x86/aacpsdsp: add SSE and SSE3 optimized functions

Between 1.5 and 2.5 times faster
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: James Almer <jamrial@gmail.com>

libavcodec/aacps.c

@@ -936,8 +936,8 @@ static void stereo_processing(PSContext *ps, INTFLOAT (*l)[32][2], INTFLOAT (*r)
             H22[0][e+1][b] = h22;
         }
         for (k = 0; k < NR_BANDS[is34]; k++) {
-            INTFLOAT h[2][4];
-            INTFLOAT h_step[2][4];
+            LOCAL_ALIGNED_16(INTFLOAT, h, [2], [4]);
+            LOCAL_ALIGNED_16(INTFLOAT, h_step, [2], [4]);
             int start = ps->border_position[e];
             int stop  = ps->border_position[e+1];
             INTFLOAT width = Q30(1.f) / (stop - start);

libavcodec/aacpsdsp.h

@@ -52,5 +52,6 @@ typedef struct PSDSPContext {
 void AAC_RENAME(ff_psdsp_init)(PSDSPContext *s);
 void ff_psdsp_init_arm(PSDSPContext *s);
 void ff_psdsp_init_mips(PSDSPContext *s);
+void ff_psdsp_init_x86(PSDSPContext *s);
 
 #endif /* LIBAVCODEC_AACPSDSP_H */

libavcodec/aacpsdsp_template.c

@@ -224,5 +224,7 @@ av_cold void AAC_RENAME(ff_psdsp_init)(PSDSPContext *s)
         ff_psdsp_init_arm(s);
     if (ARCH_MIPS)
         ff_psdsp_init_mips(s);
+    if (ARCH_X86)
+        ff_psdsp_init_x86(s);
 #endif /* !USE_FIXED */
 }

libavcodec/x86/Makefile

@@ -38,7 +38,8 @@ OBJS-$(CONFIG_VP8DSP)                  += x86/vp8dsp_init.o
 OBJS-$(CONFIG_XMM_CLOBBER_TEST)        += x86/w64xmmtest.o
 
 # decoders/encoders
-OBJS-$(CONFIG_AAC_DECODER)             += x86/sbrdsp_init.o
+OBJS-$(CONFIG_AAC_DECODER)             += x86/aacpsdsp_init.o          \
+                                          x86/sbrdsp_init.o
 OBJS-$(CONFIG_ADPCM_G722_DECODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_ADPCM_G722_ENCODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_APNG_DECODER)            += x86/pngdsp_init.o
@@ -130,7 +131,8 @@ YASM-OBJS-$(CONFIG_VP8DSP)             += x86/vp8dsp.o                  \
                                           x86/vp8dsp_loopfilter.o
 
 # decoders/encoders
-YASM-OBJS-$(CONFIG_AAC_DECODER)        += x86/sbrdsp.o
+YASM-OBJS-$(CONFIG_AAC_DECODER)        += x86/aacpsdsp.o                \
+                                          x86/sbrdsp.o
 YASM-OBJS-$(CONFIG_ADPCM_G722_DECODER) += x86/g722dsp.o
 YASM-OBJS-$(CONFIG_ADPCM_G722_ENCODER) += x86/g722dsp.o
 YASM-OBJS-$(CONFIG_APNG_DECODER)       += x86/pngdsp.o

libavcodec/x86/aacpsdsp.asm

+;******************************************************************************
+;* SIMD optimized MPEG-4 Parametric Stereo decoding functions
+;*
+;* Copyright (C) 2015 James Almer
+;*
+;* 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
+
+ps_p1m1p1m1: dd 0, 0x80000000, 0, 0x80000000
+
+SECTION_TEXT
+
+;*************************************************************************
+;void ff_ps_add_squares_<opt>(float *dst, const float (*src)[2], int n);
+;*************************************************************************
+%macro PS_ADD_SQUARES 1
+cglobal ps_add_squares, 3, 3, %1, dst, src, n
+.loop:
+    movaps m0, [srcq]
+    movaps m1, [srcq+mmsize]
+    mulps  m0, m0
+    mulps  m1, m1
+%if cpuflag(sse3)
+    haddps m0, m1
+%else
+    movaps m3, m0
+    movaps m4, m1
+    shufps m3, m3, q0301
+    shufps m4, m4, q0301
+    addps  m0, m3
+    addps  m1, m4
+    shufps m0, m1, q2020
+%endif
+    addps  m0, [dstq]
+    movaps [dstq], m0
+    add  dstq, mmsize
+    add  srcq, mmsize*2
+    sub    nd, mmsize/4
+    jg .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse
+PS_ADD_SQUARES 3
+INIT_XMM sse3
+PS_ADD_SQUARES 5
+
+;*******************************************************************
+;void ff_ps_mul_pair_single_sse(float (*dst)[2], float (*src0)[2],
+;                                   float *src1, int n);
+;*******************************************************************
+INIT_XMM sse
+cglobal ps_mul_pair_single, 4, 5, 4, dst, src1, src2, n
+    xor r4q, r4q
+
+.loop:
+    movu     m0, [src1q+r4q]
+    movu     m1, [src1q+r4q+mmsize]
+    mova     m2, [src2q]
+    mova     m3, m2
+    unpcklps m2, m2
+    unpckhps m3, m3
+    mulps    m0, m2
+    mulps    m1, m3
+    mova [dstq+r4q], m0
+    mova [dstq+r4q+mmsize], m1
+    add   src2q, mmsize
+    add     r4q, mmsize*2
+    sub      nd, mmsize/4
+    jg .loop
+    REP_RET
+
+;***********************************************************************
+;void ff_ps_stereo_interpolate_sse3(float (*l)[2], float (*r)[2],
+;                                   float h[2][4], float h_step[2][4],
+;                                   int len);
+;***********************************************************************
+INIT_XMM sse3
+cglobal ps_stereo_interpolate, 5, 5, 6, l, r, h, h_step, n
+    movaps   m0, [hq]
+    movaps   m1, [h_stepq]
+    cmp      nd, 0
+    jle .ret
+    shl      nd, 3
+    add      lq, nq
+    add      rq, nq
+    neg      nq
+
+align 16
+.loop:
+    addps    m0, m1
+    movddup  m2, [lq+nq]
+    movddup  m3, [rq+nq]
+    movaps   m4, m0
+    movaps   m5, m0
+    unpcklps m4, m4
+    unpckhps m5, m5
+    mulps    m2, m4
+    mulps    m3, m5
+    addps    m2, m3
+    movsd  [lq+nq], m2
+    movhps [rq+nq], m2
+    add      nq, 8
+    jl .loop
+.ret:
+    REP_RET
+
+;*******************************************************************
+;void ff_ps_hybrid_analysis_<opt>(float (*out)[2], float (*in)[2],
+;                                 const float (*filter)[8][2],
+;                                 int stride, int n);
+;*******************************************************************
+%macro PS_HYBRID_ANALYSIS_LOOP 3
+    movu     %1, [inq+mmsize*%3]
+    movu     m1, [inq+mmsize*(5-%3)+8]
+%if cpuflag(sse3)
+    pshufd   %2, %1, q2301
+    pshufd   m4, m1, q0123
+    pshufd   m1, m1, q1032
+    pshufd   m2, [filterq+nq+mmsize*%3], q2301
+    addsubps %2, m4
+    addsubps %1, m1
+%else
+    mova     m2, [filterq+nq+mmsize*%3]
+    mova     %2, %1
+    mova     m4, m1
+    shufps   %2, %2, q2301
+    shufps   m4, m4, q0123
+    shufps   m1, m1, q1032
+    shufps   m2, m2, q2301
+    xorps    m4, m7
+    xorps    m1, m7
+    subps    %2, m4
+    subps    %1, m1
+%endif
+    mulps    %2, m2
+    mulps    %1, m2
+%if %3
+    addps    m3, %2
+    addps    m0, %1
+%endif
+%endmacro
+
+%macro PS_HYBRID_ANALYSIS 0
+cglobal ps_hybrid_analysis, 5, 5, 8, out, in, filter, stride, n
+%if cpuflag(sse3)
+%define MOVH movsd
+%else
+%define MOVH movlps
+%endif
+    shl strided, 3
+    shl nd, 6
+    add filterq, nq
+    neg nq
+    mova m7, [ps_p1m1p1m1]
+
+align 16
+.loop:
+    PS_HYBRID_ANALYSIS_LOOP m0, m3, 0
+    PS_HYBRID_ANALYSIS_LOOP m5, m6, 1
+    PS_HYBRID_ANALYSIS_LOOP m5, m6, 2
+
+%if cpuflag(sse3)
+    pshufd   m3, m3, q2301
+    xorps    m0, m7
+    hsubps   m3, m0
+    pshufd   m1, m3, q0020
+    pshufd   m3, m3, q0031
+    addps    m1, m3
+    movsd    m2, [inq+6*8]
+%else
+    mova     m1, m3
+    mova     m2, m0
+    shufps   m1, m1, q2301
+    shufps   m2, m2, q2301
+    subps    m1, m3
+    addps    m2, m0
+    unpcklps m3, m1, m2
+    unpckhps m1, m2
+    addps    m1, m3
+    movu     m2, [inq+6*8] ; faster than movlps and no risk of overread
+%endif
+    movss    m3, [filterq+nq+8*6]
+    SPLATD   m3
+    mulps    m2, m3
+    addps    m1, m2
+    MOVH [outq], m1
+    add    outq, strideq
+    add      nq, 64
+    jl .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse
+PS_HYBRID_ANALYSIS
+INIT_XMM sse3
+PS_HYBRID_ANALYSIS

libavcodec/x86/aacpsdsp_init.c

+/*
+ * SIMD optimized MPEG-4 Parametric Stereo decoding functions
+ *
+ * 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/x86/cpu.h"
+#include "libavutil/attributes.h"
+#include "libavcodec/aacpsdsp.h"
+
+void ff_ps_add_squares_sse  (float *dst, const float (*src)[2], int n);
+void ff_ps_add_squares_sse3 (float *dst, const float (*src)[2], int n);
+void ff_ps_mul_pair_single_sse (float (*dst)[2], float (*src0)[2],
+                                float *src1, int n);
+void ff_ps_hybrid_analysis_sse (float (*out)[2], float (*in)[2],
+                                const float (*filter)[8][2],
+                                int stride, int n);
+void ff_ps_hybrid_analysis_sse3(float (*out)[2], float (*in)[2],
+                                const float (*filter)[8][2],
+                                int stride, int n);
+void ff_ps_stereo_interpolate_sse3(float (*l)[2], float (*r)[2],
+                                   float h[2][4], float h_step[2][4],
+                                   int len);
+
+av_cold void ff_psdsp_init_x86(PSDSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (EXTERNAL_SSE(cpu_flags)) {
+        s->add_squares            = ff_ps_add_squares_sse;
+        s->mul_pair_single        = ff_ps_mul_pair_single_sse;
+        s->hybrid_analysis        = ff_ps_hybrid_analysis_sse;
+    }
+    if (EXTERNAL_SSE3(cpu_flags)) {
+        s->add_squares            = ff_ps_add_squares_sse3;
+        s->stereo_interpolate[0]  = ff_ps_stereo_interpolate_sse3;
+        s->hybrid_analysis        = ff_ps_hybrid_analysis_sse3;
+    }
+}