- support of dmix asm code for x86-64.

src/pcm/Makefile.am

@@ -14,7 +14,7 @@ libpcm_la_SOURCES = atomic.c mask.c interval.c \
 		    pcm_symbols.c
 noinst_HEADERS = pcm_local.h pcm_plugin.h mask.h mask_inline.h \
 	         interval.h interval_inline.h plugin_ops.h ladspa.h \
-		 pcm_direct.h pcm_dmix_i386.h
+		 pcm_direct.h pcm_dmix_i386.h pcm_dmix_x86_64.h
 
 alsadir = $(datadir)/alsa
 

src/pcm/pcm_direct.h

@@ -33,13 +33,13 @@
 
 typedef void (mix_areas1_t)(unsigned int size,
 			volatile signed short *dst, signed short *src,
-			volatile signed int *sum, unsigned int dst_step,
+			volatile signed int *sum, size_t dst_step,
-			unsigned int src_step, unsigned int sum_step);
+			size_t src_step, size_t sum_step);
 
 typedef void (mix_areas2_t)(unsigned int size,
 			volatile signed int *dst, signed int *src,
-			volatile signed int *sum, unsigned int dst_step,
+			volatile signed int *sum, size_t dst_step,
-			unsigned int src_step, unsigned int sum_step);
+			size_t src_step, size_t sum_step);
 
 struct slave_params {
 	snd_pcm_format_t format;

src/pcm/pcm_dmix.c

@@ -138,22 +138,20 @@ static void mix_select_callbacks(snd_pcm_direct_t *dmix)
 	char line[255];
 	int smp = 0, mmx = 0;
 	
-	/* safe settings for all i386 CPUs */
-	dmix->u.dmix.mix_areas1 = mix_areas1_smp;
 	/* try to determine, if we have a MMX capable CPU */
 	in = fopen("/proc/cpuinfo", "r");
-	if (in == NULL)
+	if (in) {
-		return;
+		while (!feof(in)) {
-	while (!feof(in)) {
+			fgets(line, sizeof(line), in);
-		fgets(line, sizeof(line), in);
+			if (!strncmp(line, "processor", 9))
-		if (!strncmp(line, "processor", 9))
+				smp++;
-			smp++;
+			else if (!strncmp(line, "flags", 5)) {
-		else if (!strncmp(line, "flags", 5)) {
+				if (strstr(line, " mmx"))
-			if (strstr(line, " mmx"))
+					mmx = 1;
-				mmx = 1;
+			}
 		}
+		fclose(in);
 	}
-	fclose(in);
 	// printf("MMX: %i, SMP: %i\n", mmx, smp);
 	if (mmx) {
 		dmix->u.dmix.mix_areas1 = smp > 1 ? mix_areas1_smp_mmx : mix_areas1_mmx;
@@ -164,12 +162,54 @@ static void mix_select_callbacks(snd_pcm_direct_t *dmix)
 }
 #endif
 
+#ifdef __x86_64__
+#define ADD_AND_SATURATE
+
+#define MIX_AREAS1 mix_areas1
+#define MIX_AREAS2 mix_areas2
+#define LOCK_PREFIX ""
+#include "pcm_dmix_x86_64.h"
+#undef MIX_AREAS1
+#undef MIX_AREAS2
+#undef LOCK_PREFIX
+
+#define MIX_AREAS1 mix_areas1_smp
+#define MIX_AREAS2 mix_areas2_smp
+#define LOCK_PREFIX "lock ; "
+#include "pcm_dmix_x86_64.h"
+#undef MIX_AREAS1
+#undef MIX_AREAS2
+#undef LOCK_PREFIX
+ 
+static void mix_select_callbacks(snd_pcm_direct_t *dmix)
+{
+	FILE *in;
+	char line[255];
+	int smp = 0;
+	
+	/* try to determine, if we have SMP */
+	in = fopen("/proc/cpuinfo", "r");
+	if (in) {
+		while (!feof(in)) {
+			fgets(line, sizeof(line), in);
+			if (!strncmp(line, "processor", 9))
+				smp++;
+		}
+		fclose(in);
+	}
+	// printf("SMP: %i\n", smp);
+	dmix->u.dmix.mix_areas1 = smp > 1 ? mix_areas1_smp : mix_areas1;
+	dmix->u.dmix.mix_areas2 = smp > 1 ? mix_areas2_smp : mix_areas2;
+}
+#endif
+
+
 #ifndef ADD_AND_SATURATE
 #warning Please, recode mix_areas1() routine to your architecture...
 static void mix_areas1(unsigned int size,
 		       volatile signed short *dst, signed short *src,
-		       volatile signed int *sum, unsigned int dst_step,
+		       volatile signed int *sum, size_t dst_step,
-		       unsigned int src_step, unsigned int sum_step)
+		       size_t src_step, size_t sum_step)
 {
 	register signed int sample, old_sample;
 
@@ -198,8 +238,8 @@ static void mix_areas1(unsigned int size,
 #warning Please, recode mix_areas2() routine to your architecture...
 static void mix_areas2(unsigned int size,
 		       volatile signed int *dst, signed int *src,
-		       volatile signed int *sum, unsigned int dst_step,
+		       volatile signed int *sum, size_t dst_step,
-		       unsigned int src_step, unsigned int sum_step)
+		       size_t src_step, size_t sum_step)
 {
 	register signed int sample, old_sample;
 

src/pcm/pcm_dmix_i386.h

@@ -31,8 +31,8 @@
  */
 static void MIX_AREAS1(unsigned int size,
 		       volatile signed short *dst, signed short *src,
-		       volatile signed int *sum, unsigned int dst_step,
+		       volatile signed int *sum, size_t dst_step,
-		       unsigned int src_step, unsigned int sum_step)
+		       size_t src_step, size_t sum_step)
 {
 	/*
 	 *  ESI - src
@@ -55,7 +55,7 @@ static void MIX_AREAS1(unsigned int size,
 		/*
 		 * while (size-- > 0) {
 		 */
-		"\tcmp $0, %0\n"
+		"\tcmpl $0, %0\n"
 		"jz 6f\n"
 
 		"\t.p2align 4,,15\n"
@@ -154,8 +154,8 @@ static void MIX_AREAS1(unsigned int size,
  */
 static void MIX_AREAS1_MMX(unsigned int size,
 			   volatile signed short *dst, signed short *src,
-			   volatile signed int *sum, unsigned int dst_step,
+			   volatile signed int *sum, size_t dst_step,
-			   unsigned int src_step, unsigned int sum_step)
+			   size_t src_step, size_t sum_step)
 {
 	/*
 	 *  ESI - src
@@ -178,7 +178,7 @@ static void MIX_AREAS1_MMX(unsigned int size,
 		/*
 		 * while (size-- > 0) {
 		 */
-		"\tcmp $0, %0\n"
+		"\tcmpl $0, %0\n"
 		"jz 6f\n"
 
 		"\t.p2align 4,,15\n"
@@ -244,8 +244,8 @@ static void MIX_AREAS1_MMX(unsigned int size,
  */
 static void MIX_AREAS2(unsigned int size,
 		       volatile signed int *dst, signed int *src,
-		       volatile signed int *sum, unsigned int dst_step,
+		       volatile signed int *sum, size_t dst_step,
-		       unsigned int src_step, unsigned int sum_step)
+		       size_t src_step, size_t sum_step)
 {
 	/*
 	 *  ESI - src
@@ -268,7 +268,7 @@ static void MIX_AREAS2(unsigned int size,
 		/*
 		 * while (size-- > 0) {
 		 */
-		"\tcmp $0, %0\n"
+		"\tcmpl $0, %0\n"
 		"jz 6f\n"
 
 		"\t.p2align 4,,15\n"

src/pcm/pcm_dmix_x86_64.h

@@ -0,0 +1,228 @@
+/**
+ * \file pcm/pcm_dmix_x86_64.h
+ * \ingroup PCM_Plugins
+ * \brief PCM Direct Stream Mixing (dmix) Plugin Interface - X86-64 assembler code
+ * \author Takashi Iwai <tiwai@suse.de>
+ * \date 2003
+ */
+/*
+ *  PCM - Direct Stream Mixing
+ *  Copyright (c) 2003 by Jaroslav Kysela <perex@suse.cz>
+ *                        Takashi Iwai <tiwai@suse.de>
+ *
+ *
+ *   This library is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU Lesser General Public License as
+ *   published by the Free Software Foundation; either version 2.1 of
+ *   the License, or (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU Lesser General Public License for more details.
+ *
+ *   You should have received a copy of the GNU Lesser General Public
+ *   License along with this library; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ *
+ */
+
+/*
+ *  MMX optimized
+ */
+static void MIX_AREAS1(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)
+{
+	/*
+	 *  ESI - src
+	 *  EDI - dst
+	 *  EBX - sum
+	 *  ECX - old sample
+	 *  EAX - sample / temporary
+	 *  EDX - temporary
+	 */
+	__asm__ __volatile__ (
+		"\n"
+
+		/*
+		 *  initialization, load ESI, EDI, EBX registers
+		 */
+		"\tmovq %1, %%rdi\n"
+		"\tmovq %2, %%rsi\n"
+		"\tmovq %3, %%rbx\n"
+
+		/*
+		 * while (size-- > 0) {
+		 */
+		"\tcmpl $0, %0\n"
+		"jz 6f\n"
+
+		"\t.p2align 4,,15\n"
+
+		"1:"
+
+		/*
+		 *   sample = *src;
+		 *   sum_sample = *sum;
+		 *   if (cmpxchg(*dst, 0, 1) == 0)
+		 *     sample -= sum_sample;
+		 *   xadd(*sum, sample);
+		 */
+		"\tmovw $0, %%ax\n"
+		"\tmovw $1, %%cx\n"
+		"\tmovl (%%rbx), %%edx\n"
+		"\t" LOCK_PREFIX "cmpxchgw %%cx, (%%rdi)\n"
+		"\tmovswl (%%rsi), %%ecx\n"
+		"\tjnz 2f\n"
+		"\tsubl %%edx, %%ecx\n"
+		"2:"
+		"\t" LOCK_PREFIX "addl %%ecx, (%%rbx)\n"
+
+		/*
+		 *   do {
+		 *     sample = old_sample = *sum;
+		 *     saturate(v);
+		 *     *dst = sample;
+		 *   } while (v != *sum);
+		 */
+
+		"3:"
+		"\tmovl (%%rbx), %%ecx\n"
+		"\tmovd %%ecx, %%mm0\n"
+		"\tpackssdw %%mm1, %%mm0\n"
+		"\tmovd %%mm0, %%eax\n"
+		"\tmovw %%ax, (%%rdi)\n"
+		"\tcmpl %%ecx, (%%rbx)\n"
+		"\tjnz 3b\n"
+
+		/*
+		 * while (size-- > 0)
+		 */
+		"\tadd %4, %%rdi\n"
+		"\tadd %5, %%rsi\n"
+		"\tadd %6, %%rbx\n"
+		"\tdecl %0\n"
+		"\tjnz 1b\n"
+		"\tjmp 6f\n"
+
+		"6:"
+		
+		"\temms\n"
+
+		: /* no output regs */
+		: "m" (size), "m" (dst), "m" (src), "m" (sum), "m" (dst_step), "m" (src_step), "m" (sum_step)
+		: "rsi", "rdi", "edx", "ecx", "rbx", "eax"
+	);
+}
+
+/*
+ *  32-bit version (24-bit resolution)
+ */
+static void MIX_AREAS2(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)
+{
+	/*
+	 *  ESI - src
+	 *  EDI - dst
+	 *  EBX - sum
+	 *  ECX - old sample
+	 *  EAX - sample / temporary
+	 *  EDX - temporary
+	 */
+	__asm__ __volatile__ (
+		"\n"
+
+		/*
+		 *  initialization, load ESI, EDI, EBX registers
+		 */
+		"\tmovq %1, %%rdi\n"
+		"\tmovq %2, %%rsi\n"
+		"\tmovq %3, %%rbx\n"
+
+		/*
+		 * while (size-- > 0) {
+		 */
+		"\tcmpl $0, %0\n"
+		"jz 6f\n"
+
+		"\t.p2align 4,,15\n"
+
+		"1:"
+
+		/*
+		 *   sample = *src;
+		 *   sum_sample = *sum;
+		 *   if (cmpxchg(*dst, 0, 1) == 0)
+		 *     sample -= sum_sample;
+		 *   xadd(*sum, sample);
+		 */
+		"\tmovl $0, %%eax\n"
+		"\tmovl $1, %%ecx\n"
+		"\tmovl (%%rbx), %%edx\n"
+		"\t" LOCK_PREFIX "cmpxchgl %%ecx, (%%rdi)\n"
+		"\tjnz 2f\n"
+		"\tmovl (%%rsi), %%ecx\n"
+		/* sample >>= 8 */
+		"\tsarl $8, %%ecx\n"
+		"\tsubl %%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"
+
+		/*
+		 *   do {
+		 *     sample = old_sample = *sum;
+		 *     saturate(v);
+		 *     *dst = sample;
+		 *   } while (v != *sum);
+		 */
+
+		"3:"
+		"\tmovl (%%rbx), %%ecx\n"
+		/*
+		 *  if (sample > 0x7fff00)
+		 */
+		"\tmovl $0x7fffff, %%eax\n"
+		"\tcmpl %%eax, %%ecx\n"
+		"\tjg 4f\n"
+		/*
+		 *  if (sample < -0x800000)
+		 */
+		"\tmovl $-0x800000, %%eax\n"
+		"\tcmpl %%eax, %%ecx\n"
+		"\tjl 4f\n"
+		"\tmovl %%ecx, %%eax\n"
+		"4:"
+		/*
+		 *  sample <<= 8;
+		 */
+		"\tsall $8, %%eax\n"
+		"\tmovl %%eax, (%%rdi)\n"
+		"\tcmpl %%ecx, (%%rbx)\n"
+		"\tjnz 3b\n"
+
+		/*
+		 * while (size-- > 0)
+		 */
+		"\tadd %4, %%rdi\n"
+		"\tadd %5, %%rsi\n"
+		"\tadd %6, %%rbx\n"
+		"\tdecl %0\n"
+		"\tjnz 1b\n"
+		// "\tjmp 6f\n"
+		
+		"6:"
+		: /* no output regs */
+		: "m" (size), "m" (dst), "m" (src), "m" (sum), "m" (dst_step), "m" (src_step), "m" (sum_step)
+		: "rsi", "rdi", "edx", "ecx", "rbx", "eax"
+	);
+}
+