diff --git a/common/frame.c b/common/frame.c
index 9956527..e481668 100644
--- a/common/frame.c
+++ b/common/frame.c
@@ -804,6 +804,7 @@ void x264_deblock_init( int cpu, x264_deblock_function_t *pf )
         pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_mmxext;
         pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_mmxext;
 #endif
+#ifndef BROKEN_STACK_ALIGNMENT
         if( cpu&X264_CPU_SSE2 )
         {
             pf->deblock_v_luma = x264_deblock_v_luma_sse2;
@@ -811,6 +812,7 @@ void x264_deblock_init( int cpu, x264_deblock_function_t *pf )
             pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_sse2;
             pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_sse2;
         }
+#endif
     }
 #endif
 
diff --git a/common/osdep.h b/common/osdep.h
index 939fe78..36685be 100644
--- a/common/osdep.h
+++ b/common/osdep.h
@@ -133,4 +133,10 @@
 /* FIXME: long isn't always the native register size (e.g. win64). */
 #define WORD_SIZE sizeof(long)
 
+#if !defined(_WIN64) && !defined(__LP64__)
+#if defined(_MSC_VER) || defined(__INTEL_COMPILER)
+#define BROKEN_STACK_ALIGNMENT /* define it if stack is not mod16 */
+#endif
+#endif
+
 #endif /* X264_OSDEP_H */