Commit c516dd67 authored by Johann's avatar Johann Committed by Johann Koenig

neon hadamard 16x16

Runs about twice as fast as C

BUG=webm:1027

Change-Id: I6760d99f4e22259439ca35d746194b12a81bfa71
parent 63eb2ee0
...@@ -212,4 +212,9 @@ INSTANTIATE_TEST_CASE_P(C, Hadamard16x16Test, ...@@ -212,4 +212,9 @@ INSTANTIATE_TEST_CASE_P(C, Hadamard16x16Test,
INSTANTIATE_TEST_CASE_P(SSE2, Hadamard16x16Test, INSTANTIATE_TEST_CASE_P(SSE2, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_sse2)); ::testing::Values(&vpx_hadamard_16x16_sse2));
#endif // HAVE_SSE2 #endif // HAVE_SSE2
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_neon));
#endif // HAVE_NEON
} // namespace } // namespace
...@@ -160,3 +160,42 @@ void vpx_hadamard_8x8_neon(const int16_t *src_diff, int src_stride, ...@@ -160,3 +160,42 @@ void vpx_hadamard_8x8_neon(const int16_t *src_diff, int src_stride,
vst1q_s16(coeff + 48, a6); vst1q_s16(coeff + 48, a6);
vst1q_s16(coeff + 56, a7); vst1q_s16(coeff + 56, a7);
} }
void vpx_hadamard_16x16_neon(const int16_t *src_diff, int src_stride,
int16_t *coeff) {
int i;
/* Rearrange 16x16 to 8x32 and remove stride.
* Top left first. */
vpx_hadamard_8x8_neon(src_diff + 0 + 0 * src_stride, src_stride, coeff + 0);
/* Top right. */
vpx_hadamard_8x8_neon(src_diff + 8 + 0 * src_stride, src_stride, coeff + 64);
/* Bottom left. */
vpx_hadamard_8x8_neon(src_diff + 0 + 8 * src_stride, src_stride, coeff + 128);
/* Bottom right. */
vpx_hadamard_8x8_neon(src_diff + 8 + 8 * src_stride, src_stride, coeff + 192);
for (i = 0; i < 64; i += 8) {
const int16x8_t a0 = vld1q_s16(coeff + 0);
const int16x8_t a1 = vld1q_s16(coeff + 64);
const int16x8_t a2 = vld1q_s16(coeff + 128);
const int16x8_t a3 = vld1q_s16(coeff + 192);
const int16x8_t b0 = vhaddq_s16(a0, a1);
const int16x8_t b1 = vhsubq_s16(a0, a1);
const int16x8_t b2 = vhaddq_s16(a2, a3);
const int16x8_t b3 = vhsubq_s16(a2, a3);
const int16x8_t c0 = vaddq_s16(b0, b2);
const int16x8_t c1 = vaddq_s16(b1, b3);
const int16x8_t c2 = vsubq_s16(b0, b2);
const int16x8_t c3 = vsubq_s16(b1, b3);
vst1q_s16(coeff + 0, c0);
vst1q_s16(coeff + 64, c1);
vst1q_s16(coeff + 128, c2);
vst1q_s16(coeff + 192, c3);
coeff += 8;
}
}
...@@ -1020,7 +1020,7 @@ if ((vpx_config("CONFIG_VP9_ENCODER") eq "yes") || (vpx_config("CONFIG_VP10_ENCO ...@@ -1020,7 +1020,7 @@ if ((vpx_config("CONFIG_VP9_ENCODER") eq "yes") || (vpx_config("CONFIG_VP10_ENCO
specialize qw/vpx_hadamard_8x8 sse2 neon/, "$ssse3_x86_64_x86inc"; specialize qw/vpx_hadamard_8x8 sse2 neon/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, int src_stride, int16_t *coeff"; add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, int src_stride, int16_t *coeff";
specialize qw/vpx_hadamard_16x16 sse2/; specialize qw/vpx_hadamard_16x16 sse2 neon/;
add_proto qw/int vpx_satd/, "const int16_t *coeff, int length"; add_proto qw/int vpx_satd/, "const int16_t *coeff, int length";
specialize qw/vpx_satd sse2 neon/; specialize qw/vpx_satd sse2 neon/;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment