Redesign AVX2 2d interpolation filter
Allow flexible support to handle various filter taps in the vertical
and horizontal directions respectively.
Change-Id: Ic260cf7c29a6f464f4afbcbca78c33bd19a2a789
diff --git a/aom_dsp/x86/convolve_avx2.h b/aom_dsp/x86/convolve_avx2.h
index db0f22a..bde5557 100644
--- a/aom_dsp/x86/convolve_avx2.h
+++ b/aom_dsp/x86/convolve_avx2.h
@@ -59,6 +59,80 @@
6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14
};
+#define CONVOLVE_SR_HORIZONTAL_FILTER_4TAP \
+ for (i = 0; i < (im_h - 2); i += 2) { \
+ __m256i data = _mm256_castsi128_si256( \
+ _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + j])); \
+ data = _mm256_inserti128_si256( \
+ data, \
+ _mm_loadu_si128( \
+ (__m128i *)&src_ptr[(i * src_stride) + j + src_stride]), \
+ 1); \
+ __m256i res = convolve_lowbd_x_4tap(data, coeffs_h + 1, filt); \
+ res = \
+ _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h), round_shift_h); \
+ _mm256_store_si256((__m256i *)&im_block[i * im_stride], res); \
+ } \
+ __m256i data_1 = _mm256_castsi128_si256( \
+ _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + j])); \
+ __m256i res = convolve_lowbd_x_4tap(data_1, coeffs_h + 1, filt); \
+ res = _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h), round_shift_h); \
+ _mm256_store_si256((__m256i *)&im_block[i * im_stride], res);
+
+#define CONVOLVE_SR_VERTICAL_FILTER_4TAP \
+ __m256i s[6]; \
+ __m256i src_0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride)); \
+ __m256i src_1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride)); \
+ __m256i src_2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride)); \
+ __m256i src_3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride)); \
+ \
+ s[0] = _mm256_unpacklo_epi16(src_0, src_1); \
+ s[1] = _mm256_unpacklo_epi16(src_2, src_3); \
+ s[3] = _mm256_unpackhi_epi16(src_0, src_1); \
+ s[4] = _mm256_unpackhi_epi16(src_2, src_3); \
+ \
+ for (i = 0; i < h; i += 2) { \
+ const int16_t *data = &im_block[i * im_stride]; \
+ const __m256i s4 = _mm256_loadu_si256((__m256i *)(data + 4 * im_stride)); \
+ const __m256i s5 = _mm256_loadu_si256((__m256i *)(data + 5 * im_stride)); \
+ s[2] = _mm256_unpacklo_epi16(s4, s5); \
+ s[5] = _mm256_unpackhi_epi16(s4, s5); \
+ \
+ __m256i res_a = convolve_4tap(s, coeffs_v + 1); \
+ __m256i res_b = convolve_4tap(s + 3, coeffs_v + 1); \
+ \
+ res_a = \
+ _mm256_sra_epi32(_mm256_add_epi32(res_a, sum_round_v), sum_shift_v); \
+ res_b = \
+ _mm256_sra_epi32(_mm256_add_epi32(res_b, sum_round_v), sum_shift_v); \
+ const __m256i res_a_round = _mm256_sra_epi32( \
+ _mm256_add_epi32(res_a, round_const_v), round_shift_v); \
+ const __m256i res_b_round = _mm256_sra_epi32( \
+ _mm256_add_epi32(res_b, round_const_v), round_shift_v); \
+ const __m256i res_16bit = _mm256_packs_epi32(res_a_round, res_b_round); \
+ const __m256i res_8b = _mm256_packus_epi16(res_16bit, res_16bit); \
+ const __m128i res_0 = _mm256_castsi256_si128(res_8b); \
+ const __m128i res_1 = _mm256_extracti128_si256(res_8b, 1); \
+ \
+ __m128i *const p_0 = (__m128i *)&dst[i * dst_stride + j]; \
+ __m128i *const p_1 = (__m128i *)&dst[i * dst_stride + j + dst_stride]; \
+ if (w - j > 4) { \
+ _mm_storel_epi64(p_0, res_0); \
+ _mm_storel_epi64(p_1, res_1); \
+ } else if (w == 4) { \
+ xx_storel_32(p_0, res_0); \
+ xx_storel_32(p_1, res_1); \
+ } else { \
+ *(uint16_t *)p_0 = (uint16_t)_mm_cvtsi128_si32(res_0); \
+ *(uint16_t *)p_1 = (uint16_t)_mm_cvtsi128_si32(res_1); \
+ } \
+ \
+ s[0] = s[1]; \
+ s[1] = s[2]; \
+ s[3] = s[4]; \
+ s[4] = s[5]; \
+ }
+
#define CONVOLVE_SR_HORIZONTAL_FILTER_8TAP \
for (i = 0; i < (im_h - 2); i += 2) { \
__m256i data = _mm256_castsi128_si256( \
diff --git a/av1/common/x86/convolve_2d_avx2.c b/av1/common/x86/convolve_2d_avx2.c
index bfb57dd..4d4d450 100644
--- a/av1/common/x86/convolve_2d_avx2.c
+++ b/av1/common/x86/convolve_2d_avx2.c
@@ -27,8 +27,7 @@
const int subpel_x_qn, const int subpel_y_qn,
ConvolveParams *conv_params) {
const int bd = 8;
- int im_stride = 8;
- int i, is_horiz_4tap = 0, is_vert_4tap = 0;
+ int im_stride = 8, i;
DECLARE_ALIGNED(32, int16_t, im_block[(MAX_SB_SIZE + MAX_FILTER_TAP) * 8]);
const int bits =
FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1;
@@ -57,143 +56,37 @@
prepare_coeffs_lowbd(filter_params_x, subpel_x_qn, coeffs_h);
prepare_coeffs(filter_params_y, subpel_y_qn, coeffs_v);
- // Condition for checking valid horz_filt taps
- if (!(_mm256_extract_epi32(_mm256_or_si256(coeffs_h[0], coeffs_h[3]), 0)))
- is_horiz_4tap = 1;
+ const int16_t *const filter_x = av1_get_interp_filter_subpel_kernel(
+ filter_params_x, subpel_x_qn & SUBPEL_MASK);
+ const int16_t *const filter_y = av1_get_interp_filter_subpel_kernel(
+ filter_params_y, subpel_y_qn & SUBPEL_MASK);
- // Condition for checking valid vert_filt taps
- if (!(_mm256_extract_epi32(_mm256_or_si256(coeffs_v[0], coeffs_v[3]), 0)))
- is_vert_4tap = 1;
+ int horiz_tap = SUBPEL_TAPS;
+ int vert_tap = SUBPEL_TAPS;
- // horz_filt as 4 tap and vert_filt as 8 tap
- if (is_horiz_4tap) {
- int im_h = h + filter_params_y->taps - 1;
- const int fo_vert = filter_params_y->taps / 2 - 1;
- const int fo_horiz = 1;
- const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz;
+ if (!(filter_x[0] | filter_x[1] | filter_x[6] | filter_x[7])) horiz_tap = 4;
- // horz-filter
- for (int j = 0; j < w; j += 8) {
- for (i = 0; i < (im_h - 2); i += 2) {
- __m256i data = _mm256_castsi128_si256(
- _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + j]));
+ if (!(filter_y[0] | filter_y[1] | filter_y[6] | filter_y[7])) vert_tap = 4;
- // Load the next line
- data = _mm256_inserti128_si256(
- data,
- _mm_loadu_si128(
- (__m128i *)&src_ptr[(i * src_stride) + j + src_stride]),
- 1);
- __m256i res = convolve_lowbd_x_4tap(data, coeffs_h + 1, filt);
+ int im_h = h + vert_tap - 1;
+ const int fo_vert = vert_tap / 2 - 1;
+ const int fo_horiz = horiz_tap / 2 - 1;
+ const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz;
- res = _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h),
- round_shift_h);
- _mm256_store_si256((__m256i *)&im_block[i * im_stride], res);
- }
+ filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2));
+ filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3));
- __m256i data_1 = _mm256_castsi128_si256(
- _mm_loadu_si128((__m128i *)&src_ptr[(i * src_stride) + j]));
-
- __m256i res = convolve_lowbd_x_4tap(data_1, coeffs_h + 1, filt);
- res =
- _mm256_sra_epi16(_mm256_add_epi16(res, round_const_h), round_shift_h);
- _mm256_store_si256((__m256i *)&im_block[i * im_stride], res);
-
- // vert filter
- CONVOLVE_SR_VERTICAL_FILTER_8TAP;
+ for (int j = 0; j < w; j += 8) {
+ if (horiz_tap == 4) {
+ CONVOLVE_SR_HORIZONTAL_FILTER_4TAP
+ } else {
+ CONVOLVE_SR_HORIZONTAL_FILTER_8TAP
}
- } else if (is_vert_4tap) {
- int im_h = h + 3;
- const int fo_vert = 1;
- const int fo_horiz = filter_params_x->taps / 2 - 1;
- const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz;
- filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2));
- filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3));
-
- for (int j = 0; j < w; j += 8) {
- // horz_filter
- CONVOLVE_SR_HORIZONTAL_FILTER_8TAP;
- // vert_filter
- __m256i s[6];
- __m256i src_0 = _mm256_loadu_si256((__m256i *)(im_block + 0 * im_stride));
- __m256i src_1 = _mm256_loadu_si256((__m256i *)(im_block + 1 * im_stride));
- __m256i src_2 = _mm256_loadu_si256((__m256i *)(im_block + 2 * im_stride));
- __m256i src_3 = _mm256_loadu_si256((__m256i *)(im_block + 3 * im_stride));
-
- s[0] = _mm256_unpacklo_epi16(src_0, src_1);
- s[1] = _mm256_unpacklo_epi16(src_2, src_3);
- s[3] = _mm256_unpackhi_epi16(src_0, src_1);
- s[4] = _mm256_unpackhi_epi16(src_2, src_3);
-
- for (i = 0; i < h; i += 2) {
- const int16_t *data = &im_block[i * im_stride];
-
- const __m256i s4 =
- _mm256_loadu_si256((__m256i *)(data + 4 * im_stride));
- const __m256i s5 =
- _mm256_loadu_si256((__m256i *)(data + 5 * im_stride));
-
- s[2] = _mm256_unpacklo_epi16(s4, s5);
- s[5] = _mm256_unpackhi_epi16(s4, s5);
-
- __m256i res_a = convolve_4tap(s, coeffs_v + 1);
- __m256i res_b = convolve_4tap(s + 3, coeffs_v + 1);
-
- // Combine V round and 2F-H-V round into a single rounding
- res_a =
- _mm256_sra_epi32(_mm256_add_epi32(res_a, sum_round_v), sum_shift_v);
- res_b =
- _mm256_sra_epi32(_mm256_add_epi32(res_b, sum_round_v), sum_shift_v);
-
- const __m256i res_a_round = _mm256_sra_epi32(
- _mm256_add_epi32(res_a, round_const_v), round_shift_v);
- const __m256i res_b_round = _mm256_sra_epi32(
- _mm256_add_epi32(res_b, round_const_v), round_shift_v);
-
- /* rounding code */
- // 16 bit conversion
- const __m256i res_16bit = _mm256_packs_epi32(res_a_round, res_b_round);
- // 8 bit conversion and saturation to uint8
- const __m256i res_8b = _mm256_packus_epi16(res_16bit, res_16bit);
-
- const __m128i res_0 = _mm256_castsi256_si128(res_8b);
- const __m128i res_1 = _mm256_extracti128_si256(res_8b, 1);
-
- // Store values into the destination buffer
- __m128i *const p_0 = (__m128i *)&dst[i * dst_stride + j];
- __m128i *const p_1 = (__m128i *)&dst[i * dst_stride + j + dst_stride];
- if (w - j > 4) {
- _mm_storel_epi64(p_0, res_0);
- _mm_storel_epi64(p_1, res_1);
- } else if (w == 4) {
- xx_storel_32(p_0, res_0);
- xx_storel_32(p_1, res_1);
- } else {
- *(uint16_t *)p_0 = _mm_cvtsi128_si32(res_0);
- *(uint16_t *)p_1 = _mm_cvtsi128_si32(res_1);
- }
-
- s[0] = s[1];
- s[1] = s[2];
- s[3] = s[4];
- s[4] = s[5];
- }
- }
- } else {
- int j;
- int im_h = h + filter_params_y->taps - 1;
- const int fo_vert = filter_params_y->taps / 2 - 1;
- const int fo_horiz = filter_params_x->taps / 2 - 1;
- const uint8_t *const src_ptr = src - fo_vert * src_stride - fo_horiz;
-
- filt[2] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 2));
- filt[3] = _mm256_load_si256((__m256i const *)(filt_global_avx2 + 32 * 3));
-
- for (j = 0; j < w; j += 8) {
- CONVOLVE_SR_HORIZONTAL_FILTER_8TAP;
-
- CONVOLVE_SR_VERTICAL_FILTER_8TAP;
+ if (vert_tap == 4) {
+ CONVOLVE_SR_VERTICAL_FILTER_4TAP
+ } else {
+ CONVOLVE_SR_VERTICAL_FILTER_8TAP
}
}
}
