-; RUN: llc < %s -debug-only=isel -march=r600 -mcpu=redwood -o - 2>&1 | FileCheck %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck --check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; CHECK: rotr
-; CHECK: @rotr
-; CHECK: BIT_ALIGN_INT
-define void @rotr(i32 addrspace(1)* %in, i32 %x, i32 %y) {
+; FUNC-LABEL: @rotr_i32:
+; R600: BIT_ALIGN_INT
+
+; SI: V_ALIGNBIT_B32
+define void @rotr_i32(i32 addrspace(1)* %in, i32 %x, i32 %y) {
entry:
- %0 = sub i32 32, %y
- %1 = shl i32 %x, %0
- %2 = lshr i32 %x, %y
- %3 = or i32 %1, %2
- store i32 %3, i32 addrspace(1)* %in
+ %tmp0 = sub i32 32, %y
+ %tmp1 = shl i32 %x, %tmp0
+ %tmp2 = lshr i32 %x, %y
+ %tmp3 = or i32 %tmp1, %tmp2
+ store i32 %tmp3, i32 addrspace(1)* %in
ret void
}
-; CHECK: rotr
-; CHECK: @rotl
-; CHECK: SUB_INT {{\** T[0-9]+\.[XYZW]}}, literal.x
-; CHECK-NEXT: 32
-; CHECK: BIT_ALIGN_INT {{\** T[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW], PV.[xyzw]}}
-define void @rotl(i32 addrspace(1)* %in, i32 %x, i32 %y) {
+; FUNC-LABEL: @rotr_v2i32:
+; R600: BIT_ALIGN_INT
+; R600: BIT_ALIGN_INT
+
+; SI: V_ALIGNBIT_B32
+; SI: V_ALIGNBIT_B32
+define void @rotr_v2i32(<2 x i32> addrspace(1)* %in, <2 x i32> %x, <2 x i32> %y) {
+entry:
+ %tmp0 = sub <2 x i32> <i32 32, i32 32>, %y
+ %tmp1 = shl <2 x i32> %x, %tmp0
+ %tmp2 = lshr <2 x i32> %x, %y
+ %tmp3 = or <2 x i32> %tmp1, %tmp2
+ store <2 x i32> %tmp3, <2 x i32> addrspace(1)* %in
+ ret void
+}
+
+; FUNC-LABEL: @rotr_v4i32:
+; R600: BIT_ALIGN_INT
+; R600: BIT_ALIGN_INT
+; R600: BIT_ALIGN_INT
+; R600: BIT_ALIGN_INT
+
+; SI: V_ALIGNBIT_B32
+; SI: V_ALIGNBIT_B32
+; SI: V_ALIGNBIT_B32
+; SI: V_ALIGNBIT_B32
+define void @rotr_v4i32(<4 x i32> addrspace(1)* %in, <4 x i32> %x, <4 x i32> %y) {
entry:
- %0 = shl i32 %x, %y
- %1 = sub i32 32, %y
- %2 = lshr i32 %x, %1
- %3 = or i32 %0, %2
- store i32 %3, i32 addrspace(1)* %in
+ %tmp0 = sub <4 x i32> <i32 32, i32 32, i32 32, i32 32>, %y
+ %tmp1 = shl <4 x i32> %x, %tmp0
+ %tmp2 = lshr <4 x i32> %x, %y
+ %tmp3 = or <4 x i32> %tmp1, %tmp2
+ store <4 x i32> %tmp3, <4 x i32> addrspace(1)* %in
ret void
}