runtime cpudetect in liba52 way

git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@3909 b3059339-0415-0410-9bf9-f77b7e298cf2

5 files changed, 28 insertions, 21 deletions

diff --git a/liba52/imdct.c b/liba52/imdct.c
index 523db22e86..63036cb7a6 100644
--- a/liba52/imdct.c
+++ b/liba52/imdct.c
@@ -26,7 +26,6 @@
  */
 
 #include "config.h"
-#include "../cpudetect.h"
 
 #include <math.h>
 #include <stdio.h>
@@ -1107,9 +1106,9 @@ void imdct_init (uint32_t mm_accel)
     {
 	int i, j, k;
 
-	if(gCpuCaps.hasSSE) 		fprintf (stderr, "Using SSE optimized IMDCT transform\n");
-	else if(gCpuCaps.has3DNow) 	fprintf (stderr, "Using experimental 3DNow optimized IMDCT transform\n");
-	else		    		fprintf (stderr, "No accelerated IMDCT transform found\n");
+	if(mm_accel & MM_ACCEL_X86_SSE) 	fprintf (stderr, "Using SSE optimized IMDCT transform\n");
+	else if(mm_accel & MM_ACCEL_X86_3DNOW) 	fprintf (stderr, "Using 3DNow optimized IMDCT transform\n");
+	else		    			fprintf (stderr, "No accelerated IMDCT transform found\n");
 
 	/* Twiddle factors to turn IFFT into IMDCT */
 	for (i = 0; i < 128; i++) {
@@ -1185,8 +1184,8 @@ void imdct_init (uint32_t mm_accel)
 
 	imdct_512 = imdct_do_512;
 #ifdef ARCH_X86
-	if(gCpuCaps.hasSSE)		imdct_512 = imdct_do_512_sse;
-	else if(gCpuCaps.has3DNow)	imdct_512 = imdct_do_512_3dnow;
+	if(mm_accel & MM_ACCEL_X86_SSE)		imdct_512 = imdct_do_512_sse;
+	else if(mm_accel & MM_ACCEL_X86_3DNOW)	imdct_512 = imdct_do_512_3dnow;
 #endif // arch_x86
 	imdct_256 = imdct_do_256;
     }
diff --git a/liba52/resample.c b/liba52/resample.c
index 6d88b10a65..61a45ab5d4 100644
--- a/liba52/resample.c
+++ b/liba52/resample.c
@@ -18,8 +18,8 @@
 #include <inttypes.h>
 #include <stdio.h>
 #include "a52.h"
+#include "mm_accel.h"
 #include "../config.h"
-#include "../cpudetect.h"
 
 int (* a52_resample) (float * _f, int16_t * s16)=NULL;
 
@@ -592,12 +592,12 @@ void a52_resample_init(uint32_t mm_accel,int _flags,int _chans){
 
     if(a52_resample==NULL) // only once please ;)
     {
-	    if(gCpuCaps.hasMMX) fprintf(stderr, "Using MMX optimized resampler\n");
-	    else		fprintf(stderr, "No accelerated resampler found\n");
+	    if(mm_accel & MM_ACCEL_X86_MMX) 	fprintf(stderr, "Using MMX optimized resampler\n");
+	    else				fprintf(stderr, "No accelerated resampler found\n");
     }
     
 #ifdef ARCH_X86
-    if(gCpuCaps.hasMMX) a52_resample= a52_resample_MMX;
+    if(mm_accel & MM_ACCEL_X86_MMX) a52_resample= a52_resample_MMX;
 #else
     if(0);
 #endif
diff --git a/liba52/resample_c.c b/liba52/resample_c.c
index 6d88b10a65..61a45ab5d4 100644
--- a/liba52/resample_c.c
+++ b/liba52/resample_c.c
@@ -18,8 +18,8 @@
 #include <inttypes.h>
 #include <stdio.h>
 #include "a52.h"
+#include "mm_accel.h"
 #include "../config.h"
-#include "../cpudetect.h"
 
 int (* a52_resample) (float * _f, int16_t * s16)=NULL;
 
@@ -592,12 +592,12 @@ void a52_resample_init(uint32_t mm_accel,int _flags,int _chans){
 
     if(a52_resample==NULL) // only once please ;)
     {
-	    if(gCpuCaps.hasMMX) fprintf(stderr, "Using MMX optimized resampler\n");
-	    else		fprintf(stderr, "No accelerated resampler found\n");
+	    if(mm_accel & MM_ACCEL_X86_MMX) 	fprintf(stderr, "Using MMX optimized resampler\n");
+	    else				fprintf(stderr, "No accelerated resampler found\n");
     }
     
 #ifdef ARCH_X86
-    if(gCpuCaps.hasMMX) a52_resample= a52_resample_MMX;
+    if(mm_accel & MM_ACCEL_X86_MMX) a52_resample= a52_resample_MMX;
 #else
     if(0);
 #endif
diff --git a/liba52/resample_mmx.c b/liba52/resample_mmx.c
index 6d88b10a65..61a45ab5d4 100644
--- a/liba52/resample_mmx.c
+++ b/liba52/resample_mmx.c
@@ -18,8 +18,8 @@
 #include <inttypes.h>
 #include <stdio.h>
 #include "a52.h"
+#include "mm_accel.h"
 #include "../config.h"
-#include "../cpudetect.h"
 
 int (* a52_resample) (float * _f, int16_t * s16)=NULL;
 
@@ -592,12 +592,12 @@ void a52_resample_init(uint32_t mm_accel,int _flags,int _chans){
 
     if(a52_resample==NULL) // only once please ;)
     {
-	    if(gCpuCaps.hasMMX) fprintf(stderr, "Using MMX optimized resampler\n");
-	    else		fprintf(stderr, "No accelerated resampler found\n");
+	    if(mm_accel & MM_ACCEL_X86_MMX) 	fprintf(stderr, "Using MMX optimized resampler\n");
+	    else				fprintf(stderr, "No accelerated resampler found\n");
     }
     
 #ifdef ARCH_X86
-    if(gCpuCaps.hasMMX) a52_resample= a52_resample_MMX;
+    if(mm_accel & MM_ACCEL_X86_MMX) a52_resample= a52_resample_MMX;
 #else
     if(0);
 #endif
diff --git a/liba52/test.c b/liba52/test.c
index d77b716d35..13fca5bed7 100644
--- a/liba52/test.c
+++ b/liba52/test.c
@@ -11,6 +11,7 @@
 #include <inttypes.h>
 
 #include "a52.h"
+#include "mm_accel.h"
 #include "../cpudetect.h"
 
 static sample_t * samples;
@@ -51,6 +52,13 @@ long long t, sum=0, min=256*256*256*64;
     stdout= stderr; //EVIL HACK FIXME
     GetCpuCaps(&gCpuCaps);
     stdout= temp;
+//    gCpuCaps.hasMMX=0;
+//    gCpuCaps.hasSSE=0;
+    if(gCpuCaps.hasMMX) 	accel |= MM_ACCEL_X86_MMX;
+    if(gCpuCaps.hasMMX2) 	accel |= MM_ACCEL_X86_MMXEXT;
+    if(gCpuCaps.hasSSE) 	accel |= MM_ACCEL_X86_SSE;
+    if(gCpuCaps.has3DNow) 	accel |= MM_ACCEL_X86_3DNOW;
+//    if(gCpuCaps.has3DNowExt) 	accel |= MM_ACCEL_X86_3DNOWEXT;
     
     samples = a52_init (accel);
     if (samples == NULL) {
@@ -87,7 +95,7 @@ ENDTIMING
     buf_size=0;
 
     // decode:
-    flags=A52_STEREO; //A52_STEREO; // A52_DOLBY // A52_2F2R // A52_3F2R | A52_LFE
+    flags=A52_STEREO; //A52_STEREO; //A52_DOLBY; //A52_STEREO; // A52_DOLBY // A52_2F2R // A52_3F2R | A52_LFE
     channels=2;
     
     flags |= A52_ADJUST_LEVEL;
@@ -99,7 +107,7 @@ ENDTIMING
     // a52_dynrng (&state, NULL, NULL); // disable dynamic range compensation
 
 STARTTIMING
-    a52_resample_init(flags,channels);
+    a52_resample_init(accel,flags,channels);
     s16 = out_buf;
     for (i = 0; i < 6; i++) {
 	if (a52_block (&state, samples))
@@ -112,7 +120,7 @@ ENDTIMING
 if(sum<min) min=sum;
 sum=0;
 #endif
-    fwrite(out_buf,6*256*2*2,1,stdout);
+    fwrite(out_buf,6*256*2*channels,1,stdout);
 
 }