core: Refactor code to multiply s16 by volume

move code to function pa_mult_s16_volume() in sample-util.h
use 64 bit integers on 64 bit platforms (it's faster)

on i5, 2.5GHz (64-bit)

Running suite(s): Mult-s16
32 bit mult: 1272300 usec (avg: 12723, min = 12533, max = 18749, stddev = 620.48).
64 bit mult: 852241 usec (avg: 8522.41, min = 8420, max = 9148, stddev = 109.388).
100%: Checks: 1, Failures: 0, Errors: 0

on Pentium D, 3.4GHz (32-bit)

Running suite(s): Mult-s16
32 bit mult: 2228504 usec (avg: 22285, min = 18775, max = 29648, stddev = 3865.59).
64 bit mult: 5546861 usec (avg: 55468.6, min = 55028, max = 64924, stddev = 978.981).
100%: Checks: 1, Failures: 0, Errors: 0

on TI DM3730, Cortex-A8, 800MHz (32-bit)

Running suite(s): Mult-s16
32 bit mult: 23708900 usec (avg: 237089, min = 191864, max = 557312, stddev = 77503.6).
64 bit mult: 22190039 usec (avg: 221900, min = 177978, max = 480469, stddev = 68520.5).
100%: Checks: 1, Failures: 0, Errors: 0

there is a test program called mult-s16-test which checks that the functions compute the
same results, and compares runtime

Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>

src/pulsecore/mix.c

@@ -131,23 +131,10 @@ static void pa_mix_s16ne_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
 
         for (i = 0; i < nstreams; i++) {
             pa_mix_info *m = streams + i;
-            int32_t v, lo, hi, cv = m->linear[channel].i;
+            int32_t cv = m->linear[channel].i;
 
-            if (PA_LIKELY(cv > 0)) {
+            if (PA_LIKELY(cv > 0))
-
+                sum += pa_mult_s16_volume(*((int16_t*) m->ptr), m->linear[channel].i);
-                /* Multiplying the 32bit volume factor with the
-                 * 16bit sample might result in an 48bit value. We
-                 * want to do without 64 bit integers and hence do
-                 * the multiplication independently for the HI and
-                 * LO part of the volume. */
-
-                hi = cv >> 16;
-                lo = cv & 0xFFFF;
-
-                v = *((int16_t*) m->ptr);
-                v = ((v * lo) >> 16) + (v * hi);
-                sum += v;
-            }
             m->ptr = (uint8_t*) m->ptr + sizeof(int16_t);
         }
 
@@ -170,17 +157,10 @@ static void pa_mix_s16re_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
 
         for (i = 0; i < nstreams; i++) {
             pa_mix_info *m = streams + i;
-            int32_t v, lo, hi, cv = m->linear[channel].i;
+            int32_t cv = m->linear[channel].i;
 
-            if (PA_LIKELY(cv > 0)) {
+            if (PA_LIKELY(cv > 0))
-
+                sum += pa_mult_s16_volume(PA_INT16_SWAP(*((int16_t*) m->ptr)), cv);
-                hi = cv >> 16;
-                lo = cv & 0xFFFF;
-
-                v = PA_INT16_SWAP(*((int16_t*) m->ptr));
-                v = ((v * lo) >> 16) + (v * hi);
-                sum += v;
-            }
             m->ptr = (uint8_t*) m->ptr + sizeof(int16_t);
         }
 
@@ -207,7 +187,6 @@ static void pa_mix_s32ne_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = *((int32_t*) m->ptr);
                 v = (v * cv) >> 16;
                 sum += v;
@@ -238,7 +217,6 @@ static void pa_mix_s32re_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = PA_INT32_SWAP(*((int32_t*) m->ptr));
                 v = (v * cv) >> 16;
                 sum += v;
@@ -269,7 +247,6 @@ static void pa_mix_s24ne_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = (int32_t) (PA_READ24NE(m->ptr) << 8);
                 v = (v * cv) >> 16;
                 sum += v;
@@ -300,7 +277,6 @@ static void pa_mix_s24re_c(pa_mix_info streams[], unsigned nstreams, unsigned ch
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = (int32_t) (PA_READ24RE(m->ptr) << 8);
                 v = (v * cv) >> 16;
                 sum += v;
@@ -331,7 +307,6 @@ static void pa_mix_s24_32ne_c(pa_mix_info streams[], unsigned nstreams, unsigned
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = (int32_t) (*((uint32_t*)m->ptr) << 8);
                 v = (v * cv) >> 16;
                 sum += v;
@@ -362,7 +337,6 @@ static void pa_mix_s24_32re_c(pa_mix_info streams[], unsigned nstreams, unsigned
             int64_t v;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = (int32_t) (PA_UINT32_SWAP(*((uint32_t*) m->ptr)) << 8);
                 v = (v * cv) >> 16;
                 sum += v;
@@ -392,7 +366,6 @@ static void pa_mix_u8_c(pa_mix_info streams[], unsigned nstreams, unsigned chann
             int32_t v, cv = m->linear[channel].i;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = (int32_t) *((uint8_t*) m->ptr) - 0x80;
                 v = (v * cv) >> 16;
                 sum += v;
@@ -419,17 +392,10 @@ static void pa_mix_ulaw_c(pa_mix_info streams[], unsigned nstreams, unsigned cha
 
         for (i = 0; i < nstreams; i++) {
             pa_mix_info *m = streams + i;
-            int32_t v, hi, lo, cv = m->linear[channel].i;
+            int32_t cv = m->linear[channel].i;
 
-            if (PA_LIKELY(cv > 0)) {
+            if (PA_LIKELY(cv > 0))
-
+                sum += pa_mult_s16_volume(st_ulaw2linear16(*((uint8_t*) m->ptr)), cv);
-                hi = cv >> 16;
-                lo = cv & 0xFFFF;
-
-                v = (int32_t) st_ulaw2linear16(*((uint8_t*) m->ptr));
-                v = ((v * lo) >> 16) + (v * hi);
-                sum += v;
-            }
             m->ptr = (uint8_t*) m->ptr + 1;
         }
 
@@ -452,17 +418,10 @@ static void pa_mix_alaw_c(pa_mix_info streams[], unsigned nstreams, unsigned cha
 
         for (i = 0; i < nstreams; i++) {
             pa_mix_info *m = streams + i;
-            int32_t v, hi, lo, cv = m->linear[channel].i;
+            int32_t cv = m->linear[channel].i;
 
-            if (PA_LIKELY(cv > 0)) {
+            if (PA_LIKELY(cv > 0))
-
+                sum += pa_mult_s16_volume(st_alaw2linear16(*((uint8_t*) m->ptr)), cv);
-                hi = cv >> 16;
-                lo = cv & 0xFFFF;
-
-                v = (int32_t) st_alaw2linear16(*((uint8_t*) m->ptr));
-                v = ((v * lo) >> 16) + (v * hi);
-                sum += v;
-            }
             m->ptr = (uint8_t*) m->ptr + 1;
         }
 
@@ -488,7 +447,6 @@ static void pa_mix_float32ne_c(pa_mix_info streams[], unsigned nstreams, unsigne
             float v, cv = m->linear[channel].f;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = *((float*) m->ptr);
                 v *= cv;
                 sum += v;
@@ -517,7 +475,6 @@ static void pa_mix_float32re_c(pa_mix_info streams[], unsigned nstreams, unsigne
             float v, cv = m->linear[channel].f;
 
             if (PA_LIKELY(cv > 0)) {
-
                 v = PA_FLOAT32_SWAP(*(float*) m->ptr);
                 v *= cv;
                 sum += v;

src/pulsecore/sample-util.h

@@ -23,6 +23,9 @@
   USA.
 ***/
 
+#include <inttypes.h>
+#include <limits.h>
+
 #include <pulse/gccmacro.h>
 #include <pulse/sample.h>
 #include <pulse/volume.h>
@@ -53,6 +56,23 @@ void pa_deinterleave(const void *src, void *dst[], unsigned channels, size_t ss,
 
 void pa_sample_clamp(pa_sample_format_t format, void *dst, size_t dstr, const void *src, size_t sstr, unsigned n);
 
+static inline int32_t pa_mult_s16_volume(int16_t v, int32_t cv) {
+#if __WORDSIZE == 64 || ((ULONG_MAX) > (UINT_MAX))
+    /* Multiply with 64 bit integers on 64 bit platforms */
+    return (v * (int64_t) cv) >> 16;
+#else
+    /* Multiplying the 32 bit volume factor with the
+     * 16 bit sample might result in an 48 bit value. We
+     * want to do without 64 bit integers and hence do
+     * the multiplication independently for the HI and
+     * LO part of the volume. */
+
+    int32_t hi = cv >> 16;
+    int32_t lo = cv & 0xFFFF;
+    return ((v * lo) >> 16) + (v * hi);
+#endif
+}
+
 pa_usec_t pa_bytes_to_usec_round_up(uint64_t length, const pa_sample_spec *spec);
 size_t pa_usec_to_bytes_round_up(pa_usec_t t, const pa_sample_spec *spec);
 

src/pulsecore/svolume_c.c

@@ -24,7 +24,6 @@
 #include <config.h>
 #endif
 
-
 #include <pulsecore/macro.h>
 #include <pulsecore/g711.h>
 #include <pulsecore/endianmacros.h>
@@ -35,13 +34,8 @@ static void pa_volume_u8_c(uint8_t *samples, const int32_t *volumes, unsigned ch
     unsigned channel;
 
     for (channel = 0; length; length--) {
-        int32_t t, hi, lo;
+        int32_t t = pa_mult_s16_volume(*samples - 0x80, volumes[channel]);
 
-        hi = volumes[channel] >> 16;
-        lo = volumes[channel] & 0xFFFF;
-
-        t = (int32_t) *samples - 0x80;
-        t = ((t * lo) >> 16) + (t * hi);
         t = PA_CLAMP_UNLIKELY(t, -0x80, 0x7F);
         *samples++ = (uint8_t) (t + 0x80);
 
@@ -54,13 +48,8 @@ static void pa_volume_alaw_c(uint8_t *samples, const int32_t *volumes, unsigned
     unsigned channel;
 
     for (channel = 0; length; length--) {
-        int32_t t, hi, lo;
+        int32_t t = pa_mult_s16_volume(st_alaw2linear16(*samples), volumes[channel]);
 
-        hi = volumes[channel] >> 16;
-        lo = volumes[channel] & 0xFFFF;
-
-        t = (int32_t) st_alaw2linear16(*samples);
-        t = ((t * lo) >> 16) + (t * hi);
         t = PA_CLAMP_UNLIKELY(t, -0x8000, 0x7FFF);
         *samples++ = (uint8_t) st_13linear2alaw((int16_t) t >> 3);
 
@@ -73,13 +62,8 @@ static void pa_volume_ulaw_c(uint8_t *samples, const int32_t *volumes, unsigned
     unsigned channel;
 
     for (channel = 0; length; length--) {
-        int32_t t, hi, lo;
+        int32_t t = pa_mult_s16_volume(st_ulaw2linear16(*samples), volumes[channel]);
 
-        hi = volumes[channel] >> 16;
-        lo = volumes[channel] & 0xFFFF;
-
-        t = (int32_t) st_ulaw2linear16(*samples);
-        t = ((t * lo) >> 16) + (t * hi);
         t = PA_CLAMP_UNLIKELY(t, -0x8000, 0x7FFF);
         *samples++ = (uint8_t) st_14linear2ulaw((int16_t) t >> 2);
 
@@ -94,19 +78,8 @@ static void pa_volume_s16ne_c(int16_t *samples, const int32_t *volumes, unsigned
     length /= sizeof(int16_t);
 
     for (channel = 0; length; length--) {
-        int32_t t, hi, lo;
+        int32_t t = pa_mult_s16_volume(*samples, volumes[channel]);
 
-        /* Multiplying the 32bit volume factor with the 16bit
-         * sample might result in an 48bit value. We want to
-         * do without 64 bit integers and hence do the
-         * multiplication independently for the HI and LO part
-         * of the volume. */
-
-        hi = volumes[channel] >> 16;
-        lo = volumes[channel] & 0xFFFF;
-
-        t = (int32_t)(*samples);
-        t = ((t * lo) >> 16) + (t * hi);
         t = PA_CLAMP_UNLIKELY(t, -0x8000, 0x7FFF);
         *samples++ = (int16_t) t;
 
@@ -121,13 +94,8 @@ static void pa_volume_s16re_c(int16_t *samples, const int32_t *volumes, unsigned
     length /= sizeof(int16_t);
 
     for (channel = 0; length; length--) {
-        int32_t t, hi, lo;
+        int32_t t = pa_mult_s16_volume(PA_INT16_SWAP(*samples), volumes[channel]);
 
-        hi = volumes[channel] >> 16;
-        lo = volumes[channel] & 0xFFFF;
-
-        t = (int32_t) PA_INT16_SWAP(*samples);
-        t = ((t * lo) >> 16) + (t * hi);
         t = PA_CLAMP_UNLIKELY(t, -0x8000, 0x7FFF);
         *samples++ = PA_INT16_SWAP((int16_t) t);