comparison liba52/downmix.c @ 3740:dcbdad3fc5cd

mix31to2 & mix31toS in SSE

3739:b6040dbc751a

static void mix31to2 (sample_t * samples, sample_t bias)
{
    int i;
    sample_t common;

    for (i = 0; i < 256; i++) {
	common = samples[i + 256] + samples[i + 768] + bias;
	samples[i] += common;
	samples[i + 256] = samples[i + 512] + common;
    }
}

static void mix31toS (sample_t * samples, sample_t bias)
{
    int i;
    sample_t common, surround;

    for (i = 0; i < 256; i++) {
	common = samples[i + 256] + bias;
	surround = samples[i + 768];
	samples[i] += common - surround;
	samples[i + 256] = samples[i + 512] + common + surround;
    }
}

static void mix22toS (sample_t * samples, sample_t bias)
{
    int i;

3740:dcbdad3fc5cd

static void mix31to2 (sample_t * samples, sample_t bias)
{
    int i;
    sample_t common;

#ifdef HAVE_SSE
	asm volatile(
		"movlps %1, %%xmm7		\n\t"
		"shufps $0x00, %%xmm7, %%xmm7	\n\t"
		"movl $-1024, %%esi		\n\t"
		"1:				\n\t"
		"movaps 1024(%0, %%esi), %%xmm0	\n\t"  
		"addps 3072(%0, %%esi), %%xmm0	\n\t"  
		"addps %%xmm7, %%xmm0		\n\t" // common
		"movaps (%0, %%esi), %%xmm1	\n\t" 
		"movaps 2048(%0, %%esi), %%xmm2	\n\t"
		"addps %%xmm0, %%xmm1		\n\t"
		"addps %%xmm0, %%xmm2		\n\t"
		"movaps %%xmm1, (%0, %%esi)	\n\t"
		"movaps %%xmm2, 1024(%0, %%esi)	\n\t"
		"addl $16, %%esi		\n\t"
		" jnz 1b			\n\t"
	:: "r" (samples+256), "m" (bias)
	: "%esi"
	);
#else
    for (i = 0; i < 256; i++) {
	common = samples[i + 256] + samples[i + 768] + bias;
	samples[i] += common;
	samples[i + 256] = samples[i + 512] + common;
    }
#endif
}

static void mix31toS (sample_t * samples, sample_t bias)
{
    int i;
    sample_t common, surround;

#ifdef HAVE_SSE
	asm volatile(
		"movlps %1, %%xmm7		\n\t"
		"shufps $0x00, %%xmm7, %%xmm7	\n\t"
		"movl $-1024, %%esi		\n\t"
		"1:				\n\t"
		"movaps 1024(%0, %%esi), %%xmm0	\n\t"  
		"movaps 3072(%0, %%esi), %%xmm3	\n\t" // surround
		"addps %%xmm7, %%xmm0		\n\t" // common
		"movaps (%0, %%esi), %%xmm1	\n\t" 
		"movaps 2048(%0, %%esi), %%xmm2	\n\t"
		"addps %%xmm0, %%xmm1		\n\t"
		"addps %%xmm0, %%xmm2		\n\t"
		"subps %%xmm3, %%xmm1		\n\t"
		"addps %%xmm3, %%xmm2		\n\t"
		"movaps %%xmm1, (%0, %%esi)	\n\t"
		"movaps %%xmm2, 1024(%0, %%esi)	\n\t"
		"addl $16, %%esi		\n\t"
		" jnz 1b			\n\t"
	:: "r" (samples+256), "m" (bias)
	: "%esi"
	);
#else
    for (i = 0; i < 256; i++) {
	common = samples[i + 256] + bias;
	surround = samples[i + 768];
	samples[i] += common - surround;
	samples[i + 256] = samples[i + 512] + common + surround;
    }
#endif
}

static void mix22toS (sample_t * samples, sample_t bias)
{
    int i;