/* * FFT/MDCT transform with SSE 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 */ #include "libavutil/x86_cpu.h" #include "libavutil/common.h" #include "libavcodec/dsputil.h" #include "fft.h" DECLARE_ALIGNED(16, static const int, m1m1m1m1)[4] = { 1 << 31, 1 << 31, 1 << 31, 1 << 31 }; void ff_fft_dispatch_sse(FFTComplex *z, int nbits); void ff_fft_dispatch_interleave_sse(FFTComplex *z, int nbits); void ff_fft_calc_sse(FFTContext *s, FFTComplex *z) { int n = 1 << s->nbits; ff_fft_dispatch_interleave_sse(z, s->nbits); if(n <= 16) { x86_reg i = -8*n; __asm__ volatile( "1: \n" "movaps (%0,%1), %%xmm0 \n" "movaps %%xmm0, %%xmm1 \n" "unpcklps 16(%0,%1), %%xmm0 \n" "unpckhps 16(%0,%1), %%xmm1 \n" "movaps %%xmm0, (%0,%1) \n" "movaps %%xmm1, 16(%0,%1) \n" "add $32, %0 \n" "jl 1b \n" :"+r"(i) :"r"(z+n) :"memory" ); } } void ff_fft_permute_sse(FFTContext *s, FFTComplex *z) { int n = 1 << s->nbits; int i; for(i=0; itmp_buf[s->revtab[i]]), "=m"(s->tmp_buf[s->revtab[i+1]]) :"m"(z[i]) ); } memcpy(z, s->tmp_buf, n*sizeof(FFTComplex)); } void ff_imdct_half_sse(FFTContext *s, FFTSample *output, const FFTSample *input) { av_unused x86_reg i, j, k, l; long n = 1 << s->mdct_bits; long n2 = n >> 1; long n4 = n >> 2; long n8 = n >> 3; const uint16_t *revtab = s->revtab + n8; const FFTSample *tcos = s->tcos; const FFTSample *tsin = s->tsin; FFTComplex *z = (FFTComplex *)output; /* pre rotation */ for(k=n8-2; k>=0; k-=2) { __asm__ volatile( "movaps (%2,%1,2), %%xmm0 \n" // { z[k].re, z[k].im, z[k+1].re, z[k+1].im } "movaps -16(%2,%0,2), %%xmm1 \n" // { z[-k-2].re, z[-k-2].im, z[-k-1].re, z[-k-1].im } "movaps %%xmm0, %%xmm2 \n" "shufps $0x88, %%xmm1, %%xmm0 \n" // { z[k].re, z[k+1].re, z[-k-2].re, z[-k-1].re } "shufps $0x77, %%xmm2, %%xmm1 \n" // { z[-k-1].im, z[-k-2].im, z[k+1].im, z[k].im } "movlps (%3,%1), %%xmm4 \n" "movlps (%4,%1), %%xmm5 \n" "movhps -8(%3,%0), %%xmm4 \n" // { cos[k], cos[k+1], cos[-k-2], cos[-k-1] } "movhps -8(%4,%0), %%xmm5 \n" // { sin[k], sin[k+1], sin[-k-2], sin[-k-1] } "movaps %%xmm0, %%xmm2 \n" "movaps %%xmm1, %%xmm3 \n" "mulps %%xmm5, %%xmm0 \n" // re*sin "mulps %%xmm4, %%xmm1 \n" // im*cos "mulps %%xmm4, %%xmm2 \n" // re*cos "mulps %%xmm5, %%xmm3 \n" // im*sin "subps %%xmm0, %%xmm1 \n" // -> re "addps %%xmm3, %%xmm2 \n" // -> im "movaps %%xmm1, %%xmm0 \n" "unpcklps %%xmm2, %%xmm1 \n" // { z[k], z[k+1] } "unpckhps %%xmm2, %%xmm0 \n" // { z[-k-2], z[-k-1] } ::"r"(-4*k), "r"(4*k), "r"(input+n4), "r"(tcos+n8), "r"(tsin+n8) ); #if ARCH_X86_64 // if we have enough regs, don't let gcc make the luts latency-bound // but if not, latency is faster than spilling __asm__("movlps %%xmm0, %0 \n" "movhps %%xmm0, %1 \n" "movlps %%xmm1, %2 \n" "movhps %%xmm1, %3 \n" :"=m"(z[revtab[-k-2]]), "=m"(z[revtab[-k-1]]), "=m"(z[revtab[ k ]]), "=m"(z[revtab[ k+1]]) ); #else __asm__("movlps %%xmm0, %0" :"=m"(z[revtab[-k-2]])); __asm__("movhps %%xmm0, %0" :"=m"(z[revtab[-k-1]])); __asm__("movlps %%xmm1, %0" :"=m"(z[revtab[ k ]])); __asm__("movhps %%xmm1, %0" :"=m"(z[revtab[ k+1]])); #endif } ff_fft_dispatch_sse(z, s->nbits); /* post rotation + reinterleave + reorder */ #define CMUL(j,xmm0,xmm1)\ "movaps (%2,"#j",2), %%xmm6 \n"\ "movaps 16(%2,"#j",2), "#xmm0"\n"\ "movaps %%xmm6, "#xmm1"\n"\ "movaps "#xmm0",%%xmm7 \n"\ "mulps (%3,"#j"), %%xmm6 \n"\ "mulps (%4,"#j"), "#xmm0"\n"\ "mulps (%4,"#j"), "#xmm1"\n"\ "mulps (%3,"#j"), %%xmm7 \n"\ "subps %%xmm6, "#xmm0"\n"\ "addps %%xmm7, "#xmm1"\n" j = -n2; k = n2-16; __asm__ volatile( "1: \n" CMUL(%0, %%xmm0, %%xmm1) CMUL(%1, %%xmm4, %%xmm5) "shufps $0x1b, %%xmm1, %%xmm1 \n" "shufps $0x1b, %%xmm5, %%xmm5 \n" "movaps %%xmm4, %%xmm6 \n" "unpckhps %%xmm1, %%xmm4 \n" "unpcklps %%xmm1, %%xmm6 \n" "movaps %%xmm0, %%xmm2 \n" "unpcklps %%xmm5, %%xmm0 \n" "unpckhps %%xmm5, %%xmm2 \n" "movaps %%xmm6, (%2,%1,2) \n" "movaps %%xmm4, 16(%2,%1,2) \n" "movaps %%xmm0, (%2,%0,2) \n" "movaps %%xmm2, 16(%2,%0,2) \n" "sub $16, %1 \n" "add $16, %0 \n" "jl 1b \n" :"+&r"(j), "+&r"(k) :"r"(z+n8), "r"(tcos+n8), "r"(tsin+n8) :"memory" ); } void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input) { x86_reg j, k; long n = 1 << s->mdct_bits; long n4 = n >> 2; ff_imdct_half_sse(s, output+n4, input); j = -n; k = n-16; __asm__ volatile( "movaps %4, %%xmm7 \n" "1: \n" "movaps (%2,%1), %%xmm0 \n" "movaps (%3,%0), %%xmm1 \n" "shufps $0x1b, %%xmm0, %%xmm0 \n" "shufps $0x1b, %%xmm1, %%xmm1 \n" "xorps %%xmm7, %%xmm0 \n" "movaps %%xmm1, (%3,%1) \n" "movaps %%xmm0, (%2,%0) \n" "sub $16, %1 \n" "add $16, %0 \n" "jl 1b \n" :"+r"(j), "+r"(k) :"r"(output+n4), "r"(output+n4*3), "m"(*m1m1m1m1) ); } DECLARE_ALIGNED(16, static const float, b1)[] = { 0.500603, 0.505471, 0.515447, 0.531043, 0.553104, 0.582935, 0.622504, 0.674808, -1.169440, -0.972568, -0.839350, -0.744536, -10.190008, -3.407609, -2.057781, -1.484165, 0.502419, 0.522499, 0.566944, 0.646822, 0.788155, 1.060678, 1.722447, 5.101149, 0.509796, 0.601345, 0.899976, 2.562916, 1.000000, 1.000000, 1.306563, 0.541196, 1.000000, 0.707107, 1.000000, -0.707107 }; DECLARE_ALIGNED(16, static const int32_t, smask)[4] = { 0, 0, 0x80000000, 0x80000000 }; /* butterfly operator */ #define BUTTERFLY(a,b,c,tmp) \ "movaps %%" #a ", %%" #tmp " \n\t" \ "subps %%" #b ", %%" #a " \n\t" \ "addps %%" #tmp ", %%" #b " \n\t" \ "mulps " #c ", %%" #a " \n\t" ///* Same as BUTTERFLY when vectors a and b overlap */ #define BUTTERFLY0(val, mask, cos, tmp, shuf) \ "movaps %%" #val ", %%" #tmp " \n\t" \ "shufps " #shuf ", %%" #val ",%%" #val " \n\t" \ "xorps %%" #mask ", %%" #tmp " \n\t" /* flip signs */ \ "addps %%" #tmp ", %%" #val " \n\t" \ "mulps %%" #cos ", %%" #val " \n\t" #define BUTTERFLY2(val, mask, cos, tmp) BUTTERFLY0(val, mask, cos, tmp, $0x1b) #define BUTTERFLY3(val, mask, cos, tmp) BUTTERFLY0(val, mask, cos, tmp, $0xb1) void ff_dct32_float_sse(FFTSample *out, const FFTSample *in) { int32_t tmp1 = 0; __asm__ volatile( /* pass 1 */ "movaps (%4), %%xmm0 \n\t" "movaps 112(%4), %%xmm1 \n\t" "shufps $0x1b, %%xmm1, %%xmm1 \n\t" BUTTERFLY(xmm0, xmm1, (%2), xmm3) "movaps 64(%4), %%xmm7 \n\t" "movaps 48(%4), %%xmm4 \n\t" "shufps $0x1b, %%xmm4, %%xmm4 \n\t" BUTTERFLY(xmm7, xmm4, 48(%2), xmm3) /* pass 2 */ "movaps 64(%2), %%xmm2 \n\t" BUTTERFLY(xmm1, xmm4, %%xmm2, xmm3) "movaps %%xmm1, 48(%1) \n\t" "movaps %%xmm4, (%1) \n\t" /* pass 1 */ "movaps 16(%4), %%xmm1 \n\t" "movaps 96(%4), %%xmm6 \n\t" "shufps $0x1b, %%xmm6, %%xmm6 \n\t" BUTTERFLY(xmm1, xmm6, 16(%2), xmm3) "movaps 80(%4), %%xmm4 \n\t" "movaps 32(%4), %%xmm5 \n\t" "shufps $0x1b, %%xmm5, %%xmm5 \n\t" BUTTERFLY(xmm4, xmm5, 32(%2), xmm3) /* pass 2 */ BUTTERFLY(xmm0, xmm7, %%xmm2, xmm3) "movaps 80(%2), %%xmm2 \n\t" BUTTERFLY(xmm6, xmm5, %%xmm2, xmm3) BUTTERFLY(xmm1, xmm4, %%xmm2, xmm3) /* pass 3 */ "movaps 96(%2), %%xmm2 \n\t" "shufps $0x1b, %%xmm1, %%xmm1 \n\t" BUTTERFLY(xmm0, xmm1, %%xmm2, xmm3) "movaps %%xmm0, 112(%1) \n\t" "movaps %%xmm1, 96(%1) \n\t" "movaps 0(%1), %%xmm0 \n\t" "shufps $0x1b, %%xmm5, %%xmm5 \n\t" BUTTERFLY(xmm0, xmm5, %%xmm2, xmm3) "movaps 48(%1), %%xmm1 \n\t" "shufps $0x1b, %%xmm6, %%xmm6 \n\t" BUTTERFLY(xmm1, xmm6, %%xmm2, xmm3) "movaps %%xmm1, 48(%1) \n\t" "shufps $0x1b, %%xmm4, %%xmm4 \n\t" BUTTERFLY(xmm7, xmm4, %%xmm2, xmm3) /* pass 4 */ "movaps (%3), %%xmm3 \n\t" "movaps 112(%2), %%xmm2 \n\t" BUTTERFLY2(xmm5, xmm3, xmm2, xmm1) BUTTERFLY2(xmm0, xmm3, xmm2, xmm1) "movaps %%xmm0, 16(%1) \n\t" BUTTERFLY2(xmm6, xmm3, xmm2, xmm1) "movaps %%xmm6, 32(%1) \n\t" "movaps 48(%1), %%xmm0 \n\t" BUTTERFLY2(xmm0, xmm3, xmm2, xmm1) "movaps %%xmm0, 48(%1) \n\t" BUTTERFLY2(xmm4, xmm3, xmm2, xmm1) BUTTERFLY2(xmm7, xmm3, xmm2, xmm1) "movaps 96(%1), %%xmm6 \n\t" BUTTERFLY2(xmm6, xmm3, xmm2, xmm1) "movaps 112(%1), %%xmm0 \n\t" BUTTERFLY2(xmm0, xmm3, xmm2, xmm1) /* pass 5 */ "movaps 128(%2), %%xmm2 \n\t" "shufps $0xCC, %%xmm3,%%xmm3 \n\t" BUTTERFLY3(xmm5, xmm3, xmm2, xmm1) "movaps %%xmm5, (%1) \n\t" "movaps 16(%1), %%xmm1 \n\t" BUTTERFLY3(xmm1, xmm3, xmm2, xmm5) "movaps %%xmm1, 16(%1) \n\t" BUTTERFLY3(xmm4, xmm3, xmm2, xmm5) "movaps %%xmm4, 64(%1) \n\t" BUTTERFLY3(xmm7, xmm3, xmm2, xmm5) "movaps %%xmm7, 80(%1) \n\t" "movaps 32(%1), %%xmm5 \n\t" BUTTERFLY3(xmm5, xmm3, xmm2, xmm7) "movaps %%xmm5, 32(%1) \n\t" "movaps 48(%1), %%xmm4 \n\t" BUTTERFLY3(xmm4, xmm3, xmm2, xmm7) "movaps %%xmm4, 48(%1) \n\t" BUTTERFLY3(xmm6, xmm3, xmm2, xmm7) "movaps %%xmm6, 96(%1) \n\t" BUTTERFLY3(xmm0, xmm3, xmm2, xmm7) "movaps %%xmm0, 112(%1) \n\t" /* pass 6, no SIMD... */ "movss 56(%1), %%xmm3 \n\t" "movl 4(%1), %0 \n\t" "addss 60(%1), %%xmm3 \n\t" "movss 72(%1), %%xmm7 \n\t" "addss %%xmm3, %%xmm4 \n\t" "movss 52(%1), %%xmm2 \n\t" "addss %%xmm3, %%xmm2 \n\t" "movss 24(%1), %%xmm3 \n\t" "addss 28(%1), %%xmm3 \n\t" "addss 76(%1), %%xmm7 \n\t" "addss %%xmm3, %%xmm1 \n\t" "addss %%xmm4, %%xmm5 \n\t" "movss %%xmm1, 16(%1) \n\t" "movss 20(%1), %%xmm1 \n\t" "addss %%xmm3, %%xmm1 \n\t" "movss 40(%1), %%xmm3 \n\t" "movss %%xmm1, 48(%1) \n\t" "addss 44(%1), %%xmm3 \n\t" "movss 20(%1), %%xmm1 \n\t" "addss %%xmm3, %%xmm4 \n\t" "addss %%xmm2, %%xmm3 \n\t" "addss 28(%1), %%xmm1 \n\t" "movss %%xmm3, 40(%1) \n\t" "addss 36(%1), %%xmm2 \n\t" "movss 8(%1), %%xmm3 \n\t" "movss %%xmm2, 56(%1) \n\t" "addss 12(%1), %%xmm3 \n\t" "movss %%xmm5, 8(%1) \n\t" "movss %%xmm3, 32(%1) \n\t" "movss 52(%1), %%xmm2 \n\t" "movss 80(%1), %%xmm3 \n\t" "movss 120(%1), %%xmm5 \n\t" "movss %%xmm1, 80(%1) \n\t" "movss %%xmm4, 24(%1) \n\t" "addss 124(%1), %%xmm5 \n\t" "movss 64(%1), %%xmm1 \n\t" "addss 60(%1), %%xmm2 \n\t" "addss %%xmm5, %%xmm0 \n\t" "addss 116(%1), %%xmm5 \n\t" "movl %0, 64(%1) \n\t" "addss %%xmm0, %%xmm6 \n\t" "addss %%xmm6, %%xmm1 \n\t" "movl 12(%1), %0 \n\t" "movss %%xmm1, 4(%1) \n\t" "movss 88(%1), %%xmm1 \n\t" "movl %0, 96(%1) \n\t" "addss 92(%1), %%xmm1 \n\t" "movss 104(%1), %%xmm4 \n\t" "movl 28(%1), %0 \n\t" "addss 108(%1), %%xmm4 \n\t" "addss %%xmm4, %%xmm0 \n\t" "addss %%xmm1, %%xmm3 \n\t" "addss 84(%1), %%xmm1 \n\t" "addss %%xmm5, %%xmm4 \n\t" "addss %%xmm3, %%xmm6 \n\t" "addss %%xmm0, %%xmm3 \n\t" "addss %%xmm7, %%xmm0 \n\t" "addss 100(%1), %%xmm5 \n\t" "addss %%xmm4, %%xmm7 \n\t" "movl %0, 112(%1) \n\t" "movss %%xmm0, 28(%1) \n\t" "movss 36(%1), %%xmm0 \n\t" "movss %%xmm7, 36(%1) \n\t" "addss %%xmm1, %%xmm4 \n\t" "movss 116(%1), %%xmm7 \n\t" "addss %%xmm2, %%xmm0 \n\t" "addss 124(%1), %%xmm7 \n\t" "movss %%xmm0, 72(%1) \n\t" "movss 44(%1), %%xmm0 \n\t" "movss %%xmm6, 12(%1) \n\t" "movss %%xmm3, 20(%1) \n\t" "addss %%xmm0, %%xmm2 \n\t" "movss %%xmm4, 44(%1) \n\t" "movss %%xmm2, 88(%1) \n\t" "addss 60(%1), %%xmm0 \n\t" "movl 60(%1), %0 \n\t" "movl %0, 120(%1) \n\t" "movss %%xmm0, 104(%1) \n\t" "addss %%xmm5, %%xmm1 \n\t" "addss 68(%1), %%xmm5 \n\t" "movss %%xmm1, 52(%1) \n\t" "movss %%xmm5, 60(%1) \n\t" "movss 68(%1), %%xmm1 \n\t" "movss 100(%1), %%xmm5 \n\t" "addss %%xmm7, %%xmm5 \n\t" "addss 108(%1), %%xmm7 \n\t" "addss %%xmm5, %%xmm1 \n\t" "movss 84(%1), %%xmm2 \n\t" "addss 92(%1), %%xmm2 \n\t" "addss %%xmm2, %%xmm5 \n\t" "movss %%xmm1, 68(%1) \n\t" "addss %%xmm7, %%xmm2 \n\t" "movss 76(%1), %%xmm1 \n\t" "movss %%xmm2, 84(%1) \n\t" "movss %%xmm5, 76(%1) \n\t" "movss 108(%1), %%xmm2 \n\t" "addss %%xmm1, %%xmm7 \n\t" "addss 124(%1), %%xmm2 \n\t" "addss %%xmm2, %%xmm1 \n\t" "addss 92(%1), %%xmm2 \n\t" "movss %%xmm1, 100(%1) \n\t" "movss %%xmm2, 108(%1) \n\t" "movss 92(%1), %%xmm2 \n\t" "movss %%xmm7, 92(%1) \n\t" "addss 124(%1), %%xmm2 \n\t" "movss %%xmm2, 116(%1) \n\t" :"+&r"(tmp1) :"r"(out), "r"(b1), "r"(smask), "r"(in) :"memory" ); }