Index: src/OFMatrix4x4.h
==================================================================
--- src/OFMatrix4x4.h
+++ src/OFMatrix4x4.h
@@ -21,11 +21,11 @@
  * @brief A 4x4 matrix of floats.
  */
 OF_SUBCLASSING_RESTRICTED
 @interface OFMatrix4x4: OFObject <OFCopying>
 {
-	float _values[4][4];
+	float _values[4][4] OF_ALIGN(16);
 }
 
 #ifdef OF_HAVE_CLASS_PROPERTIES
 @property (readonly, class) OFMatrix4x4 *identityMatrix;
 #endif

Index: src/OFMatrix4x4.m
==================================================================
--- src/OFMatrix4x4.m
+++ src/OFMatrix4x4.m
@@ -28,10 +28,54 @@
 	{ 0, 0, 0, 1 }
 };
 
 @implementation OFMatrix4x4
 #if (defined(OF_AMD64) || defined(OF_X86)) && defined(__GNUC__)
+# ifndef __clang__
+#  pragma GCC push_options
+#  pragma GCC target("sse4.1")
+# endif
+static void
+transformVectors_SSE41(OFMatrix4x4 *self, SEL _cmd, OFVector4D *vectors,
+    size_t count)
+{
+	__asm__ __volatile__ (
+	    "test	%0, %0\n\t"
+	    "jz		0f\n"
+	    "\n\t"
+	    "movaps	(%2), %%xmm0\n\t"
+	    "movaps	16(%2), %%xmm1\n\t"
+	    "movaps	32(%2), %%xmm2\n\t"
+	    "movaps	48(%2), %%xmm3\n"
+	    "\n\t"
+	    "0:\n\t"
+	    "movaps	(%1), %%xmm4\n\t"
+	    "movaps	%%xmm4, %%xmm5\n\t"
+	    "dpps	$0xFF, %%xmm0, %%xmm4\n\t"
+	    "movaps	%%xmm5, %%xmm6\n\t"
+	    "dpps	$0xFF, %%xmm1, %%xmm5\n\t"
+	    "movaps	%%xmm6, %%xmm7\n\t"
+	    "dpps	$0xFF, %%xmm2, %%xmm6\n\t"
+	    "dpps	$0xFF, %%xmm3, %%xmm7\n\t"
+	    "insertps	$0x10, %%xmm5, %%xmm4\n\t"
+	    "insertps	$0x20, %%xmm6, %%xmm4\n\t"
+	    "insertps	$0x30, %%xmm7, %%xmm4\n\t"
+	    "movaps	%%xmm4, (%1)\n"
+	    "\n\t"
+	    "add	$16, %1\n\t"
+	    "dec	%0\n\t"
+	    "jnz	0b\n"
+	    : "+r"(count), "+r"(vectors)
+	    : "r"(&self->_values)
+	    : "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7",
+	      "memory"
+	);
+}
+# ifndef __clang__
+#  pragma GCC pop_options
+# endif
+
 # ifndef __clang__
 #  pragma GCC push_options
 #  pragma GCC target("3dnow,3dnowa")
 # endif
 static void
@@ -264,11 +308,14 @@
 # define REPLACE(selector, func)					\
 	typeEncoding = method_getTypeEncoding(				\
 	    class_getInstanceMethod(self, selector));			\
 	class_replaceMethod(self, selector, (IMP)func, typeEncoding);
 
-	if ([OFSystemInfo supports3DNow]) {
+	if ([OFSystemInfo supportsSSE41]) {
+		REPLACE(@selector(transformVectors:count:),
+		    transformVectors_SSE41)
+	} else if ([OFSystemInfo supports3DNow]) {
 		if ([OFSystemInfo supportsEnhanced3DNow]) {
 			REPLACE(@selector(multiplyWithMatrix:),
 			    multiplyWithMatrix_enhanced3DNow)
 			REPLACE(@selector(transformVectors:count:),
 			    transformVectors_enhanced3DNow)

Index: src/OFObject.h
==================================================================
--- src/OFObject.h
+++ src/OFObject.h
@@ -347,11 +347,11 @@
 /**
  * @struct OFVector4D OFObject.h ObjFW/OFObject.h
  *
  * @brief A vector in 4D space.
  */
-typedef struct OF_BOXABLE OFVector4D {
+typedef struct OF_BOXABLE OF_ALIGN(16) OFVector4D {
 	/** The x coordinate of the vector */
 	float x;
 	/** The y coordinate of the vector */
 	float y;
 	/** The z coordinate of the vector */

Index: src/macros.h
==================================================================
--- src/macros.h
+++ src/macros.h
@@ -99,17 +99,21 @@
 # define OF_UNLIKELY(cond) (cond)
 # define OF_CONST_FUNC
 # define OF_NO_RETURN_FUNC
 # define OF_WEAK_REF(sym)
 #endif
+
+#ifdef __GNUC__
+# define OF_ALIGN(alignment) __attribute__((__aligned__(alignment)))
+#endif
 
 #if __STDC_VERSION__ >= 201112L
 # define OF_ALIGNOF(type) _Alignof(type)
 # define OF_ALIGNAS(type) _Alignas(type)
 #else
 # define OF_ALIGNOF(type) __alignof__(type)
-# define OF_ALIGNAS(type) __attribute__((__aligned__(__alignof__(type))))
+# define OF_ALIGNAS(type) OF_ALIGN(OF_ALIGNOF(type))
 #endif
 
 #if __STDC_VERSION__ >= 201112L && defined(OF_HAVE_MAX_ALIGN_T)
 # define OF_BIGGEST_ALIGNMENT _Alignof(max_align_t)
 #else