Add SSE2 variant of hbd quantize_b_64x64_adaptive
Added SSE2 variant for 'aom_highbd_quantize_b_64x64_adaptive_c'
Module gains improved by factor of ~2.8x w.r.t C module
Change-Id: I4a0c2f6610f5bd6e6b0646306d618de849ea4022
diff --git a/aom_dsp/aom_dsp.cmake b/aom_dsp/aom_dsp.cmake
index 6cfbd32..449f30c 100644
--- a/aom_dsp/aom_dsp.cmake
+++ b/aom_dsp/aom_dsp.cmake
@@ -194,6 +194,7 @@
"${AOM_ROOT}/aom_dsp/x86/highbd_variance_sse2.c"
"${AOM_ROOT}/aom_dsp/x86/quantize_sse2.c"
"${AOM_ROOT}/aom_dsp/x86/adaptive_quantize_sse2.c"
+ "${AOM_ROOT}/aom_dsp/x86/highbd_adaptive_quantize_sse2.c"
"${AOM_ROOT}/aom_dsp/x86/quantize_x86.h"
"${AOM_ROOT}/aom_dsp/x86/sum_squares_sse2.c"
"${AOM_ROOT}/aom_dsp/x86/variance_sse2.c")
diff --git a/aom_dsp/aom_dsp_rtcd_defs.pl b/aom_dsp/aom_dsp_rtcd_defs.pl
index 8cffcf3..bddc6e0 100755
--- a/aom_dsp/aom_dsp_rtcd_defs.pl
+++ b/aom_dsp/aom_dsp_rtcd_defs.pl
@@ -546,6 +546,9 @@
add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
specialize qw/aom_highbd_quantize_b_64x64 sse2/;
+
+ add_proto qw/void aom_highbd_quantize_b_64x64_adaptive/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+ specialize qw/aom_highbd_quantize_b_64x64_adaptive sse2/;
} # CONFIG_AV1_ENCODER
#
diff --git a/aom_dsp/x86/highbd_adaptive_quantize_sse2.c b/aom_dsp/x86/highbd_adaptive_quantize_sse2.c
new file mode 100644
index 0000000..269aa19
--- /dev/null
+++ b/aom_dsp/x86/highbd_adaptive_quantize_sse2.c
@@ -0,0 +1,308 @@
+/*
+ * Copyright (c) 2019, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <emmintrin.h>
+#include "config/aom_dsp_rtcd.h"
+
+#include "aom/aom_integer.h"
+#include "aom_dsp/x86/quantize_x86.h"
+#include "av1/encoder/av1_quantize.h"
+
+static INLINE __m128i highbd_invert_sign_64bit_sse2(__m128i a, __m128i sign) {
+ a = _mm_xor_si128(a, sign);
+ return _mm_sub_epi64(a, sign);
+}
+
+static INLINE void highbd_mul_shift_sse2(const __m128i *x, const __m128i *y,
+ __m128i *p, const int shift) {
+ __m128i sign = _mm_srai_epi32(*y, 31);
+ __m128i sign_lo = _mm_unpacklo_epi32(sign, sign);
+ __m128i sign_hi = _mm_unpackhi_epi32(sign, sign);
+ __m128i abs_y = invert_sign_32_sse2(*y, sign);
+ __m128i prod_lo = _mm_mul_epu32(*x, abs_y);
+ __m128i prod_hi = _mm_srli_epi64(*x, 32);
+ const __m128i mult_hi = _mm_srli_epi64(abs_y, 32);
+ prod_hi = _mm_mul_epu32(prod_hi, mult_hi);
+ prod_lo = highbd_invert_sign_64bit_sse2(prod_lo, sign_lo);
+ prod_hi = highbd_invert_sign_64bit_sse2(prod_hi, sign_hi);
+
+ prod_lo = _mm_srli_epi64(prod_lo, shift);
+ const __m128i mask = _mm_set_epi32(0, -1, 0, -1);
+ prod_lo = _mm_and_si128(prod_lo, mask);
+ prod_hi = _mm_srli_epi64(prod_hi, shift);
+
+ prod_hi = _mm_slli_epi64(prod_hi, 32);
+ *p = _mm_or_si128(prod_lo, prod_hi);
+}
+
+static INLINE void highbd_calculate_qcoeff(__m128i *coeff, const __m128i *round,
+ const __m128i *quant,
+ const __m128i *shift,
+ const int *log_scale) {
+ __m128i tmp, qcoeff;
+ qcoeff = _mm_add_epi32(*coeff, *round);
+ highbd_mul_shift_sse2(&qcoeff, quant, &tmp, 16);
+ qcoeff = _mm_add_epi32(tmp, qcoeff);
+ highbd_mul_shift_sse2(&qcoeff, shift, coeff, 16 - *log_scale);
+}
+
+static INLINE void highbd_update_mask1(__m128i *cmp_mask0,
+ const int16_t *iscan_ptr, int *is_found,
+ __m128i *mask) {
+ __m128i temp_mask = _mm_setzero_si128();
+ if (_mm_movemask_epi8(*cmp_mask0)) {
+ __m128i iscan0 = _mm_load_si128((const __m128i *)(iscan_ptr));
+ __m128i mask0 = _mm_and_si128(*cmp_mask0, iscan0);
+ temp_mask = mask0;
+ *is_found = 1;
+ }
+ *mask = _mm_max_epi16(temp_mask, *mask);
+}
+
+static INLINE void highbd_update_mask0(__m128i *qcoeff0, __m128i *qcoeff1,
+ __m128i *threshold,
+ const int16_t *iscan_ptr, int *is_found,
+ __m128i *mask) {
+ __m128i coeff[2], cmp_mask0, cmp_mask1;
+
+ coeff[0] = _mm_slli_epi32(*qcoeff0, AOM_QM_BITS);
+ cmp_mask0 = _mm_cmpgt_epi32(coeff[0], threshold[0]);
+ coeff[1] = _mm_slli_epi32(*qcoeff1, AOM_QM_BITS);
+ cmp_mask1 = _mm_cmpgt_epi32(coeff[1], threshold[1]);
+
+ cmp_mask0 = _mm_packs_epi32(cmp_mask0, cmp_mask1);
+
+ highbd_update_mask1(&cmp_mask0, iscan_ptr, is_found, mask);
+}
+
+static INLINE __m128i highbd_calculate_dqcoeff(__m128i qcoeff, __m128i dequant,
+ const int log_scale) {
+ __m128i coeff_sign = _mm_srai_epi32(qcoeff, 31);
+ __m128i abs_coeff = invert_sign_32_sse2(qcoeff, coeff_sign);
+ highbd_mul_shift_sse2(&abs_coeff, &dequant, &abs_coeff, log_scale);
+ return invert_sign_32_sse2(abs_coeff, coeff_sign);
+}
+
+void aom_highbd_quantize_b_64x64_adaptive_sse2(
+ const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr,
+ const int16_t *round_ptr, const int16_t *quant_ptr,
+ const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr,
+ const int16_t *scan, const int16_t *iscan) {
+ int index = 8;
+ const int log_scale = 2;
+ int non_zero_count = 0;
+ int non_zero_count_prescan_add_zero = 0;
+ int is_found0 = 0, is_found1 = 0;
+ int eob = -1;
+ const __m128i zero = _mm_setzero_si128();
+ const __m128i one = _mm_set1_epi32(1);
+ const __m128i log_scale_vec = _mm_set1_epi32(log_scale);
+ __m128i zbin, round, quant, dequant, shift;
+ __m128i coeff0, coeff1, coeff0_sign, coeff1_sign;
+ __m128i qcoeff0, qcoeff1;
+ __m128i cmp_mask0, cmp_mask1, cmp_mask;
+ __m128i all_zero;
+ __m128i mask0 = zero, mask1 = zero;
+
+ const int zbins[2] = { ROUND_POWER_OF_TWO(zbin_ptr[0], log_scale),
+ ROUND_POWER_OF_TWO(zbin_ptr[1], log_scale) };
+ int prescan_add[2];
+ int thresh[4];
+ const qm_val_t wt = (1 << AOM_QM_BITS);
+ for (int i = 0; i < 2; ++i) {
+ prescan_add[i] = ROUND_POWER_OF_TWO(dequant_ptr[i] * EOB_FACTOR, 7);
+ thresh[i] = (zbins[i] * wt + prescan_add[i]) - 1;
+ }
+ thresh[2] = thresh[3] = thresh[1];
+ __m128i threshold[2];
+ threshold[0] = _mm_loadu_si128((__m128i *)&thresh[0]);
+ threshold[1] = _mm_unpackhi_epi64(threshold[0], threshold[0]);
+
+#if SKIP_EOB_FACTOR_ADJUST
+ int first = -1;
+#endif
+ // Setup global values.
+ zbin = _mm_load_si128((const __m128i *)zbin_ptr);
+ round = _mm_load_si128((const __m128i *)round_ptr);
+ quant = _mm_load_si128((const __m128i *)quant_ptr);
+ dequant = _mm_load_si128((const __m128i *)dequant_ptr);
+ shift = _mm_load_si128((const __m128i *)quant_shift_ptr);
+
+ __m128i zbin_sign = _mm_srai_epi16(zbin, 15);
+ __m128i round_sign = _mm_srai_epi16(round, 15);
+ __m128i quant_sign = _mm_srai_epi16(quant, 15);
+ __m128i dequant_sign = _mm_srai_epi16(dequant, 15);
+ __m128i shift_sign = _mm_srai_epi16(shift, 15);
+
+ zbin = _mm_unpacklo_epi16(zbin, zbin_sign);
+ round = _mm_unpacklo_epi16(round, round_sign);
+ quant = _mm_unpacklo_epi16(quant, quant_sign);
+ dequant = _mm_unpacklo_epi16(dequant, dequant_sign);
+ shift = _mm_unpacklo_epi16(shift, shift_sign);
+
+ // Shift with rounding.
+ zbin = _mm_add_epi32(zbin, log_scale_vec);
+ round = _mm_add_epi32(round, log_scale_vec);
+ zbin = _mm_srli_epi32(zbin, log_scale);
+ round = _mm_srli_epi32(round, log_scale);
+ zbin = _mm_sub_epi32(zbin, one);
+
+ // Do DC and first 15 AC.
+ coeff0 = _mm_load_si128((__m128i *)(coeff_ptr));
+ coeff1 = _mm_load_si128((__m128i *)(coeff_ptr + 4));
+
+ coeff0_sign = _mm_srai_epi32(coeff0, 31);
+ coeff1_sign = _mm_srai_epi32(coeff1, 31);
+ qcoeff0 = invert_sign_32_sse2(coeff0, coeff0_sign);
+ qcoeff1 = invert_sign_32_sse2(coeff1, coeff1_sign);
+
+ highbd_update_mask0(&qcoeff0, &qcoeff1, threshold, iscan, &is_found0, &mask0);
+
+ cmp_mask0 = _mm_cmpgt_epi32(qcoeff0, zbin);
+ zbin = _mm_unpackhi_epi64(zbin, zbin); // Switch DC to AC
+ cmp_mask1 = _mm_cmpgt_epi32(qcoeff1, zbin);
+ cmp_mask = _mm_packs_epi32(cmp_mask0, cmp_mask1);
+ highbd_update_mask1(&cmp_mask, iscan, &is_found1, &mask1);
+
+ threshold[0] = threshold[1];
+ all_zero = _mm_or_si128(cmp_mask0, cmp_mask1);
+ if (_mm_movemask_epi8(all_zero) == 0) {
+ _mm_store_si128((__m128i *)(qcoeff_ptr), zero);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + 4), zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr), zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + 4), zero);
+
+ round = _mm_unpackhi_epi64(round, round);
+ quant = _mm_unpackhi_epi64(quant, quant);
+ shift = _mm_unpackhi_epi64(shift, shift);
+ dequant = _mm_unpackhi_epi64(dequant, dequant);
+ } else {
+ highbd_calculate_qcoeff(&qcoeff0, &round, &quant, &shift, &log_scale);
+
+ round = _mm_unpackhi_epi64(round, round);
+ quant = _mm_unpackhi_epi64(quant, quant);
+ shift = _mm_unpackhi_epi64(shift, shift);
+ highbd_calculate_qcoeff(&qcoeff1, &round, &quant, &shift, &log_scale);
+
+ // Reinsert signs
+ qcoeff0 = invert_sign_32_sse2(qcoeff0, coeff0_sign);
+ qcoeff1 = invert_sign_32_sse2(qcoeff1, coeff1_sign);
+
+ // Mask out zbin threshold coeffs
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i *)(qcoeff_ptr), qcoeff0);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + 4), qcoeff1);
+
+ coeff0 = highbd_calculate_dqcoeff(qcoeff0, dequant, log_scale);
+ dequant = _mm_unpackhi_epi64(dequant, dequant);
+ coeff1 = highbd_calculate_dqcoeff(qcoeff1, dequant, log_scale);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr), coeff0);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + 4), coeff1);
+ }
+
+ // AC only loop.
+ while (index < n_coeffs) {
+ coeff0 = _mm_load_si128((__m128i *)(coeff_ptr + index));
+ coeff1 = _mm_load_si128((__m128i *)(coeff_ptr + index + 4));
+
+ coeff0_sign = _mm_srai_epi32(coeff0, 31);
+ coeff1_sign = _mm_srai_epi32(coeff1, 31);
+ qcoeff0 = invert_sign_32_sse2(coeff0, coeff0_sign);
+ qcoeff1 = invert_sign_32_sse2(coeff1, coeff1_sign);
+
+ highbd_update_mask0(&qcoeff0, &qcoeff1, threshold, iscan + index,
+ &is_found0, &mask0);
+
+ cmp_mask0 = _mm_cmpgt_epi32(qcoeff0, zbin);
+ cmp_mask1 = _mm_cmpgt_epi32(qcoeff1, zbin);
+ cmp_mask = _mm_packs_epi32(cmp_mask0, cmp_mask1);
+ highbd_update_mask1(&cmp_mask, iscan + index, &is_found1, &mask1);
+
+ all_zero = _mm_or_si128(cmp_mask0, cmp_mask1);
+ if (_mm_movemask_epi8(all_zero) == 0) {
+ _mm_store_si128((__m128i *)(qcoeff_ptr + index), zero);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + index + 4), zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + index), zero);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + index + 4), zero);
+ index += 8;
+ continue;
+ }
+ highbd_calculate_qcoeff(&qcoeff0, &round, &quant, &shift, &log_scale);
+ highbd_calculate_qcoeff(&qcoeff1, &round, &quant, &shift, &log_scale);
+
+ qcoeff0 = invert_sign_32_sse2(qcoeff0, coeff0_sign);
+ qcoeff1 = invert_sign_32_sse2(qcoeff1, coeff1_sign);
+
+ qcoeff0 = _mm_and_si128(qcoeff0, cmp_mask0);
+ qcoeff1 = _mm_and_si128(qcoeff1, cmp_mask1);
+
+ _mm_store_si128((__m128i *)(qcoeff_ptr + index), qcoeff0);
+ _mm_store_si128((__m128i *)(qcoeff_ptr + index + 4), qcoeff1);
+
+ coeff0 = highbd_calculate_dqcoeff(qcoeff0, dequant, log_scale);
+ coeff1 = highbd_calculate_dqcoeff(qcoeff1, dequant, log_scale);
+
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + index), coeff0);
+ _mm_store_si128((__m128i *)(dqcoeff_ptr + index + 4), coeff1);
+
+ index += 8;
+ }
+ if (is_found0) non_zero_count = calculate_non_zero_count(mask0);
+ if (is_found1)
+ non_zero_count_prescan_add_zero = calculate_non_zero_count(mask1);
+
+ for (int i = non_zero_count_prescan_add_zero - 1; i >= non_zero_count; i--) {
+ const int rc = scan[i];
+ qcoeff_ptr[rc] = 0;
+ dqcoeff_ptr[rc] = 0;
+ }
+
+ for (int i = non_zero_count - 1; i >= 0; i--) {
+ const int rc = scan[i];
+ if (qcoeff_ptr[rc]) {
+ eob = i;
+ break;
+ }
+ }
+
+ *eob_ptr = eob + 1;
+#if SKIP_EOB_FACTOR_ADJUST
+ // TODO(Aniket): Experiment the following loop with intrinsic by combining
+ // with the quantization loop above
+ for (int i = 0; i < non_zero_count; i++) {
+ const int rc = scan[i];
+ const int qcoeff = qcoeff_ptr[rc];
+ if (qcoeff) {
+ first = i;
+ break;
+ }
+ }
+ if ((*eob_ptr - 1) >= 0 && first == (*eob_ptr - 1)) {
+ const int rc = scan[(*eob_ptr - 1)];
+ if (qcoeff_ptr[rc] == 1 || qcoeff_ptr[rc] == -1) {
+ const int coeff = coeff_ptr[rc] * wt;
+ const int coeff_sign = (coeff >> 31);
+ const int abs_coeff = (coeff ^ coeff_sign) - coeff_sign;
+ const int factor = EOB_FACTOR + SKIP_EOB_FACTOR_ADJUST;
+ const int prescan_add_val =
+ ROUND_POWER_OF_TWO(dequant_ptr[rc != 0] * factor, 7);
+ if (abs_coeff < (zbins[rc != 0] * (1 << AOM_QM_BITS) + prescan_add_val)) {
+ qcoeff_ptr[rc] = 0;
+ dqcoeff_ptr[rc] = 0;
+ *eob_ptr = 0;
+ }
+ }
+ }
+#endif
+}
diff --git a/av1/encoder/av1_quantize.c b/av1/encoder/av1_quantize.c
index 22c15e6..12ae05c 100644
--- a/av1/encoder/av1_quantize.c
+++ b/av1/encoder/av1_quantize.c
@@ -444,7 +444,7 @@
eob_ptr, sc->scan, sc->iscan);
break;
case 2:
- aom_highbd_quantize_b_64x64_adaptive_c(
+ aom_highbd_quantize_b_64x64_adaptive(
coeff_ptr, n_coeffs, p->zbin_QTX, p->round_QTX, p->quant_QTX,
p->quant_shift_QTX, qcoeff_ptr, dqcoeff_ptr, p->dequant_QTX,
eob_ptr, sc->scan, sc->iscan);
diff --git a/test/quantize_func_test.cc b/test/quantize_func_test.cc
index b2977f2..d0c660e 100644
--- a/test/quantize_func_test.cc
+++ b/test/quantize_func_test.cc
@@ -431,6 +431,15 @@
static_cast<TX_SIZE>(TX_64X64), TYPE_B, AOM_BITS_10),
make_tuple(&aom_highbd_quantize_b_64x64_c, &aom_highbd_quantize_b_64x64_sse2,
static_cast<TX_SIZE>(TX_64X64), TYPE_B, AOM_BITS_12),
+ make_tuple(&aom_highbd_quantize_b_64x64_adaptive_c,
+ &aom_highbd_quantize_b_64x64_adaptive_sse2,
+ static_cast<TX_SIZE>(TX_64X64), TYPE_B, AOM_BITS_8),
+ make_tuple(&aom_highbd_quantize_b_64x64_adaptive_c,
+ &aom_highbd_quantize_b_64x64_adaptive_sse2,
+ static_cast<TX_SIZE>(TX_64X64), TYPE_B, AOM_BITS_10),
+ make_tuple(&aom_highbd_quantize_b_64x64_adaptive_c,
+ &aom_highbd_quantize_b_64x64_adaptive_sse2,
+ static_cast<TX_SIZE>(TX_64X64), TYPE_B, AOM_BITS_12),
make_tuple(&aom_quantize_b_adaptive_c, &aom_quantize_b_adaptive_sse2,
static_cast<TX_SIZE>(TX_16X16), TYPE_B, AOM_BITS_8),
make_tuple(&aom_quantize_b_adaptive_c, &aom_quantize_b_adaptive_sse2,
