; PTX: bar.sync 0;
; load gep cast
- %4 = load float, float* getelementptr inbounds ([10 x float]* addrspacecast ([10 x float] addrspace(3)* @array to [10 x float]*), i32 0, i32 5), align 4
+ %4 = load float, float* getelementptr inbounds ([10 x float], [10 x float]* addrspacecast ([10 x float] addrspace(3)* @array to [10 x float]*), i32 0, i32 5), align 4
; PTX: ld.shared.f32 %f{{[0-9]+}}, [array+20];
; store gep cast
- store float %v, float* getelementptr inbounds ([10 x float]* addrspacecast ([10 x float] addrspace(3)* @array to [10 x float]*), i32 0, i32 5), align 4
+ store float %v, float* getelementptr inbounds ([10 x float], [10 x float]* addrspacecast ([10 x float] addrspace(3)* @array to [10 x float]*), i32 0, i32 5), align 4
; PTX: st.shared.f32 [array+20], %f{{[0-9]+}};
call void @llvm.cuda.syncthreads()
; PTX: bar.sync 0;