#endif
#include <pulse/timeval.h>
+#include <pulse/rtclock.h>
+
#include <pulsecore/random.h>
#include <pulsecore/macro.h>
#include <pulsecore/g711.h>
#include <pulsecore/core-util.h>
+#include <pulsecore/endianmacros.h>
#include "cpu-x86.h"
#include "sample-util.h"
-#include "endianmacros.h"
#if defined (__i386__) || defined (__amd64__)
/* in s: 2 int16_t samples
" punpcklwd %%mm4, "#s" \n\t" /* .. | 0 | p0 | */ \
" pcmpgtw "#v", %%mm4 \n\t" /* .. | 0 | s(vl) | */ \
" pand "#s", %%mm4 \n\t" /* .. | 0 | (p0) | (vl >> 15) & p */ \
- " movq %%mm6, %%mm5 \n\t" /* .. | ffff | 0 | */ \
- " pand "#v", %%mm5 \n\t" /* .. | vh | 0 | */ \
- " por %%mm5, %%mm4 \n\t" /* .. | vh | (p0) | */ \
- " pmulhw "#s", "#v" \n\t" /* .. | 0 | vl*p0 | */ \
- " paddw %%mm4, "#v" \n\t" /* .. | vh | vl*p0 | vh + sign correct */ \
- " pslld $16, "#s" \n\t" /* .. | p0 | 0 | */ \
- " por %%mm7, "#s" \n\t" /* .. | p0 | 1 | */ \
- " pmaddwd "#s", "#v" \n\t" /* .. | p0 * v0 | */ \
+ " movq "#s", %%mm5 \n\t" \
+ " pmulhw "#v", "#s" \n\t" /* .. | 0 | vl*p0 | */ \
+ " paddw %%mm4, "#s" \n\t" /* .. | 0 | vl*p0 | + sign correct */ \
+ " pslld $16, "#s" \n\t" /* .. | vl*p0 | 0 | */ \
+ " psrld $16, "#v" \n\t" /* .. | 0 | vh | */ \
+ " psrad $16, "#s" \n\t" /* .. | vl*p0 | sign extend */ \
+ " pmaddwd %%mm5, "#v" \n\t" /* .. | p0 * vh | */ \
+ " paddd "#s", "#v" \n\t" /* .. | p0 * v0 | */ \
" packssdw "#v", "#v" \n\t" /* .. | p1*v1 | p0*v0 | */
/* approximately advances %3 = (%3 + a) % b. This function requires that
" por %%mm4, "#s1" \n\t" /* .. | l h | */ \
" por %%mm5, "#s2" \n\t"
-static void
-pa_volume_s16ne_mmx (int16_t *samples, int32_t *volumes, unsigned channels, unsigned length)
-{
+static void pa_volume_s16ne_mmx(int16_t *samples, int32_t *volumes, unsigned channels, unsigned length) {
pa_reg_x86 channel, temp;
- /* the max number of samples we process at a time, this is also the max amount
- * we overread the volume array, which should have enough padding. */
- channels = PA_MAX (4U, channels);
+ /* Channels must be at least 4, and always a multiple of the original number.
+ * This is also the max amount we overread the volume array, which should
+ * have enough padding. */
+ channels = channels == 3 ? 6 : PA_MAX (4U, channels);
__asm__ __volatile__ (
" xor %3, %3 \n\t"
" sar $1, %2 \n\t" /* length /= sizeof (int16_t) */
- " pcmpeqw %%mm6, %%mm6 \n\t" /* .. | ffff | ffff | */
- " pcmpeqw %%mm7, %%mm7 \n\t" /* .. | ffff | ffff | */
- " pslld $16, %%mm6 \n\t" /* .. | ffff | 0 | */
- " psrld $31, %%mm7 \n\t" /* .. | 0 | 1 | */
" test $1, %2 \n\t" /* check for odd samples */
" je 2f \n\t"
" emms \n\t"
: "+r" (samples), "+r" (volumes), "+r" (length), "=D" ((pa_reg_x86)channel), "=&r" (temp)
- : "r" ((pa_reg_x86)channels)
+ : "rm" ((pa_reg_x86)channels)
: "cc"
);
}
-static void
-pa_volume_s16re_mmx (int16_t *samples, int32_t *volumes, unsigned channels, unsigned length)
-{
+static void pa_volume_s16re_mmx(int16_t *samples, int32_t *volumes, unsigned channels, unsigned length) {
pa_reg_x86 channel, temp;
- /* the max number of samples we process at a time, this is also the max amount
- * we overread the volume array, which should have enough padding. */
- channels = PA_MAX (4U, channels);
+ /* Channels must be at least 4, and always a multiple of the original number.
+ * This is also the max amount we overread the volume array, which should
+ * have enough padding. */
+ channels = channels == 3 ? 6 : PA_MAX (4U, channels);
__asm__ __volatile__ (
" xor %3, %3 \n\t"
" emms \n\t"
: "+r" (samples), "+r" (volumes), "+r" (length), "=D" ((pa_reg_x86)channel), "=&r" (temp)
- : "r" ((pa_reg_x86)channels)
+ : "rm" ((pa_reg_x86)channels)
: "cc"
);
}
#ifdef RUN_TEST
#define CHANNELS 2
-#define SAMPLES 1021
+#define SAMPLES 1022
#define TIMES 1000
#define PADDING 16
-static void run_test (void) {
+static void run_test(void) {
int16_t samples[SAMPLES];
int16_t samples_ref[SAMPLES];
int16_t samples_orig[SAMPLES];
pa_do_volume_func_t func;
pa_usec_t start, stop;
- func = pa_get_volume_func (PA_SAMPLE_S16NE);
+ func = pa_get_volume_func(PA_SAMPLE_S16NE);
- printf ("checking MMX %zd\n", sizeof (samples));
+ printf("checking MMX %zd\n", sizeof(samples));
- pa_random (samples, sizeof (samples));
- memcpy (samples_ref, samples, sizeof (samples));
- memcpy (samples_orig, samples, sizeof (samples));
+ pa_random(samples, sizeof(samples));
+ /* for (i = 0; i < SAMPLES; i++)
+ samples[i] = -1; */
+ memcpy(samples_ref, samples, sizeof(samples));
+ memcpy(samples_orig, samples, sizeof(samples));
for (i = 0; i < CHANNELS; i++)
- volumes[i] = rand() >> 1;
+ volumes[i] = PA_CLAMP_VOLUME(rand() >> 1);
+ /* volumes[i] = 0x0000ffff; */
for (padding = 0; padding < PADDING; padding++, i++)
volumes[i] = volumes[padding];
- func (samples_ref, volumes, CHANNELS, sizeof (samples));
- pa_volume_s16ne_mmx (samples, volumes, CHANNELS, sizeof (samples));
+ func(samples_ref, volumes, CHANNELS, sizeof(samples));
+ pa_volume_s16ne_mmx(samples, volumes, CHANNELS, sizeof(samples));
for (i = 0; i < SAMPLES; i++) {
if (samples[i] != samples_ref[i]) {
- printf ("%d: %04x != %04x (%04x * %04x)\n", i, samples[i], samples_ref[i],
+ printf("%d: %04x != %04x (%04x * %08x)\n", i, samples[i], samples_ref[i],
samples_orig[i], volumes[i % CHANNELS]);
}
}
start = pa_rtclock_now();
for (j = 0; j < TIMES; j++) {
- memcpy (samples, samples_orig, sizeof (samples));
- pa_volume_s16ne_mmx (samples, volumes, CHANNELS, sizeof (samples));
+ memcpy(samples, samples_orig, sizeof(samples));
+ pa_volume_s16ne_mmx(samples, volumes, CHANNELS, sizeof(samples));
}
stop = pa_rtclock_now();
pa_log_info("MMX: %llu usec.", (long long unsigned int)(stop - start));
start = pa_rtclock_now();
for (j = 0; j < TIMES; j++) {
- memcpy (samples_ref, samples_orig, sizeof (samples));
- func (samples_ref, volumes, CHANNELS, sizeof (samples));
+ memcpy(samples_ref, samples_orig, sizeof(samples));
+ func(samples_ref, volumes, CHANNELS, sizeof(samples));
}
stop = pa_rtclock_now();
pa_log_info("ref: %llu usec.", (long long unsigned int)(stop - start));
+
+ pa_assert_se(memcmp(samples_ref, samples, sizeof(samples)) == 0);
}
#endif
#endif /* defined (__i386__) || defined (__amd64__) */
-void pa_volume_func_init_mmx (pa_cpu_x86_flag_t flags) {
+void pa_volume_func_init_mmx(pa_cpu_x86_flag_t flags) {
#if defined (__i386__) || defined (__amd64__)
- pa_log_info("Initialising MMX optimized functions.");
#ifdef RUN_TEST
- run_test ();
+ run_test();
#endif
- pa_set_volume_func (PA_SAMPLE_S16NE, (pa_do_volume_func_t) pa_volume_s16ne_mmx);
- pa_set_volume_func (PA_SAMPLE_S16RE, (pa_do_volume_func_t) pa_volume_s16re_mmx);
+ if ((flags & PA_CPU_X86_MMX) && (flags & PA_CPU_X86_CMOV)) {
+ pa_log_info("Initialising MMX optimized functions.");
+
+ pa_set_volume_func(PA_SAMPLE_S16NE, (pa_do_volume_func_t) pa_volume_s16ne_mmx);
+ pa_set_volume_func(PA_SAMPLE_S16RE, (pa_do_volume_func_t) pa_volume_s16re_mmx);
+ }
#endif /* defined (__i386__) || defined (__amd64__) */
}