RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2025-02-17 12:27:18 +00:00
Code Issues Projects Releases Wiki Activity

Merge commit 'd2f4846591727fedcc2b452b688da8da09ee8305' into release/0.10

Browse Source 
* commit 'd2f4846591727fedcc2b452b688da8da09ee8305':
  Prepare for 0.8.7 Release
  x86: fft: Remove 3DNow! optimizations, they break FATE
  x86: ac3dsp: Drop mmx variant of ac3_max_msb_abs_int16

Conflicts:
	RELEASE
	libavcodec/x86/fft_3dn.c
	libavcodec/x86/fft_3dn2.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2013-11-04 19:34:04 +01:00
commit 8528feb13c
6 changed files with 0 additions and 220 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/x86/Makefile
View File

@ -3,8 +3,6 @@ OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o
YASM-OBJS-$(CONFIG_DCT) += x86/dct32_sse.o
YASM-OBJS-FFT-$(HAVE_AMD3DNOW) += x86/fft_3dn.o
YASM-OBJS-FFT-$(HAVE_AMD3DNOWEXT) += x86/fft_3dn2.o
YASM-OBJS-FFT-$(HAVE_SSE) += x86/fft_sse.o
YASM-OBJS-$(CONFIG_FFT) += x86/fft_mmx.o \
$(YASM-OBJS-FFT-yes)

9
libavcodec/x86/ac3dsp.asm
View File

@ -86,7 +86,6 @@ AC3_EXPONENT_MIN sse2
; This function uses 2 different methods to calculate a valid result.
; 1) logical 'or' of abs of each element
; This is used for ssse3 because of the pabsw instruction.
; It is also used for mmx because of the lack of min/max instructions.
; 2) calculate min/max for the array, then or(abs(min),abs(max))
; This is used for mmxext and sse2 because they have pminsw/pmaxsw.
;-----------------------------------------------------------------------------
@ -104,15 +103,9 @@ cglobal ac3_max_msb_abs_int16_%1, 2,2,5, src, len
pmaxsw m3, m0
pmaxsw m3, m1
%else ; or_abs
%ifidn %1, mmx
mova m0, [srcq]
mova m1, [srcq+mmsize]
ABS2 m0, m1, m3, m4
%else ; ssse3
; using memory args is faster for ssse3
pabsw m0, [srcq]
pabsw m1, [srcq+mmsize]
%endif
por m2, m0
por m2, m1
%endif
@ -137,9 +130,7 @@ cglobal ac3_max_msb_abs_int16_%1, 2,2,5, src, len
%endmacro
INIT_MMX
%define ABS2 ABS2_MMX
%define PSHUFLW pshufw
AC3_MAX_MSB_ABS_INT16 mmx, or_abs
%define ABS2 ABS2_MMX2
AC3_MAX_MSB_ABS_INT16 mmxext, min_max
INIT_XMM

2
libavcodec/x86/ac3dsp_mmx.c
View File

@ -27,7 +27,6 @@ extern void ff_ac3_exponent_min_mmx (uint8_t *exp, int num_reuse_blocks, int n
extern void ff_ac3_exponent_min_mmxext(uint8_t *exp, int num_reuse_blocks, int nb_coefs);
extern void ff_ac3_exponent_min_sse2 (uint8_t *exp, int num_reuse_blocks, int nb_coefs);
extern int ff_ac3_max_msb_abs_int16_mmx (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_mmxext(const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_sse2 (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_ssse3 (const int16_t *src, int len);
@ -55,7 +54,6 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
if (mm_flags & AV_CPU_FLAG_MMX) {
c->ac3_exponent_min = ff_ac3_exponent_min_mmx;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx;
c->ac3_lshift_int16 = ff_ac3_lshift_int16_mmx;
c->ac3_rshift_int32 = ff_ac3_rshift_int32_mmx;
}

10
libavcodec/x86/fft.c
View File

@ -39,16 +39,6 @@ av_cold void ff_fft_init_mmx(FFTContext *s)
s->fft_permute = ff_fft_permute_sse;
s->fft_calc = ff_fft_calc_sse;
s->fft_permutation = FF_FFT_PERM_SWAP_LSBS;
} else if (has_vectors & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT) {
/* 3DNowEx for K7 */
s->imdct_calc = ff_imdct_calc_3dn2;
s->imdct_half = ff_imdct_half_3dn2;
s->fft_calc = ff_fft_calc_3dn2;
} else if (has_vectors & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) {
/* 3DNow! for K6-2/3 */
s->imdct_calc = ff_imdct_calc_3dn;
s->imdct_half = ff_imdct_half_3dn;
s->fft_calc = ff_fft_calc_3dn;
}
#endif
}

23
libavcodec/x86/fft_3dn.c
View File

@ -1,23 +0,0 @@
/*
* FFT/MDCT transform with 3DNow! optimizations
* Copyright (c) 2008 Loren Merritt
*
* 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
*/
#define EMULATE_3DNOWEXT
#include "fft_3dn2.c"

174
libavcodec/x86/fft_3dn2.c
View File

@ -1,174 +0,0 @@
/*
* FFT/MDCT transform with Extended 3DNow! optimizations
* Copyright (c) 2006-2008 Zuxy MENG Jie, Loren Merritt
*
* 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_cpu.h"
#include "libavcodec/dsputil.h"
#include "fft.h"
DECLARE_ALIGNED(8, static const unsigned int, m1m1)[2] = { 1U<<31, 1U<<31 };
#ifdef EMULATE_3DNOWEXT
#define PSWAPD(s,d)\
"movq "#s","#d"\n"\
"psrlq $32,"#d"\n"\
"punpckldq "#s","#d"\n"
#define ff_fft_calc_3dn2 ff_fft_calc_3dn
#define ff_fft_dispatch_3dn2 ff_fft_dispatch_3dn
#define ff_fft_dispatch_interleave_3dn2 ff_fft_dispatch_interleave_3dn
#define ff_imdct_calc_3dn2 ff_imdct_calc_3dn
#define ff_imdct_half_3dn2 ff_imdct_half_3dn
#else
#define PSWAPD(s,d) "pswapd "#s","#d"\n"
#endif
void ff_fft_dispatch_3dn2(FFTComplex *z, int nbits);
void ff_fft_dispatch_interleave_3dn2(FFTComplex *z, int nbits);
void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z)
{
int n = 1<<s->nbits;
int i;
ff_fft_dispatch_interleave_3dn2(z, s->nbits);
__asm__ volatile("femms");
if(n <= 8)
for(i=0; i<n; i+=2)
FFSWAP(FFTSample, z[i].im, z[i+1].re);
}
void ff_imdct_half_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input)
{
x86_reg j, k;
long n = s->mdct_size;
long n2 = n >> 1;
long n4 = n >> 2;
long n8 = n >> 3;
const uint16_t *revtab = s->revtab;
const FFTSample *tcos = s->tcos;
const FFTSample *tsin = s->tsin;
const FFTSample *in1, *in2;
FFTComplex *z = (FFTComplex *)output;
/* pre rotation */
in1 = input;
in2 = input + n2 - 1;
#ifdef EMULATE_3DNOWEXT
__asm__ volatile("movd %0, %%mm7" ::"r"(1U<<31));
#endif
for(k = 0; k < n4; k++) {
// FIXME a single block is faster, but gcc 2.95 and 3.4.x on 32bit can't compile it
__asm__ volatile(
"movd %0, %%mm0 \n"
"movd %2, %%mm1 \n"
"punpckldq %1, %%mm0 \n"
"punpckldq %3, %%mm1 \n"
"movq %%mm0, %%mm2 \n"
PSWAPD( %%mm1, %%mm3 )
"pfmul %%mm1, %%mm0 \n"
"pfmul %%mm3, %%mm2 \n"
#ifdef EMULATE_3DNOWEXT
"movq %%mm0, %%mm1 \n"
"punpckhdq %%mm2, %%mm0 \n"
"punpckldq %%mm2, %%mm1 \n"
"pxor %%mm7, %%mm0 \n"
"pfadd %%mm1, %%mm0 \n"
#else
"pfpnacc %%mm2, %%mm0 \n"
#endif
::"m"(in2[-2*k]), "m"(in1[2*k]),
"m"(tcos[k]), "m"(tsin[k])
);
__asm__ volatile(
"movq %%mm0, %0 \n\t"
:"=m"(z[revtab[k]])
);
}
ff_fft_dispatch_3dn2(z, s->nbits);
#define CMUL(j,mm0,mm1)\
"movq (%2,"#j",2), %%mm6 \n"\
"movq 8(%2,"#j",2), "#mm0"\n"\
"movq %%mm6, "#mm1"\n"\
"movq "#mm0",%%mm7 \n"\
"pfmul (%3,"#j"), %%mm6 \n"\
"pfmul (%4,"#j"), "#mm0"\n"\
"pfmul (%4,"#j"), "#mm1"\n"\
"pfmul (%3,"#j"), %%mm7 \n"\
"pfsub %%mm6, "#mm0"\n"\
"pfadd %%mm7, "#mm1"\n"
/* post rotation */
j = -n2;
k = n2-8;
__asm__ volatile(
"1: \n"
CMUL(%0, %%mm0, %%mm1)
CMUL(%1, %%mm2, %%mm3)
"movd %%mm0, (%2,%0,2) \n"
"movd %%mm1,12(%2,%1,2) \n"
"movd %%mm2, (%2,%1,2) \n"
"movd %%mm3,12(%2,%0,2) \n"
"psrlq $32, %%mm0 \n"
"psrlq $32, %%mm1 \n"
"psrlq $32, %%mm2 \n"
"psrlq $32, %%mm3 \n"
"movd %%mm0, 8(%2,%0,2) \n"
"movd %%mm1, 4(%2,%1,2) \n"
"movd %%mm2, 8(%2,%1,2) \n"
"movd %%mm3, 4(%2,%0,2) \n"
"sub $8, %1 \n"
"add $8, %0 \n"
"jl 1b \n"
:"+r"(j), "+r"(k)
:"r"(z+n8), "r"(tcos+n8), "r"(tsin+n8)
:"memory"
);
__asm__ volatile("femms");
}
void ff_imdct_calc_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input)
{
x86_reg j, k;
long n = s->mdct_size;
long n4 = n >> 2;
ff_imdct_half_3dn2(s, output+n4, input);
j = -n;
k = n-8;
__asm__ volatile(
"movq %4, %%mm7 \n"
"1: \n"
PSWAPD((%2,%1), %%mm0)
PSWAPD((%3,%0), %%mm1)
"pxor %%mm7, %%mm0 \n"
"movq %%mm1, (%3,%1) \n"
"movq %%mm0, (%2,%0) \n"
"sub $8, %1 \n"
"add $8, %0 \n"
"jl 1b \n"
:"+r"(j), "+r"(k)
:"r"(output+n4), "r"(output+n4*3),
"m"(*m1m1)
);
__asm__ volatile("femms");
}