Index: configure.ac
==================================================================
--- configure.ac
+++ configure.ac
@@ -332,19 +332,10 @@
 	flag="-Xclang -fno-constant-cfstrings"
 	OBJCFLAGS="$OBJCFLAGS $flag"
 	OBJFW_OBJCFLAGS="$OBJFW_OBJCFLAGS $flag"
 ])
 
-case "$host_cpu" in
-i*86|x86_64)
-	AX_CHECK_COMPILER_FLAGS([-masm=intel], [
-		OBJCFLAGS="$OBJCFLAGS -masm=intel"
-		AC_DEFINE(HAVE_INTEL_SYNTAX, 1, [Whether asm syntax is Intel])
-	])
-	;;
-esac
-
 AX_CHECK_COMPILER_FLAGS([-Wsign-compare -Werror],
 	[OBJCFLAGS="$OBJCFLAGS -Wsign-compare"])
 AS_IF([test x"$with_nds" != x"yes"], [
 	AX_CHECK_COMPILER_FLAGS([-Wshadow -Werror],
 		[OBJCFLAGS="$OBJCFLAGS -Wshadow"])

Index: src/OFMatrix4x4.m
==================================================================
--- src/OFMatrix4x4.m
+++ src/OFMatrix4x4.m
@@ -27,30 +27,30 @@
 	{ 0, 0, 1, 0 },
 	{ 0, 0, 0, 1 }
 };
 
 @implementation OFMatrix4x4
-#if (defined(OF_AMD64) || defined(OF_X86)) && defined(HAVE_INTEL_SYNTAX)
+#if defined(OF_AMD64) || defined(OF_X86)
 static void
 multiplyWithMatrix_3DNow(OFMatrix4x4 *self, SEL _cmd, OFMatrix4x4 *matrix)
 {
-	float result[4][4] = {{ 0 }};
+	float result[4][4];
 
 	for (uint_fast8_t i = 0; i < 4; i++) {
 		for (uint_fast8_t j = 0; j < 4; j++) {
 			__asm__ (
-			    "movd	mm0, [%2]\n\t"
-			    "punpckldq	mm0, [%2 + 16]\n\t"
-			    "pfmul	mm0, [%1]\n\t"
-			    "movd	mm1, [%2 + 32]\n\t"
-			    "punpckldq	mm1, [%2 + 48]\n\t"
-			    "pfmul	mm1, [%1 + 8]\n\t"
-			    "pfadd	mm0, mm1\n\t"
-			    "movq	mm1, mm0\n\t"
-			    "psrlq	mm1, 32\n\t"
-			    "pfadd	mm0, mm1\n\t"
-			    "movd	%0, mm0"
+			    "movd	(%2), %%mm0\n\t"
+			    "punpckldq	16(%2), %%mm0\n\t"
+			    "pfmul	(%1), %%mm0\n\t"
+			    "movd	32(%2), %%mm1\n\t"
+			    "punpckldq	48(%2), %%mm1\n\t"
+			    "pfmul	8(%1), %%mm1\n\t"
+			    "pfadd	%%mm1, %%mm0\n\t"
+			    "movq	%%mm0, %%mm1\n\t"
+			    "psrlq	$32, %%mm1\n\t"
+			    "pfadd	%%mm1, %%mm0\n\t"
+			    "movd	%%mm0, %0"
 			    :: "m"(result[i][j]), "r"(&matrix->_values[i][0]),
 			       "r"(&self->_values[0][j])
 			    : "mm0", "mm1", "memory"
 			);
 		}
@@ -65,52 +65,52 @@
 transformedVector_3DNow(OFMatrix4x4 *self, SEL _cmd, OFVector4D vector)
 {
 	OFVector4D result;
 
 	__asm__ (
-	    "movq	mm0, [%2]\n\t"
-	    "movq	mm1, [%2 + 8]\n"
-	    "\n\t"
-	    "movq	mm2, mm0\n\t"
-	    "movq	mm3, mm1\n\t"
-	    "pfmul	mm2, [%1]\n\t"
-	    "pfmul	mm3, [%1 + 8]\n\t"
-	    "pfadd	mm2, mm3\n\t"
-	    "movq	mm3, mm2\n\t"
-	    "psrlq	mm3, 32\n\t"
-	    "pfadd	mm2, mm3\n"
-	    "\n\t"
-	    "movq	mm3, mm0\n\t"
-	    "movq	mm4, mm1\n\t"
-	    "pfmul	mm3, [%1 + 16]\n\t"
-	    "pfmul	mm4, [%1 + 24]\n\t"
-	    "pfadd	mm3, mm4\n\t"
-	    "movq	mm4, mm3\n\t"
-	    "psrlq	mm4, 32\n\t"
-	    "pfadd	mm3, mm4\n"
-	    "\n\t"
-	    "punpckldq	mm2, mm3\n\t"
-	    "movq	[%0], mm2\n"
-	    "\n\t"
-	    "movq	mm2, mm0\n\t"
-	    "movq	mm3, mm1\n\t"
-	    "pfmul	mm2, [%1 + 32]\n\t"
-	    "pfmul	mm3, [%1 + 40]\n\t"
-	    "pfadd	mm2, mm3\n\t"
-	    "movq	mm3, mm2\n\t"
-	    "psrlq	mm3, 32\n\t"
-	    "pfadd	mm2, mm3\n"
-	    "\n\t"
-	    "pfmul	mm0, [%1 + 48]\n\t"
-	    "pfmul	mm1, [%1 + 56]\n\t"
-	    "pfadd	mm0, mm1\n\t"
-	    "movq	mm1, mm0\n\t"
-	    "psrlq	mm1, 32\n\t"
-	    "pfadd	mm0, mm1\n"
-	    "\n\t"
-	    "punpckldq	mm2, mm0\n\t"
-	    "movq	[%0 + 8], mm2\n"
+	    "movq	(%2), %%mm0\n\t"
+	    "movq	8(%2), %%mm1\n"
+	    "\n\t"
+	    "movq	%%mm0, %%mm2\n\t"
+	    "movq	%%mm1, %%mm3\n\t"
+	    "pfmul	(%1), %%mm2\n\t"
+	    "pfmul	8(%1), %%mm3\n\t"
+	    "pfadd	%%mm3, %%mm2\n\t"
+	    "movq	%%mm2, %%mm3\n\t"
+	    "psrlq	$32, %%mm3\n\t"
+	    "pfadd	%%mm3, %%mm2\n"
+	    "\n\t"
+	    "movq	%%mm0, %%mm3\n\t"
+	    "movq	%%mm1, %%mm4\n\t"
+	    "pfmul	16(%1), %%mm3\n\t"
+	    "pfmul	24(%1), %%mm4\n\t"
+	    "pfadd	%%mm4, %%mm3\n\t"
+	    "movq	%%mm3, %%mm4\n\t"
+	    "psrlq	$32, %%mm4\n\t"
+	    "pfadd	%%mm4, %%mm3\n"
+	    "\n\t"
+	    "punpckldq	%%mm3, %%mm2\n\t"
+	    "movq	%%mm2, (%0)\n"
+	    "\n\t"
+	    "movq	%%mm0, %%mm2\n\t"
+	    "movq	%%mm1, %%mm3\n\t"
+	    "pfmul	32(%1), %%mm2\n\t"
+	    "pfmul	40(%1), %%mm3\n\t"
+	    "pfadd	%%mm3, %%mm2\n\t"
+	    "movq	%%mm2, %%mm3\n\t"
+	    "psrlq	$32, %%mm3\n\t"
+	    "pfadd	%%mm3, %%mm2\n"
+	    "\n\t"
+	    "pfmul	48(%1), %%mm0\n\t"
+	    "pfmul	56(%1), %%mm1\n\t"
+	    "pfadd	%%mm1, %%mm0\n\t"
+	    "movq	%%mm0, %%mm1\n\t"
+	    "psrlq	$32, %%mm1\n\t"
+	    "pfadd	%%mm1, %%mm0\n"
+	    "\n\t"
+	    "punpckldq	%%mm0, %%mm2\n\t"
+	    "movq	%%mm2, 8(%0)\n"
 	    "\n\t"
 	    "femms"
 	    :: "r"(&result), "r"(&self->_values), "r"(&vector)
 	    : "mm0", "mm1", "mm2", "mm3", "mm4", "memory"
 	);

Index: src/OFSystemInfo.m
==================================================================
--- src/OFSystemInfo.m
+++ src/OFSystemInfo.m
@@ -296,13 +296,13 @@
 	 * as ebx is a special register in PIC code. Yes, GCC is indeed not
 	 * able to just push a register onto the stack before the __asm__ block
 	 * and to pop it afterwards.
 	 */
 	__asm__ (
-	    "xchg{l}	{ %%ebx, %%edi | edi, ebx }\n\t"
+	    "xchgl	%%ebx, %%edi\n\t"
 	    "cpuid\n\t"
-	    "xchg{l}	{ %%edi, %%ebx | ebx, edi }"
+	    "xchgl	%%edi, %%ebx"
 	    : "=a"(regs.eax), "=D"(regs.ebx), "=c"(regs.ecx), "=d"(regs.edx)
 	    : "a"(eax), "c"(ecx)
 	);
 # else
 	memset(&regs, 0, sizeof(regs));