[FFmpeg-devel] [PATCH 2/2] aacenc: add SIMD optimizations for abs_pow34 and quantization

Rostislav Pehlivanov atomnuker at gmail.com
Sat Oct 8 18:20:31 EEST 2016


Performance improvements:

quant_bands:
with:     681 decicycles in quant_bands, 8388453 runs,    155 skips
without: 1190 decicycles in quant_bands, 8388386 runs,    222 skips
Around 42% for the function

Twoloop coder:

abs_pow34:
with/without: 7.82s/8.17s
Around 4% for the entire encoder

Both:
with/without: 7.15s/8.17s
Around 12% for the entire encoder

Fast coder:

abs_pow34:
with/without: 3.40s/3.77s
Around 10% for the entire encoder

Both:
with/without: 3.02s/3.77s
Around 20% faster for the entire encoder

Signed-off-by: Rostislav Pehlivanov <atomnuker at gmail.com>
---
 libavcodec/aaccoder.c            |  22 ++++----
 libavcodec/aaccoder_trellis.h    |   2 +-
 libavcodec/aaccoder_twoloop.h    |   2 +-
 libavcodec/aacenc.c              |   4 ++
 libavcodec/aacenc.h              |   6 +++
 libavcodec/aacenc_is.c           |   6 +--
 libavcodec/aacenc_ltp.c          |   4 +-
 libavcodec/aacenc_pred.c         |   6 +--
 libavcodec/aacenc_quantization.h |   4 +-
 libavcodec/aacenc_utils.h        |   2 +-
 libavcodec/x86/Makefile          |   2 +
 libavcodec/x86/aacencdsp.asm     | 111 +++++++++++++++++++++++++++++++++++++++
 libavcodec/x86/aacencdsp_init.c  |  57 ++++++++++++++++++++
 13 files changed, 204 insertions(+), 24 deletions(-)
 create mode 100644 libavcodec/x86/aacencdsp.asm
 create mode 100644 libavcodec/x86/aacencdsp_init.c

diff --git a/libavcodec/aaccoder.c b/libavcodec/aaccoder.c
index 35787e8..6760a70 100644
--- a/libavcodec/aaccoder.c
+++ b/libavcodec/aaccoder.c
@@ -88,7 +88,7 @@ static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce
     float next_minrd = INFINITY;
     int next_mincb = 0;
 
-    abs_pow34_v(s->scoefs, sce->coeffs, 1024);
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
     start = win*128;
     for (cb = 0; cb < CB_TOT_ALL; cb++) {
         path[0][cb].cost     = 0.0f;
@@ -299,7 +299,7 @@ static void search_for_quantizers_anmr(AVCodecContext *avctx, AACEncContext *s,
         }
     }
     idx = 1;
-    abs_pow34_v(s->scoefs, sce->coeffs, 1024);
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
     for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
         start = w*128;
         for (g = 0; g < sce->ics.num_swb; g++) {
@@ -446,7 +446,7 @@ static void search_for_quantizers_fast(AVCodecContext *avctx, AACEncContext *s,
 
     if (!allz)
         return;
-    abs_pow34_v(s->scoefs, sce->coeffs, 1024);
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
     ff_quantize_band_cost_cache_init(s);
 
     for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
@@ -652,8 +652,8 @@ static void search_for_pns(AACEncContext *s, AVCodecContext *avctx, SingleChanne
                 s->fdsp->vector_fmul_scalar(PNS, PNS, scale, sce->ics.swb_sizes[g]);
                 pns_senergy = s->fdsp->scalarproduct_float(PNS, PNS, sce->ics.swb_sizes[g]);
                 pns_energy += pns_senergy;
-                abs_pow34_v(NOR34, &sce->coeffs[start_c], sce->ics.swb_sizes[g]);
-                abs_pow34_v(PNS34, PNS, sce->ics.swb_sizes[g]);
+                s->abs_pow34(NOR34, &sce->coeffs[start_c], sce->ics.swb_sizes[g]);
+                s->abs_pow34(PNS34, PNS, sce->ics.swb_sizes[g]);
                 dist1 += quantize_band_cost(s, &sce->coeffs[start_c],
                                             NOR34,
                                             sce->ics.swb_sizes[g],
@@ -789,8 +789,8 @@ static void search_for_ms(AACEncContext *s, ChannelElement *cpe)
                         S[i] =  M[i]
                               - sce1->coeffs[start+(w+w2)*128+i];
                     }
-                    abs_pow34_v(M34, M, sce0->ics.swb_sizes[g]);
-                    abs_pow34_v(S34, S, sce0->ics.swb_sizes[g]);
+                    s->abs_pow34(M34, M, sce0->ics.swb_sizes[g]);
+                    s->abs_pow34(S34, S, sce0->ics.swb_sizes[g]);
                     for (i = 0; i < sce0->ics.swb_sizes[g]; i++ ) {
                         Mmax = FFMAX(Mmax, M34[i]);
                         Smax = FFMAX(Smax, S34[i]);
@@ -833,10 +833,10 @@ static void search_for_ms(AACEncContext *s, ChannelElement *cpe)
                                   - sce1->coeffs[start+(w+w2)*128+i];
                         }
 
-                        abs_pow34_v(L34, sce0->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
-                        abs_pow34_v(R34, sce1->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
-                        abs_pow34_v(M34, M,                         sce0->ics.swb_sizes[g]);
-                        abs_pow34_v(S34, S,                         sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(L34, sce0->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(R34, sce1->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(M34, M,                         sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(S34, S,                         sce0->ics.swb_sizes[g]);
                         dist1 += quantize_band_cost(s, &sce0->coeffs[start + (w+w2)*128],
                                                     L34,
                                                     sce0->ics.swb_sizes[g],
diff --git a/libavcodec/aaccoder_trellis.h b/libavcodec/aaccoder_trellis.h
index 0230052..940ebf0 100644
--- a/libavcodec/aaccoder_trellis.h
+++ b/libavcodec/aaccoder_trellis.h
@@ -70,7 +70,7 @@ static void codebook_trellis_rate(AACEncContext *s, SingleChannelElement *sce,
     float next_minbits = INFINITY;
     int next_mincb = 0;
 
-    abs_pow34_v(s->scoefs, sce->coeffs, 1024);
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
     start = win*128;
     for (cb = 0; cb < CB_TOT_ALL; cb++) {
         path[0][cb].cost     = run_bits+4;
diff --git a/libavcodec/aaccoder_twoloop.h b/libavcodec/aaccoder_twoloop.h
index 42aea52..fb9849e 100644
--- a/libavcodec/aaccoder_twoloop.h
+++ b/libavcodec/aaccoder_twoloop.h
@@ -291,7 +291,7 @@ static void search_for_quantizers_twoloop(AVCodecContext *avctx,
 
     if (!allz)
         return;
-    abs_pow34_v(s->scoefs, sce->coeffs, 1024);
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
     ff_quantize_band_cost_cache_init(s);
 
     for (i = 0; i < sizeof(minsf) / sizeof(minsf[0]); ++i)
diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c
index ee3cbf8..f30691a 100644
--- a/libavcodec/aacenc.c
+++ b/libavcodec/aacenc.c
@@ -1033,6 +1033,10 @@ static av_cold int aac_encode_init(AVCodecContext *avctx)
     ff_lpc_init(&s->lpc, 2*avctx->frame_size, TNS_MAX_ORDER, FF_LPC_TYPE_LEVINSON);
     s->random_state = 0x1f2e3d4c;
 
+    s->abs_pow34   = &abs_pow34_v;
+    s->quant_bands = &quantize_bands;
+    ff_aac_dsp_init_x86(s);
+
     if (HAVE_MIPSDSP)
         ff_aac_coder_init_mips(s);
 
diff --git a/libavcodec/aacenc.h b/libavcodec/aacenc.h
index 1ace00d..f768af8 100644
--- a/libavcodec/aacenc.h
+++ b/libavcodec/aacenc.h
@@ -127,11 +127,17 @@ typedef struct AACEncContext {
     uint16_t quantize_band_cost_cache_generation;
     AACQuantizeBandCostCacheEntry quantize_band_cost_cache[256][128]; ///< memoization area for quantize_band_cost
 
+    void (*abs_pow34)(float *out, const float *in, const int64_t size);
+    void (*quant_bands)(int *out, const float *in, const float *scaled,
+                        int size, float Q34, int is_signed, int maxval,
+                        const float rounding);
+
     struct {
         float *samples;
     } buffer;
 } AACEncContext;
 
+void ff_aac_dsp_init_x86(AACEncContext *s);
 void ff_aac_coder_init_mips(AACEncContext *c);
 void ff_quantize_band_cost_cache_init(struct AACEncContext *s);
 
diff --git a/libavcodec/aacenc_is.c b/libavcodec/aacenc_is.c
index 473897b..2f5b7eb 100644
--- a/libavcodec/aacenc_is.c
+++ b/libavcodec/aacenc_is.c
@@ -59,9 +59,9 @@ struct AACISError ff_aac_is_encoding_err(AACEncContext *s, ChannelElement *cpe,
         float minthr = FFMIN(band0->threshold, band1->threshold);
         for (i = 0; i < sce0->ics.swb_sizes[g]; i++)
             IS[i] = (L[start+(w+w2)*128+i] + phase*R[start+(w+w2)*128+i])*sqrt(ener0/ener01);
-        abs_pow34_v(L34, &L[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
-        abs_pow34_v(R34, &R[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
-        abs_pow34_v(I34, IS,                   sce0->ics.swb_sizes[g]);
+        s->abs_pow34(L34, &L[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
+        s->abs_pow34(R34, &R[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
+        s->abs_pow34(I34, IS,                   sce0->ics.swb_sizes[g]);
         maxval = find_max_val(1, sce0->ics.swb_sizes[g], I34);
         is_band_type = find_min_book(maxval, is_sf_idx);
         dist1 += quantize_band_cost(s, &L[start + (w+w2)*128], L34,
diff --git a/libavcodec/aacenc_ltp.c b/libavcodec/aacenc_ltp.c
index b9d43b4..1bec85b 100644
--- a/libavcodec/aacenc_ltp.c
+++ b/libavcodec/aacenc_ltp.c
@@ -190,8 +190,8 @@ void ff_aac_search_for_ltp(AACEncContext *s, SingleChannelElement *sce,
                 FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
                 for (i = 0; i < sce->ics.swb_sizes[g]; i++)
                     PCD[i] = sce->coeffs[start+(w+w2)*128+i] - sce->lcoeffs[start+(w+w2)*128+i];
-                abs_pow34_v(C34,  &sce->coeffs[start+(w+w2)*128],  sce->ics.swb_sizes[g]);
-                abs_pow34_v(PCD34, PCD, sce->ics.swb_sizes[g]);
+                s->abs_pow34(C34,  &sce->coeffs[start+(w+w2)*128],  sce->ics.swb_sizes[g]);
+                s->abs_pow34(PCD34, PCD, sce->ics.swb_sizes[g]);
                 dist1 += quantize_band_cost(s, &sce->coeffs[start+(w+w2)*128], C34, sce->ics.swb_sizes[g],
                                             sce->sf_idx[(w+w2)*16+g], sce->band_type[(w+w2)*16+g],
                                             s->lambda/band->threshold, INFINITY, &bits_tmp1, NULL, 0);
diff --git a/libavcodec/aacenc_pred.c b/libavcodec/aacenc_pred.c
index e77a3de..d111192 100644
--- a/libavcodec/aacenc_pred.c
+++ b/libavcodec/aacenc_pred.c
@@ -270,7 +270,7 @@ void ff_aac_search_for_pred(AACEncContext *s, SingleChannelElement *sce)
             continue;
 
         /* Normal coefficients */
-        abs_pow34_v(O34, &sce->coeffs[start_coef], num_coeffs);
+        s->abs_pow34(O34, &sce->coeffs[start_coef], num_coeffs);
         dist1 = quantize_and_encode_band_cost(s, NULL, &sce->coeffs[start_coef], NULL,
                                               O34, num_coeffs, sce->sf_idx[sfb],
                                               cb_n, s->lambda / band->threshold, INFINITY, &cost1, NULL, 0);
@@ -279,7 +279,7 @@ void ff_aac_search_for_pred(AACEncContext *s, SingleChannelElement *sce)
         /* Encoded coefficients - needed for #bits, band type and quant. error */
         for (i = 0; i < num_coeffs; i++)
             SENT[i] = sce->coeffs[start_coef + i] - sce->prcoeffs[start_coef + i];
-        abs_pow34_v(S34, SENT, num_coeffs);
+        s->abs_pow34(S34, SENT, num_coeffs);
         if (cb_n < RESERVED_BT)
             cb_p = av_clip(find_min_book(find_max_val(1, num_coeffs, S34), sce->sf_idx[sfb]), cb_min, cb_max);
         else
@@ -291,7 +291,7 @@ void ff_aac_search_for_pred(AACEncContext *s, SingleChannelElement *sce)
         /* Reconstructed coefficients - needed for distortion measurements */
         for (i = 0; i < num_coeffs; i++)
             sce->prcoeffs[start_coef + i] += QERR[i] != 0.0f ? (sce->prcoeffs[start_coef + i] - QERR[i]) : 0.0f;
-        abs_pow34_v(P34, &sce->prcoeffs[start_coef], num_coeffs);
+        s->abs_pow34(P34, &sce->prcoeffs[start_coef], num_coeffs);
         if (cb_n < RESERVED_BT)
             cb_p = av_clip(find_min_book(find_max_val(1, num_coeffs, P34), sce->sf_idx[sfb]), cb_min, cb_max);
         else
diff --git a/libavcodec/aacenc_quantization.h b/libavcodec/aacenc_quantization.h
index 4250407..59a9734 100644
--- a/libavcodec/aacenc_quantization.h
+++ b/libavcodec/aacenc_quantization.h
@@ -74,10 +74,10 @@ static av_always_inline float quantize_and_encode_band_cost_template(
         return cost * lambda;
     }
     if (!scaled) {
-        abs_pow34_v(s->scoefs, in, size);
+        s->abs_pow34(s->scoefs, in, size);
         scaled = s->scoefs;
     }
-    quantize_bands(s->qcoefs, in, scaled, size, Q34, !BT_UNSIGNED, aac_cb_maxval[cb], ROUNDING);
+    s->quant_bands(s->qcoefs, in, scaled, size, Q34, !BT_UNSIGNED, aac_cb_maxval[cb], ROUNDING);
     if (BT_UNSIGNED) {
         off = 0;
     } else {
diff --git a/libavcodec/aacenc_utils.h b/libavcodec/aacenc_utils.h
index ff9188a..d4211a5 100644
--- a/libavcodec/aacenc_utils.h
+++ b/libavcodec/aacenc_utils.h
@@ -37,7 +37,7 @@
 #define ROUND_TO_ZERO 0.1054f
 #define C_QUANT 0.4054f
 
-static inline void abs_pow34_v(float *out, const float *in, const int size)
+static inline void abs_pow34_v(float *out, const float *in, const int64_t size)
 {
     int i;
     for (i = 0; i < size; i++) {
diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 522b6c2..1db1137 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -42,6 +42,7 @@ OBJS-$(CONFIG_XMM_CLOBBER_TEST)        += x86/w64xmmtest.o
 # decoders/encoders
 OBJS-$(CONFIG_AAC_DECODER)             += x86/aacpsdsp_init.o          \
                                           x86/sbrdsp_init.o
+OBJS-$(CONFIG_AAC_ENCODER)             += x86/aacencdsp_init.o
 OBJS-$(CONFIG_ADPCM_G722_DECODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_ADPCM_G722_ENCODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_ALAC_DECODER)            += x86/alacdsp_init.o
@@ -132,6 +133,7 @@ YASM-OBJS-$(CONFIG_VP8DSP)             += x86/vp8dsp.o                  \
 # decoders/encoders
 YASM-OBJS-$(CONFIG_AAC_DECODER)        += x86/aacpsdsp.o                \
                                           x86/sbrdsp.o
+YASM-OBJS-$(CONFIG_AAC_ENCODER)        += x86/aacencdsp.o
 YASM-OBJS-$(CONFIG_ADPCM_G722_DECODER) += x86/g722dsp.o
 YASM-OBJS-$(CONFIG_ADPCM_G722_ENCODER) += x86/g722dsp.o
 YASM-OBJS-$(CONFIG_ALAC_DECODER)       += x86/alacdsp.o
diff --git a/libavcodec/x86/aacencdsp.asm b/libavcodec/x86/aacencdsp.asm
new file mode 100644
index 0000000..79e70c6
--- /dev/null
+++ b/libavcodec/x86/aacencdsp.asm
@@ -0,0 +1,111 @@
+;******************************************************************************
+;* SIMD optimized AAC encoder DSP functions
+;*
+;* Copyright (C) 2016 Rostislav Pehlivanov <atomnuker at gmail.com>
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* FFmpeg is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with FFmpeg; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+SECTION_RODATA
+
+float_abs_mask:  times 4 dd 0x7fffffff
+float_sign_mask: times 4 dd 0x80000000
+
+SECTION .text
+
+;*******************************************************************
+;void ff_abs_pow34_sse(float *out, const float *in, const int64_t size);
+;*******************************************************************
+INIT_XMM sse
+cglobal abs_pow34, 3, 3, 3, out, in, size
+    mova   m2, [float_abs_mask]
+
+    shl    sizeq, 2
+    add    inq, sizeq
+    add    outq, sizeq
+    neg    sizeq
+
+.loop:
+    mova   m0, [inq+sizeq]
+
+    andps  m0, m2
+
+    sqrtps m1, m0
+    mulps  m0, m1
+    sqrtps m0, m0
+
+    mova   [outq+sizeq], m0
+
+    add    sizeq, mmsize
+    jl     .loop
+    RET
+
+
+%if ARCH_X86_64
+
+;*******************************************************************
+;void ff_aac_quantize_bands_sse(int *out, const float *in, const float *scaled,
+;                               int size, const float *Q34, int is_signed, const float *maxval,
+;                               const float *rounding)
+;*******************************************************************
+INIT_XMM sse
+cglobal aac_quantize_bands, 8, 8, 7, out, in, scaled, size, Q34, is_signed, maxval, rounding
+
+    mova    m1, [Q34q]
+    SPLATD  m1
+
+    mova    m2, [roundingq]
+    SPLATD  m2
+
+    mova    m3, [maxvalq]
+    SPLATD  m3
+
+    shl     is_signedd, 31
+    movd    m4, is_signedd
+    SPLATD  m4
+
+    mova    m6, [float_sign_mask]
+
+    shl     sizeq,   2
+    add     inq,     sizeq
+    add     outq,    sizeq
+    add     scaledq, sizeq
+    neg     sizeq
+
+.loop:
+    mova   m0, [scaledq+sizeq]
+
+    mulps  m0, m1
+    addps  m0, m2
+    minps  m0, m3
+
+    mova   m5, [inq+sizeq]
+    andps  m5, m6
+    pand   m5, m4
+
+    orps   m0, m5
+
+    cvttps2dq m0, m0
+    mova   [outq+sizeq], m0
+
+    add    sizeq, mmsize
+    jl     .loop
+    RET
+
+%endif
diff --git a/libavcodec/x86/aacencdsp_init.c b/libavcodec/x86/aacencdsp_init.c
new file mode 100644
index 0000000..6219d4c
--- /dev/null
+++ b/libavcodec/x86/aacencdsp_init.c
@@ -0,0 +1,57 @@
+/*
+ * AAC encoder assembly optimizations
+ * Copyright (C) 2016 Rostislav Pehlivanov <atomnuker at gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/float_dsp.h"
+#include "libavutil/x86/cpu.h"
+#include "libavcodec/aacenc.h"
+
+void ff_abs_pow34_sse(float *out, const float *in, const int64_t size);
+
+#define QUANT_BANDS_WRAPPER(SET, RSIZE_BYTES)                                                         \
+    static void ff_aac_quantize_bands_wrapper_ ## SET(int *out, const float *in, const float *scaled, \
+                                                      int size, float Q34, int is_signed, int maxval, \
+                                                      const float rounding)                           \
+{                                                                                                     \
+    void ff_aac_quantize_bands_ ## SET(int *out, const float *in, const float *scaled,                \
+                               int size, const float *Q34, int is_signed, const float *maxval,        \
+                               const float *rounding);                                                \
+                                                                                                      \
+    /* Can't pass floats to external assembly directly */                                             \
+    ff_aac_quantize_bands_ ## SET(out, in, scaled, size, (const float [RSIZE_BYTES]){Q34},            \
+                                  is_signed, (const float [RSIZE_BYTES]){(float)maxval},              \
+                                  (const float [RSIZE_BYTES]){rounding});                             \
+}
+
+QUANT_BANDS_WRAPPER(sse, 4)
+
+av_cold void ff_aac_dsp_init_x86(AACEncContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (EXTERNAL_SSE(cpu_flags)) {
+        s->abs_pow34   = &ff_abs_pow34_sse;
+#ifdef ARCH_X86_64
+        s->quant_bands = &ff_aac_quantize_bands_wrapper_sse;
+#endif
+    }
+}
-- 
2.9.3



More information about the ffmpeg-devel mailing list