ppc Add vpx_sad32x16/32/64_vsx

Change-Id: I3c4f9d595275669580413a71b3c3c810e7ddcacd
This commit is contained in:
Alexandra Hájková 2017-05-08 11:56:17 +00:00 committed by James Zern
parent a12ea1d5e9
commit 0b15bf1e54
3 changed files with 32 additions and 4 deletions

View File

@ -924,8 +924,12 @@ INSTANTIATE_TEST_CASE_P(MSA, SADx4Test, ::testing::ValuesIn(x4d_msa_tests));
// VSX functions
#if HAVE_VSX
const SadMxNParam vsx_tests[] = {
SadMxNParam(32, 64, &vpx_sad32x64_vsx),
SadMxNParam(32, 32, &vpx_sad32x32_vsx),
SadMxNParam(32, 16, &vpx_sad32x16_vsx),
SadMxNParam(16, 32, &vpx_sad16x32_vsx),
SadMxNParam(16, 16, &vpx_sad16x16_vsx), SadMxNParam(16, 8, &vpx_sad16x8_vsx),
SadMxNParam(16, 16, &vpx_sad16x16_vsx),
SadMxNParam(16, 8, &vpx_sad16x8_vsx),
};
INSTANTIATE_TEST_CASE_P(VSX, SADTest, ::testing::ValuesIn(vsx_tests));
#endif // HAVE_VSX

View File

@ -48,6 +48,30 @@
return sad[3] + sad[2] + sad[1] + sad[0]; \
}
#define SAD32(height) \
unsigned int vpx_sad32x##height##_vsx(const uint8_t *a, int a_stride, \
const uint8_t *b, int b_stride) { \
int y; \
unsigned int sad[4]; \
uint8x16_t v_a, v_b; \
int16x8_t v_ah, v_al, v_bh, v_bl, v_absh, v_absl, v_subh, v_subl; \
int32x4_t v_sad = vec_splat_s32(0); \
\
for (y = 0; y < height; y++) { \
PROCESS16(0); \
PROCESS16(16); \
\
a += a_stride; \
b += b_stride; \
} \
vec_vsx_st((uint32x4_t)v_sad, 0, sad); \
\
return sad[3] + sad[2] + sad[1] + sad[0]; \
}
SAD16(8);
SAD16(16);
SAD16(32);
SAD32(16);
SAD32(32);
SAD32(64);

View File

@ -702,13 +702,13 @@ add_proto qw/unsigned int vpx_sad64x32/, "const uint8_t *src_ptr, int src_stride
specialize qw/vpx_sad64x32 avx2 msa sse2/;
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 vsx/;