generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge "mips msa vp9 avg subpel variance optimization"

Browse Source
This commit is contained in:
James Zern 2015-07-01 20:05:49 +00:00 committed by Gerrit Code Review
commit 4f7e7c4d49
3 changed files with 1290 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
test/variance_test.cc
View File

@ -2095,6 +2095,47 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(5, 6, subpel_variance32x64_msa, 0),
make_tuple(6, 5, subpel_variance64x32_msa, 0),
make_tuple(6, 6, subpel_variance64x64_msa, 0)));
const vp9_subp_avg_variance_fn_t subpel_avg_variance4x4_msa =
vp9_sub_pixel_avg_variance4x4_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance4x8_msa =
vp9_sub_pixel_avg_variance4x8_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance8x4_msa =
vp9_sub_pixel_avg_variance8x4_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance8x8_msa =
vp9_sub_pixel_avg_variance8x8_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance8x16_msa =
vp9_sub_pixel_avg_variance8x16_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance16x8_msa =
vp9_sub_pixel_avg_variance16x8_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance16x16_msa =
vp9_sub_pixel_avg_variance16x16_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance16x32_msa =
vp9_sub_pixel_avg_variance16x32_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance32x16_msa =
vp9_sub_pixel_avg_variance32x16_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance32x32_msa =
vp9_sub_pixel_avg_variance32x32_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance32x64_msa =
vp9_sub_pixel_avg_variance32x64_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance64x32_msa =
vp9_sub_pixel_avg_variance64x32_msa;
const vp9_subp_avg_variance_fn_t subpel_avg_variance64x64_msa =
vp9_sub_pixel_avg_variance64x64_msa;
INSTANTIATE_TEST_CASE_P(
MSA, VP9SubpelAvgVarianceTest,
::testing::Values(make_tuple(2, 2, subpel_avg_variance4x4_msa, 0),
make_tuple(2, 3, subpel_avg_variance4x8_msa, 0),
make_tuple(3, 2, subpel_avg_variance8x4_msa, 0),
make_tuple(3, 3, subpel_avg_variance8x8_msa, 0),
make_tuple(3, 4, subpel_avg_variance8x16_msa, 0),
make_tuple(4, 3, subpel_avg_variance16x8_msa, 0),
make_tuple(4, 4, subpel_avg_variance16x16_msa, 0),
make_tuple(4, 5, subpel_avg_variance16x32_msa, 0),
make_tuple(5, 4, subpel_avg_variance32x16_msa, 0),
make_tuple(5, 5, subpel_avg_variance32x32_msa, 0),
make_tuple(5, 6, subpel_avg_variance32x64_msa, 0),
make_tuple(6, 5, subpel_avg_variance64x32_msa, 0),
make_tuple(6, 6, subpel_avg_variance64x64_msa, 0)));
#endif // CONFIG_VP9_ENCODER
#endif // HAVE_MSA
} // namespace

26
vp9/common/vp9_rtcd_defs.pl
View File

@ -801,81 +801,81 @@ add_proto qw/unsigned int vp9_sub_pixel_variance64x64/, "const uint8_t *src_ptr,
specialize qw/vp9_sub_pixel_variance64x64 avx2 neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance64x64 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance64x64 avx2 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance32x64 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance32x64/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance32x64 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance64x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance64x32/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance64x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance32x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance32x16/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance32x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance16x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance16x32/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance16x32 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance32x32 avx2 neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance32x32 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance32x32 avx2 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance16x16 neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance16x16/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance16x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance8x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance8x16/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance8x16 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance16x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance16x8/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance16x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance8x8 neon msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance8x8/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance8x8 msa/, "$sse2_x86inc", "$ssse3_x86inc";
# TODO(jingning): need to convert 8x4/4x8 functions into mmx/sse form
add_proto qw/unsigned int vp9_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance8x4 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance8x4/, "$sse2_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance8x4 msa/, "$sse2_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance4x8 msa/, "$sse_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance4x8/, "$sse_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance4x8 msa/, "$sse_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
specialize qw/vp9_sub_pixel_variance4x4 msa/, "$sse_x86inc", "$ssse3_x86inc";
#vp9_sub_pixel_variance4x4_sse2=vp9_sub_pixel_variance4x4_wmt
add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
specialize qw/vp9_sub_pixel_avg_variance4x4/, "$sse_x86inc", "$ssse3_x86inc";
specialize qw/vp9_sub_pixel_avg_variance4x4 msa/, "$sse_x86inc", "$ssse3_x86inc";
add_proto qw/unsigned int vp9_avg_8x8/, "const uint8_t *, int p";
specialize qw/vp9_avg_8x8 sse2 neon msa/;

1237
vp9/encoder/mips/msa/vp9_variance_msa.c
View File

File diff suppressed because it is too large Load Diff