Commit bbf6186e authored by Yi Luo's avatar Yi Luo

Highbd intrapred DC_LEFT/TOP/128 sse2 optimization

Also extend intra pred speed test to rectangular block.
Speedup (i7-6700)
predictor      sse2 v. C
left 4x4       ~5.6x
top  4x4       ~7.2x
128  4x4       ~6.9x
left 4x8       ~7.7x
top  4x8       ~10.1x
128  4x8       ~10.0x

left 8x4       ~8.1x
top  8x4       ~9.1x
128  8x4       ~10.1x
left 8x8       ~10.3x
top  8x8       ~13.6x
128  8x8       ~14.8x
left 8x16      ~12.6x
top  8x16      ~14.0x
128  8x16      ~15.5x

left 16x8      ~6.3x
top  16x8      ~7.0x
128  16x8      ~6.5x
left 16x16     ~6.5x
top  16x16     ~7.1x
128  16x16     ~8.2x
left 16x32     ~5.1x
top  16x32     ~6.4x
128  16x32     ~5.6x

left 32x16     ~4.2x
top  32x16     ~4.3x
128  32x16     ~4.5x
left 32x32     ~3.8x
top  32x32     ~3.7x
128  32x32     ~3.9x

Change-Id: Ie7fcc85b9ded3030ee904623c40e9edeec1695ae
parent 94a504af
......@@ -133,6 +133,36 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
specialize qw/aom_highbd_h_predictor_16x32 sse2/;
specialize qw/aom_highbd_h_predictor_32x16 sse2/;
specialize qw/aom_highbd_h_predictor_32x32 sse2/;
specialize qw/aom_highbd_dc_left_predictor_4x4 sse2/;
specialize qw/aom_highbd_dc_top_predictor_4x4 sse2/;
specialize qw/aom_highbd_dc_128_predictor_4x4 sse2/;
specialize qw/aom_highbd_dc_left_predictor_4x8 sse2/;
specialize qw/aom_highbd_dc_top_predictor_4x8 sse2/;
specialize qw/aom_highbd_dc_128_predictor_4x8 sse2/;
specialize qw/aom_highbd_dc_left_predictor_8x4 sse2/;
specialize qw/aom_highbd_dc_top_predictor_8x4 sse2/;
specialize qw/aom_highbd_dc_128_predictor_8x4 sse2/;
specialize qw/aom_highbd_dc_left_predictor_8x8 sse2/;
specialize qw/aom_highbd_dc_top_predictor_8x8 sse2/;
specialize qw/aom_highbd_dc_128_predictor_8x8 sse2/;
specialize qw/aom_highbd_dc_left_predictor_8x16 sse2/;
specialize qw/aom_highbd_dc_top_predictor_8x16 sse2/;
specialize qw/aom_highbd_dc_128_predictor_8x16 sse2/;
specialize qw/aom_highbd_dc_left_predictor_16x8 sse2/;
specialize qw/aom_highbd_dc_top_predictor_16x8 sse2/;
specialize qw/aom_highbd_dc_128_predictor_16x8 sse2/;
specialize qw/aom_highbd_dc_left_predictor_16x16 sse2/;
specialize qw/aom_highbd_dc_top_predictor_16x16 sse2/;
specialize qw/aom_highbd_dc_128_predictor_16x16 sse2/;
specialize qw/aom_highbd_dc_left_predictor_16x32 sse2/;
specialize qw/aom_highbd_dc_top_predictor_16x32 sse2/;
specialize qw/aom_highbd_dc_128_predictor_16x32 sse2/;
specialize qw/aom_highbd_dc_left_predictor_32x16 sse2/;
specialize qw/aom_highbd_dc_top_predictor_32x16 sse2/;
specialize qw/aom_highbd_dc_128_predictor_32x16 sse2/;
specialize qw/aom_highbd_dc_left_predictor_32x32 sse2/;
specialize qw/aom_highbd_dc_top_predictor_32x32 sse2/;
specialize qw/aom_highbd_dc_128_predictor_32x32 sse2/;
} # CONFIG_HIGHBITDEPTH
#
......
......@@ -240,3 +240,444 @@ void aom_highbd_h_predictor_32x32_sse2(uint16_t *dst, ptrdiff_t stride,
dst += stride << 3;
}
}
// -----------------------------------------------------------------------------
// DC_TOP, DC_LEFT, DC_128
// 4x4
static INLINE __m128i dc_sum_4(const uint16_t *ref) {
const __m128i _dcba = _mm_loadl_epi64((const __m128i *)ref);
const __m128i _xxdc = _mm_shufflelo_epi16(_dcba, 0xe);
const __m128i a = _mm_add_epi16(_dcba, _xxdc);
return _mm_add_epi16(a, _mm_shufflelo_epi16(a, 0x1));
}
static INLINE void dc_store_4x4(uint16_t *dst, ptrdiff_t stride,
const __m128i *dc) {
const __m128i dc_dup = _mm_shufflelo_epi16(*dc, 0x0);
int i;
for (i = 0; i < 4; ++i, dst += stride) {
_mm_storel_epi64((__m128i *)dst, dc_dup);
}
}
void aom_highbd_dc_left_predictor_4x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i two = _mm_cvtsi32_si128(2);
const __m128i sum = dc_sum_4(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, two), 2);
(void)above;
(void)bd;
dc_store_4x4(dst, stride, &dc);
}
void aom_highbd_dc_top_predictor_4x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i two = _mm_cvtsi32_si128(2);
const __m128i sum = dc_sum_4(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, two), 2);
(void)left;
(void)bd;
dc_store_4x4(dst, stride, &dc);
}
void aom_highbd_dc_128_predictor_4x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_4x4(dst, stride, &dc_dup);
}
// -----------------------------------------------------------------------------
// 4x8
static INLINE void dc_store_4x8(uint16_t *dst, ptrdiff_t stride,
const __m128i *dc) {
const __m128i dc_dup = _mm_shufflelo_epi16(*dc, 0x0);
int i;
for (i = 0; i < 8; ++i, dst += stride) {
_mm_storel_epi64((__m128i *)dst, dc_dup);
}
}
// Shared with DC 8xh
static INLINE __m128i dc_sum_8(const uint16_t *ref) {
const __m128i ref_u16 = _mm_load_si128((const __m128i *)ref);
const __m128i _dcba = _mm_add_epi16(ref_u16, _mm_srli_si128(ref_u16, 8));
const __m128i _xxdc = _mm_shufflelo_epi16(_dcba, 0xe);
const __m128i a = _mm_add_epi16(_dcba, _xxdc);
return _mm_add_epi16(a, _mm_shufflelo_epi16(a, 0x1));
}
void aom_highbd_dc_left_predictor_4x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i sum = dc_sum_8(left);
const __m128i four = _mm_cvtsi32_si128(4);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, four), 3);
(void)above;
(void)bd;
dc_store_4x8(dst, stride, &dc);
}
void aom_highbd_dc_top_predictor_4x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i two = _mm_cvtsi32_si128(2);
const __m128i sum = dc_sum_4(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, two), 2);
(void)left;
(void)bd;
dc_store_4x8(dst, stride, &dc);
}
void aom_highbd_dc_128_predictor_4x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_4x8(dst, stride, &dc_dup);
}
// -----------------------------------------------------------------------------
// 8xh
static INLINE void dc_store_8xh(uint16_t *dst, ptrdiff_t stride, int height,
const __m128i *dc) {
const __m128i dc_dup_lo = _mm_shufflelo_epi16(*dc, 0);
const __m128i dc_dup = _mm_unpacklo_epi64(dc_dup_lo, dc_dup_lo);
int i;
for (i = 0; i < height; ++i, dst += stride) {
_mm_store_si128((__m128i *)dst, dc_dup);
}
}
// -----------------------------------------------------------------------------
// DC_TOP
static INLINE void dc_top_predictor_8xh(uint16_t *dst, ptrdiff_t stride,
int height, const uint16_t *above) {
const __m128i four = _mm_cvtsi32_si128(4);
const __m128i sum = dc_sum_8(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, four), 3);
dc_store_8xh(dst, stride, height, &dc);
}
void aom_highbd_dc_top_predictor_8x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)left;
(void)bd;
dc_top_predictor_8xh(dst, stride, 4, above);
}
void aom_highbd_dc_top_predictor_8x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)left;
(void)bd;
dc_top_predictor_8xh(dst, stride, 8, above);
}
void aom_highbd_dc_top_predictor_8x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)left;
(void)bd;
dc_top_predictor_8xh(dst, stride, 16, above);
}
// -----------------------------------------------------------------------------
// DC_LEFT
void aom_highbd_dc_left_predictor_8x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i two = _mm_cvtsi32_si128(2);
const __m128i sum = dc_sum_4(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, two), 2);
(void)above;
(void)bd;
dc_store_8xh(dst, stride, 4, &dc);
}
void aom_highbd_dc_left_predictor_8x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i four = _mm_cvtsi32_si128(4);
const __m128i sum = dc_sum_8(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, four), 3);
(void)above;
(void)bd;
dc_store_8xh(dst, stride, 8, &dc);
}
// Shared with DC 16xh
static INLINE __m128i dc_sum_16(const uint16_t *ref) {
const __m128i sum_lo = dc_sum_8(ref);
const __m128i sum_hi = dc_sum_8(ref + 8);
return _mm_add_epi16(sum_lo, sum_hi);
}
void aom_highbd_dc_left_predictor_8x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)above;
(void)bd;
dc_store_8xh(dst, stride, 16, &dc);
}
// -----------------------------------------------------------------------------
// DC_128
static INLINE void dc_128_predictor_8xh(uint16_t *dst, ptrdiff_t stride,
int height, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
dc_store_8xh(dst, stride, height, &dc_dup);
}
void aom_highbd_dc_128_predictor_8x4_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)above;
(void)left;
dc_128_predictor_8xh(dst, stride, 4, bd);
}
void aom_highbd_dc_128_predictor_8x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)above;
(void)left;
dc_128_predictor_8xh(dst, stride, 8, bd);
}
void aom_highbd_dc_128_predictor_8x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
(void)above;
(void)left;
dc_128_predictor_8xh(dst, stride, 16, bd);
}
// -----------------------------------------------------------------------------
// 16xh
static INLINE void dc_store_16xh(uint16_t *dst, ptrdiff_t stride, int height,
const __m128i *dc) {
const __m128i dc_dup_lo = _mm_shufflelo_epi16(*dc, 0);
const __m128i dc_dup = _mm_unpacklo_epi64(dc_dup_lo, dc_dup_lo);
int i;
for (i = 0; i < height; ++i, dst += stride) {
_mm_store_si128((__m128i *)dst, dc_dup);
_mm_store_si128((__m128i *)(dst + 8), dc_dup);
}
}
// -----------------------------------------------------------------------------
// DC_LEFT
void aom_highbd_dc_left_predictor_16x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i four = _mm_cvtsi32_si128(4);
const __m128i sum = dc_sum_8(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, four), 3);
(void)above;
(void)bd;
dc_store_16xh(dst, stride, 8, &dc);
}
void aom_highbd_dc_left_predictor_16x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)above;
(void)bd;
dc_store_16xh(dst, stride, 16, &dc);
}
// Shared with 32xh
static INLINE __m128i dc_sum_32(const uint16_t *ref) {
const __m128i zero = _mm_setzero_si128();
const __m128i sum_a = dc_sum_16(ref);
const __m128i sum_b = dc_sum_16(ref + 16);
// 12 bit bd will outrange, so expand to 32 bit before adding final total
return _mm_add_epi32(_mm_unpacklo_epi16(sum_a, zero),
_mm_unpacklo_epi16(sum_b, zero));
}
void aom_highbd_dc_left_predictor_16x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i sixteen = _mm_cvtsi32_si128(16);
const __m128i sum = dc_sum_32(left);
const __m128i dc = _mm_srli_epi32(_mm_add_epi32(sum, sixteen), 5);
(void)above;
(void)bd;
dc_store_16xh(dst, stride, 32, &dc);
}
// -----------------------------------------------------------------------------
// DC_TOP
void aom_highbd_dc_top_predictor_16x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)left;
(void)bd;
dc_store_16xh(dst, stride, 8, &dc);
}
void aom_highbd_dc_top_predictor_16x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)left;
(void)bd;
dc_store_16xh(dst, stride, 16, &dc);
}
void aom_highbd_dc_top_predictor_16x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(above);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)left;
(void)bd;
dc_store_16xh(dst, stride, 32, &dc);
}
// -----------------------------------------------------------------------------
// DC_128
void aom_highbd_dc_128_predictor_16x8_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_16xh(dst, stride, 8, &dc_dup);
}
void aom_highbd_dc_128_predictor_16x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_16xh(dst, stride, 16, &dc_dup);
}
void aom_highbd_dc_128_predictor_16x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_16xh(dst, stride, 32, &dc_dup);
}
// -----------------------------------------------------------------------------
// 32xh
static INLINE void dc_store_32xh(uint16_t *dst, ptrdiff_t stride, int height,
const __m128i *dc) {
const __m128i dc_dup_lo = _mm_shufflelo_epi16(*dc, 0);
const __m128i dc_dup = _mm_unpacklo_epi64(dc_dup_lo, dc_dup_lo);
int i;
for (i = 0; i < height; ++i, dst += stride) {
_mm_store_si128((__m128i *)dst, dc_dup);
_mm_store_si128((__m128i *)(dst + 8), dc_dup);
_mm_store_si128((__m128i *)(dst + 16), dc_dup);
_mm_store_si128((__m128i *)(dst + 24), dc_dup);
}
}
void aom_highbd_dc_left_predictor_32x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i eight = _mm_cvtsi32_si128(8);
const __m128i sum = dc_sum_16(left);
const __m128i dc = _mm_srli_epi16(_mm_add_epi16(sum, eight), 4);
(void)above;
(void)bd;
dc_store_32xh(dst, stride, 16, &dc);
}
void aom_highbd_dc_left_predictor_32x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i sixteen = _mm_cvtsi32_si128(16);
const __m128i sum = dc_sum_32(left);
const __m128i dc = _mm_srli_epi32(_mm_add_epi32(sum, sixteen), 5);
(void)above;
(void)bd;
dc_store_32xh(dst, stride, 32, &dc);
}
void aom_highbd_dc_top_predictor_32x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i sixteen = _mm_cvtsi32_si128(16);
const __m128i sum = dc_sum_32(above);
const __m128i dc = _mm_srli_epi32(_mm_add_epi32(sum, sixteen), 5);
(void)left;
(void)bd;
dc_store_32xh(dst, stride, 16, &dc);
}
void aom_highbd_dc_128_predictor_32x16_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_32xh(dst, stride, 16, &dc_dup);
}
void aom_highbd_dc_top_predictor_32x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i sixteen = _mm_cvtsi32_si128(16);
const __m128i sum = dc_sum_32(above);
const __m128i dc = _mm_srli_epi32(_mm_add_epi32(sum, sixteen), 5);
(void)left;
(void)bd;
dc_store_32xh(dst, stride, 32, &dc);
}
void aom_highbd_dc_128_predictor_32x32_sse2(uint16_t *dst, ptrdiff_t stride,
const uint16_t *above,
const uint16_t *left, int bd) {
const __m128i dc = _mm_cvtsi32_si128(1 << (bd - 1));
const __m128i dc_dup = _mm_shufflelo_epi16(dc, 0x0);
(void)above;
(void)left;
dc_store_32xh(dst, stride, 32, &dc_dup);
}
......@@ -136,45 +136,171 @@ TEST_P(AV1IntraPredTest, IntraPredTests) {
#if HAVE_SSE2
#if CONFIG_HIGHBITDEPTH
const IntraPredFunc IntraPredTestVector8[] = {
highbd_entry(dc, 4, 4, sse2, 8), highbd_entry(dc, 8, 8, sse2, 8),
highbd_entry(dc, 16, 16, sse2, 8), highbd_entry(dc, 32, 32, sse2, 8),
highbd_entry(v, 4, 4, sse2, 8), highbd_entry(v, 8, 8, sse2, 8),
highbd_entry(v, 16, 16, sse2, 8), highbd_entry(v, 32, 32, sse2, 8),
highbd_entry(h, 4, 4, sse2, 8), highbd_entry(h, 4, 8, sse2, 8),
highbd_entry(h, 8, 4, sse2, 8), highbd_entry(h, 8, 8, sse2, 8),
highbd_entry(h, 8, 16, sse2, 8), highbd_entry(h, 16, 8, sse2, 8),
highbd_entry(h, 16, 16, sse2, 8), highbd_entry(h, 16, 32, sse2, 8),
highbd_entry(h, 32, 16, sse2, 8), highbd_entry(h, 32, 32, sse2, 8),
highbd_entry(dc, 4, 4, sse2, 8),
highbd_entry(dc, 8, 8, sse2, 8),
highbd_entry(dc, 16, 16, sse2, 8),
highbd_entry(dc, 32, 32, sse2, 8),
highbd_entry(dc_left, 4, 4, sse2, 8),
highbd_entry(dc_left, 4, 8, sse2, 8),
highbd_entry(dc_top, 4, 4, sse2, 8),
highbd_entry(dc_top, 4, 8, sse2, 8),
highbd_entry(dc_128, 4, 4, sse2, 8),
highbd_entry(dc_128, 4, 8, sse2, 8),
highbd_entry(dc_left, 8, 4, sse2, 8),
highbd_entry(dc_top, 8, 4, sse2, 8),
highbd_entry(dc_128, 8, 4, sse2, 8),
highbd_entry(dc_left, 8, 8, sse2, 8),
highbd_entry(dc_top, 8, 8, sse2, 8),
highbd_entry(dc_128, 8, 8, sse2, 8),
highbd_entry(dc_left, 8, 16, sse2, 8),
highbd_entry(dc_top, 8, 16, sse2, 8),
highbd_entry(dc_128, 8, 16, sse2, 8),
highbd_entry(dc_left, 16, 8, sse2, 8),
highbd_entry(dc_top, 16, 8, sse2, 8),
highbd_entry(dc_128, 16, 8, sse2, 8),
highbd_entry(dc_left, 16, 16, sse2, 8),
highbd_entry(dc_top, 16, 16, sse2, 8),
highbd_entry(dc_128, 16, 16, sse2, 8),
highbd_entry(dc_left, 16, 32, sse2, 8),
highbd_entry(dc_top, 16, 32, sse2, 8),
highbd_entry(dc_128, 16, 32, sse2, 8),
highbd_entry(dc_left, 32, 16, sse2, 8),
highbd_entry(dc_top, 32, 16, sse2, 8),
highbd_entry(dc_128, 32, 16, sse2, 8),
highbd_entry(dc_left, 32, 32, sse2, 8),
highbd_entry(dc_top, 32, 32, sse2, 8),
highbd_entry(dc_128, 32, 32, sse2, 8),
highbd_entry(v, 4, 4, sse2, 8),
highbd_entry(v, 8, 8, sse2, 8),
highbd_entry(v, 16, 16, sse2, 8),
highbd_entry(v, 32, 32, sse2, 8),
highbd_entry(h, 4, 4, sse2, 8),
highbd_entry(h, 4, 8, sse2, 8),
highbd_entry(h, 8, 4, sse2, 8),
highbd_entry(h, 8, 8, sse2, 8),
highbd_entry(h, 8, 16, sse2, 8),
highbd_entry(h, 16, 8, sse2, 8),
highbd_entry(h, 16, 16, sse2, 8),
highbd_entry(h, 16, 32, sse2, 8),
highbd_entry(h, 32, 16, sse2, 8),
highbd_entry(h, 32, 32, sse2, 8),
};
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_8, AV1IntraPredTest,
::testing::ValuesIn(IntraPredTestVector8));
const IntraPredFunc IntraPredTestVector10[] = {
highbd_entry(dc, 4, 4, sse2, 10), highbd_entry(dc, 8, 8, sse2, 10),
highbd_entry(dc, 16, 16, sse2, 10), highbd_entry(dc, 32, 32, sse2, 10),
highbd_entry(v, 4, 4, sse2, 10), highbd_entry(v, 8, 8, sse2, 10),
highbd_entry(v, 16, 16, sse2, 10), highbd_entry(v, 32, 32, sse2, 10),
highbd_entry(h, 4, 4, sse2, 10), highbd_entry(h, 4, 8, sse2, 10),
highbd_entry(h, 8, 4, sse2, 10), highbd_entry(h, 8, 8, sse2, 10),
highbd_entry(h, 8, 16, sse2, 10), highbd_entry(h, 16, 8, sse2, 10),
highbd_entry(h, 16, 16, sse2, 10), highbd_entry(h, 16, 32, sse2, 10),
highbd_entry(h, 32, 16, sse2, 10), highbd_entry(h, 32, 32, sse2, 10),
highbd_entry(dc, 4, 4, sse2, 10),
highbd_entry(dc, 8, 8, sse2, 10),
highbd_entry(dc, 16, 16, sse2, 10),
highbd_entry(dc, 32, 32, sse2, 10),
highbd_entry(dc_left, 4, 4, sse2, 10),
highbd_entry(dc_left, 4, 8, sse2, 10),
highbd_entry(dc_top, 4, 4, sse2, 10),
highbd_entry(dc_top, 4, 8, sse2, 10),
highbd_entry(dc_128, 4, 4, sse2, 10),
highbd_entry(dc_128, 4, 8, sse2, 10),
highbd_entry(dc_left, 8, 4, sse2, 10),
highbd_entry(dc_top, 8, 4, sse2, 10),
highbd_entry(dc_128, 8, 4, sse2, 10),
highbd_entry(dc_left, 8, 8, sse2, 10),
highbd_entry(dc_top, 8, 8, sse2, 10),
highbd_entry(dc_128, 8, 8, sse2, 10),
highbd_entry(dc_left, 8, 16, sse2, 10),