Merge "ppc: Add vpx_sadnxmx4d_vsx for n,m = {8, 16, 32 ,64}"

[libvpx] / vpx_dsp / vpx_dsp_rtcd_defs.pl

diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index 7368d90ea827bdfdceb55a2626a3822aea7665f7..9f005b284058092a643f64ee400dbcffc520fbb3 100644 (file)
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -484,7 +484,7 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
 if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {
 if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
   add_proto qw/void vpx_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
-  specialize qw/vpx_fdct4x4 sse2/;
+  specialize qw/vpx_fdct4x4 neon sse2/;
 
   add_proto qw/void vpx_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
   specialize qw/vpx_fdct4x4_1 sse2/;
@@ -532,7 +532,7 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
   add_proto qw/void vpx_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
 } else {
   add_proto qw/void vpx_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
-  specialize qw/vpx_fdct4x4 sse2 msa/;
+  specialize qw/vpx_fdct4x4 neon sse2 msa/;
 
   add_proto qw/void vpx_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
   specialize qw/vpx_fdct4x4_1 sse2/;
@@ -629,18 +629,18 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
 
   add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct4x4_1_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
-  specialize qw/vpx_highbd_idct4x4_1_add neon/;
+  specialize qw/vpx_highbd_idct4x4_1_add neon sse2/;
 
   add_proto qw/void vpx_highbd_idct8x8_64_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct8x8_12_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct8x8_1_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
-  specialize qw/vpx_highbd_idct8x8_1_add neon/;
+  specialize qw/vpx_highbd_idct8x8_1_add neon sse2/;
 
   add_proto qw/void vpx_highbd_idct16x16_256_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct16x16_38_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct16x16_10_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct16x16_1_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
-  specialize qw/vpx_highbd_idct16x16_1_add neon/;
+  specialize qw/vpx_highbd_idct16x16_1_add neon sse2/;
 
   add_proto qw/void vpx_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
   add_proto qw/void vpx_highbd_idct32x32_135_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
@@ -696,28 +696,28 @@ specialize qw/vpx_subtract_block neon msa sse2/;
 # Single block SAD
 #
 add_proto qw/unsigned int vpx_sad64x64/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad64x64 avx2 neon msa sse2/;
+specialize qw/vpx_sad64x64 avx2 neon msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad64x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad64x32 avx2 msa sse2/;
+specialize qw/vpx_sad64x32 avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x64/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad32x64 avx2 msa sse2/;
+specialize qw/vpx_sad32x64 avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad32x32 avx2 neon msa sse2/;
+specialize qw/vpx_sad32x32 avx2 neon msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad32x16 avx2 msa sse2/;
+specialize qw/vpx_sad32x16 avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad16x32 msa sse2/;
+specialize qw/vpx_sad16x32 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad16x16 neon msa sse2/;
+specialize qw/vpx_sad16x16 neon msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-specialize qw/vpx_sad16x8 neon msa sse2/;
+specialize qw/vpx_sad16x8 neon msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad8x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
 specialize qw/vpx_sad8x16 neon msa sse2/;
@@ -778,28 +778,28 @@ if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {
 }  # CONFIG_VP9_ENCODER
 
 add_proto qw/unsigned int vpx_sad64x64_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad64x64_avg avx2 msa sse2/;
+specialize qw/vpx_sad64x64_avg avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad64x32_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad64x32_avg avx2 msa sse2/;
+specialize qw/vpx_sad64x32_avg avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x64_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad32x64_avg avx2 msa sse2/;
+specialize qw/vpx_sad32x64_avg avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x32_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad32x32_avg avx2 msa sse2/;
+specialize qw/vpx_sad32x32_avg avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad32x16_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad32x16_avg avx2 msa sse2/;
+specialize qw/vpx_sad32x16_avg avx2 msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x32_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad16x32_avg msa sse2/;
+specialize qw/vpx_sad16x32_avg msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x16_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad16x16_avg msa sse2/;
+specialize qw/vpx_sad16x16_avg msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad16x8_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
-specialize qw/vpx_sad16x8_avg msa sse2/;
+specialize qw/vpx_sad16x8_avg msa sse2 vsx/;
 
 add_proto qw/unsigned int vpx_sad8x16_avg/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
 specialize qw/vpx_sad8x16_avg msa sse2/;
@@ -873,28 +873,28 @@ specialize qw/vpx_sad4x4x8 sse4_1 msa/;
 # Multi-block SAD, comparing a reference to N independent blocks
 #
 add_proto qw/void vpx_sad64x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad64x64x4d avx2 neon msa sse2/;
+specialize qw/vpx_sad64x64x4d avx2 neon msa sse2 vsx/;
 
 add_proto qw/void vpx_sad64x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad64x32x4d msa sse2/;
+specialize qw/vpx_sad64x32x4d msa sse2 vsx/;
 
 add_proto qw/void vpx_sad32x64x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad32x64x4d msa sse2/;
+specialize qw/vpx_sad32x64x4d msa sse2 vsx/;
 
 add_proto qw/void vpx_sad32x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad32x32x4d avx2 neon msa sse2/;
+specialize qw/vpx_sad32x32x4d avx2 neon msa sse2 vsx/;
 
 add_proto qw/void vpx_sad32x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad32x16x4d msa sse2/;
+specialize qw/vpx_sad32x16x4d msa sse2 vsx/;
 
 add_proto qw/void vpx_sad16x32x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad16x32x4d msa sse2/;
+specialize qw/vpx_sad16x32x4d msa sse2 vsx/;
 
 add_proto qw/void vpx_sad16x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad16x16x4d neon msa sse2/;
+specialize qw/vpx_sad16x16x4d neon msa sse2 vsx/;
 
 add_proto qw/void vpx_sad16x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
-specialize qw/vpx_sad16x8x4d msa sse2/;
+specialize qw/vpx_sad16x8x4d msa sse2 vsx/;
 
 add_proto qw/void vpx_sad8x16x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
 specialize qw/vpx_sad8x16x4d msa sse2/;
@@ -1142,12 +1142,10 @@ add_proto qw/unsigned int vpx_variance8x4/, "const uint8_t *src_ptr, int source_
   specialize qw/vpx_variance8x4 sse2 neon msa/;
 
 add_proto qw/unsigned int vpx_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-# TODO(johannkoenig): neon
-  specialize qw/vpx_variance4x8 sse2 msa/;
+  specialize qw/vpx_variance4x8 sse2 neon msa/;
 
 add_proto qw/unsigned int vpx_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-# TODO(johannkoenig): neon
-  specialize qw/vpx_variance4x4 sse2 msa/;
+  specialize qw/vpx_variance4x4 sse2 neon msa/;
 
 #
 # Specialty Variance
@@ -1171,13 +1169,13 @@ add_proto qw/unsigned int vpx_mse8x8/, "const uint8_t *src_ptr, int  source_stri
   specialize qw/vpx_mse8x8 sse2 msa/;
 
 add_proto qw/unsigned int vpx_get_mb_ss/, "const int16_t *";
-  specialize qw/vpx_get_mb_ss sse2 msa/;
+  specialize qw/vpx_get_mb_ss sse2 msa vsx/;
 
 add_proto qw/unsigned int vpx_get4x4sse_cs/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride";
   specialize qw/vpx_get4x4sse_cs neon msa vsx/;
 
 add_proto qw/void vpx_comp_avg_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride";
-  specialize qw/vpx_comp_avg_pred sse2 vsx/;
+  specialize qw/vpx_comp_avg_pred neon sse2 vsx/;
 
 #
 # Subpixel Variance
@@ -1216,49 +1214,49 @@ add_proto qw/uint32_t vpx_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int s
   specialize qw/vpx_sub_pixel_variance8x4 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
-  specialize qw/vpx_sub_pixel_variance4x8 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_variance4x8 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
-  specialize qw/vpx_sub_pixel_variance4x4 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_variance4x4 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance64x64 avx2 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance64x64 neon avx2 msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance64x32 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance64x32 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance32x64 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance32x64 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance32x32 avx2 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance32x32 neon avx2 msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance32x16 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance32x16 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance16x32 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance16x32 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance16x16 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance16x16 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance16x8 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance16x8 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance8x16 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance8x16 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance8x8 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance8x8 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance8x4 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance8x4 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance4x8 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance4x8 neon msa sse2 ssse3/;
 
 add_proto qw/uint32_t vpx_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
-  specialize qw/vpx_sub_pixel_avg_variance4x4 msa sse2 ssse3/;
+  specialize qw/vpx_sub_pixel_avg_variance4x4 neon msa sse2 ssse3/;
 
 if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
   add_proto qw/unsigned int vpx_highbd_12_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";