|
; dequant: q[i] = q[i] * dq[i]
|
|
vmul.i16 q2, q2, q0
|
|
vmul.i16 q3, q3, q1
|
|
vmul.i16 q4, q4, q0
|
|
vmul.i16 q5, q5, q1
|
|
|
|
vld1.16 {d0}, [r2]
|
|
@@ -185,14 +185,14 @@
|
|
vst1.32 {d3[0]}, [r3]
|
|
vst1.32 {d3[1]}, [r2]
|
|
|
|
bx lr
|
|
|
|
ENDP ; |idct_dequant_full_2x_neon|
|
|
|
|
; Constant Pool
|
|
-_CONSTANTS_ DCD cospi8sqrt2minus1
|
|
+_CONSTANTS_ EQU cospi8sqrt2minus1
|
|
cospi8sqrt2minus1 DCD 0x4e7b
|
|
; because the lowest bit in 0x8a8c is 0, we can pre-shift this
|
|
sinpi8sqrt2 DCD 0x4546
|
|
|
|
END
|