Removing unused 8x4 transform from the encoder.

Change-Id: Icbcf68b5b685a56f255ebc3859c9692accdadf9e

vp9/common/vp9_rtcd_defs.sh

@@ -701,9 +701,6 @@ specialize vp9_short_fdct8x8 sse2
 prototype void vp9_short_fdct4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
 specialize vp9_short_fdct4x4 sse2
 
-prototype void vp9_short_fdct8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
-specialize vp9_short_fdct8x4 sse2
-
 prototype void vp9_short_fdct32x32 "int16_t *InputData, int16_t *OutputData, int pitch"
 specialize vp9_short_fdct32x32 sse2
 
@@ -716,9 +713,6 @@ specialize vp9_short_fdct16x16 sse2
 prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
 specialize vp9_short_walsh4x4
 
-prototype void vp9_short_walsh8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
-specialize vp9_short_walsh8x4
-
 #
 # Motion search
 #

vp9/encoder/vp9_block.h

@@ -172,7 +172,6 @@ struct macroblock {
   BLOCK_SIZE sb64_partitioning;
 
   void (*fwd_txm4x4)(int16_t *input, int16_t *output, int pitch);
-  void (*fwd_txm8x4)(int16_t *input, int16_t *output, int pitch);
   void (*fwd_txm8x8)(int16_t *input, int16_t *output, int pitch);
   void (*fwd_txm16x16)(int16_t *input, int16_t *output, int pitch);
   void (*quantize_b_4x4)(MACROBLOCK *x, int b_idx, TX_TYPE tx_type,

vp9/encoder/vp9_dct.c

@@ -178,11 +178,6 @@ void vp9_short_fht4x4_c(int16_t *input, int16_t *output,
   }
 }
 
-void vp9_short_fdct8x4_c(int16_t *input, int16_t *output, int pitch) {
-    vp9_short_fdct4x4_c(input, output, pitch);
-    vp9_short_fdct4x4_c(input + 4, output + 16, pitch);
-}
-
 static void fdct8(const int16_t *input, int16_t *output) {
   /*canbe16*/ int s0, s1, s2, s3, s4, s5, s6, s7;
   /*needs32*/ int t0, t1, t2, t3;
@@ -647,12 +642,6 @@ void vp9_short_walsh4x4_c(int16_t *input, int16_t *output, int pitch) {
   }
 }
 
-void vp9_short_walsh8x4_c(int16_t *input, int16_t *output, int pitch) {
-  vp9_short_walsh4x4_c(input,   output,    pitch);
-  vp9_short_walsh4x4_c(input + 4, output + 16, pitch);
-}
-
-
 // Rewrote to use same algorithm as others.
 static void fdct16(const int16_t in[16], int16_t out[16]) {
   /*canbe16*/ int step1[8];

vp9/encoder/vp9_encodeframe.c

@@ -1861,7 +1861,6 @@ static void init_encode_frame_mb_context(VP9_COMP *cpi) {
 static void switch_lossless_mode(VP9_COMP *cpi, int lossless) {
   if (lossless) {
     // printf("Switching to lossless\n");
-    cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4;
     cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4;
     cpi->mb.e_mbd.itxm_add = vp9_iwht4x4_add;
     cpi->mb.optimize = 0;
@@ -1870,7 +1869,6 @@ static void switch_lossless_mode(VP9_COMP *cpi, int lossless) {
     cpi->common.tx_mode = ONLY_4X4;
   } else {
     // printf("Not lossless\n");
-    cpi->mb.fwd_txm8x4 = vp9_short_fdct8x4;
     cpi->mb.fwd_txm4x4 = vp9_short_fdct4x4;
     cpi->mb.e_mbd.itxm_add = vp9_idct4x4_add;
   }

vp9/encoder/vp9_onyx_if.c

@@ -955,10 +955,8 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
 
   cpi->mb.fwd_txm16x16  = vp9_short_fdct16x16;
   cpi->mb.fwd_txm8x8    = vp9_short_fdct8x8;
-  cpi->mb.fwd_txm8x4    = vp9_short_fdct8x4;
   cpi->mb.fwd_txm4x4    = vp9_short_fdct4x4;
   if (cpi->oxcf.lossless || cpi->mb.e_mbd.lossless) {
-    cpi->mb.fwd_txm8x4    = vp9_short_walsh8x4;
     cpi->mb.fwd_txm4x4    = vp9_short_walsh4x4;
   }
 

vp9/encoder/x86/vp9_dct_sse2.c

@@ -112,11 +112,6 @@ void vp9_short_fdct4x4_sse2(int16_t *input, int16_t *output, int pitch) {
   }
 }
 
-void vp9_short_fdct8x4_sse2(int16_t *input, int16_t *output, int pitch) {
-  vp9_short_fdct4x4_sse2(input, output, pitch);
-  vp9_short_fdct4x4_sse2(input + 4, output + 16, pitch);
-}
-
 static INLINE void load_buffer_4x4(int16_t *input, __m128i *in, int stride) {
   const __m128i k__nonzero_bias_a = _mm_setr_epi16(0, 1, 1, 1, 1, 1, 1, 1);
   const __m128i k__nonzero_bias_b = _mm_setr_epi16(1, 0, 0, 0, 0, 0, 0, 0);