diff --git a/builtins/svml.m4 b/builtins/svml.m4
index 9608dea6..71a6a709 100644
--- a/builtins/svml.m4
+++ b/builtins/svml.m4
@@ -83,11 +83,43 @@ define(`svml_define',`
 
 
 ;; define x2 __svml calls
-define(`svml_define_x2',`
-   svml_stubs($1,$3,$4)
+define(`svml_define_x',`
+  define <$5 x $1> @__svml_sin$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_sin$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_asin$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_asin$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_cos$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_cos$2, %0)
+    ret <$5 x $1> %ret
+  }
+  declare void @__svml_sincos$4(<$5 x $1>,<$5 x $1>*,<$5 x $1>*) nounwind readnone alwaysinline 
+  define <$5 x $1> @__svml_tan$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_tan$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_atan$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_atan$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_atan2$4(<$5 x $1>,<$5 x $1>) nounwind readnone alwaysinline {
+    binary$3to$5(ret, $1, @__svml_atan2$2, %0, %1)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_exp$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_exp$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_log$4(<$5 x $1>) nounwind readnone alwaysinline {
+    unary$3to$5(ret, $1, @__svml_log$2, %0)
+    ret <$5 x $1> %ret
+  }
+  define <$5 x $1> @__svml_pow$4(<$5 x $1>,<$5 x $1>) nounwind readnone alwaysinline {
+    binary$3to$5(ret, $1, @__svml_pow$2, %0, %1)
+    ret <$5 x $1> %ret
+  }
 ')
 
-;; define x4 __svml calls
-define(`svml_define_x4',`
-   svml_stubs($1,$3,$4)
-')
diff --git a/builtins/target-avx-x2.ll b/builtins/target-avx-x2.ll
index f3f1590a..f8fd5cd5 100644
--- a/builtins/target-avx-x2.ll
+++ b/builtins/target-avx-x2.ll
@@ -140,11 +140,11 @@ define <16 x float> @__sqrt_varying_float(<16 x float>) nounwind readonly always
 include(`svml.m4')
 ;; single precision
 svml_declare(float,f8,8)
-svml_define_x2(float,f8,8,f,16)
+svml_define_x(float,f8,8,f,16)
 
 ;; double precision
 svml_declare(double,4,4)
-svml_define_x2(double,4,4,d,16)
+svml_define_x(double,4,4,d,16)
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;; float min/max
diff --git a/builtins/target-avx.ll b/builtins/target-avx.ll
index 7e7ab330..196e5ea4 100644
--- a/builtins/target-avx.ll
+++ b/builtins/target-avx.ll
@@ -144,7 +144,7 @@ svml_define(float,f8,8,f)
 
 ;; double precision
 svml_declare(double,4,4)
-svml_define_x2(double,4,4,d,8)
+svml_define_x(double,4,4,d,8)
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;; float min/max
diff --git a/builtins/target-sse2-x2.ll b/builtins/target-sse2-x2.ll
index 9fa607a4..77bf1a9d 100644
--- a/builtins/target-sse2-x2.ll
+++ b/builtins/target-sse2-x2.ll
@@ -108,86 +108,11 @@ define <8 x float> @__sqrt_varying_float(<8 x float>) nounwind readonly alwaysin
 include(`svml.m4')
 ;; single precision
 svml_declare(float,f4,4)
+svml_define_x(float,f4,4,f,8)
 
 ;; double precision
 svml_declare(double,2,2)
-svml_define_x4(double,2,2,d,8)
-
-define <8 x float> @__svml_sinf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_sinf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_asinf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_asinf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_cosf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_cosf4, %0)
-  ret <8 x float> %ret
-}
-
-define void @__svml_sincosf(<8 x float>, <8 x float> *,
-                                    <8 x float> *) nounwind readnone alwaysinline {
-  ; call svml_sincosf4 two times with the two 4-wide sub-vectors
-  %a = shufflevector <8 x float> %0, <8 x float> undef,
-         <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %b = shufflevector <8 x float> %0, <8 x float> undef,
-         <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-
-  %cospa = alloca <4 x float>
-  %sa = call <4 x float> @__svml_sincosf4(<4 x float> * %cospa, <4 x float> %a)
-
-  %cospb = alloca <4 x float>
-  %sb = call <4 x float> @__svml_sincosf4(<4 x float> * %cospb, <4 x float> %b)
-
-  %sin = shufflevector <4 x float> %sa, <4 x float> %sb,
-         <8 x i32> <i32 0, i32 1, i32 2, i32 3,
-                    i32 4, i32 5, i32 6, i32 7>
-  store <8 x float> %sin, <8 x float> * %1
-
-  %cosa = load <4 x float> * %cospa
-  %cosb = load <4 x float> * %cospb
-  %cos = shufflevector <4 x float> %cosa, <4 x float> %cosb,
-         <8 x i32> <i32 0, i32 1, i32 2, i32 3,
-                    i32 4, i32 5, i32 6, i32 7>
-  store <8 x float> %cos, <8 x float> * %2
-
-  ret void
-}
-
-define <8 x float> @__svml_tanf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_tanf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_atanf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_atanf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_atan2f(<8 x float>,
-                                          <8 x float>) nounwind readnone alwaysinline {
-  binary4to8(ret, float, @__svml_atan2f4, %0, %1)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_expf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_expf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_logf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_logf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_powf(<8 x float>,
-                                        <8 x float>) nounwind readnone alwaysinline {
-  binary4to8(ret, float, @__svml_powf4, %0, %1)
-  ret <8 x float> %ret
-}
+svml_define_x(double,2,2,d,8)
 
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
diff --git a/builtins/target-sse2.ll b/builtins/target-sse2.ll
index c858ccb6..e42d4990 100644
--- a/builtins/target-sse2.ll
+++ b/builtins/target-sse2.ll
@@ -503,7 +503,7 @@ svml_define(float,f4,4,f)
 
 ;; double precision
 svml_declare(double,2,2)
-svml_define_x2(double,2,2,d,4)
+svml_define_x(double,2,2,d,4)
 
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
diff --git a/builtins/target-sse4-x2.ll b/builtins/target-sse4-x2.ll
index c45966e3..842db53f 100644
--- a/builtins/target-sse4-x2.ll
+++ b/builtins/target-sse4-x2.ll
@@ -108,87 +108,11 @@ define <8 x float> @__sqrt_varying_float(<8 x float>) nounwind readonly alwaysin
 include(`svml.m4')
 ;; single precision
 svml_declare(float,f4,4)
+svml_define_x(float,f4,4,f,8)
 
 ;; double precision
 svml_declare(double,2,2)
-svml_define_x4(double,2,2,d,8)
-
-
-define <8 x float> @__svml_sinf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_sinf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_asinf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_asinf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_cosf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_cosf4, %0)
-  ret <8 x float> %ret
-}
-
-define void @__svml_sincosf(<8 x float>, <8 x float> *,
-                                    <8 x float> *) nounwind readnone alwaysinline {
-  ; call svml_sincosf4 two times with the two 4-wide sub-vectors
-  %a = shufflevector <8 x float> %0, <8 x float> undef,
-         <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %b = shufflevector <8 x float> %0, <8 x float> undef,
-         <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-
-  %cospa = alloca <4 x float>
-  %sa = call <4 x float> @__svml_sincosf4(<4 x float> * %cospa, <4 x float> %a)
-
-  %cospb = alloca <4 x float>
-  %sb = call <4 x float> @__svml_sincosf4(<4 x float> * %cospb, <4 x float> %b)
-
-  %sin = shufflevector <4 x float> %sa, <4 x float> %sb,
-         <8 x i32> <i32 0, i32 1, i32 2, i32 3,
-                    i32 4, i32 5, i32 6, i32 7>
-  store <8 x float> %sin, <8 x float> * %1
-
-  %cosa = load <4 x float> * %cospa
-  %cosb = load <4 x float> * %cospb
-  %cos = shufflevector <4 x float> %cosa, <4 x float> %cosb,
-         <8 x i32> <i32 0, i32 1, i32 2, i32 3,
-                    i32 4, i32 5, i32 6, i32 7>
-  store <8 x float> %cos, <8 x float> * %2
-
-  ret void
-}
-
-define <8 x float> @__svml_tanf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_tanf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_atanf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_atanf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_atan2f(<8 x float>,
-                                          <8 x float>) nounwind readnone alwaysinline {
-  binary4to8(ret, float, @__svml_atan2f4, %0, %1)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_expf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_expf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_logf(<8 x float>) nounwind readnone alwaysinline {
-  unary4to8(ret, float, @__svml_logf4, %0)
-  ret <8 x float> %ret
-}
-
-define <8 x float> @__svml_powf(<8 x float>,
-                                        <8 x float>) nounwind readnone alwaysinline {
-  binary4to8(ret, float, @__svml_powf4, %0, %1)
-  ret <8 x float> %ret
-}
+svml_define_x(double,2,2,d,8)
 
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
diff --git a/builtins/target-sse4.ll b/builtins/target-sse4.ll
index eb82ab9a..88be6c59 100644
--- a/builtins/target-sse4.ll
+++ b/builtins/target-sse4.ll
@@ -216,7 +216,7 @@ svml_define(float,f4,4,f)
 
 ;; double precision
 svml_declare(double,2,2)
-svml_define_x2(double,2,2,d,4)
+svml_define_x(double,2,2,d,4)
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ; horizontal ops / reductions