if (vpx_config("CONFIG_VP9") eq "yes") {
add_proto qw/void vpx_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct16x16_38_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct32x32_135_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int stride";
-
add_proto qw/void vpx_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int stride";
if (vpx_config("CONFIG_EMULATE_HARDWARE") ne "yes") {
+ # Note that there are more specializations appended when CONFIG_VP9_HIGHBITDEPTH is off.
specialize qw/vpx_idct4x4_16_add neon sse2/;
-
specialize qw/vpx_idct4x4_1_add neon sse2/;
-
specialize qw/vpx_idct8x8_64_add neon sse2 ssse3/;
-
specialize qw/vpx_idct8x8_12_add neon sse2 ssse3/;
-
specialize qw/vpx_idct8x8_1_add neon sse2/;
-
specialize qw/vpx_idct16x16_256_add neon sse2/;
-
specialize qw/vpx_idct16x16_38_add neon sse2/;
$vpx_idct16x16_38_add_sse2=vpx_idct16x16_256_add_sse2;
-
specialize qw/vpx_idct16x16_10_add neon sse2/;
-
specialize qw/vpx_idct16x16_1_add neon sse2/;
-
specialize qw/vpx_idct32x32_1024_add neon sse2 ssse3/;
-
specialize qw/vpx_idct32x32_135_add neon sse2 ssse3/;
$vpx_idct32x32_135_add_sse2=vpx_idct32x32_1024_add_sse2;
-
specialize qw/vpx_idct32x32_34_add neon sse2 ssse3/;
-
specialize qw/vpx_idct32x32_1_add neon sse2/;
+
+ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") ne "yes") {
+ # Note that these specializations appends to the above ones.
+ specialize qw/vpx_idct4x4_16_add dspr2 msa/;
+ specialize qw/vpx_idct4x4_1_add dspr2 msa/;
+ specialize qw/vpx_idct8x8_64_add dspr2 msa/;
+ specialize qw/vpx_idct8x8_12_add dspr2 msa/;
+ specialize qw/vpx_idct8x8_1_add dspr2 msa/;
+ specialize qw/vpx_idct16x16_256_add dspr2 msa/;
+ specialize qw/vpx_idct16x16_38_add dspr2 msa/;
+ $vpx_idct16x16_38_add_dspr2=vpx_idct16x16_256_add_dspr2;
+ $vpx_idct16x16_38_add_msa=vpx_idct16x16_256_add_msa;
+ specialize qw/vpx_idct16x16_10_add dspr2 msa/;
+ specialize qw/vpx_idct16x16_1_add dspr2 msa/;
+ specialize qw/vpx_idct32x32_1024_add dspr2 msa/;
+ specialize qw/vpx_idct32x32_135_add dspr2 msa/;
+ $vpx_idct32x32_135_add_dspr2=vpx_idct32x32_1024_add_dspr2;
+ $vpx_idct32x32_135_add_msa=vpx_idct32x32_1024_add_msa;
+ specialize qw/vpx_idct32x32_34_add dspr2 msa/;
+ specialize qw/vpx_idct32x32_1_add dspr2 msa/;
+ specialize qw/vpx_iwht4x4_16_add msa sse2/;
+ specialize qw/vpx_iwht4x4_1_add msa/;
+ } # !CONFIG_VP9_HIGHBITDEPTH
} # !CONFIG_EMULATE_HARDWARE
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
# that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
specialize qw/vpx_iwht4x4_16_add sse2/;
- add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
+ 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/;
- add_proto qw/void vpx_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
+ 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/;
- add_proto qw/void vpx_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct16x16_38_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
+ 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/;
- add_proto qw/void vpx_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct32x32_135_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
+ 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";
+ add_proto qw/void vpx_highbd_idct32x32_34_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
+ add_proto qw/void vpx_highbd_idct32x32_1_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
specialize qw/vpx_highbd_idct32x32_1_add neon sse2/;
- add_proto qw/void vpx_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
-
- add_proto qw/void vpx_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int stride, int bd";
+ add_proto qw/void vpx_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
+ add_proto qw/void vpx_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint16_t *dest, int stride, int bd";
if (vpx_config("CONFIG_EMULATE_HARDWARE") ne "yes") {
specialize qw/vpx_highbd_idct4x4_16_add neon sse2/;
-
specialize qw/vpx_highbd_idct8x8_64_add neon sse2/;
-
specialize qw/vpx_highbd_idct8x8_12_add neon sse2/;
-
specialize qw/vpx_highbd_idct16x16_256_add neon sse2/;
-
specialize qw/vpx_highbd_idct16x16_38_add neon sse2/;
$vpx_highbd_idct16x16_38_add_sse2=vpx_highbd_idct16x16_256_add_sse2;
-
specialize qw/vpx_highbd_idct16x16_10_add neon sse2/;
-
specialize qw/vpx_highbd_idct32x32_1024_add neon/;
-
specialize qw/vpx_highbd_idct32x32_135_add neon/;
-
specialize qw/vpx_highbd_idct32x32_34_add neon/;
} # !CONFIG_EMULATE_HARDWARE
-} else {
- if (vpx_config("CONFIG_EMULATE_HARDWARE") ne "yes") {
- specialize qw/vpx_idct4x4_16_add dspr2 msa/;
-
- specialize qw/vpx_idct4x4_1_add dspr2 msa/;
-
- specialize qw/vpx_idct8x8_64_add dspr2 msa/;
-
- specialize qw/vpx_idct8x8_12_add dspr2 msa/;
-
- specialize qw/vpx_idct8x8_1_add dspr2 msa/;
-
- specialize qw/vpx_idct16x16_256_add dspr2 msa/;
-
- specialize qw/vpx_idct16x16_38_add dspr2 msa/;
- $vpx_idct16x16_38_add_dspr2=vpx_idct16x16_256_add_dspr2;
- $vpx_idct16x16_38_add_msa=vpx_idct16x16_256_add_msa;
-
- specialize qw/vpx_idct16x16_10_add dspr2 msa/;
-
- specialize qw/vpx_idct16x16_1_add dspr2 msa/;
-
- specialize qw/vpx_idct32x32_1024_add dspr2 msa/;
-
- specialize qw/vpx_idct32x32_135_add dspr2 msa/;
- $vpx_idct32x32_135_add_dspr2=vpx_idct32x32_1024_add_dspr2;
- $vpx_idct32x32_135_add_msa=vpx_idct32x32_1024_add_msa;
-
- specialize qw/vpx_idct32x32_34_add dspr2 msa/;
-
- specialize qw/vpx_idct32x32_1_add dspr2 msa/;
-
- specialize qw/vpx_iwht4x4_16_add msa sse2/;
-
- specialize qw/vpx_iwht4x4_1_add msa/;
- } # !CONFIG_EMULATE_HARDWARE
} # CONFIG_VP9_HIGHBITDEPTH
} # CONFIG_VP9
specialize qw/vpx_variance32x32 sse2 avx2 neon msa/;
add_proto qw/unsigned int vpx_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vpx_variance32x16 sse2 avx2 msa/;
+ specialize qw/vpx_variance32x16 sse2 avx2 neon msa/;
add_proto qw/unsigned int vpx_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vpx_variance16x32 sse2 msa/;
+ specialize qw/vpx_variance16x32 sse2 neon msa/;
add_proto qw/unsigned int vpx_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vpx_variance16x16 sse2 avx2 neon msa/;
specialize qw/vpx_variance8x8 sse2 neon msa/;
add_proto qw/unsigned int vpx_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
- specialize qw/vpx_variance8x4 sse2 msa/;
+ 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/;
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_get4x4sse_cs neon msa/;
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/;
+ specialize qw/vpx_comp_avg_pred sse2 vsx/;
#
# Subpixel Variance