* Check that OS support SSE optimization for PIII (to avoid illegal hardware instruct...

diff --git a/include/ac3_imdct.h b/include/ac3_imdct.h
index 4720653c15a96a50708bff19ae3da3e1df0d2ddb..f77afa183d0b388f5fe1c85481f8550e916cad5e 100644 (file)
--- a/include/ac3_imdct.h
+++ b/include/ac3_imdct.h
@@ -2,7 +2,7 @@
  * ac3_imdct.h : AC3 IMDCT types
  *****************************************************************************
  * Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct.h,v 1.3 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_imdct.h,v 1.4 2001/06/12 00:30:41 reno Exp $
  *
  * Authors: Michel Kaempf <maxx@via.ecp.fr>
  *          Renaud Dartus <reno@videolan.org>
@@ -31,29 +31,29 @@ typedef struct complex_s {
 
 typedef struct imdct_s
 {
-    complex_t buf[N/4];
+    complex_t buf[N/4] __attribute__ ((aligned(16)));
 
     /* Delay buffer for time domain interleaving */
-    float delay[6][256];
-    float delay1[6][256];
+    float delay[6][256] __attribute__ ((aligned(16)));
+    float delay1[6][256] __attribute__ ((aligned(16)));
 
     /* Twiddle factors for IMDCT */
-    float xcos1[N/4];
-    float xsin1[N/4];
-    float xcos2[N/8];
-    float xsin2[N/8];
+    float xcos1[N/4] __attribute__ ((aligned(16)));
+    float xsin1[N/4] __attribute__ ((aligned(16)));
+    float xcos2[N/8] __attribute__ ((aligned(16)));
+    float xsin2[N/8] __attribute__ ((aligned(16)));
    
     /* Twiddle factor LUT */
-    complex_t *w[7];
-    complex_t w_1[1];
-    complex_t w_2[2];
-    complex_t w_4[4];
-    complex_t w_8[8];
-    complex_t w_16[16];
-    complex_t w_32[32];
-    complex_t w_64[64];
+    complex_t *w[7] __attribute__ ((aligned(16)));
+    complex_t w_1[1] __attribute__ ((aligned(16)));
+    complex_t w_2[2] __attribute__ ((aligned(16)));
+    complex_t w_4[4] __attribute__ ((aligned(16)));
+    complex_t w_8[8] __attribute__ ((aligned(16)));
+    complex_t w_16[16] __attribute__ ((aligned(16)));
+    complex_t w_32[32] __attribute__ ((aligned(16)));
+    complex_t w_64[64] __attribute__ ((aligned(16)));
 
-    float xcos_sin_sse[128 * 4] __attribute__((aligned(16)));
+    float xcos_sin_sse[128 * 4] __attribute__ ((aligned(16)));
     
     /* Module used and shortcuts */
     struct module_s * p_module;

diff --git a/plugins/imdct/imdct3dn.c b/plugins/imdct/imdct3dn.c
index c4623a943054f06bd5117f5e07d59c697d625585..92c76cb0e432016d9db37441e325bbd942c933ee 100644 (file)
--- a/plugins/imdct/imdct3dn.c
+++ b/plugins/imdct/imdct3dn.c
@@ -2,7 +2,7 @@
  * imdct3dn.c : accelerated 3D Now! IMDCT module
  *****************************************************************************
  * Copyright (C) 1999, 2000 VideoLAN
- * $Id: imdct3dn.c,v 1.4 2001/06/03 12:47:21 sam Exp $
+ * $Id: imdct3dn.c,v 1.5 2001/06/12 00:30:41 reno Exp $
  *
  * Authors: Gaël Hendryckx <jimmy@via.ecp.fr>
  *
@@ -98,7 +98,7 @@ static int imdct_Probe( probedata_t *p_data )
         return( 0 );
     }
 
-    if( TestMethod( DOWNMIX_METHOD_VAR, "imdct3dn" ) )
+    if( TestMethod( IMDCT_METHOD_VAR, "imdct3dn" ) )
     {
         return( 999 );
     }

diff --git a/plugins/imdct/imdctsse.c b/plugins/imdct/imdctsse.c
index 8ac58c38328d53b11b02ddba2c032dc8975416d7..d5a244863c65011135c8cc0925a2116f9d2d4421 100644 (file)
--- a/plugins/imdct/imdctsse.c
+++ b/plugins/imdct/imdctsse.c
@@ -2,7 +2,7 @@
  * imdctsse.c : accelerated SSE IMDCT module
  *****************************************************************************
  * Copyright (C) 1999, 2000 VideoLAN
- * $Id: imdctsse.c,v 1.4 2001/06/03 12:47:21 sam Exp $
+ * $Id: imdctsse.c,v 1.5 2001/06/12 00:30:41 reno Exp $
  *
  * Authors: Gaël Hendryckx <jimmy@via.ecp.fr>
  *
@@ -98,7 +98,7 @@ static int imdct_Probe( probedata_t *p_data )
         return( 0 );
     }
 
-    if( TestMethod( IDCT_METHOD_VAR, "imdctsse" ) )
+    if( TestMethod( IMDCT_METHOD_VAR, "imdctsse" ) )
     {
         return( 999 );
     }

diff --git a/src/ac3_decoder/ac3_decoder.h b/src/ac3_decoder/ac3_decoder.h
index 9237d126524d854629e2131bd3099d2aee0fdc92..301de3664b7d3a451ac818d115aaf0249f76bbe0 100644 (file)
--- a/src/ac3_decoder/ac3_decoder.h
+++ b/src/ac3_decoder/ac3_decoder.h
@@ -2,7 +2,7 @@
  * ac3_decoder.h : ac3 decoder interface
  *****************************************************************************
  * Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_decoder.h,v 1.9 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_decoder.h,v 1.10 2001/06/12 00:30:41 reno Exp $
  *
  * Authors: Michel Kaempf <maxx@via.ecp.fr>
  *          Renaud Dartus <reno@videolan.org>
@@ -370,7 +370,7 @@ struct ac3dec_s
     bsi_t               bsi;
     audblk_t            audblk;
 
-    float               samples[6][256];
+    float               samples[6][256] __attribute__ ((aligned(16)));
     dm_par_t            dm_par;
 
     bit_allocate_t      bit_allocate;

diff --git a/src/interface/main.c b/src/interface/main.c
index 3f86e43a6317797bd6124466ed1bcb4c9a552eb1..ae93e1290d13a0f1dbb1fcceb4af7965b432f974 100644 (file)
--- a/src/interface/main.c
+++ b/src/interface/main.c
@@ -4,7 +4,7 @@
  * and spawn threads.
  *****************************************************************************
  * Copyright (C) 1998, 1999, 2000 VideoLAN
- * $Id: main.c,v 1.99 2001/05/31 01:37:08 sam Exp $
+ * $Id: main.c,v 1.100 2001/06/12 00:30:41 reno Exp $
  *
  * Authors: Vincent Seguin <seguin@via.ecp.fr>
  *          Samuel Hocevar <sam@zoy.org>
@@ -31,6 +31,7 @@
 
 #include <signal.h>                               /* SIGHUP, SIGINT, SIGKILL */
 #include <stdio.h>                                              /* sprintf() */
+#include <setjmp.h>                                       /* longjmp, setjmp */
 
 #ifdef HAVE_GETOPT_LONG
 #   ifdef HAVE_GETOPT_H
@@ -213,8 +214,10 @@ static void Version                 ( void );
 static void InitSignalHandler       ( void );
 static void SimpleSignalHandler     ( int i_signal );
 static void FatalSignalHandler      ( int i_signal );
-
+static void InstructionSignalHandler( int i_signal );
 static int  CPUCapabilities         ( void );
+static jmp_buf env;
+static int  i_illegal;
 
 /*****************************************************************************
  * main: parse command line, start interface and spawn threads
@@ -895,6 +898,31 @@ static void FatalSignalHandler( int i_signal )
     p_main->p_intf->b_die = 1;
 }
 
+/*****************************************************************************
+ * InstructionSignalHandler: system signal handler
+ *****************************************************************************
+ * This function is called when a illegal instruction signal is received by
+ * the program.
+ * We use this function to test OS and CPU_Capabilities
+ *****************************************************************************/
+static void InstructionSignalHandler( int i_signal )
+{
+    /* Once a signal has been trapped, the termination sequence will be
+     * armed and following signals will be ignored to avoid sending messages
+     * to an interface having been destroyed */
+
+    /* Acknowledge the signal received */
+    fprintf(stderr,"illegal instruction : optimization disable\n");
+
+    i_illegal = 1;
+    
+    sigrelse( i_signal );
+    longjmp( env, 1 );
+}
+
+
+
+
 /*****************************************************************************
  * CPUCapabilities: list the processors MMX support and other capabilities
  *****************************************************************************
@@ -942,6 +970,8 @@ static int CPUCapabilities( void )
     unsigned int  i_eax, i_ebx, i_ecx, i_edx;
     boolean_t     b_amd;
 
+    signal( SIGILL,  InstructionSignalHandler );
+    
 #   define cpuid( a )              \
     asm volatile ( "cpuid"         \
                  : "=a" ( i_eax ), \
@@ -967,6 +997,7 @@ static int CPUCapabilities( void )
 
     if( i_eax == i_ebx )
     {
+        signal( SIGILL, NULL );     
         return( i_capabilities );
     }
 
@@ -977,6 +1008,7 @@ static int CPUCapabilities( void )
 
     if( !i_eax )
     {
+        signal( SIGILL, NULL );     
         return( i_capabilities );
     }
 
@@ -992,6 +1024,7 @@ static int CPUCapabilities( void )
 
     if( ! (i_edx & 0x00800000) )
     {
+        signal( SIGILL, NULL );     
         return( i_capabilities );
     }
 
@@ -999,8 +1032,19 @@ static int CPUCapabilities( void )
 
     if( i_edx & 0x02000000 )
     {
+        
         i_capabilities |= CPU_CAPABILITY_MMXEXT;
-        i_capabilities |= CPU_CAPABILITY_SSE;
+
+        /* We test if OS support the SSE instructions */
+        i_illegal = 0;
+        if(setjmp(env)==0) { /* Test a SSE instruction */
+            __asm__ __volatile__ (
+            "xorps %%xmm0,%%xmm0\n"
+            ::);
+        }
+        
+        if( i_illegal != 1 )
+            i_capabilities |= CPU_CAPABILITY_SSE;
     }
     
     /* test for additional capabilities */
@@ -1008,6 +1052,7 @@ static int CPUCapabilities( void )
 
     if( i_eax < 0x80000001 )
     {
+        signal( SIGILL, NULL );     
         return( i_capabilities );
     }
 
@@ -1016,7 +1061,16 @@ static int CPUCapabilities( void )
 
     if( i_edx & 0x80000000 )
     {
-        i_capabilities |= CPU_CAPABILITY_3DNOW;
+        i_illegal = 0;
+        if(setjmp(env)==0) { /* Test a 3D Now! instruction */
+            __asm__ __volatile__ (
+            "pfadd %%mm0,%%mm0\n"
+            "femms\n"
+            ::);
+        }
+        
+        if( i_illegal != 1 ) 
+            i_capabilities |= CPU_CAPABILITY_3DNOW;
     }
 
     if( b_amd && ( i_edx & 0x00400000 ) )
@@ -1027,6 +1081,7 @@ static int CPUCapabilities( void )
     /* default behaviour */
 
 #endif
+    signal( SIGILL, NULL );     
     return( i_capabilities );
 }