X-Git-Url: https://code.delx.au/pulseaudio/blobdiff_plain/6076cef2092391d8b46aa84f86857cffebce4583..877ad8dcf8d3997f65ffe8ac4f544f6bdd8952e7:/src/pulsecore/remap_mmx.c diff --git a/src/pulsecore/remap_mmx.c b/src/pulsecore/remap_mmx.c index bfcae6c5..51a6da7e 100644 --- a/src/pulsecore/remap_mmx.c +++ b/src/pulsecore/remap_mmx.c @@ -24,9 +24,8 @@ #include #endif -#include - #include +#include #include #include @@ -51,7 +50,7 @@ " punpckl"#s" %%mm4, %%mm4 \n\t" \ " punpckh"#s" %%mm5, %%mm5 \n\t" \ " punpckl"#s" %%mm6, %%mm6 \n\t" \ - " punpckh"#s" %%mm7, %%mm7 \n\t" \ + " punpckh"#s" %%mm7, %%mm7 \n\t" #define STORE_SAMPLES \ " movq %%mm0, (%0) \n\t" \ @@ -65,84 +64,92 @@ " add $32, %1 \n\t" \ " add $64, %0 \n\t" -#define HANDLE_SINGLE(s) \ +#define HANDLE_SINGLE_dq() \ " movd (%1), %%mm0 \n\t" \ - " movq %%mm0, %%mm1 \n\t" \ - " punpckl"#s" %%mm0, %%mm0 \n\t" \ + " punpckldq %%mm0, %%mm0 \n\t" \ " movq %%mm0, (%0) \n\t" \ " add $4, %1 \n\t" \ " add $8, %0 \n\t" -#define MONO_TO_STEREO(s) \ - " mov %3, %2 \n\t" \ - " sar $3, %2 \n\t" \ - " cmp $0, %2 \n\t" \ - " je 2f \n\t" \ - "1: \n\t" \ - LOAD_SAMPLES \ - UNPACK_SAMPLES(s) \ - STORE_SAMPLES \ - " dec %2 \n\t" \ - " jne 1b \n\t" \ - "2: \n\t" \ - " mov %3, %2 \n\t" \ - " and $7, %2 \n\t" \ - " je 4f \n\t" \ - "3: \n\t" \ - HANDLE_SINGLE(s) \ - " dec %2 \n\t" \ - " jne 3b \n\t" \ - "4: \n\t" \ +#define HANDLE_SINGLE_wd() \ + " movw (%1), %w3 \n\t" \ + " movd %3, %%mm0 \n\t" \ + " punpcklwd %%mm0, %%mm0 \n\t" \ + " movd %%mm0, (%0) \n\t" \ + " add $2, %1 \n\t" \ + " add $4, %0 \n\t" + +#define MONO_TO_STEREO(s,shift,mask) \ + " mov %4, %2 \n\t" \ + " sar $"#shift", %2 \n\t" \ + " cmp $0, %2 \n\t" \ + " je 2f \n\t" \ + "1: \n\t" \ + LOAD_SAMPLES \ + UNPACK_SAMPLES(s) \ + STORE_SAMPLES \ + " dec %2 \n\t" \ + " jne 1b \n\t" \ + "2: \n\t" \ + " mov %4, %2 \n\t" \ + " and $"#mask", %2 \n\t" \ + " je 4f \n\t" \ + "3: \n\t" \ + HANDLE_SINGLE_##s() \ + " dec %2 \n\t" \ + " jne 3b \n\t" \ + "4: \n\t" \ " emms \n\t" -static void remap_mono_to_stereo_mmx (pa_remap_t *m, void *dst, const void *src, unsigned n) { - pa_reg_x86 temp; - - switch (*m->format) { - case PA_SAMPLE_FLOAT32NE: - { - __asm__ __volatile__ ( - MONO_TO_STEREO(dq) /* do doubles to quads */ - : "+r" (dst), "+r" (src), "=&r" (temp) - : "r" ((pa_reg_x86)n) - : "cc" - ); - break; - } - case PA_SAMPLE_S16NE: - { - __asm__ __volatile__ ( - MONO_TO_STEREO(wd) /* do words to doubles */ - : "+r" (dst), "+r" (src), "=&r" (temp) - : "r" ((pa_reg_x86)n) - : "cc" - ); - break; - } - default: - pa_assert_not_reached(); - } +#if defined (__i386__) || defined (__amd64__) +static void remap_mono_to_stereo_s16ne_mmx(pa_remap_t *m, int16_t *dst, const int16_t *src, unsigned n) { + pa_reg_x86 temp, temp2; + + __asm__ __volatile__ ( + MONO_TO_STEREO(wd,4,15) /* do words to doubles */ + : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2) + : "r" ((pa_reg_x86)n) + : "cc" + ); +} + +static void remap_mono_to_stereo_float32ne_mmx(pa_remap_t *m, float *dst, const float *src, unsigned n) { + pa_reg_x86 temp, temp2; + + __asm__ __volatile__ ( + MONO_TO_STEREO(dq,3,7) /* do doubles to quads */ + : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2) + : "r" ((pa_reg_x86)n) + : "cc" + ); } /* set the function that will execute the remapping based on the matrices */ -static void init_remap_mmx (pa_remap_t *m) { +static void init_remap_mmx(pa_remap_t *m) { unsigned n_oc, n_ic; - n_oc = m->o_ss->channels; - n_ic = m->i_ss->channels; + n_oc = m->o_ss.channels; + n_ic = m->i_ss.channels; /* find some common channel remappings, fall back to full matrix operation. */ if (n_ic == 1 && n_oc == 2 && - m->map_table_f[0][0] >= 1.0 && m->map_table_f[1][0] >= 1.0) { - m->do_remap = (pa_do_remap_func_t) remap_mono_to_stereo_mmx; + m->map_table_i[0][0] == 0x10000 && m->map_table_i[1][0] == 0x10000) { + pa_log_info("Using MMX mono to stereo remapping"); + pa_set_remap_func(m, (pa_do_remap_func_t) remap_mono_to_stereo_s16ne_mmx, + (pa_do_remap_func_t) remap_mono_to_stereo_float32ne_mmx); } } +#endif /* defined (__i386__) || defined (__amd64__) */ -void pa_remap_func_init_mmx (pa_cpu_x86_flag_t flags) { +void pa_remap_func_init_mmx(pa_cpu_x86_flag_t flags) { #if defined (__i386__) || defined (__amd64__) - pa_log_info("Initialising MMX optimized remappers."); - pa_set_init_remap_func ((pa_init_remap_func_t) init_remap_mmx); + if (flags & PA_CPU_X86_MMX) { + pa_log_info("Initialising MMX optimized remappers."); + + pa_set_init_remap_func((pa_init_remap_func_t) init_remap_mmx); + } + #endif /* defined (__i386__) || defined (__amd64__) */ }