comparison x86/fft_sse.c @ 10199:38ab367d4231 libavcodec

Merge FFTContext and MDCTContext

10198:78af613fc316

        );
    }
    memcpy(z, s->tmp_buf, n*sizeof(FFTComplex));
}

void ff_imdct_half_sse(MDCTContext *s, FFTSample *output, const FFTSample *input)
{
    av_unused x86_reg i, j, k, l;
    long n = 1 << s->nbits;
    long n2 = n >> 1;
    long n4 = n >> 2;
    long n8 = n >> 3;
    const uint16_t *revtab = s->fft.revtab + n8;
    const FFTSample *tcos = s->tcos;
    const FFTSample *tsin = s->tsin;
    FFTComplex *z = (FFTComplex *)output;

    /* pre rotation */

        __asm__("movlps %%xmm1, %0" :"=m"(z[revtab[ k  ]]));
        __asm__("movhps %%xmm1, %0" :"=m"(z[revtab[ k+1]]));
#endif
    }

    ff_fft_dispatch_sse(z, s->fft.nbits);

    /* post rotation + reinterleave + reorder */

#define CMUL(j,xmm0,xmm1)\
        "movaps   (%2,"#j",2), %%xmm6 \n"\

        :"r"(z+n8), "r"(tcos+n8), "r"(tsin+n8)
        :"memory"
    );
}

void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output, const FFTSample *input)
{
    x86_reg j, k;
    long n = 1 << s->nbits;
    long n4 = n >> 2;

    ff_imdct_half_sse(s, output+n4, input);

    j = -n;

10199:38ab367d4231

        );
    }
    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 */

        __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"\

        :"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;