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

Add AVX2 version of vpx_convolve8_avg.

Browse Source 
vpx_convolve8_avg works by first running a normal horizontal filter then a
vertical filter averages at the end.

The added vpx_convolve8_avg_avx2 calls pre-existing AVX2 code for the
horizontal step.

vpx_convolve8_avg_vert_avx2 is also added, but only uses ssse3 code.

Change-Id: If5160c0c8e778e10de61ee9bf42ee4be5975c983
This commit is contained in:
Kyle Siefring
2017-10-07 16:02:02 -04:00
parent 807248ec81
commit 9ca06bcdd2
3 changed files with 50 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
vpx_dsp/x86/vpx_subpixel_8t_intrin_avx2.c
View File

@@ -539,6 +539,12 @@ filter8_1dfunction vpx_filter_block1d4_h8_ssse3;
#define vpx_filter_block1d8_h8_avx2 vpx_filter_block1d8_h8_ssse3
#define vpx_filter_block1d4_h8_avx2 vpx_filter_block1d4_h8_ssse3
#endif // ARCH_X86_64
filter8_1dfunction vpx_filter_block1d16_v8_avg_ssse3;
filter8_1dfunction vpx_filter_block1d8_v8_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_v8_avg_ssse3;
#define vpx_filter_block1d16_v8_avg_avx2 vpx_filter_block1d16_v8_avg_ssse3
#define vpx_filter_block1d8_v8_avg_avx2 vpx_filter_block1d8_v8_avg_ssse3
#define vpx_filter_block1d4_v8_avg_avx2 vpx_filter_block1d4_v8_avg_ssse3
filter8_1dfunction vpx_filter_block1d16_v2_ssse3;
filter8_1dfunction vpx_filter_block1d16_h2_ssse3;
filter8_1dfunction vpx_filter_block1d8_v2_ssse3;
@@ -552,6 +558,12 @@ filter8_1dfunction vpx_filter_block1d4_h2_ssse3;
#define vpx_filter_block1d8_h2_avx2 vpx_filter_block1d8_h2_ssse3
#define vpx_filter_block1d4_v2_avx2 vpx_filter_block1d4_v2_ssse3
#define vpx_filter_block1d4_h2_avx2 vpx_filter_block1d4_h2_ssse3
filter8_1dfunction vpx_filter_block1d16_v2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d8_v2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_v2_avg_ssse3;
#define vpx_filter_block1d16_v2_avg_avx2 vpx_filter_block1d16_v2_avg_ssse3
#define vpx_filter_block1d8_v2_avg_avx2 vpx_filter_block1d8_v2_avg_ssse3
#define vpx_filter_block1d4_v2_avg_avx2 vpx_filter_block1d4_v2_avg_ssse3
// void vpx_convolve8_horiz_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
@@ -562,13 +574,25 @@ filter8_1dfunction vpx_filter_block1d4_h2_ssse3;
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
// void vpx_convolve8_avg_vert_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4,
// int y_step_q4, int w, int h);
FUN_CONV_1D(horiz, x0_q4, x_step_q4, h, src, , avx2);
FUN_CONV_1D(vert, y0_q4, y_step_q4, v, src - src_stride * 3, , avx2);
FUN_CONV_1D(avg_vert, y0_q4, y_step_q4, v, src - src_stride * 3, avg_, avx2);
// void vpx_convolve8_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
// void vpx_convolve8_avg_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
FUN_CONV_2D(, avx2);
FUN_CONV_2D(avg_, avx2);
#endif // HAVE_AX2 && HAVE_SSSE3