ref: b1434f31255c6857bb0f06abc942bbf07bb156d5
parent: 55a2abfe31b82af13a77c3dc89913774f667b898
author: Luc Trudeau <[email protected]>
date: Thu Jun 7 11:30:23 EDT 2018
VSX Version of vp9_quantize_fp Low bit depth version only. Passes the VP9QuantizeTest test suite. VP9QuantizeTest Speed Test (POWER8 Model 2.1) 4x4 C time = 86.3 ms (±0.7 ms), VSX time = 18.2 ms (±0.0 ms) [ 4.7x] 8x8 C time = 57.7 ms (±0.3 ms), VSX time = 7.6 ms (±0.0 ms) [ 7.6x] 16x16 C time = 50.7 ms (±0.1 ms), VSX time = 4.9 ms (±0.0 ms) [10.3x] Change-Id: Ic09bc786c57cc89bba14624064216b52996075eb
--- a/test/vp9_quantize_test.cc
+++ b/test/vp9_quantize_test.cc
@@ -562,13 +562,16 @@
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH
#if HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(VSX, VP9QuantizeTest,
- ::testing::Values(make_tuple(&vpx_quantize_b_vsx,
- &vpx_quantize_b_c,
- VPX_BITS_8, 16, false),
- make_tuple(&vpx_quantize_b_32x32_vsx,
- &vpx_quantize_b_32x32_c,
- VPX_BITS_8, 32, false)));
+INSTANTIATE_TEST_CASE_P(
+ VSX, VP9QuantizeTest,
+ ::testing::Values(make_tuple(&vpx_quantize_b_vsx, &vpx_quantize_b_c,
+ VPX_BITS_8, 16, false),
+ make_tuple(&vpx_quantize_b_32x32_vsx,
+ &vpx_quantize_b_32x32_c, VPX_BITS_8, 32,
+ false),
+ make_tuple(&QuantFPWrapper<vp9_quantize_fp_vsx>,
+ &QuantFPWrapper<vp9_quantize_fp_c>, VPX_BITS_8,
+ 16, true)));
#endif // HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH
// Only useful to compare "Speed" test results.
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -129,7 +129,7 @@
add_proto qw/int64_t vp9_block_error_fp/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, int block_size";
add_proto qw/void vp9_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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/vp9_quantize_fp neon sse2 avx2/, "$ssse3_x86_64";
+specialize qw/vp9_quantize_fp neon sse2 avx2 vsx/, "$ssse3_x86_64";
add_proto qw/void vp9_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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/vp9_quantize_fp_32x32 neon/, "$ssse3_x86_64";
--- /dev/null
+++ b/vp9/encoder/ppc/vp9_quantize_vsx.c
@@ -1,0 +1,144 @@
+/*
+ * Copyright (c) 2018 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "./vpx_config.h"
+
+#include "./vp9_rtcd.h"
+#include "vpx_dsp/ppc/types_vsx.h"
+
+// Multiply the packed 16-bit integers in a and b, producing intermediate 32-bit
+// integers, and return the high 16 bits of the intermediate integers.
+// (a * b) >> 16
+// Note: Because this is done in 2 operations, a and b cannot both be UINT16_MIN
+static INLINE int16x8_t vec_mulhi(int16x8_t a, int16x8_t b) {+ // madds does ((A * B) >> 15) + C, we need >> 16, so we perform an extra right
+ // shift.
+ return vec_sra(vec_madds(a, b, vec_zeros_s16), vec_ones_u16);
+}
+
+// Negate 16-bit integers in a when the corresponding signed 16-bit
+// integer in b is negative.
+static INLINE int16x8_t vec_sign(int16x8_t a, int16x8_t b) {+ const int16x8_t mask = vec_sra(b, vec_shift_sign_s16);
+ return vec_xor(vec_add(a, mask), mask);
+}
+
+// Compare packed 16-bit integers across a, and return the maximum value in
+// every element. Returns a vector containing the biggest value across vector a.
+static INLINE int16x8_t vec_max_across(int16x8_t a) {+ a = vec_max(a, vec_perm(a, a, vec_perm64));
+ a = vec_max(a, vec_perm(a, a, vec_perm32));
+ return vec_max(a, vec_perm(a, a, vec_perm16));
+}
+
+void vp9_quantize_fp_vsx(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
+ int skip_block, const int16_t *round_ptr,
+ const int16_t *quant_ptr, tran_low_t *qcoeff_ptr,
+ tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr,
+ uint16_t *eob_ptr, const int16_t *scan_ptr,
+ const int16_t *iscan_ptr) {+ int16x8_t qcoeff0, qcoeff1, dqcoeff0, dqcoeff1, eob;
+ bool16x8_t zero_coeff0, zero_coeff1;
+
+ int16x8_t round = vec_vsx_ld(0, round_ptr);
+ int16x8_t quant = vec_vsx_ld(0, quant_ptr);
+ int16x8_t dequant = vec_vsx_ld(0, dequant_ptr);
+ int16x8_t coeff0 = vec_vsx_ld(0, coeff_ptr);
+ int16x8_t coeff1 = vec_vsx_ld(16, coeff_ptr);
+ int16x8_t scan0 = vec_vsx_ld(0, iscan_ptr);
+ int16x8_t scan1 = vec_vsx_ld(16, iscan_ptr);
+
+ (void)scan_ptr;
+ (void)skip_block;
+ assert(!skip_block);
+
+ // First set of 8 coeff starts with DC + 7 AC
+ qcoeff0 = vec_mulhi(vec_vaddshs(vec_abs(coeff0), round), quant);
+ zero_coeff0 = vec_cmpeq(qcoeff0, vec_zeros_s16);
+ qcoeff0 = vec_sign(qcoeff0, coeff0);
+ vec_vsx_st(qcoeff0, 0, qcoeff_ptr);
+
+ dqcoeff0 = vec_mladd(qcoeff0, dequant, vec_zeros_s16);
+ vec_vsx_st(dqcoeff0, 0, dqcoeff_ptr);
+
+ // Remove DC value from round and quant
+ round = vec_splat(round, 1);
+ quant = vec_splat(quant, 1);
+
+ // Remove DC value from dequant
+ dequant = vec_splat(dequant, 1);
+
+ // Second set of 8 coeff starts with (all AC)
+ qcoeff1 = vec_mulhi(vec_vaddshs(vec_abs(coeff1), round), quant);
+ zero_coeff1 = vec_cmpeq(qcoeff1, vec_zeros_s16);
+ qcoeff1 = vec_sign(qcoeff1, coeff1);
+ vec_vsx_st(qcoeff1, 16, qcoeff_ptr);
+
+ dqcoeff1 = vec_mladd(qcoeff1, dequant, vec_zeros_s16);
+ vec_vsx_st(dqcoeff1, 16, dqcoeff_ptr);
+
+ eob = vec_max(vec_or(scan0, zero_coeff0), vec_or(scan1, zero_coeff1));
+
+ // We quantize 16 coeff up front (enough for a 4x4) and process 24 coeff per
+ // loop iteration.
+ // for 8x8: 16 + 2 x 24 = 64
+ // for 16x16: 16 + 10 x 24 = 256
+ if (n_coeffs > 16) {+ int16x8_t coeff2, qcoeff2, dqcoeff2, eob2, scan2;
+ bool16x8_t zero_coeff2;
+
+ int index = 16;
+ int off0 = 32;
+ int off1 = 48;
+ int off2 = 64;
+
+ do {+ coeff0 = vec_vsx_ld(off0, coeff_ptr);
+ coeff1 = vec_vsx_ld(off1, coeff_ptr);
+ coeff2 = vec_vsx_ld(off2, coeff_ptr);
+ scan0 = vec_vsx_ld(off0, iscan_ptr);
+ scan1 = vec_vsx_ld(off1, iscan_ptr);
+ scan2 = vec_vsx_ld(off2, iscan_ptr);
+
+ qcoeff0 = vec_mulhi(vec_vaddshs(vec_abs(coeff0), round), quant);
+ zero_coeff0 = vec_cmpeq(qcoeff0, vec_zeros_s16);
+ qcoeff0 = vec_sign(qcoeff0, coeff0);
+ vec_vsx_st(qcoeff0, off0, qcoeff_ptr);
+ dqcoeff0 = vec_mladd(qcoeff0, dequant, vec_zeros_s16);
+ vec_vsx_st(dqcoeff0, off0, dqcoeff_ptr);
+
+ qcoeff1 = vec_mulhi(vec_vaddshs(vec_abs(coeff1), round), quant);
+ zero_coeff1 = vec_cmpeq(qcoeff1, vec_zeros_s16);
+ qcoeff1 = vec_sign(qcoeff1, coeff1);
+ vec_vsx_st(qcoeff1, off1, qcoeff_ptr);
+ dqcoeff1 = vec_mladd(qcoeff1, dequant, vec_zeros_s16);
+ vec_vsx_st(dqcoeff1, off1, dqcoeff_ptr);
+
+ qcoeff2 = vec_mulhi(vec_vaddshs(vec_abs(coeff2), round), quant);
+ zero_coeff2 = vec_cmpeq(qcoeff2, vec_zeros_s16);
+ qcoeff2 = vec_sign(qcoeff2, coeff2);
+ vec_vsx_st(qcoeff2, off2, qcoeff_ptr);
+ dqcoeff2 = vec_mladd(qcoeff2, dequant, vec_zeros_s16);
+ vec_vsx_st(dqcoeff2, off2, dqcoeff_ptr);
+
+ eob = vec_max(eob, vec_or(scan0, zero_coeff0));
+ eob2 = vec_max(vec_or(scan1, zero_coeff1), vec_or(scan2, zero_coeff2));
+ eob = vec_max(eob, eob2);
+
+ index += 24;
+ off0 += 48;
+ off1 += 48;
+ off2 += 48;
+ } while (index < n_coeffs);
+ }
+
+ eob = vec_max_across(eob);
+ *eob_ptr = eob[0] + 1;
+}
--- a/vp9/vp9cx.mk
+++ b/vp9/vp9cx.mk
@@ -140,6 +140,8 @@
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct16x16_msa.c
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct_msa.h
+VP9_CX_SRCS-$(HAVE_VSX) += encoder/ppc/vp9_quantize_vsx.c
+
# Strip unnecessary files with CONFIG_REALTIME_ONLY
VP9_CX_SRCS_REMOVE-$(CONFIG_REALTIME_ONLY) += encoder/vp9_firstpass.c
VP9_CX_SRCS_REMOVE-$(CONFIG_REALTIME_ONLY) += encoder/vp9_mbgraph.c