[FFmpeg-devel] [PATCH 3/6] cavs: convert idct from inline asm to yasm.

Ronald S. Bultje rsbultje at gmail.com
Tue Apr 4 19:48:15 EEST 2017


---
 libavcodec/x86/Makefile     |   2 +-
 libavcodec/x86/cavsdsp.c    | 165 ++------------------------------------------
 libavcodec/x86/cavsidct.asm | 165 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 171 insertions(+), 161 deletions(-)
 create mode 100644 libavcodec/x86/cavsidct.asm

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 0295a9f..56df530 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -48,7 +48,7 @@ OBJS-$(CONFIG_ADPCM_G722_DECODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_ADPCM_G722_ENCODER)      += x86/g722dsp_init.o
 OBJS-$(CONFIG_ALAC_DECODER)            += x86/alacdsp_init.o
 OBJS-$(CONFIG_APNG_DECODER)            += x86/pngdsp_init.o
-OBJS-$(CONFIG_CAVS_DECODER)            += x86/cavsdsp.o
+OBJS-$(CONFIG_CAVS_DECODER)            += x86/cavsdsp.o x86/cavsidct.o
 OBJS-$(CONFIG_DCA_DECODER)             += x86/dcadsp_init.o x86/synth_filter_init.o
 OBJS-$(CONFIG_DNXHD_ENCODER)           += x86/dnxhdenc_init.o
 OBJS-$(CONFIG_HEVC_DECODER)            += x86/hevcdsp_init.o
diff --git a/libavcodec/x86/cavsdsp.c b/libavcodec/x86/cavsdsp.c
index ecb9b23..add4536 100644
--- a/libavcodec/x86/cavsdsp.c
+++ b/libavcodec/x86/cavsdsp.c
@@ -34,172 +34,19 @@
 #include "idctdsp.h"
 #include "config.h"
 
-#if HAVE_MMX_INLINE
 
-/* in/out: mma=mma+mmb, mmb=mmb-mma */
-#define SUMSUB_BA( a, b ) \
-    "paddw "#b", "#a" \n\t"\
-    "paddw "#b", "#b" \n\t"\
-    "psubw "#a", "#b" \n\t"
-
-/*****************************************************************************
- *
- * inverse transform
- *
- ****************************************************************************/
-
-static inline void cavs_idct8_1d(int16_t *block, uint64_t bias)
-{
-    __asm__ volatile(
-        "movq 112(%0), %%mm4  \n\t" /* mm4 = src7 */
-        "movq  16(%0), %%mm5  \n\t" /* mm5 = src1 */
-        "movq  80(%0), %%mm2  \n\t" /* mm2 = src5 */
-        "movq  48(%0), %%mm7  \n\t" /* mm7 = src3 */
-        "movq   %%mm4, %%mm0  \n\t"
-        "movq   %%mm5, %%mm3  \n\t"
-        "movq   %%mm2, %%mm6  \n\t"
-        "movq   %%mm7, %%mm1  \n\t"
-
-        "paddw  %%mm4, %%mm4  \n\t" /* mm4 = 2*src7 */
-        "paddw  %%mm3, %%mm3  \n\t" /* mm3 = 2*src1 */
-        "paddw  %%mm6, %%mm6  \n\t" /* mm6 = 2*src5 */
-        "paddw  %%mm1, %%mm1  \n\t" /* mm1 = 2*src3 */
-        "paddw  %%mm4, %%mm0  \n\t" /* mm0 = 3*src7 */
-        "paddw  %%mm3, %%mm5  \n\t" /* mm5 = 3*src1 */
-        "paddw  %%mm6, %%mm2  \n\t" /* mm2 = 3*src5 */
-        "paddw  %%mm1, %%mm7  \n\t" /* mm7 = 3*src3 */
-        "psubw  %%mm4, %%mm5  \n\t" /* mm5 = 3*src1 - 2*src7 = a0 */
-        "paddw  %%mm6, %%mm7  \n\t" /* mm7 = 3*src3 + 2*src5 = a1 */
-        "psubw  %%mm2, %%mm1  \n\t" /* mm1 = 2*src3 - 3*src5 = a2 */
-        "paddw  %%mm0, %%mm3  \n\t" /* mm3 = 2*src1 + 3*src7 = a3 */
-
-        "movq   %%mm5, %%mm4  \n\t"
-        "movq   %%mm7, %%mm6  \n\t"
-        "movq   %%mm3, %%mm0  \n\t"
-        "movq   %%mm1, %%mm2  \n\t"
-        SUMSUB_BA( %%mm7, %%mm5 )   /* mm7 = a0 + a1  mm5 = a0 - a1 */
-        "paddw  %%mm3, %%mm7  \n\t" /* mm7 = a0 + a1 + a3 */
-        "paddw  %%mm1, %%mm5  \n\t" /* mm5 = a0 - a1 + a2 */
-        "paddw  %%mm7, %%mm7  \n\t"
-        "paddw  %%mm5, %%mm5  \n\t"
-        "paddw  %%mm6, %%mm7  \n\t" /* mm7 = b4 */
-        "paddw  %%mm4, %%mm5  \n\t" /* mm5 = b5 */
-
-        SUMSUB_BA( %%mm1, %%mm3 )   /* mm1 = a3 + a2  mm3 = a3 - a2 */
-        "psubw  %%mm1, %%mm4  \n\t" /* mm4 = a0 - a2 - a3 */
-        "movq   %%mm4, %%mm1  \n\t" /* mm1 = a0 - a2 - a3 */
-        "psubw  %%mm6, %%mm3  \n\t" /* mm3 = a3 - a2 - a1 */
-        "paddw  %%mm1, %%mm1  \n\t"
-        "paddw  %%mm3, %%mm3  \n\t"
-        "psubw  %%mm2, %%mm1  \n\t" /* mm1 = b7 */
-        "paddw  %%mm0, %%mm3  \n\t" /* mm3 = b6 */
-
-        "movq  32(%0), %%mm2  \n\t" /* mm2 = src2 */
-        "movq  96(%0), %%mm6  \n\t" /* mm6 = src6 */
-        "movq   %%mm2, %%mm4  \n\t"
-        "movq   %%mm6, %%mm0  \n\t"
-        "psllw  $2,    %%mm4  \n\t" /* mm4 = 4*src2 */
-        "psllw  $2,    %%mm6  \n\t" /* mm6 = 4*src6 */
-        "paddw  %%mm4, %%mm2  \n\t" /* mm2 = 5*src2 */
-        "paddw  %%mm6, %%mm0  \n\t" /* mm0 = 5*src6 */
-        "paddw  %%mm2, %%mm2  \n\t"
-        "paddw  %%mm0, %%mm0  \n\t"
-        "psubw  %%mm0, %%mm4  \n\t" /* mm4 = 4*src2 - 10*src6 = a7 */
-        "paddw  %%mm2, %%mm6  \n\t" /* mm6 = 4*src6 + 10*src2 = a6 */
-
-        "movq    (%0), %%mm2  \n\t" /* mm2 = src0 */
-        "movq  64(%0), %%mm0  \n\t" /* mm0 = src4 */
-        SUMSUB_BA( %%mm0, %%mm2 )   /* mm0 = src0+src4  mm2 = src0-src4 */
-        "psllw  $3,    %%mm0  \n\t"
-        "psllw  $3,    %%mm2  \n\t"
-        "paddw  %1,    %%mm0  \n\t" /* add rounding bias */
-        "paddw  %1,    %%mm2  \n\t" /* add rounding bias */
-
-        SUMSUB_BA( %%mm6, %%mm0 )   /* mm6 = a4 + a6  mm0 = a4 - a6 */
-        SUMSUB_BA( %%mm4, %%mm2 )   /* mm4 = a5 + a7  mm2 = a5 - a7 */
-        SUMSUB_BA( %%mm7, %%mm6 )   /* mm7 = dst0  mm6 = dst7 */
-        SUMSUB_BA( %%mm5, %%mm4 )   /* mm5 = dst1  mm4 = dst6 */
-        SUMSUB_BA( %%mm3, %%mm2 )   /* mm3 = dst2  mm2 = dst5 */
-        SUMSUB_BA( %%mm1, %%mm0 )   /* mm1 = dst3  mm0 = dst4 */
-        :: "r"(block), "m"(bias)
-    );
-}
-
-#define SBUTTERFLY(a,b,t,n,m)\
-    "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
-    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
-    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
+#if HAVE_MMX_EXTERNAL
 
-#define TRANSPOSE4(a,b,c,d,t)\
-    SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
-    SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
-    SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
-    SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
+void ff_cavs_idct8_mmx(int16_t *out, const int16_t *in);
 
 static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, ptrdiff_t stride)
 {
-    int i;
     LOCAL_ALIGNED(16, int16_t, b2, [64]);
-
-    for(i=0; i<2; i++){
-        cavs_idct8_1d(block + 4 * i, ff_pw_4.a);
-
-        __asm__ volatile(
-            "psraw     $3, %%mm7  \n\t"
-            "psraw     $3, %%mm6  \n\t"
-            "psraw     $3, %%mm5  \n\t"
-            "psraw     $3, %%mm4  \n\t"
-            "psraw     $3, %%mm3  \n\t"
-            "psraw     $3, %%mm2  \n\t"
-            "psraw     $3, %%mm1  \n\t"
-            "psraw     $3, %%mm0  \n\t"
-            "movq   %%mm7,  (%0)  \n\t"
-            TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
-            "movq   %%mm0,  8(%0)  \n\t"
-            "movq   %%mm6, 24(%0)  \n\t"
-            "movq   %%mm7, 40(%0)  \n\t"
-            "movq   %%mm4, 56(%0)  \n\t"
-            "movq    (%0),  %%mm7  \n\t"
-            TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
-            "movq   %%mm7,   (%0)  \n\t"
-            "movq   %%mm1, 16(%0)  \n\t"
-            "movq   %%mm0, 32(%0)  \n\t"
-            "movq   %%mm3, 48(%0)  \n\t"
-            :
-            : "r"(b2 + 32 * i)
-            : "memory"
-        );
-    }
-
-    for(i=0; i<2; i++){
-        cavs_idct8_1d(b2+4*i, ff_pw_64.a);
-
-        __asm__ volatile(
-            "psraw     $7, %%mm7  \n\t"
-            "psraw     $7, %%mm6  \n\t"
-            "psraw     $7, %%mm5  \n\t"
-            "psraw     $7, %%mm4  \n\t"
-            "psraw     $7, %%mm3  \n\t"
-            "psraw     $7, %%mm2  \n\t"
-            "psraw     $7, %%mm1  \n\t"
-            "psraw     $7, %%mm0  \n\t"
-            "movq   %%mm7,    (%0)  \n\t"
-            "movq   %%mm5,  16(%0)  \n\t"
-            "movq   %%mm3,  32(%0)  \n\t"
-            "movq   %%mm1,  48(%0)  \n\t"
-            "movq   %%mm0,  64(%0)  \n\t"
-            "movq   %%mm2,  80(%0)  \n\t"
-            "movq   %%mm4,  96(%0)  \n\t"
-            "movq   %%mm6, 112(%0)  \n\t"
-            :: "r"(b2+4*i)
-            : "memory"
-        );
-    }
-
+    ff_cavs_idct8_mmx(b2, block);
     ff_add_pixels_clamped(b2, dst, stride);
 }
 
-#endif /* HAVE_MMX_INLINE */
+#endif /* HAVE_MMX_EXTERNAL */
 
 #if (HAVE_MMXEXT_INLINE || HAVE_AMD3DNOW_INLINE)
 
@@ -529,12 +376,10 @@ static av_cold void cavsdsp_init_mmx(CAVSDSPContext *c,
     c->put_cavs_qpel_pixels_tab[1][0] = put_cavs_qpel8_mc00_mmx;
     c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_mmx;
     c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmx;
-#endif
 
-#if HAVE_MMX_INLINE
     c->cavs_idct8_add = cavs_idct8_add_mmx;
     c->idct_perm      = FF_IDCT_PERM_TRANSPOSE;
-#endif /* HAVE_MMX_INLINE */
+#endif /* HAVE_MMX_EXTERNAL */
 }
 
 #define DSPFUNC(PFX, IDX, NUM, EXT)                                                       \
diff --git a/libavcodec/x86/cavsidct.asm b/libavcodec/x86/cavsidct.asm
new file mode 100644
index 0000000..5421196
--- /dev/null
+++ b/libavcodec/x86/cavsidct.asm
@@ -0,0 +1,165 @@
+; Chinese AVS video (AVS1-P2, JiZhun profile) decoder
+; Copyright (c) 2006  Stefan Gehrer <stefan.gehrer at gmx.de>
+;
+; MMX-optimized DSP functions, based on H.264 optimizations by
+; Michael Niedermayer and Loren Merritt
+; Conversion from gcc syntax to x264asm syntax with modifications
+; by Ronald S. Bultje <rsbultje 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"
+
+cextern pw_4
+cextern pw_64
+
+SECTION .text
+
+%macro CAVS_IDCT8_1D 2 ; source, round
+    mova            m4, [%1+7*16]       ; m4 = src7
+    mova            m5, [%1+1*16]       ; m5 = src1
+    mova            m2, [%1+5*16]       ; m2 = src5
+    mova            m7, [%1+3*16]       ; m7 = src3
+    mova            m0, m4
+    mova            m3, m5
+    mova            m6, m2
+    mova            m1, m7
+
+    paddw           m4, m4              ; m4 = 2*src7
+    paddw           m3, m3              ; m3 = 2*src1
+    paddw           m6, m6              ; m6 = 2*src5
+    paddw           m1, m1              ; m1 = 2*src3
+    paddw           m0, m4              ; m0 = 3*src7
+    paddw           m5, m3              ; m5 = 3*src1
+    paddw           m2, m6              ; m2 = 3*src5
+    paddw           m7, m1              ; m7 = 3*src3
+    psubw           m5, m4              ; m5 = 3*src1 - 2*src7 = a0
+    paddw           m7, m6              ; m7 = 3*src3 - 2*src5 = a1
+    psubw           m1, m2              ; m1 = 2*src3 - 3*src5 = a2
+    paddw           m3, m0              ; m3 = 2*src1 - 3*src7 = a3
+
+    mova            m4, m5
+    mova            m6, m7
+    mova            m0, m3
+    mova            m2, m1
+    SUMSUB_BA     w, 7, 5               ; m7 = a0 + a1, m5 = a0 - a1
+    paddw           m7, m3              ; m7 = a0 + a1 + a3
+    paddw           m5, m1              ; m5 = a0 - a1 + a2
+    paddw           m7, m7
+    paddw           m5, m5
+    paddw           m7, m6              ; m7 = b4
+    paddw           m5, m4              ; m5 = b5
+
+    SUMSUB_BA     w, 1, 3               ; m1 = a3 + a2, m3 = a3 - a2
+    psubw           m4, m1              ; m4 = a0 - a2 - a3
+    mova            m1, m4              ; m1 = a0 - a2 - a3
+    psubw           m3, m6              ; m3 = a3 - a2 - a1
+    paddw           m1, m1
+    paddw           m3, m3
+    psubw           m1, m2              ; m1 = b7
+    paddw           m3, m0              ; m3 = b6
+
+    mova            m2, [%1+2*16]       ; m2 = src2
+    mova            m6, [%1+6*16]       ; m6 = src6
+    mova            m4, m2
+    mova            m0, m6
+    psllw           m4, 2               ; m4 = 4*src2
+    psllw           m6, 2               ; m6 = 4*src6
+    paddw           m2, m4              ; m2 = 5*src2
+    paddw           m0, m6              ; m0 = 5*src6
+    paddw           m2, m2
+    paddw           m0, m0
+    psubw           m4, m0              ; m4 = 4*src2 - 10*src6 = a7
+    paddw           m6, m2              ; m6 = 4*src6 + 10*src2 = a6
+
+    mova            m2, [%1+0*16]       ; m2 = src0
+    mova            m0, [%1+4*16]       ; m0 = src4
+    SUMSUB_BA     w, 0, 2               ; m0 = src0 + src4, m2 = src0 - src4
+    psllw           m0, 3
+    psllw           m2, 3
+    paddw           m0, %2              ; add rounding bias
+    paddw           m2, %2              ; add rounding bias
+
+    SUMSUB_BA     w, 6, 0               ; m6 = a4 + a6, m0 = a4 - a6
+    SUMSUB_BA     w, 4, 2               ; m4 = a5 + a7, m2 = a5 - a7
+    SUMSUB_BA     w, 7, 6               ; m7 = dst0, m6 = dst7
+    SUMSUB_BA     w, 5, 4               ; m5 = dst1, m4 = dst6
+    SUMSUB_BA     w, 3, 2               ; m3 = dst2, m2 = dst5
+    SUMSUB_BA     w, 1, 0               ; m1 = dst3, m0 = dst4
+%endmacro
+
+INIT_MMX mmx
+cglobal cavs_idct8, 2, 4, 8, 8 * 16, out, in, cnt, tmp
+    mov           cntd, 2
+    mov           tmpq, rsp
+
+.loop_1:
+    CAVS_IDCT8_1D  inq, [pw_4]
+    psraw           m7, 3
+    psraw           m6, 3
+    psraw           m5, 3
+    psraw           m4, 3
+    psraw           m3, 3
+    psraw           m2, 3
+    psraw           m1, 3
+    psraw           m0, 3
+    mova        [tmpq], m7
+    TRANSPOSE4x4W    0, 2, 4, 6, 7
+    mova    [tmpq+1*8], m0
+    mova    [tmpq+3*8], m2
+    mova    [tmpq+5*8], m4
+    mova    [tmpq+7*8], m6
+    mova            m7, [tmpq]
+    TRANSPOSE4x4W    7, 5, 3, 1, 0
+    mova    [tmpq+0*8], m7
+    mova    [tmpq+2*8], m5
+    mova    [tmpq+4*8], m3
+    mova    [tmpq+6*8], m1
+
+    add            inq, mmsize
+    add           tmpq, 64
+    dec           cntd
+    jg .loop_1
+
+    mov           cntd, 2
+    mov           tmpq, rsp
+.loop_2:
+    CAVS_IDCT8_1D tmpq, [pw_64]
+    psraw           m7, 7
+    psraw           m6, 7
+    psraw           m5, 7
+    psraw           m4, 7
+    psraw           m3, 7
+    psraw           m2, 7
+    psraw           m1, 7
+    psraw           m0, 7
+
+    mova   [outq+0*16], m7
+    mova   [outq+1*16], m5
+    mova   [outq+2*16], m3
+    mova   [outq+3*16], m1
+    mova   [outq+4*16], m0
+    mova   [outq+5*16], m2
+    mova   [outq+6*16], m4
+    mova   [outq+7*16], m6
+
+    add           outq, mmsize
+    add           tmpq, mmsize
+    dec           cntd
+    jg .loop_2
+
+    RET
-- 
2.8.1



More information about the ffmpeg-devel mailing list