Implemented snd_pcm_rewind() for the dmix plugin

src/pcm/pcm_direct.h

@@ -162,6 +162,10 @@ struct snd_pcm_direct {
 			mix_areas_32_t *mix_areas_32;
 			mix_areas_24_t *mix_areas_24;
 			mix_areas_u8_t *mix_areas_u8;
+			mix_areas_16_t *remix_areas_16;
+			mix_areas_32_t *remix_areas_32;
+			mix_areas_24_t *remix_areas_24;
+			mix_areas_u8_t *remix_areas_u8;
 		} dmix;
 		struct {
 		} dsnoop;

src/pcm/pcm_dmix.c

@@ -213,6 +213,70 @@ static void mix_areas(snd_pcm_direct_t *dmix,
 	}
 }
 
+static void remix_areas(snd_pcm_direct_t *dmix,
+			const snd_pcm_channel_area_t *src_areas,
+			const snd_pcm_channel_area_t *dst_areas,
+			snd_pcm_uframes_t src_ofs,
+			snd_pcm_uframes_t dst_ofs,
+			snd_pcm_uframes_t size)
+{
+	unsigned int src_step, dst_step;
+	unsigned int chn, dchn, channels, sample_size;
+	mix_areas_t *do_remix_areas;
+	
+	channels = dmix->channels;
+	switch (dmix->shmptr->s.format) {
+	case SND_PCM_FORMAT_S16_LE:
+	case SND_PCM_FORMAT_S16_BE:
+		sample_size = 2;
+		do_remix_areas = (mix_areas_t *)dmix->u.dmix.remix_areas_16;
+		break;
+	case SND_PCM_FORMAT_S32_LE:
+	case SND_PCM_FORMAT_S32_BE:
+		sample_size = 4;
+		do_remix_areas = (mix_areas_t *)dmix->u.dmix.remix_areas_32;
+		break;
+	case SND_PCM_FORMAT_S24_3LE:
+		sample_size = 3;
+		do_remix_areas = (mix_areas_t *)dmix->u.dmix.remix_areas_24;
+		break;
+	case SND_PCM_FORMAT_U8:
+		sample_size = 1;
+		do_remix_areas = (mix_areas_t *)dmix->u.dmix.remix_areas_u8;
+		break;
+	default:
+		return;
+	}
+	if (dmix->interleaved) {
+		/*
+		 * process all areas in one loop
+		 * it optimizes the memory accesses for this case
+		 */
+		do_remix_areas(size * channels,
+			       (unsigned char *)dst_areas[0].addr + sample_size * dst_ofs * channels,
+			       (unsigned char *)src_areas[0].addr + sample_size * src_ofs * channels,
+			       dmix->u.dmix.sum_buffer + dst_ofs * channels,
+			       sample_size,
+			       sample_size,
+			       sizeof(signed int));
+		return;
+	}
+	for (chn = 0; chn < channels; chn++) {
+		dchn = dmix->bindings ? dmix->bindings[chn] : chn;
+		if (dchn >= dmix->shmptr->s.channels)
+			continue;
+		src_step = src_areas[chn].step / 8;
+		dst_step = dst_areas[dchn].step / 8;
+		do_remix_areas(size,
+			       ((unsigned char *)dst_areas[dchn].addr + dst_areas[dchn].first / 8) + dst_ofs * dst_step,
+			       ((unsigned char *)src_areas[chn].addr + src_areas[chn].first / 8) + src_ofs * src_step,
+			       dmix->u.dmix.sum_buffer + channels * dst_ofs + chn,
+			       dst_step,
+			       src_step,
+			       channels * sizeof(signed int));
+	}
+}
+
 /*
  * if no concurrent access is allowed in the mixing routines, we need to protect
  * the area via semaphore
@@ -234,7 +298,7 @@ static void snd_pcm_dmix_sync_area(snd_pcm_t *pcm)
 {
 	snd_pcm_direct_t *dmix = pcm->private_data;
 	snd_pcm_uframes_t slave_hw_ptr, slave_appl_ptr, slave_size;
-	snd_pcm_uframes_t appl_ptr, size;
+	snd_pcm_uframes_t appl_ptr, size, transfer;
 	const snd_pcm_channel_area_t *src_areas, *dst_areas;
 	
 	/* calculate the size to transfer */
@@ -247,6 +311,27 @@ static void snd_pcm_dmix_sync_area(snd_pcm_t *pcm)
 	if (size >= pcm->boundary / 2)
 		size = pcm->boundary - size;
 
+	/* the slave_app_ptr can be far behing the slave_hw_ptr */
+	/* reduce mixing and errors here - just skip not catched writes */
+	if (dmix->slave_hw_ptr < dmix->slave_appl_ptr)
+		slave_size = dmix->slave_appl_ptr - dmix->slave_hw_ptr;
+	else
+		slave_size = dmix->slave_appl_ptr + (dmix->slave_boundary - dmix->slave_hw_ptr);
+	if (slave_size > dmix->slave_buffer_size) {
+		transfer = dmix->slave_buffer_size - slave_size;
+		if (transfer > size)
+			transfer = size;
+		dmix->last_appl_ptr += transfer;
+		dmix->last_appl_ptr %= pcm->boundary;
+		dmix->slave_appl_ptr += transfer;
+		dmix->slave_appl_ptr %= dmix->slave_boundary;
+		size = dmix->appl_ptr - dmix->last_appl_ptr;
+		if (! size)
+			return;
+		if (size >= pcm->boundary / 2)
+			size = pcm->boundary - size;
+	}
+
 	/* check the available size in the slave PCM buffer */
 	slave_hw_ptr = dmix->slave_hw_ptr;
 	/* don't write on the last active period - this area may be cleared
@@ -276,7 +361,7 @@ static void snd_pcm_dmix_sync_area(snd_pcm_t *pcm)
 	dmix->slave_appl_ptr %= dmix->slave_boundary;
 	dmix_down_sem(dmix);
 	for (;;) {
-		snd_pcm_uframes_t transfer = size;
+		transfer = size;
 		if (appl_ptr + transfer > pcm->buffer_size)
 			transfer = pcm->buffer_size - appl_ptr;
 		if (slave_appl_ptr + transfer > dmix->slave_buffer_size)
@@ -564,15 +649,78 @@ static int snd_pcm_dmix_pause(snd_pcm_t *pcm ATTRIBUTE_UNUSED, int enable ATTRIB
 	return -EIO;
 }
 
-static snd_pcm_sframes_t snd_pcm_dmix_rewind(snd_pcm_t *pcm ATTRIBUTE_UNUSED, snd_pcm_uframes_t frames ATTRIBUTE_UNUSED)
+static snd_pcm_sframes_t snd_pcm_dmix_rewind(snd_pcm_t *pcm, snd_pcm_uframes_t frames)
 {
-#if 0
+	snd_pcm_direct_t *dmix = pcm->private_data;
-	/* FIXME: substract samples from the mix ring buffer, too? */
+	snd_pcm_uframes_t slave_appl_ptr, slave_size;
+	snd_pcm_uframes_t appl_ptr, size, transfer, result;
+	const snd_pcm_channel_area_t *src_areas, *dst_areas;
+
+	if (dmix->state == SND_PCM_STATE_RUNNING ||
+	    dmix->state == SND_PCM_STATE_DRAINING)
+	    	return snd_pcm_dmix_hwsync(pcm);
+
+	if (dmix->last_appl_ptr < dmix->appl_ptr)
+		size = dmix->appl_ptr - dmix->last_appl_ptr;
+	else
+		size = dmix->appl_ptr + (pcm->boundary - dmix->last_appl_ptr);
+	if (frames < size)
+		size = frames;
+	snd_pcm_mmap_appl_backward(pcm, size);
+	frames -= size;
+	if (!frames)
+		return size;
+	result = size;
+
+	if (dmix->hw_ptr < dmix->appl_ptr)
+		size = dmix->appl_ptr - dmix->hw_ptr;
+	else
+		size = dmix->appl_ptr + (pcm->boundary - dmix->hw_ptr);
+	if (size > frames)
+		size = frames;
+	if (dmix->slave_hw_ptr < dmix->slave_appl_ptr)
+		slave_size = dmix->slave_appl_ptr - dmix->slave_hw_ptr;
+	else
+		slave_size = dmix->slave_appl_ptr + (pcm->boundary - dmix->slave_hw_ptr);
+	if (slave_size < size)
+		size = slave_size;
+	frames -= size;
+	result += size;
+		
+	/* add sample areas here */
+	src_areas = snd_pcm_mmap_areas(pcm);
+	dst_areas = snd_pcm_mmap_areas(dmix->spcm);
+	dmix->last_appl_ptr -= size;
+	dmix->last_appl_ptr %= pcm->boundary;
+	appl_ptr = dmix->last_appl_ptr % pcm->buffer_size;
+	dmix->slave_appl_ptr -= size;
+	dmix->slave_appl_ptr %= dmix->slave_boundary;
+	slave_appl_ptr = dmix->slave_appl_ptr % dmix->slave_buffer_size;
+	dmix_down_sem(dmix);
+	for (;;) {
+		transfer = size;
+		if (appl_ptr + transfer > pcm->buffer_size)
+			transfer = pcm->buffer_size - appl_ptr;
+		if (slave_appl_ptr + transfer > dmix->slave_buffer_size)
+			transfer = dmix->slave_buffer_size - slave_appl_ptr;
+		remix_areas(dmix, src_areas, dst_areas, appl_ptr, slave_appl_ptr, transfer);
+		size -= transfer;
+		if (! size)
+			break;
+		slave_appl_ptr += transfer;
+		slave_appl_ptr %= dmix->slave_buffer_size;
+		appl_ptr += transfer;
+		appl_ptr %= pcm->buffer_size;
+	}
+	dmix->last_appl_ptr -= frames;
+	dmix->last_appl_ptr %= pcm->boundary;
+	dmix->slave_appl_ptr -= frames;
+	dmix->slave_appl_ptr %= dmix->slave_boundary;
+	dmix_up_sem(dmix);
+
 	snd_pcm_mmap_appl_backward(pcm, frames);
-	return frames;
+
-#else
+	return result + frames;
-	return -EIO;
-#endif
 }
 
 static snd_pcm_sframes_t snd_pcm_dmix_forward(snd_pcm_t *pcm, snd_pcm_uframes_t frames)

src/pcm/pcm_dmix_generic.c

@@ -87,7 +87,7 @@ static void mix_areas_32(unsigned int size,
 	register signed int sample, old_sample;
 
 	for (;;) {
-		sample = *src / 256;
+		sample = *src >> 8;
 		old_sample = *sum;
 		if (ARCH_CMPXCHG(dst, 0, 1) == 0)
 			sample -= old_sample;
@@ -159,6 +159,37 @@ static void generic_mix_areas_16_native(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_16_native(unsigned int size,
+					  volatile signed short *dst,
+					  signed short *src,
+					  volatile signed int *sum,
+					  size_t dst_step,
+					  size_t src_step,
+					  size_t sum_step)
+{
+	register signed int sample;
+
+	for (;;) {
+		sample = *src;
+		if (! *dst) {
+			*sum = -sample;
+			*dst = *src;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7fff)
+				sample = 0x7fff;
+			else if (sample < -0x8000)
+				sample = -0x8000;
+			*dst = sample;
+		}
+		if (!--size)
+			return;
+		src = (signed short *) ((char *)src + src_step);
+		dst = (signed short *) ((char *)dst + dst_step);
+		sum = (signed int *)   ((char *)sum + sum_step);
+	}
+}
+
 static void generic_mix_areas_32_native(unsigned int size,
 					volatile signed int *dst,
 					signed int *src,
@@ -170,7 +201,7 @@ static void generic_mix_areas_32_native(unsigned int size,
 	register signed int sample;
 
 	for (;;) {
-		sample = *src / 256;
+		sample = *src >> 8;
 		if (! *dst) {
 			*sum = sample;
 			*dst = *src;
@@ -193,6 +224,39 @@ static void generic_mix_areas_32_native(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_32_native(unsigned int size,
+					  volatile signed int *dst,
+					  signed int *src,
+					  volatile signed int *sum,
+					  size_t dst_step,
+					  size_t src_step,
+					  size_t sum_step)
+{
+	register signed int sample;
+
+	for (;;) {
+		sample = *src >> 8;
+		if (! *dst) {
+			*sum = -sample;
+			*dst = *src;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7fffff)
+				sample = 0x7fffffff;
+			else if (sample < -0x800000)
+				sample = -0x80000000;
+			else
+				sample *= 256;
+			*dst = sample;
+		}
+		if (!--size)
+			return;
+		src = (signed int *) ((char *)src + src_step);
+		dst = (signed int *) ((char *)dst + dst_step);
+		sum = (signed int *) ((char *)sum + sum_step);
+	}
+}
+
 static void generic_mix_areas_16_swap(unsigned int size,
 				      volatile signed short *dst,
 				      signed short *src,
@@ -225,6 +289,37 @@ static void generic_mix_areas_16_swap(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_16_swap(unsigned int size,
+				        volatile signed short *dst,
+				        signed short *src,
+				        volatile signed int *sum,
+				        size_t dst_step,
+				        size_t src_step,
+				        size_t sum_step)
+{
+	register signed int sample;
+
+	for (;;) {
+		sample = (signed short) bswap_16(*src);
+		if (! *dst) {
+			*sum = -sample;
+			*dst = *src;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7fff)
+				sample = 0x7fff;
+			else if (sample < -0x8000)
+				sample = -0x8000;
+			*dst = (signed short) bswap_16((signed short) sample);
+		}
+		if (!--size)
+			return;
+		src = (signed short *) ((char *)src + src_step);
+		dst = (signed short *) ((char *)dst + dst_step);
+		sum = (signed int *)   ((char *)sum + sum_step);
+	}
+}
+
 static void generic_mix_areas_32_swap(unsigned int size,
 				      volatile signed int *dst,
 				      signed int *src,
@@ -236,7 +331,7 @@ static void generic_mix_areas_32_swap(unsigned int size,
 	register signed int sample;
 
 	for (;;) {
-		sample = bswap_32(*src) / 256;
+		sample = bswap_32(*src) >> 8;
 		if (! *dst) {
 			*sum = sample;
 			*dst = *src;
@@ -259,6 +354,39 @@ static void generic_mix_areas_32_swap(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_32_swap(unsigned int size,
+				        volatile signed int *dst,
+				        signed int *src,
+				        volatile signed int *sum,
+				        size_t dst_step,
+				        size_t src_step,
+				        size_t sum_step)
+{
+	register signed int sample;
+
+	for (;;) {
+		sample = bswap_32(*src) >> 8;
+		if (! *dst) {
+			*sum = -sample;
+			*dst = *src;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7fffff)
+				sample = 0x7fffffff;
+			else if (sample < -0x800000)
+				sample = -0x80000000;
+			else
+				sample *= 256;
+			*dst = bswap_32(sample);
+		}
+		if (!--size)
+			return;
+		src = (signed int *) ((char *)src + src_step);
+		dst = (signed int *) ((char *)dst + dst_step);
+		sum = (signed int *) ((char *)sum + sum_step);
+	}
+}
+
 /* always little endian */
 static void generic_mix_areas_24(unsigned int size,
 				 volatile unsigned char *dst,
@@ -293,6 +421,38 @@ static void generic_mix_areas_24(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_24(unsigned int size,
+				   volatile unsigned char *dst,
+				   unsigned char *src,
+				   volatile signed int *sum,
+				   size_t dst_step,
+				   size_t src_step,
+				   size_t sum_step)
+{
+	register signed int sample;
+
+	for (;;) {
+		sample = src[0] | (src[1] << 8) | (((signed char *)src)[2] << 16);
+		if (!(dst[0] | dst[1] | dst[2])) {
+			*sum = -sample;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7fffff)
+				sample = 0x7fffff;
+			else if (sample < -0x800000)
+				sample = -0x800000;
+		}
+		dst[0] = sample;
+		dst[1] = sample >> 8;
+		dst[2] = sample >> 16;
+		if (!--size)
+			return;
+		dst += dst_step;
+		src += src_step;
+		sum = (signed int *) ((char *)sum + sum_step);
+	}
+}
+
 static void generic_mix_areas_u8(unsigned int size,
 				 volatile unsigned char *dst,
 				 unsigned char *src,
@@ -322,18 +482,52 @@ static void generic_mix_areas_u8(unsigned int size,
 	}
 }
 
+static void generic_remix_areas_u8(unsigned int size,
+				   volatile unsigned char *dst,
+				   unsigned char *src,
+				   volatile signed int *sum,
+				   size_t dst_step,
+				   size_t src_step,
+				   size_t sum_step)
+{
+	for (;;) {
+		register int sample = *src - 0x80;
+		if (*dst == 0x80) {
+			*sum = -sample;
+		} else {
+			*sum = sample = *sum - sample;
+			if (sample > 0x7f)
+				sample = 0x7f;
+			else if (sample < -0x80)
+				sample = -0x80;
+		}
+		*dst = sample + 0x80;
+		if (!--size)
+			return;
+		dst += dst_step;
+		src += src_step;
+		sum = (signed int *) ((char *)sum + sum_step);
+	}
+}
+
 
 static void generic_mix_select_callbacks(snd_pcm_direct_t *dmix)
 {
 	if (snd_pcm_format_cpu_endian(dmix->shmptr->s.format)) {
 		dmix->u.dmix.mix_areas_16 = generic_mix_areas_16_native;
 		dmix->u.dmix.mix_areas_32 = generic_mix_areas_32_native;
+		dmix->u.dmix.remix_areas_16 = generic_remix_areas_16_native;
+		dmix->u.dmix.remix_areas_32 = generic_remix_areas_32_native;
 	} else {
 		dmix->u.dmix.mix_areas_16 = generic_mix_areas_16_swap;
 		dmix->u.dmix.mix_areas_32 = generic_mix_areas_32_swap;
+		dmix->u.dmix.remix_areas_16 = generic_remix_areas_16_swap;
+		dmix->u.dmix.remix_areas_32 = generic_remix_areas_32_swap;
 	}
 	dmix->u.dmix.mix_areas_24 = generic_mix_areas_24;
 	dmix->u.dmix.mix_areas_u8 = generic_mix_areas_u8;
+	dmix->u.dmix.remix_areas_24 = generic_remix_areas_24;
+	dmix->u.dmix.remix_areas_u8 = generic_remix_areas_u8;
 }
 
 #endif

src/pcm/pcm_dmix_i386.c

@@ -8,6 +8,8 @@
 #define MIX_AREAS_24 mix_areas_24
 #define MIX_AREAS_24_CMOV mix_areas_24_cmov
 #define LOCK_PREFIX ""
+#define XADD "addl"
+#define XSUB "subl"
 #include "pcm_dmix_i386.h"
 #undef MIX_AREAS_16
 #undef MIX_AREAS_16_MMX
@@ -15,6 +17,26 @@
 #undef MIX_AREAS_24
 #undef MIX_AREAS_24_CMOV
 #undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
+
+#define MIX_AREAS_16 remix_areas_16
+#define MIX_AREAS_16_MMX remix_areas_16_mmx
+#define MIX_AREAS_32 remix_areas_32
+#define MIX_AREAS_24 remix_areas_24
+#define MIX_AREAS_24_CMOV remix_areas_24_cmov
+#define LOCK_PREFIX ""
+#define XADD "subl"
+#define XSUB "addl"
+#include "pcm_dmix_i386.h"
+#undef MIX_AREAS_16
+#undef MIX_AREAS_16_MMX
+#undef MIX_AREAS_32
+#undef MIX_AREAS_24
+#undef MIX_AREAS_24_CMOV
+#undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
 
 #define MIX_AREAS_16 mix_areas_16_smp
 #define MIX_AREAS_16_MMX mix_areas_16_smp_mmx
@@ -22,6 +44,8 @@
 #define MIX_AREAS_24 mix_areas_24_smp
 #define MIX_AREAS_24_CMOV mix_areas_24_smp_cmov
 #define LOCK_PREFIX "lock ; "
+#define XADD "addl"
+#define XSUB "subl"
 #include "pcm_dmix_i386.h"
 #undef MIX_AREAS_16
 #undef MIX_AREAS_16_MMX
@@ -29,6 +53,26 @@
 #undef MIX_AREAS_24
 #undef MIX_AREAS_24_CMOV
 #undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
+ 
+#define MIX_AREAS_16 remix_areas_16_smp
+#define MIX_AREAS_16_MMX remix_areas_16_smp_mmx
+#define MIX_AREAS_32 remix_areas_32_smp
+#define MIX_AREAS_24 remix_areas_24_smp
+#define MIX_AREAS_24_CMOV remix_areas_24_smp_cmov
+#define LOCK_PREFIX "lock ; "
+#define XADD "subl"
+#define XSUB "addl"
+#include "pcm_dmix_i386.h"
+#undef MIX_AREAS_16
+#undef MIX_AREAS_16_MMX
+#undef MIX_AREAS_32
+#undef MIX_AREAS_24
+#undef MIX_AREAS_24_CMOV
+#undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
  
 #define i386_dmix_supported_format \
 	((1ULL << SND_PCM_FORMAT_S16_LE) |\
@@ -71,13 +115,18 @@ static void mix_select_callbacks(snd_pcm_direct_t *dmix)
 	
 	if (mmx) {
 		dmix->u.dmix.mix_areas_16 = smp > 1 ? mix_areas_16_smp_mmx : mix_areas_16_mmx;
+		dmix->u.dmix.remix_areas_16 = smp > 1 ? remix_areas_16_smp_mmx : remix_areas_16_mmx;
 	} else {
 		dmix->u.dmix.mix_areas_16 = smp > 1 ? mix_areas_16_smp : mix_areas_16;
+		dmix->u.dmix.remix_areas_16 = smp > 1 ? remix_areas_16_smp : remix_areas_16;
 	}
 	dmix->u.dmix.mix_areas_32 = smp > 1 ? mix_areas_32_smp : mix_areas_32;
+	dmix->u.dmix.remix_areas_32 = smp > 1 ? remix_areas_32_smp : remix_areas_32;
 	if (cmov) {
 		dmix->u.dmix.mix_areas_24 = smp > 1 ? mix_areas_24_smp_cmov : mix_areas_24_cmov;
+		dmix->u.dmix.remix_areas_24 = smp > 1 ? remix_areas_24_smp_cmov : remix_areas_24_cmov;
 	} else {
 		dmix->u.dmix.mix_areas_24 = smp > 1 ? mix_areas_24_smp: mix_areas_24;
+		dmix->u.dmix.remix_areas_24 = smp > 1 ? remix_areas_24_smp: remix_areas_24;
 	}
 }

src/pcm/pcm_dmix_i386.h

@@ -83,9 +83,9 @@ static void MIX_AREAS_16(unsigned int size,
 		"\t" LOCK_PREFIX "cmpxchgw %%cx, (%%edi)\n"
 		"\tmovswl (%%esi), %%ecx\n"
 		"\tjnz 3f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"3:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%ebx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%ebx)\n"
 
 		/*
 		 *   do {
@@ -202,9 +202,9 @@ static void MIX_AREAS_16_MMX(unsigned int size,
 		"\t" LOCK_PREFIX "cmpxchgw %%cx, (%%edi)\n"
 		"\tmovswl (%%esi), %%ecx\n"
 		"\tjnz 3f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"3:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%ebx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%ebx)\n"
 
 		/*
 		 *   do {
@@ -291,14 +291,14 @@ static void MIX_AREAS_32(unsigned int size,
 		"\tmovl (%%esi), %%ecx\n"
 		/* sample >>= 8 */
 		"\tsarl $8, %%ecx\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"\tjmp 21f\n"
 		"2:"
 		"\tmovl (%%esi), %%ecx\n"
 		/* sample >>= 8 */
 		"\tsarl $8, %%ecx\n"
 		"21:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%ebx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%ebx)\n"
 
 		/*
 		 *   do {
@@ -403,9 +403,9 @@ static void MIX_AREAS_24(unsigned int size,
 		"\t" LOCK_PREFIX "btsl $0, (%%edi)\n"
 		"\tleal (%%ecx,%%eax,1), %%ecx\n"
 		"\tjc 2f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"2:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%ebx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%ebx)\n"
 
 		/*
 		 *   do {
@@ -508,9 +508,9 @@ static void MIX_AREAS_24_CMOV(unsigned int size,
 		"\t" LOCK_PREFIX "btsl $0, (%%edi)\n"
 		"\tleal (%%ecx,%%eax,1), %%ecx\n"
 		"\tjc 2f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"2:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%ebx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%ebx)\n"
 
 		/*
 		 *   do {

src/pcm/pcm_dmix_x86_64.c

@@ -6,21 +6,57 @@
 #define MIX_AREAS_32 mix_areas_32
 #define MIX_AREAS_24 mix_areas_24
 #define LOCK_PREFIX ""
+#define XADD "addl"
+#define XSUB "subl"
 #include "pcm_dmix_x86_64.h"
 #undef MIX_AREAS_16
 #undef MIX_AREAS_32
 #undef MIX_AREAS_24
 #undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
+
+#define MIX_AREAS_16 remix_areas_16
+#define MIX_AREAS_32 remix_areas_32
+#define MIX_AREAS_24 remix_areas_24
+#define LOCK_PREFIX ""
+#define XADD "subl"
+#define XSUB "addl"
+#include "pcm_dmix_x86_64.h"
+#undef MIX_AREAS_16
+#undef MIX_AREAS_32
+#undef MIX_AREAS_24
+#undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
 
 #define MIX_AREAS_16 mix_areas_16_smp
 #define MIX_AREAS_32 mix_areas_32_smp
 #define MIX_AREAS_24 mix_areas_24_smp
 #define LOCK_PREFIX "lock ; "
+#define XADD "addl"
+#define XSUB "subl"
 #include "pcm_dmix_x86_64.h"
 #undef MIX_AREAS_16
 #undef MIX_AREAS_32
 #undef MIX_AREAS_24
 #undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
+ 
+#define MIX_AREAS_16 remix_areas_16_smp
+#define MIX_AREAS_32 remix_areas_32_smp
+#define MIX_AREAS_24 remix_areas_24_smp
+#define LOCK_PREFIX "lock ; "
+#define XADD "subl"
+#define XSUB "addl"
+#include "pcm_dmix_x86_64.h"
+#undef MIX_AREAS_16
+#undef MIX_AREAS_32
+#undef MIX_AREAS_24
+#undef LOCK_PREFIX
+#undef XADD
+#undef XSUB
  
 #define x86_64_dmix_supported_format \
 	((1ULL << SND_PCM_FORMAT_S16_LE) |\

src/pcm/pcm_dmix_x86_64.h

@@ -79,9 +79,9 @@ static void MIX_AREAS_16(unsigned int size,
 		"\t" LOCK_PREFIX "cmpxchgw %%cx, (%%rdi)\n"
 		"\tmovswl (%%rsi), %%ecx\n"
 		"\tjnz 2f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"2:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%rbx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%rbx)\n"
 
 		/*
 		 *   do {
@@ -176,14 +176,14 @@ static void MIX_AREAS_32(unsigned int size,
 		"\tmovl (%%rsi), %%ecx\n"
 		/* sample >>= 8 */
 		"\tsarl $8, %%ecx\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"\tjmp 21f\n"
 		"2:"
 		"\tmovl (%%rsi), %%ecx\n"
 		/* sample >>= 8 */
 		"\tsarl $8, %%ecx\n"
 		"21:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%rbx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%rbx)\n"
 
 		/*
 		 *   do {
@@ -290,9 +290,9 @@ static void MIX_AREAS_24(unsigned int size,
 		"\t" LOCK_PREFIX "btsl $0, (%%rdi)\n"
 		"\t.byte 0x67, 0x8d, 0x0c, 0x01\n"
 		"\tjc 2f\n"
-		"\tsubl %%edx, %%ecx\n"
+		"\t" XSUB " %%edx, %%ecx\n"
 		"2:"
-		"\t" LOCK_PREFIX "addl %%ecx, (%%rbx)\n"
+		"\t" LOCK_PREFIX XADD " %%ecx, (%%rbx)\n"
 
 		/*
 		 *   do {