changeset 1393:5eef9e69b145

Beginning sse optimization of dct code.
author atmos4
date Fri, 27 Jul 2001 17:21:23 +0000
parents 20c1ff7da89c
children d9e3f91d6da9
files mp3lib/Makefile mp3lib/sr1.c
diffstat 2 files changed, 8 insertions(+), 3 deletions(-) [+]
line wrap: on
line diff
--- a/mp3lib/Makefile	Fri Jul 27 01:16:54 2001 +0000
+++ b/mp3lib/Makefile	Fri Jul 27 17:21:23 2001 +0000
@@ -12,6 +12,10 @@
 SRCS += decode_MMX.s dct64_MMX.s tabinit_MMX.s
 OBJS += decode_MMX.o dct64_MMX.o tabinit_MMX.o
 endif
+ifeq ($(TARGET_SSE),yes)
+SRCS += dct64_sse.s
+OBJS += dct64_sse.o
+endif
 ifeq ($(TARGET_3DNOW),yes)
 SRCS += dct36_3dnow.s dct64_3dnow.s
 OBJS += dct36_3dnow.o dct64_3dnow.o
--- a/mp3lib/sr1.c	Fri Jul 27 01:16:54 2001 +0000
+++ b/mp3lib/sr1.c	Fri Jul 27 17:21:23 2001 +0000
@@ -367,6 +367,7 @@
 extern void dct64_MMX( void );
 extern void dct64_MMX_3dnow( void );
 extern void dct64_MMX_3dnowex( void );
+extern void dct64_MMX_sse( void );
 void (*dct64_MMX_func)( void );
 
 // Init decoder tables.  Call first, once!
@@ -405,8 +406,8 @@
        Note: It's ok, Since K8 will have SSE2 support and will much faster
        of P4 ;) 
      */
-//      printf( "mp3lib: Using SSE%s! optimized decore.\n",(_isse>1?"2":""));
-      printf( "mp3lib: Using Pentium%s optimized decore.\n",(_i586>1?"-MMX":""));
+      printf( "mp3lib: Using SSE%s! optimized decore.\n",(_isse>1?"2":""));
+//      printf( "mp3lib: Using Pentium%s optimized decore.\n",(_i586>1?"-MMX":""));
     else
     if(_3dnow)
       printf( "mp3lib: Using AMD 3dnow%s! optimized decore.\n",(_3dnow>1?"-dsp(k7)":""));
@@ -443,7 +444,7 @@
   if(_isse)
   {
     synth_func=synth_1to1_MMX;
-    dct64_MMX_func=dct64_MMX;
+    dct64_MMX_func=dct64_MMX_sse;
   }    
   else
 #endif