diff --git a/builtins/target-generic-common.ll b/builtins/target-generic-common.ll
index 81127897..533def68 100644
--- a/builtins/target-generic-common.ll
+++ b/builtins/target-generic-common.ll
@@ -233,7 +233,7 @@ declare <WIDTH x i32> @__masked_load_32(i8 * nocapture, <WIDTH x i1> %mask) noun
 declare <WIDTH x i64> @__masked_load_64(i8 * nocapture, <WIDTH x i1> %mask) nounwind readonly
 
 declare void @__masked_store_8(<WIDTH x i8>* nocapture, <WIDTH x i8>, 
-                                <WIDTH x i1>) nounwind 
+                               <WIDTH x i1>) nounwind 
 declare void @__masked_store_16(<WIDTH x i16>* nocapture, <WIDTH x i16>, 
                                 <WIDTH x i1>) nounwind 
 declare void @__masked_store_32(<WIDTH x i32>* nocapture, <WIDTH x i32>, 
@@ -241,8 +241,9 @@ declare void @__masked_store_32(<WIDTH x i32>* nocapture, <WIDTH x i32>,
 declare void @__masked_store_64(<WIDTH x i64>* nocapture, <WIDTH x i64>,
                                 <WIDTH x i1> %mask) nounwind 
 
+ifelse(LLVM_VERSION, `LLVM_3_1svn',`
 define void @__masked_store_blend_8(<WIDTH x i8>* nocapture, <WIDTH x i8>, 
-                                     <WIDTH x i1>) nounwind {
+                                     <WIDTH x i1>) nounwind alwaysinline {
   %v = load <WIDTH x i8> * %0
   %v1 = select <WIDTH x i1> %2, <WIDTH x i8> %1, <WIDTH x i8> %v
   store <WIDTH x i8> %v1, <WIDTH x i8> * %0
@@ -250,7 +251,7 @@ define void @__masked_store_blend_8(<WIDTH x i8>* nocapture, <WIDTH x i8>,
 }
 
 define void @__masked_store_blend_16(<WIDTH x i16>* nocapture, <WIDTH x i16>, 
-                                     <WIDTH x i1>) nounwind {
+                                     <WIDTH x i1>) nounwind alwaysinline {
   %v = load <WIDTH x i16> * %0
   %v1 = select <WIDTH x i1> %2, <WIDTH x i16> %1, <WIDTH x i16> %v
   store <WIDTH x i16> %v1, <WIDTH x i16> * %0
@@ -258,7 +259,7 @@ define void @__masked_store_blend_16(<WIDTH x i16>* nocapture, <WIDTH x i16>,
 }
 
 define void @__masked_store_blend_32(<WIDTH x i32>* nocapture, <WIDTH x i32>, 
-                                     <WIDTH x i1>) nounwind {
+                                     <WIDTH x i1>) nounwind alwaysinline {
   %v = load <WIDTH x i32> * %0
   %v1 = select <WIDTH x i1> %2, <WIDTH x i32> %1, <WIDTH x i32> %v
   store <WIDTH x i32> %v1, <WIDTH x i32> * %0
@@ -266,12 +267,22 @@ define void @__masked_store_blend_32(<WIDTH x i32>* nocapture, <WIDTH x i32>,
 }
 
 define void @__masked_store_blend_64(<WIDTH x i64>* nocapture,
-                                     <WIDTH x i64>, <WIDTH x i1>) nounwind {
+                            <WIDTH x i64>, <WIDTH x i1>) nounwind alwaysinline {
   %v = load <WIDTH x i64> * %0
   %v1 = select <WIDTH x i1> %2, <WIDTH x i64> %1, <WIDTH x i64> %v
   store <WIDTH x i64> %v1, <WIDTH x i64> * %0
   ret void
 }
+',`
+declare void @__masked_store_blend_8(<WIDTH x i8>* nocapture, <WIDTH x i8>, 
+                                     <WIDTH x i1>) nounwind 
+declare void @__masked_store_blend_16(<WIDTH x i16>* nocapture, <WIDTH x i16>, 
+                                      <WIDTH x i1>) nounwind 
+declare void @__masked_store_blend_32(<WIDTH x i32>* nocapture, <WIDTH x i32>, 
+                                      <WIDTH x i1>) nounwind 
+declare void @__masked_store_blend_64(<WIDTH x i64>* nocapture, <WIDTH x i64>,
+                                      <WIDTH x i1> %mask) nounwind 
+')
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;; gather/scatter
diff --git a/examples/intrinsics/generic-16.h b/examples/intrinsics/generic-16.h
index 3ac77a41..c7600918 100644
--- a/examples/intrinsics/generic-16.h
+++ b/examples/intrinsics/generic-16.h
@@ -1010,6 +1010,26 @@ static FORCEINLINE void __masked_store_64(void *p, __vec16_i64 val,
             ptr[i] = val.v[i];
 }
 
+static FORCEINLINE void __masked_store_blend_8(void *p, __vec16_i8 val,
+                                               __vec16_i1 mask) {
+    __masked_store_8(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_16(void *p, __vec16_i16 val,
+                                                __vec16_i1 mask) {
+    __masked_store_16(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_32(void *p, __vec16_i32 val,
+                                                __vec16_i1 mask) {
+    __masked_store_32(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_64(void *p, __vec16_i64 val,
+                                                __vec16_i1 mask) {
+    __masked_store_64(p, val, mask);
+}
+
 ///////////////////////////////////////////////////////////////////////////
 // gather/scatter
 
diff --git a/examples/intrinsics/sse4.h b/examples/intrinsics/sse4.h
index 5e5131f3..81444ecb 100644
--- a/examples/intrinsics/sse4.h
+++ b/examples/intrinsics/sse4.h
@@ -2439,7 +2439,8 @@ static FORCEINLINE void __masked_store_8(void *p, __vec4_i8 val,
         ptr[3] = _mm_extract_epi8(val.v, 3);
 }
 
-static FORCEINLINE void __masked_store_16(void *p, __vec4_i16 val, __vec4_i1 mask) {
+static FORCEINLINE void __masked_store_16(void *p, __vec4_i16 val,
+                                          __vec4_i1 mask) {
     int16_t *ptr = (int16_t *)p;
 
     uint32_t m = _mm_extract_ps(mask.v, 0);
@@ -2499,6 +2500,28 @@ static FORCEINLINE void __masked_store_64(void *p, __vec4_i64 val,
         ptr[3] = _mm_extract_epi64(val.v[1], 1);
 }
 
+static FORCEINLINE void __masked_store_blend_8(void *p, __vec4_i8 val, 
+                                               __vec4_i1 mask) {
+    __masked_store_8(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_16(void *p, __vec4_i16 val, 
+                                                __vec4_i1 mask) {
+    __masked_store_16(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_32(void *p, __vec4_i32 val, 
+                                                __vec4_i1 mask) {
+    // FIXME: do a load, blendvps, store here...
+    __masked_store_32(p, val, mask);
+}
+
+static FORCEINLINE void __masked_store_blend_64(void *p, __vec4_i64 val, 
+                                                __vec4_i1 mask) {
+    // FIXME: do a 2x (load, blendvps, store) here...
+    __masked_store_64(p, val, mask);
+}
+
 ///////////////////////////////////////////////////////////////////////////
 // gather/scatter
 // offsets * offsetScale is in bytes (for all of these)