# HG changeset patch # User mru # Date 1253049254 0 # Node ID 5cf49858179af8cdb2570278d94e573f817d5f95 # Parent 89cd870ca180e5107161c229c9782b6cfa8c6eaa Move per-arch fft init bits into the corresponding subdirs diff -r 89cd870ca180 -r 5cf49858179a Makefile --- a/Makefile Tue Sep 15 16:18:16 2009 +0000 +++ b/Makefile Tue Sep 15 21:14:14 2009 +0000 @@ -465,6 +465,7 @@ x86/dnxhd_mmx.o \ x86/dsputil_mmx.o \ x86/fdct_mmx.o \ + x86/fft.o \ x86/idct_mmx_xvid.o \ x86/idct_sse2_xvid.o \ x86/motion_est_mmx.o \ @@ -480,6 +481,7 @@ OBJS-$(ARCH_ARM) += arm/dsputil_arm.o \ arm/dsputil_arm_s.o \ + arm/fft_init_arm.o \ arm/jrevdct_arm.o \ arm/mpegvideo_arm.o \ arm/simple_idct_arm.o \ diff -r 89cd870ca180 -r 5cf49858179a arm/fft_init_arm.c --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/arm/fft_init_arm.c Tue Sep 15 21:14:14 2009 +0000 @@ -0,0 +1,39 @@ +/* + * Copyright (c) 2009 Mans Rullgard + * + * 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 "libavcodec/dsputil.h" + +void ff_fft_permute_neon(FFTContext *s, FFTComplex *z); +void ff_fft_calc_neon(FFTContext *s, FFTComplex *z); + +void ff_imdct_calc_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_half_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_mdct_calc_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); + +av_cold void ff_fft_init_arm(FFTContext *s) +{ + if (HAVE_NEON) { + s->fft_permute = ff_fft_permute_neon; + s->fft_calc = ff_fft_calc_neon; + s->imdct_calc = ff_imdct_calc_neon; + s->imdct_half = ff_imdct_half_neon; + s->mdct_calc = ff_mdct_calc_neon; + } +} diff -r 89cd870ca180 -r 5cf49858179a dsputil.h --- a/dsputil.h Tue Sep 15 16:18:16 2009 +0000 +++ b/dsputil.h Tue Sep 15 21:14:14 2009 +0000 @@ -683,6 +683,7 @@ void (*imdct_calc)(struct MDCTContext *s, FFTSample *output, const FFTSample *input); void (*imdct_half)(struct MDCTContext *s, FFTSample *output, const FFTSample *input); void (*mdct_calc)(struct MDCTContext *s, FFTSample *output, const FFTSample *input); + int split_radix; } FFTContext; extern FFTSample* const ff_cos_tabs[13]; @@ -694,14 +695,11 @@ */ int ff_fft_init(FFTContext *s, int nbits, int inverse); void ff_fft_permute_c(FFTContext *s, FFTComplex *z); -void ff_fft_permute_sse(FFTContext *s, FFTComplex *z); -void ff_fft_permute_neon(FFTContext *s, FFTComplex *z); void ff_fft_calc_c(FFTContext *s, FFTComplex *z); -void ff_fft_calc_sse(FFTContext *s, FFTComplex *z); -void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z); -void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z); -void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z); -void ff_fft_calc_neon(FFTContext *s, FFTComplex *z); + +void ff_fft_init_altivec(FFTContext *s); +void ff_fft_init_mmx(FFTContext *s); +void ff_fft_init_neon(FFTContext *s); /** * Do the permutation needed BEFORE calling ff_fft_calc(). @@ -774,15 +772,6 @@ void ff_imdct_calc_c(MDCTContext *s, FFTSample *output, const FFTSample *input); void ff_imdct_half_c(MDCTContext *s, FFTSample *output, const FFTSample *input); void ff_mdct_calc_c(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_calc_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_half_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_calc_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_half_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_half_sse(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_calc_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_imdct_half_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); -void ff_mdct_calc_neon(MDCTContext *s, FFTSample *output, const FFTSample *input); void ff_mdct_end(MDCTContext *s); /* Real Discrete Fourier Transform */ diff -r 89cd870ca180 -r 5cf49858179a fft.c --- a/fft.c Tue Sep 15 16:18:16 2009 +0000 +++ b/fft.c Tue Sep 15 21:14:14 2009 +0000 @@ -62,7 +62,6 @@ { int i, j, m, n; float alpha, c1, s1, s2; - int split_radix = 1; int av_unused has_vectors; if (nbits < 2 || nbits > 16) @@ -87,41 +86,13 @@ s->imdct_half = ff_imdct_half_c; s->mdct_calc = ff_mdct_calc_c; s->exptab1 = NULL; + s->split_radix = 1; -#if HAVE_MMX && HAVE_YASM - has_vectors = mm_support(); - if (has_vectors & FF_MM_SSE && HAVE_SSE) { - /* SSE for P3/P4/K8 */ - s->imdct_calc = ff_imdct_calc_sse; - s->imdct_half = ff_imdct_half_sse; - s->fft_permute = ff_fft_permute_sse; - s->fft_calc = ff_fft_calc_sse; - } else if (has_vectors & FF_MM_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 & FF_MM_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; - } -#elif HAVE_ALTIVEC - has_vectors = mm_support(); - if (has_vectors & FF_MM_ALTIVEC) { - s->fft_calc = ff_fft_calc_altivec; - split_radix = 0; - } -#elif HAVE_NEON - s->fft_permute = ff_fft_permute_neon; - s->fft_calc = ff_fft_calc_neon; - s->imdct_calc = ff_imdct_calc_neon; - s->imdct_half = ff_imdct_half_neon; - s->mdct_calc = ff_mdct_calc_neon; -#endif + if (ARCH_ARM) ff_fft_init_arm(s); + if (HAVE_ALTIVEC) ff_fft_init_altivec(s); + if (HAVE_MMX) ff_fft_init_mmx(s); - if (split_radix) { + if (s->split_radix) { for(j=4; j<=nbits; j++) { int m = 1<nbits >= 6); } + +av_cold void ff_fft_init_altivec(FFTContext *s) +{ + s->fft_calc = ff_fft_calc_altivec; + s->split_radix = 0; +} diff -r 89cd870ca180 -r 5cf49858179a x86/fft.c --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/x86/fft.c Tue Sep 15 21:14:14 2009 +0000 @@ -0,0 +1,44 @@ +/* + * 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 "libavcodec/dsputil.h" +#include "fft.h" + +av_cold void ff_fft_init_mmx(FFTContext *s) +{ +#if HAVE_YASM + int has_vectors = mm_support(); + if (has_vectors & FF_MM_SSE && HAVE_SSE) { + /* SSE for P3/P4/K8 */ + s->imdct_calc = ff_imdct_calc_sse; + s->imdct_half = ff_imdct_half_sse; + s->fft_permute = ff_fft_permute_sse; + s->fft_calc = ff_fft_calc_sse; + } else if (has_vectors & FF_MM_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 & FF_MM_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 +} diff -r 89cd870ca180 -r 5cf49858179a x86/fft.h --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/x86/fft.h Tue Sep 15 21:14:14 2009 +0000 @@ -0,0 +1,36 @@ +/* + * 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 + */ + +#ifndef AVCODEC_X86_FFT_H +#define AVCODEC_X86_FFT_H + +#include "libavcodec/dsputil.h" + +void ff_fft_permute_sse(FFTContext *s, FFTComplex *z); +void ff_fft_calc_sse(FFTContext *s, FFTComplex *z); +void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z); +void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z); + +void ff_imdct_calc_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_half_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_calc_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_half_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output, const FFTSample *input); +void ff_imdct_half_sse(MDCTContext *s, FFTSample *output, const FFTSample *input); + +#endif diff -r 89cd870ca180 -r 5cf49858179a x86/fft_3dn2.c --- a/x86/fft_3dn2.c Tue Sep 15 16:18:16 2009 +0000 +++ b/x86/fft_3dn2.c Tue Sep 15 21:14:14 2009 +0000 @@ -21,6 +21,7 @@ #include "libavutil/x86_cpu.h" #include "libavcodec/dsputil.h" +#include "fft.h" DECLARE_ALIGNED_8(static const int, m1m1[2]) = { 1<<31, 1<<31 }; diff -r 89cd870ca180 -r 5cf49858179a x86/fft_sse.c --- a/x86/fft_sse.c Tue Sep 15 16:18:16 2009 +0000 +++ b/x86/fft_sse.c Tue Sep 15 21:14:14 2009 +0000 @@ -21,6 +21,7 @@ #include "libavutil/x86_cpu.h" #include "libavcodec/dsputil.h" +#include "fft.h" DECLARE_ALIGNED(16, static const int, m1m1m1m1[4]) = { 1 << 31, 1 << 31, 1 << 31, 1 << 31 };