--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/CodeGen/ARM/a15-SD-dep.ll	Thu Dec 12 13:56:28 2013 +0900
@@ -0,0 +1,58 @@
+; RUN: llc -O1 -mcpu=cortex-a15 -mtriple=armv7-linux-gnueabi -disable-a15-sd-optimization -verify-machineinstrs < %s  | FileCheck -check-prefix=CHECK-DISABLED %s
+; RUN: llc -O1 -mcpu=cortex-a15 -mtriple=armv7-linux-gnueabi -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK-ENABLED %s
+
+; CHECK-ENABLED-LABEL: t1:
+; CHECK-DISABLED-LABEL: t1:
+define <2 x float> @t1(float %f) {
+  ; CHECK-ENABLED: vdup.32 d{{[0-9]*}}, d0[0]
+  ; CHECK-DISABLED-NOT: vdup.32 d{{[0-9]*}}, d0[0]
+  %i1 = insertelement <2 x float> undef, float %f, i32 1
+  %i2 = fadd <2 x float> %i1, %i1
+  ret <2 x float> %i2
+}
+
+; CHECK-ENABLED-LABEL: t2:
+; CHECK-DISABLED-LABEL: t2:
+define <4 x float> @t2(float %g, float %f) {
+  ; CHECK-ENABLED: vdup.32 q{{[0-9]*}}, d0[0]
+  ; CHECK-DISABLED-NOT: vdup.32 d{{[0-9]*}}, d0[0]
+  %i1 = insertelement <4 x float> undef, float %f, i32 1
+  %i2 = fadd <4 x float> %i1, %i1
+  ret <4 x float> %i2
+}
+
+; CHECK-ENABLED-LABEL: t3:
+; CHECK-DISABLED-LABEL: t3:
+define arm_aapcs_vfpcc <2 x float> @t3(float %f) {
+  ; CHECK-ENABLED: vdup.32 d{{[0-9]*}}, d0[0] 
+  ; CHECK-DISABLED-NOT: vdup.32 d{{[0-9]*}}, d0[0]
+  %i1 = insertelement <2 x float> undef, float %f, i32 1
+  %i2 = fadd <2 x float> %i1, %i1
+  ret <2 x float> %i2
+}
+
+; CHECK-ENABLED-LABEL: t4:
+; CHECK-DISABLED-LABEL: t4:
+define <2 x float> @t4(float %f) {
+  ; CHECK-ENABLED: vdup.32 d{{[0-9]*}}, d0[0]
+  ; CHECK-DISABLED-NOT: vdup
+  %i1 = insertelement <2 x float> undef, float %f, i32 1
+  br label %b
+
+  ; Block %b has an S-reg as live-in.
+b:
+  %i2 = fadd <2 x float> %i1, %i1
+  ret <2 x float> %i2
+}
+
+; CHECK-ENABLED-LABEL: t5:
+; CHECK-DISABLED-LABEL: t5:
+define arm_aapcs_vfpcc <4 x float> @t5(<4 x float> %q, float %f) {
+  ; CHECK-ENABLED: vdup.32 d{{[0-9]*}}, d{{[0-9]*}}[0]
+  ; CHECK-ENABLED: vadd.f32
+  ; CHECK-ENABLED-NEXT: bx lr
+  ; CHECK-DISABLED-NOT: vdup
+  %i1 = insertelement <4 x float> %q, float %f, i32 1
+  %i2 = fadd <4 x float> %i1, %i1
+  ret <4 x float> %i2
+}