diff --git a/TOOLS/fastmemcpybench.c b/TOOLS/fastmemcpybench.c
index 1ce5434bb8..35f7914285 100644
--- a/TOOLS/fastmemcpybench.c
+++ b/TOOLS/fastmemcpybench.c
@@ -86,7 +86,7 @@ static unsigned int GetTimer(){
 static inline unsigned long long int read_tsc( void )
 {
   unsigned long long int retval;
-  __asm__ __volatile ("rdtsc":"=A"(retval)::"memory");
+  __asm__ volatile ("rdtsc":"=A"(retval)::"memory");
   return retval;
 }
 
diff --git a/configure b/configure
index 818b4d8539..8f2fe0796f 100755
--- a/configure
+++ b/configure
@@ -1595,7 +1595,7 @@ if x86 && test "$_runtime_cpudetection" = no ; then
 void catch() { exit(1); }
 int main(void) {
   signal(SIGILL, catch);
-  __asm__ __volatile__ ("$3":::"memory"); return 0;
+  __asm__ volatile ("$3":::"memory"); return 0;
 }
 EOF
 
@@ -2443,7 +2443,7 @@ if arm ; then
   echocheck "ARMv5TE (Enhanced DSP Extensions)"
   if test $_armv5te = "auto" ; then
     cat > $TMPC << EOF
-int main(void) { __asm__ __volatile__ ("qadd r0, r0, r0"); return 0; }
+int main(void) { __asm__ volatile ("qadd r0, r0, r0"); return 0; }
 EOF
     _armv5te=no
     cc_check && _armv5te=yes
@@ -2453,7 +2453,7 @@ EOF
   echocheck "ARMv6 (SIMD instructions)"
   if test $_armv6 = "auto" ; then
     cat > $TMPC << EOF
-int main(void) { __asm__ __volatile__ ("sadd16 r0, r0, r0"); return 0; }
+int main(void) { __asm__ volatile ("sadd16 r0, r0, r0"); return 0; }
 EOF
     _armv6=no
     cc_check && _armv6=yes
@@ -2463,7 +2463,7 @@ EOF
   echocheck "ARM VFP"
   if test $_armvfp = "auto" ; then
     cat > $TMPC << EOF
-int main(void) { __asm__ __volatile__ ("fadds s0, s0, s0"); return 0; }
+int main(void) { __asm__ volatile ("fadds s0, s0, s0"); return 0; }
 EOF
     _armvfp=no
     cc_check && _armvfp=yes
@@ -2473,7 +2473,7 @@ EOF
   echocheck "iWMMXt (Intel XScale SIMD instructions)"
   if test $_iwmmxt = "auto" ; then
     cat > $TMPC << EOF
-int main(void) { __asm__ __volatile__ ("wunpckelub wr6, wr4"); return 0; }
+int main(void) { __asm__ volatile ("wunpckelub wr6, wr4"); return 0; }
 EOF
     _iwmmxt=no
     cc_check && _iwmmxt=yes
diff --git a/cpudetect.c b/cpudetect.c
index 488061cd85..43f5cc2a7d 100644
--- a/cpudetect.c
+++ b/cpudetect.c
@@ -57,7 +57,7 @@ static int has_cpuid(void)
 	long a, c;
 
 // code from libavcodec:
-    __asm__ __volatile__ (
+    __asm__ volatile (
                           /* See if CPUID instruction is supported ... */
                           /* ... Get copies of EFLAGS into eax and ecx */
                           "pushf\n\t"
@@ -85,14 +85,14 @@ static void
 do_cpuid(unsigned int ax, unsigned int *p)
 {
 #if 0
-	__asm__ __volatile(
+	__asm__ volatile(
 	"cpuid;"
 	: "=a" (p[0]), "=b" (p[1]), "=c" (p[2]), "=d" (p[3])
 	:  "0" (ax)
 	);
 #else
 // code from libavcodec:
-    __asm__ __volatile__
+    __asm__ volatile
 	("mov %%"REG_b", %%"REG_S"\n\t"
          "cpuid\n\t"
          "xchg %%"REG_b", %%"REG_S
@@ -400,7 +400,7 @@ static void check_os_katmai_support( void )
    if ( gCpuCaps.hasSSE ) {
       mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
       exc_fil = SetUnhandledExceptionFilter(win32_sig_handler_sse);
-      __asm__ __volatile ("xorps %xmm0, %xmm0");
+      __asm__ volatile ("xorps %xmm0, %xmm0");
       SetUnhandledExceptionFilter(exc_fil);
       mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
    }
@@ -409,7 +409,7 @@ static void check_os_katmai_support( void )
    if ( gCpuCaps.hasSSE ) {
       mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
       DosSetExceptionHandler( &RegRec );
-      __asm__ __volatile ("xorps %xmm0, %xmm0");
+      __asm__ volatile ("xorps %xmm0, %xmm0");
       DosUnsetExceptionHandler( &RegRec );
       mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
    }
@@ -432,8 +432,8 @@ static void check_os_katmai_support( void )
    if ( gCpuCaps.hasSSE ) {
       mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
 
-//      __asm__ __volatile ("xorps %%xmm0, %%xmm0");
-      __asm__ __volatile ("xorps %xmm0, %xmm0");
+//      __asm__ volatile ("xorps %%xmm0, %%xmm0");
+      __asm__ volatile ("xorps %xmm0, %xmm0");
 
       mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
    }
diff --git a/drivers/radeon_vid.c b/drivers/radeon_vid.c
index f07b7bbeb6..5514f1cb6c 100644
--- a/drivers/radeon_vid.c
+++ b/drivers/radeon_vid.c
@@ -129,7 +129,7 @@ static int swap_fourcc __initdata = 0;
 static inline double FastSin(double x)
 {
    register double res;
-   __asm__ __volatile("fsin":"=t"(res):"0"(x));
+   __asm__ volatile("fsin":"=t"(res):"0"(x));
    return res;
 }
 #undef sin
@@ -138,7 +138,7 @@ static inline double FastSin(double x)
 static inline double FastCos(double x)
 {
    register double res;
-   __asm__ __volatile("fcos":"=t"(res):"0"(x));
+   __asm__ volatile("fcos":"=t"(res):"0"(x));
    return res;
 }
 #undef cos
diff --git a/liba52/imdct_3dnow.h b/liba52/imdct_3dnow.h
index 4a54fde26e..2291158950 100644
--- a/liba52/imdct_3dnow.h
+++ b/liba52/imdct_3dnow.h
@@ -45,7 +45,7 @@ static void FFT_4_3DNOW(complex_t *x)
   /* delta_p = 1 here */
   /* x[k] = sum_{i=0..3} x[i] * w^{i*k}, w=e^{-2*pi/4} 
    */
-  __asm__ __volatile__(
+  __asm__ volatile(
 	"movq	24(%1), %%mm3\n\t"
 	"movq	8(%1), %%mm1\n\t"
 	"pxor	%2, %%mm3\n\t" /* mm3.re | -mm3.im */
@@ -90,7 +90,7 @@ static void FFT_8_3DNOW(complex_t *x)
    */
   complex_t wT1, wB1, wB2;
   
-  __asm__ __volatile__(
+  __asm__ volatile(
 	"movq	8(%2), %%mm0\n\t"
 	"movq	24(%2), %%mm1\n\t"
 	"movq	%%mm0, %0\n\t"  /* wT1 = x[1]; */
@@ -99,7 +99,7 @@ static void FFT_8_3DNOW(complex_t *x)
 	:"r"(x)
 	:"memory");
 
-  __asm__ __volatile__(
+  __asm__ volatile(
 	"movq	16(%0), %%mm2\n\t"
 	"movq	32(%0), %%mm3\n\t"
 	"movq	%%mm2, 8(%0)\n\t"  /* x[1] = x[2]; */
@@ -114,7 +114,7 @@ static void FFT_8_3DNOW(complex_t *x)
   
   /* x[0] x[4] x[2] x[6] */
   
-  __asm__ __volatile__(
+  __asm__ volatile(
       "movq	40(%1), %%mm0\n\t"
       "movq	%%mm0,	%%mm3\n\t"
       "movq	56(%1),	%%mm1\n\t"
@@ -153,7 +153,7 @@ static void FFT_8_3DNOW(complex_t *x)
       :"memory");
   
   /* x[1] x[5] */
-  __asm__ __volatile__ (
+  __asm__ volatile (
 	"movq	%6,	%%mm6\n\t"
 	"movq	%5,	%%mm7\n\t"
 	"movq	%1,	%%mm0\n\t"
@@ -203,7 +203,7 @@ static void FFT_8_3DNOW(complex_t *x)
 
 
   /* x[3] x[7] */
-  __asm__ __volatile__(
+  __asm__ volatile(
 	"movq	%1,	%%mm0\n\t"
 #ifdef HAVE_3DNOWEX
 	"pswapd	%3,	%%mm1\n\t"
@@ -358,13 +358,13 @@ imdct_do_512_3dnow
 	
     /* Pre IFFT complex multiply plus IFFT cmplx conjugate & reordering*/
 #if 1
-      __asm__ __volatile__ (
+      __asm__ volatile (
 	"movq %0, %%mm7\n\t"
 	::"m"(x_plus_minus_3dnow)
 	:"memory");
 	for( i=0; i < 128; i++) {
 		int j = pm128[i];
-	__asm__ __volatile__ (
+	__asm__ volatile (
 		"movd	%1, %%mm0\n\t"
 		"movd	%3, %%mm1\n\t"
 		"punpckldq %2, %%mm0\n\t" /* mm0 = data[256-2*j-1] | data[2*j]*/
@@ -394,7 +394,7 @@ imdct_do_512_3dnow
 		buf[i].im = (data[256-2*j-1] * xsin1[j] + data[2*j] * xcos1[j])*(-1.0);*/
 	}
 #else
-  __asm__ __volatile__ ("femms":::"memory");
+  __asm__ volatile ("femms":::"memory");
     for( i=0; i < 128; i++) {
 	/* z[i] = (X[256-2*i-1] + j * X[2*i]) * (xcos1[i] + j * xsin1[i]) ; */ 
 	int j= pm128[i];
@@ -435,14 +435,14 @@ imdct_do_512_3dnow
     
     /* Post IFFT complex multiply  plus IFFT complex conjugate*/
 #if 1  
-  __asm__ __volatile__ (
+  __asm__ volatile (
 	"movq %0, %%mm7\n\t"
 	"movq %1, %%mm6\n\t"
 	::"m"(x_plus_minus_3dnow),
 	"m"(x_minus_plus_3dnow)
 	:"eax","memory");
 	for (i=0; i < 128; i++) {
-	    __asm__ __volatile__ (
+	    __asm__ volatile (
 		"movq %1, %%mm0\n\t" /* ac3_buf[i].re | ac3_buf[i].im */
 		"movq %%mm0, %%mm1\n\t" /* ac3_buf[i].re | ac3_buf[i].im */
 #ifndef HAVE_3DNOWEX
@@ -473,7 +473,7 @@ imdct_do_512_3dnow
 		ac3_buf[i].im =(tmp_a_r * ac3_xsin1[i])  -  (tmp_a_i  * ac3_xcos1[i]);*/
 	}
 #else    
-  __asm__ __volatile__ ("femms":::"memory");
+  __asm__ volatile ("femms":::"memory");
     for( i=0; i < 128; i++) {
 	/* y[n] = z[n] * (xcos1[n] + j * xsin1[n]) ; */
 	tmp_a_r =        buf[i].real;
@@ -496,7 +496,7 @@ imdct_do_512_3dnow
 	);
 	for (i=0; i< 64; i++) {
 /* merge two loops in one to enable working of 2 decoders */
-	__asm__ __volatile__ (
+	__asm__ volatile (
 		"movd	516(%1), %%mm0\n\t"
 		"movd	(%1), %%mm1\n\t" /**data_ptr++=-buf[64+i].im**window_ptr+++*delay_ptr++;*/
 		"punpckldq (%2), %%mm0\n\t"/*data_ptr[128]=-buf[i].re*window_ptr[128]+delay_ptr[128];*/
@@ -520,7 +520,7 @@ imdct_do_512_3dnow
 	}
 	window_ptr += 128;
 #else    
-  __asm__ __volatile__ ("femms":::"memory");
+  __asm__ volatile ("femms":::"memory");
     for(i=0; i< 64; i++) { 
 	*data_ptr++   = -buf[64+i].imag   * *window_ptr++ + *delay_ptr++ + bias; 
 	*data_ptr++   =  buf[64-i-1].real * *window_ptr++ + *delay_ptr++ + bias; 
@@ -538,7 +538,7 @@ imdct_do_512_3dnow
 	for(i=0; i< 64; i++) {
 /* merge two loops in one to enable working of 2 decoders */
 	    window_ptr -=2;
-	    __asm__ __volatile__(
+	    __asm__ volatile(
 		"movd	508(%1), %%mm0\n\t"
 		"movd	(%1), %%mm1\n\t"
 		"punpckldq (%2), %%mm0\n\t"
@@ -565,9 +565,9 @@ imdct_do_512_3dnow
 		:"memory");
 		delay_ptr += 2;
 	}
-  __asm__ __volatile__ ("femms":::"memory");
+  __asm__ volatile ("femms":::"memory");
 #else    
-  __asm__ __volatile__ ("femms":::"memory");
+  __asm__ volatile ("femms":::"memory");
     for(i=0; i< 64; i++) { 
 	*delay_ptr++  = -buf[64+i].real   * *--window_ptr; 
 	*delay_ptr++  =  buf[64-i-1].imag * *--window_ptr; 
diff --git a/liba52/srfftp_3dnow.h b/liba52/srfftp_3dnow.h
index c5e6dd1c24..9f26c59b43 100644
--- a/liba52/srfftp_3dnow.h
+++ b/liba52/srfftp_3dnow.h
@@ -39,7 +39,7 @@ typedef struct
 }i_cmplx_t;
 
 #define TRANS_FILL_MM6_MM7_3DNOW()\
-    __asm__ __volatile__(\
+    __asm__ volatile(\
 	"movq	%1, %%mm7\n\t"\
 	"movq	%0, %%mm6\n\t"\
 	::"m"(x_plus_minus_3dnow),\
@@ -66,7 +66,7 @@ typedef struct
 
 #define TRANSZERO_3DNOW(A0,A4,A8,A12) \
 { \
-    __asm__ __volatile__(\
+    __asm__ volatile(\
 	"movq	%4, %%mm0\n\t" /* mm0 = wTB[0]*/\
 	"movq	%5, %%mm1\n\t" /* mm1 = wTB[k*2]*/ \
 	"movq	%%mm0, %%mm5\n\t"/*u.re = wTB[0].re + wTB[k*2].re;*/\
@@ -95,7 +95,7 @@ typedef struct
 
 #define TRANSHALF_16_3DNOW(A2,A6,A10,A14)\
 {\
-    __asm__ __volatile__(\
+    __asm__ volatile(\
 	"movq	%4, %%mm0\n\t"/*u.re = wTB[2].im + wTB[2].re;*/\
 	"movq	%%mm0, %%mm1\n\t"\
 	"pxor	%%mm7, %%mm1\n\t"\
@@ -136,7 +136,7 @@ typedef struct
 
 #define TRANS_3DNOW(A1,A5,A9,A13,WT,WB,D,D3)\
 { \
-    __asm__ __volatile__(\
+    __asm__ volatile(\
 	"movq	%1,	%%mm4\n\t"\
 	"movq	%%mm4,	%%mm5\n\t"\
 	"punpckldq %%mm4, %%mm4\n\t"/*mm4 = D.re | D.re */\
@@ -166,7 +166,7 @@ typedef struct
 	:\
 	:"m"(WT), "m"(D), "m"(WB), "m"(D3)\
 	:"memory");\
-    __asm__ __volatile__(\
+    __asm__ volatile(\
 	"movq	%4, %%mm0\n\t"/* a1 = A1*/\
 	"movq	%5, %%mm2\n\t"/* a1 = A5*/\
 	"movq	%%mm0, %%mm1\n\t"\
diff --git a/libmpcodecs/dec_video.c b/libmpcodecs/dec_video.c
index 3c31a01a7b..7d38891c1a 100644
--- a/libmpcodecs/dec_video.c
+++ b/libmpcodecs/dec_video.c
@@ -374,10 +374,10 @@ void *decode_video(sh_video_t *sh_video, unsigned char *start, int in_size,
     // some codecs are broken, and doesn't restore MMX state :(
     // it happens usually with broken/damaged files.
     if (gCpuCaps.has3DNow) {
-	__asm__ __volatile ("femms\n\t":::"memory");
+	__asm__ volatile ("femms\n\t":::"memory");
     }
     else if (gCpuCaps.hasMMX) {
-	__asm__ __volatile ("emms\n\t":::"memory");
+	__asm__ volatile ("emms\n\t":::"memory");
     }
 #endif
 
diff --git a/libmpcodecs/native/mmx.h b/libmpcodecs/native/mmx.h
index d9a6fb65cb..1661b318a4 100644
--- a/libmpcodecs/native/mmx.h
+++ b/libmpcodecs/native/mmx.h
@@ -57,24 +57,24 @@ typedef union {
 
 
 #define mmx_i2r(op,imm,reg) \
-        __asm__ __volatile__ (#op " %0, %%" #reg \
+        __asm__ volatile (#op " %0, %%" #reg \
                               : /* nothing */ \
                               : "i" (imm) )
 
 #define mmx_m2r(op, mem, reg) \
-        __asm__ __volatile__ (#op " %0, %%" #reg \
+        __asm__ volatile (#op " %0, %%" #reg \
                               : /* nothing */ \
                               : "m" (mem))
 
 #define mmx_r2m(op, reg, mem) \
-        __asm__ __volatile__ (#op " %%" #reg ", %0" \
+        __asm__ volatile (#op " %%" #reg ", %0" \
                               : "=m" (mem) \
                               : /* nothing */ )
 
 #define mmx_r2r(op, regs, regd) \
-        __asm__ __volatile__ (#op " %" #regs ", %" #regd)
+        __asm__ volatile (#op " %" #regs ", %" #regd)
 
 
-#define emms() __asm__ __volatile__ ("emms")
+#define emms() __asm__ volatile ("emms")
 
 #endif /* MPLAYER_MMX_H */
diff --git a/libvo/aclib_template.c b/libvo/aclib_template.c
index 353e7594cd..7288831172 100644
--- a/libvo/aclib_template.c
+++ b/libvo/aclib_template.c
@@ -84,7 +84,7 @@ If you have questions please contact with me: Nick Kurshev: nickols_k@mail.ru.
 #define small_memcpy(to,from,n)\
 {\
 register unsigned long int dummy;\
-__asm__ __volatile__(\
+__asm__ volatile(\
 	"rep; movsb"\
 	:"=&D"(to), "=&S"(from), "=&c"(dummy)\
 /* It's most portable way to notify compiler */\
@@ -153,7 +153,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 #endif
 #ifndef HAVE_ONLY_MMX1
         /* PREFETCH has effect even for MOVSB instruction ;) */
-	__asm__ __volatile__ (
+	__asm__ volatile (
 	        PREFETCH" (%0)\n"
 	        PREFETCH" 64(%0)\n"
 	        PREFETCH" 128(%0)\n"
@@ -188,7 +188,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 	/* if SRC is misaligned */
 	for(; i>0; i--)
 	{
-		__asm__ __volatile__ (
+		__asm__ volatile (
 		PREFETCH" 320(%0)\n"
 		"movups (%0), %%xmm0\n"
 		"movups 16(%0), %%xmm1\n"
@@ -210,7 +210,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 	*/
 	for(; i>0; i--)
 	{
-		__asm__ __volatile__ (
+		__asm__ volatile (
 		PREFETCH" 320(%0)\n"
 		"movaps (%0), %%xmm0\n"
 		"movaps 16(%0), %%xmm1\n"
@@ -228,7 +228,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 	// Align destination at BLOCK_SIZE boundary
 	for(; ((int)to & (BLOCK_SIZE-1)) && i>0; i--)
 	{
-		__asm__ __volatile__ (
+		__asm__ volatile (
 #ifndef HAVE_ONLY_MMX1
         	PREFETCH" 320(%0)\n"
 #endif
@@ -317,7 +317,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 
 	for(; i>0; i--)
 	{
-		__asm__ __volatile__ (
+		__asm__ volatile (
 #ifndef HAVE_ONLY_MMX1
         	PREFETCH" 320(%0)\n"
 #endif
@@ -346,11 +346,11 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
 #ifdef HAVE_MMX2
                 /* since movntq is weakly-ordered, a "sfence"
 		 * is needed to become ordered again. */
-		__asm__ __volatile__ ("sfence":::"memory");
+		__asm__ volatile ("sfence":::"memory");
 #endif
 #ifndef HAVE_SSE
 		/* enables to use FPU */
-		__asm__ __volatile__ (EMMS:::"memory");
+		__asm__ volatile (EMMS:::"memory");
 #endif
 	}
 	/*
@@ -405,7 +405,7 @@ static void * RENAME(mem2agpcpy)(void * to, const void * from, size_t len)
         */
 	for(; i>0; i--)
 	{
-		__asm__ __volatile__ (
+		__asm__ volatile (
         	PREFETCH" 320(%0)\n"
 		"movq (%0), %%mm0\n"
 		"movq 8(%0), %%mm1\n"
@@ -430,10 +430,10 @@ static void * RENAME(mem2agpcpy)(void * to, const void * from, size_t len)
 #ifdef HAVE_MMX2
                 /* since movntq is weakly-ordered, a "sfence"
 		 * is needed to become ordered again. */
-		__asm__ __volatile__ ("sfence":::"memory");
+		__asm__ volatile ("sfence":::"memory");
 #endif
 		/* enables to use FPU */
-		__asm__ __volatile__ (EMMS:::"memory");
+		__asm__ volatile (EMMS:::"memory");
 	}
 	/*
 	 *	Now do the tail of the block
diff --git a/loader/drv.c b/loader/drv.c
index 5d4f0dcb93..b8fecf7467 100644
--- a/loader/drv.c
+++ b/loader/drv.c
@@ -52,7 +52,7 @@ extern char* def_path;
 #else
 // this asm code is no longer needed
 #define STORE_ALL \
-    __asm__ __volatile__ ( \
+    __asm__ volatile ( \
     "push %%ebx\n\t" \
     "push %%ecx\n\t" \
     "push %%edx\n\t" \
@@ -60,7 +60,7 @@ extern char* def_path;
     "push %%edi\n\t"::)
 
 #define REST_ALL \
-    __asm__ __volatile__ ( \
+    __asm__ volatile ( \
     "pop %%edi\n\t" \
     "pop %%esi\n\t" \
     "pop %%edx\n\t" \
@@ -98,7 +98,7 @@ LRESULT WINAPI SendDriverMessage(HDRVR hDriver, UINT message,
 #endif
     if (!module || !module->hDriverModule || !module->DriverProc) return -1;
 #ifndef __svr4__
-    __asm__ __volatile__ ("fsave (%0)\n\t": :"r"(&qw));
+    __asm__ volatile ("fsave (%0)\n\t": :"r"(&qw));
 #endif
 
 #ifdef WIN32_LOADER
@@ -110,7 +110,7 @@ LRESULT WINAPI SendDriverMessage(HDRVR hDriver, UINT message,
     REST_ALL;
 
 #ifndef __svr4__
-    __asm__ __volatile__ ("frstor (%0)\n\t": :"r"(&qw));
+    __asm__ volatile ("frstor (%0)\n\t": :"r"(&qw));
 #endif
 
 #ifdef DETAILED_OUT
diff --git a/loader/ldt_keeper.c b/loader/ldt_keeper.c
index ef9386435e..e0d4447e60 100644
--- a/loader/ldt_keeper.c
+++ b/loader/ldt_keeper.c
@@ -138,7 +138,7 @@ void Setup_FS_Segment(void)
 {
     unsigned int ldt_desc = LDT_SEL(fs_ldt);
 
-    __asm__ __volatile__(
+    __asm__ volatile(
 	"movl %0,%%eax; movw %%ax, %%fs" : : "r" (ldt_desc)
 	:"eax"
     );
@@ -154,7 +154,7 @@ static int LDT_Modify( int func, struct modify_ldt_ldt_s *ptr,
 {
     int res;
 #ifdef __PIC__
-    __asm__ __volatile__( "pushl %%ebx\n\t"
+    __asm__ volatile( "pushl %%ebx\n\t"
 			  "movl %2,%%ebx\n\t"
 			  "int $0x80\n\t"
 			  "popl %%ebx"
@@ -165,7 +165,7 @@ static int LDT_Modify( int func, struct modify_ldt_ldt_s *ptr,
 			  "d"(16)//sizeof(*ptr) from kernel point of view
 			  :"esi"     );
 #else
-    __asm__ __volatile__("int $0x80"
+    __asm__ volatile("int $0x80"
 			 : "=a" (res)
 			 : "0" (__NR_modify_ldt),
 			 "b" (func),
diff --git a/loader/qtx/qtxload.c b/loader/qtx/qtxload.c
index 772ee5b0c5..38e0decda2 100644
--- a/loader/qtx/qtxload.c
+++ b/loader/qtx/qtxload.c
@@ -104,7 +104,7 @@ int main(int argc, char *argv[]){
     printf("params: flags: %d, paramSize: %d, what: %d, params[0] = %x\n",
         params->flags, params->paramSize, params->what, params->params[0]);
 
-//    __asm__ __volatile__ ("movl %%esp, %0\n\t" : "=a" (esp) :: "memory" );
+//    __asm__ volatile ("movl %%esp, %0\n\t" : "=a" (esp) :: "memory" );
 //    printf("ESP=%p\n",esp);
 
     *((void**)0x62b7d640) = &x_table[0]; //malloc(0x00001837 * 4); // ugly hack?
@@ -113,7 +113,7 @@ int main(int argc, char *argv[]){
 
     ret = dispatcher(params, &globals);
 
-//    __asm__ __volatile__ ("movl %%esp, %0\n\t" : "=a" (esp) :: "memory" );
+//    __asm__ volatile ("movl %%esp, %0\n\t" : "=a" (esp) :: "memory" );
 //    printf("ESP=%p\n",esp);
 
     printf("!!! CDComponentDispatch() => %d  glob=%p\n",ret,globals);
diff --git a/loader/win32.c b/loader/win32.c
index b3b157bb5d..576e3bd945 100644
--- a/loader/win32.c
+++ b/loader/win32.c
@@ -79,7 +79,7 @@ char* def_path = WIN32_PATH;
 
 static void do_cpuid(unsigned int ax, unsigned int *regs)
 {
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "pushl %%ebx; pushl %%ecx; pushl %%edx;"
 	 ".byte  0x0f, 0xa2;"
@@ -95,7 +95,7 @@ static void do_cpuid(unsigned int ax, unsigned int *regs)
 static unsigned int c_localcount_tsc()
 {
     int a;
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "rdtsc\n\t"
 	 :"=a"(a)
@@ -106,7 +106,7 @@ static unsigned int c_localcount_tsc()
 }
 static void c_longcount_tsc(long long* z)
 {
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "pushl %%ebx\n\t"
 	 "movl %%eax, %%ebx\n\t"
@@ -2867,7 +2867,7 @@ static int WINAPI expIsBadStringPtrA(const char* string, int nchars)
 static long WINAPI expInterlockedExchangeAdd( long* dest, long incr )
 {
     long ret;
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "lock; xaddl %0,(%1)"
 	 : "=r" (ret)
@@ -4006,7 +4006,7 @@ static int exp_initterm(INITTERMFUNC *start, INITTERMFUNC *end)
 	    // ok this trick with push/pop is necessary as otherwice
 	    // edi/esi registers are being trashed
 	    void* p = *start;
-	    __asm__ __volatile__
+	    __asm__ volatile
 		(
 		 "pushl %%ebx		\n\t"
 		 "pushl %%ecx		\n\t"
@@ -4272,7 +4272,7 @@ static double expcos(double x)
 
 static void explog10(void)
 {
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "fldl 8(%esp)	\n\t"
 	 "fldln2	\n\t"
@@ -4283,7 +4283,7 @@ static void explog10(void)
 
 static void expcos(void)
 {
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "fldl 8(%esp)	\n\t"
 	 "fcos		\n\t"
@@ -4300,7 +4300,7 @@ static void expcos(void)
 
 static void exp_ftol(void)
 {
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "sub $12, %esp		\n\t"
 	 "fstcw   -2(%ebp)	\n\t"
@@ -4319,8 +4319,8 @@ static void exp_ftol(void)
 }
 
 #define FPU_DOUBLES(var1,var2) double var1,var2; \
-  __asm__ __volatile__( "fstpl %0;fwait" : "=m" (var2) : ); \
-  __asm__ __volatile__( "fstpl %0;fwait" : "=m" (var1) : )
+  __asm__ volatile( "fstpl %0;fwait" : "=m" (var2) : ); \
+  __asm__ volatile( "fstpl %0;fwait" : "=m" (var1) : )
 
 static double exp_CIpow(void)
 {
@@ -4361,7 +4361,7 @@ static int exp_setjmp3(void* jmpbuf, int x)
 {
     //dbgprintf("!!!!UNIMPLEMENTED: setjmp3(%p, %d) => 0\n", jmpbuf, x);
     //return 0;
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 //"mov 4(%%esp), %%edx	\n\t"
 	 "mov (%%esp), %%eax   \n\t"
@@ -4383,7 +4383,7 @@ static int exp_setjmp3(void* jmpbuf, int x)
 	 : "eax"
 	);
 #if 1
-    __asm__ __volatile__
+    __asm__ volatile
 	(
 	 "mov %%fs:0, %%eax	\n\t" // unsure
 	 "mov %%eax, 24(%%edx)	\n\t"
@@ -4762,7 +4762,7 @@ static double expfloor(double x)
 }
 
 #define FPU_DOUBLE(var) double var; \
-  __asm__ __volatile__( "fstpl %0;fwait" : "=m" (var) : )
+  __asm__ volatile( "fstpl %0;fwait" : "=m" (var) : )
 
 static double exp_CIcos(void)
 {
diff --git a/loader/wine/windef.h b/loader/wine/windef.h
index 7b836e12a4..9d90a42257 100644
--- a/loader/wine/windef.h
+++ b/loader/wine/windef.h
@@ -94,7 +94,7 @@ extern "C" {
 #  ifndef _EGCS_ 
 #define __stdcall __attribute__((__stdcall__))
 #define __cdecl   __attribute__((__cdecl__))
-#  define RESTORE_ES  __asm__ __volatile__("pushl %ds\n\tpopl %es")
+#  define RESTORE_ES  __asm__ volatile("pushl %ds\n\tpopl %es")
 #  endif
 # else
 // #  error You need gcc >= 2.7 to build Wine on a 386
diff --git a/mp3lib/dct36_3dnow.c b/mp3lib/dct36_3dnow.c
index 66c9f7b148..02d23975b4 100644
--- a/mp3lib/dct36_3dnow.c
+++ b/mp3lib/dct36_3dnow.c
@@ -39,7 +39,7 @@ void dct36_3dnow(real *inbuf, real *o1,
     real *o2, real *wintab, real *tsbuf)
 #endif
 {
-    __asm__ __volatile__(
+    __asm__ volatile(
 	"movq (%%eax),%%mm0\n\t"
 	"movq 4(%%eax),%%mm1\n\t"
 	"pfadd %%mm1,%%mm0\n\t"
diff --git a/mp3lib/dct64_3dnow.c b/mp3lib/dct64_3dnow.c
index 8df9295119..919771a91d 100644
--- a/mp3lib/dct64_3dnow.c
+++ b/mp3lib/dct64_3dnow.c
@@ -18,7 +18,7 @@ static float attribute_used plus_1f = 1.0;
 void dct64_MMX_3dnow(short *a,short *b,real *c)
 {
   char tmp[256];
-    __asm__ __volatile(
+    __asm__ volatile(
 "	movl %2,%%eax\n\t"
 
 "	leal 128+%3,%%edx\n\t"
diff --git a/mp3lib/dct64_k7.c b/mp3lib/dct64_k7.c
index 058028ebd1..2d79d7a621 100644
--- a/mp3lib/dct64_k7.c
+++ b/mp3lib/dct64_k7.c
@@ -18,7 +18,7 @@ static float attribute_used plus_1f = 1.0;
 void dct64_MMX_3dnowex(short *a,short *b,real *c)
 {
   char tmp[256];
-    __asm__ __volatile(
+    __asm__ volatile(
 "	movl %2,%%eax\n\t"
 
 "	leal 128+%3,%%edx\n\t"
diff --git a/mp3lib/dct64_mmx.c b/mp3lib/dct64_mmx.c
index 317dbb0549..4cce96bd7d 100644
--- a/mp3lib/dct64_mmx.c
+++ b/mp3lib/dct64_mmx.c
@@ -9,7 +9,7 @@
 void dct64_MMX(short *a,short *b,real *c)
 {
     char tmp[256];
-    __asm__ __volatile(
+    __asm__ volatile(
 "	movl %2,%%eax\n\t"
 /* Phase 1*/
 "	flds     (%%eax)\n\t"
diff --git a/mp3lib/decode_i586.c b/mp3lib/decode_i586.c
index 9ca1aeb360..8d005735d3 100644
--- a/mp3lib/decode_i586.c
+++ b/mp3lib/decode_i586.c
@@ -41,7 +41,7 @@ int synth_1to1_pent(real *bandPtr, int channel, short *samples)
 {
   real tmp[3];
   register int retval;
-    __asm__ __volatile(
+    __asm__ volatile(
 "        movl %%ebp,"MANGLE(saved_ebp)"\n\t"
 "        movl %1,%%eax\n\t"/*bandPtr*/
 "        movl %3,%%esi\n\t"
diff --git a/mp3lib/decode_mmx.c b/mp3lib/decode_mmx.c
index 1d4e3caba9..55dc6a9081 100644
--- a/mp3lib/decode_mmx.c
+++ b/mp3lib/decode_mmx.c
@@ -217,7 +217,7 @@ int synth_1to1_MMX(real *bandPtr, int channel, short *samples)
     dct64_MMX_func(a, b, bandPtr);
     window = mp3lib_decwins + 16 - bo1;
     //printf("DEBUG: channel %d, bo %d, off %d\n", channel, bo, 16 - bo1);
-__asm__ __volatile(
+__asm__ volatile(
 ASMALIGN(4)
 ".L03:\n\t"
         "movq  (%1),%%mm0\n\t"
diff --git a/vidix/nvidia_vid.c b/vidix/nvidia_vid.c
index a045b397fe..767a358f37 100644
--- a/vidix/nvidia_vid.c
+++ b/vidix/nvidia_vid.c
@@ -372,7 +372,7 @@ static int nv_probe(int verbose, int force){
  * PCI-Memory IO access macros.
  */
 
-#define MEM_BARRIER() __asm__ __volatile__ ("" : : : "memory")
+#define MEM_BARRIER() __asm__ volatile ("" : : : "memory")
 
 #undef	VID_WR08
 #define VID_WR08(p,i,val) ({ MEM_BARRIER(); ((uint8_t *)(p))[(i)]=(val); })
diff --git a/vidix/s3_regs.h b/vidix/s3_regs.h
index 0add1bea54..4fd0e427cf 100644
--- a/vidix/s3_regs.h
+++ b/vidix/s3_regs.h
@@ -196,7 +196,7 @@ do { \
 #ifndef USE_RMW_CYCLES
 
 /* Can be used to inhibit READ-MODIFY-WRITE cycles. On by default. */
-#define MEM_BARRIER() __asm__ __volatile__ ("" : : : "memory")
+#define MEM_BARRIER() __asm__ volatile ("" : : : "memory")
 
 #undef  VID_WR08
 #define VID_WR08(p,i,val) ({ MEM_BARRIER(); ((uint8_t *)(p))[(i)]=(val); })
diff --git a/vidix/sysdep/AsmMacros_powerpc.h b/vidix/sysdep/AsmMacros_powerpc.h
index 7120446afe..0ae2fd57d2 100644
--- a/vidix/sysdep/AsmMacros_powerpc.h
+++ b/vidix/sysdep/AsmMacros_powerpc.h
@@ -68,7 +68,7 @@ extern unsigned char *ioBase;
 
 static __inline__ volatile void eieio()
 {
-	__asm__ __volatile__ ("eieio");
+	__asm__ volatile ("eieio");
 }
 
 static __inline__ void outb(short port, unsigned char value)
diff --git a/vidix/sysdep/AsmMacros_sparc.h b/vidix/sysdep/AsmMacros_sparc.h
index 2358adea6f..8ffdde26e1 100644
--- a/vidix/sysdep/AsmMacros_sparc.h
+++ b/vidix/sysdep/AsmMacros_sparc.h
@@ -68,37 +68,37 @@
 
 static __inline__ void outb(unsigned long port, char val)
 {
-  __asm__ __volatile__("stba %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
+  __asm__ volatile("stba %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
 }
 
 static __inline__ void outw(unsigned long port, char val)
 {
-  __asm__ __volatile__("stha %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
+  __asm__ volatile("stha %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
 }
 
 static __inline__ void outl(unsigned long port, char val)
 {
-  __asm__ __volatile__("sta %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
+  __asm__ volatile("sta %0, [%1] %2" : : "r" (val), "r" (port), "i" (ASI_PL));
 }
 
 static __inline__ unsigned int inb(unsigned long port)
 {
    unsigned char ret;
-   __asm__ __volatile__("lduba [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
+   __asm__ volatile("lduba [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
    return ret;
 }
 
 static __inline__ unsigned int inw(unsigned long port)
 {
    unsigned char ret;
-   __asm__ __volatile__("lduha [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
+   __asm__ volatile("lduha [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
    return ret;
 }
 
 static __inline__ unsigned int inl(unsigned long port)
 {
    unsigned char ret;
-   __asm__ __volatile__("lda [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
+   __asm__ volatile("lda [%1] %2, %0" : "=r" (ret) : "r" (port), "i" (ASI_PL));
    return ret;
 }
 
diff --git a/vidix/sysdep/AsmMacros_x86.h b/vidix/sysdep/AsmMacros_x86.h
index c0b1afc105..6bebeaa505 100644
--- a/vidix/sysdep/AsmMacros_x86.h
+++ b/vidix/sysdep/AsmMacros_x86.h
@@ -171,7 +171,7 @@ static __inline__ void outb(short port,char val)
     }
     else
 #endif
-   __asm__ __volatile__("outb %0,%1" : :"a" (val), "d" (port));
+   __asm__ volatile("outb %0,%1" : :"a" (val), "d" (port));
     return;
 }
 
@@ -199,7 +199,7 @@ static __inline__ void outw(short port,short val)
     }
     else
 #endif
-   __asm__ __volatile__("outw %0,%1" : :"a" (val), "d" (port));
+   __asm__ volatile("outw %0,%1" : :"a" (val), "d" (port));
     return;
 }
 
@@ -227,7 +227,7 @@ static __inline__ void outl(short port,unsigned int val)
     }
     else
 #endif
-   __asm__ __volatile__("outl %0,%1" : :"a" (val), "d" (port));
+   __asm__ volatile("outl %0,%1" : :"a" (val), "d" (port));
     return;
 }
 
@@ -255,7 +255,7 @@ static __inline__ unsigned int inb(short port)
     }
     else
 #endif
-   __asm__ __volatile__("inb %1,%0" :
+   __asm__ volatile("inb %1,%0" :
        "=a" (ret) :
        "d" (port));
    return ret;
@@ -285,7 +285,7 @@ static __inline__ unsigned int inw(short port)
     }
     else
 #endif
-   __asm__ __volatile__("inw %1,%0" :
+   __asm__ volatile("inw %1,%0" :
        "=a" (ret) :
        "d" (port));
    return ret;
@@ -315,7 +315,7 @@ static __inline__ unsigned int inl(short port)
     }
     else
 #endif
-   __asm__ __volatile__("inl %1,%0" :
+   __asm__ volatile("inl %1,%0" :
        "=a" (ret) :
        "d" (port));
    return ret;
@@ -327,7 +327,7 @@ static __inline__ void intr_disable()
     if (svgahelper_initialized == 1)
 	return;
 #endif
-  __asm__ __volatile__("cli");
+  __asm__ volatile("cli");
 }
 
 static __inline__ void intr_enable()
@@ -336,7 +336,7 @@ static __inline__ void intr_enable()
     if (svgahelper_initialized == 1)
 	return;
 #endif
-  __asm__ __volatile__("sti");
+  __asm__ volatile("sti");
 }
 
 #endif /* MPLAYER_ASMMACROS_X86_H */