Merge "SSE2 16x16 inverse ADST/DCT hybrid transform"

vp9/common/vp9_idct.c

@@ -534,6 +534,7 @@ static void idct16_1d(int16_t *input, int16_t *output) {
   step1[14] = -step2[14] + step2[15];
   step1[15] = step2[14] + step2[15];
 
+  // stage 4
   temp1 = (step1[0] + step1[1]) * cospi_16_64;
   temp2 = (step1[0] - step1[1]) * cospi_16_64;
   step2[0] = dct_const_round_shift(temp1);

vp9/common/vp9_rtcd_defs.sh

@@ -331,7 +331,7 @@ prototype void vp9_short_iht8x8_add "int16_t *input, uint8_t *dest, int dest_str
 specialize vp9_short_iht8x8_add sse2
 
 prototype void vp9_short_iht16x16_add "int16_t *input, uint8_t *output, int pitch, int tx_type"
-specialize vp9_short_iht16x16_add
+specialize vp9_short_iht16x16_add sse2
 
 prototype void vp9_idct4_1d "int16_t *input, int16_t *output"
 specialize vp9_idct4_1d sse2