Add correct NEON encodings for vsra and vrsra.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@117458 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index d9c9d2840093e6db6029177abe20c7953cdc01a5..cbc49bab5cf85d3e21f9a40fb387ec21e79f2301 100644 (file)
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -1730,18 +1730,18 @@ class N2VNSh<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6, bit op4,
 // both double- and quad-register.
 class N2VDShAdd<bit op24, bit op23, bits<4> op11_8, bit op7, bit op4,
                 string OpcodeStr, string Dt, ValueType Ty, SDNode ShOp>
-  : N2VImm<op24, op23, op11_8, op7, 0, op4, (outs DPR:$dst),
-           (ins DPR:$src1, DPR:$src2, i32imm:$SIMM), N2RegVShRFrm, IIC_VPALiD,
-           OpcodeStr, Dt, "$dst, $src2, $SIMM", "$src1 = $dst",
-           [(set DPR:$dst, (Ty (add DPR:$src1,
-                                (Ty (ShOp DPR:$src2, (i32 imm:$SIMM))))))]>;
+  : N2VImm<op24, op23, op11_8, op7, 0, op4, (outs DPR:$Vd),
+           (ins DPR:$src1, DPR:$Vm, i32imm:$SIMM), N2RegVShRFrm, IIC_VPALiD,
+           OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "$src1 = $Vd",
+           [(set DPR:$Vd, (Ty (add DPR:$src1,
+                                (Ty (ShOp DPR:$Vm, (i32 imm:$SIMM))))))]>;
 class N2VQShAdd<bit op24, bit op23, bits<4> op11_8, bit op7, bit op4,
                 string OpcodeStr, string Dt, ValueType Ty, SDNode ShOp>
-  : N2VImm<op24, op23, op11_8, op7, 1, op4, (outs QPR:$dst),
-           (ins QPR:$src1, QPR:$src2, i32imm:$SIMM), N2RegVShRFrm, IIC_VPALiD,
-           OpcodeStr, Dt, "$dst, $src2, $SIMM", "$src1 = $dst",
-           [(set QPR:$dst, (Ty (add QPR:$src1,
-                                (Ty (ShOp QPR:$src2, (i32 imm:$SIMM))))))]>;
+  : N2VImm<op24, op23, op11_8, op7, 1, op4, (outs QPR:$Vd),
+           (ins QPR:$src1, QPR:$Vm, i32imm:$SIMM), N2RegVShRFrm, IIC_VPALiD,
+           OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "$src1 = $Vd",
+           [(set QPR:$Vd, (Ty (add QPR:$src1,
+                                (Ty (ShOp QPR:$Vm, (i32 imm:$SIMM))))))]>;
 
 // Shift by immediate and insert,
 // both double- and quad-register.

diff --git a/test/MC/ARM/neon-shiftaccum-encoding.ll b/test/MC/ARM/neon-shiftaccum-encoding.ll
new file mode 100644 (file)
index 0000000..f969241
--- /dev/null
+++ b/test/MC/ARM/neon-shiftaccum-encoding.ll
@@ -0,0 +1,309 @@
+; RUN: llc -show-mc-encoding -march=arm -mcpu=cortex-a8 -mattr=+neon < %s | FileCheck %s
+
+define <8 x i8> @vsras8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+       %tmp1 = load <8 x i8>* %A
+       %tmp2 = load <8 x i8>* %B
+; CHECK: vsra.s8       d16, d17, #8            @ encoding: [0x31,0x01,0xc8,0xf2]
+       %tmp3 = ashr <8 x i8> %tmp2, < i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8 >
+        %tmp4 = add <8 x i8> %tmp1, %tmp3
+       ret <8 x i8> %tmp4
+}
+
+define <4 x i16> @vsras16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+       %tmp1 = load <4 x i16>* %A
+       %tmp2 = load <4 x i16>* %B
+; CHECK: vsra.s16      d16, d17, #16   @ encoding: [0x31,0x01,0xd0,0xf2]
+       %tmp3 = ashr <4 x i16> %tmp2, < i16 16, i16 16, i16 16, i16 16 >
+        %tmp4 = add <4 x i16> %tmp1, %tmp3
+       ret <4 x i16> %tmp4
+}
+
+define <2 x i32> @vsras32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = load <2 x i32>* %B
+; CHECK: vsra.s32      d16, d17, #32   @ encoding: [0x31,0x01,0xe0,0xf2]
+       %tmp3 = ashr <2 x i32> %tmp2, < i32 32, i32 32 >
+        %tmp4 = add <2 x i32> %tmp1, %tmp3
+       ret <2 x i32> %tmp4
+}
+
+define <1 x i64> @vsras64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+       %tmp1 = load <1 x i64>* %A
+       %tmp2 = load <1 x i64>* %B
+; CHECK: vsra.s64      d16, d17, #64   @ encoding: [0xb1,0x01,0xc0,0xf2]
+       %tmp3 = ashr <1 x i64> %tmp2, < i64 64 >
+        %tmp4 = add <1 x i64> %tmp1, %tmp3
+       ret <1 x i64> %tmp4
+}
+
+define <16 x i8> @vsraQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+       %tmp1 = load <16 x i8>* %A
+       %tmp2 = load <16 x i8>* %B
+; CHECK: vsra.s8       q9, q8, #8              @ encoding: [0x70,0x21,0xc8,0xf2]
+       %tmp3 = ashr <16 x i8> %tmp2, < i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8 >
+        %tmp4 = add <16 x i8> %tmp1, %tmp3
+       ret <16 x i8> %tmp4
+}
+
+define <8 x i16> @vsraQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+       %tmp1 = load <8 x i16>* %A
+       %tmp2 = load <8 x i16>* %B
+; CHECK: vsra.s16      q9, q8, #16     @ encoding: [0x70,0x21,0xd0,0xf2]
+       %tmp3 = ashr <8 x i16> %tmp2, < i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16 >
+        %tmp4 = add <8 x i16> %tmp1, %tmp3
+       ret <8 x i16> %tmp4
+}
+
+define <4 x i32> @vsraQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = load <4 x i32>* %B
+; CHECK: vsra.s32      q9, q8, #32     @ encoding: [0x70,0x21,0xe0,0xf2]
+       %tmp3 = ashr <4 x i32> %tmp2, < i32 32, i32 32, i32 32, i32 32 >
+        %tmp4 = add <4 x i32> %tmp1, %tmp3
+       ret <4 x i32> %tmp4
+}
+
+define <2 x i64> @vsraQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+       %tmp1 = load <2 x i64>* %A
+       %tmp2 = load <2 x i64>* %B
+; CHECK: vsra.s64      q9, q8, #64     @ encoding: [0xf0,0x21,0xc0,0xf2]
+       %tmp3 = ashr <2 x i64> %tmp2, < i64 64, i64 64 >
+        %tmp4 = add <2 x i64> %tmp1, %tmp3
+       ret <2 x i64> %tmp4
+}
+
+define <8 x i8> @vsrau8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+       %tmp1 = load <8 x i8>* %A
+       %tmp2 = load <8 x i8>* %B
+; CHECK: vsra.u8       d16, d17, #8            @ encoding: [0x31,0x01,0xc8,0xf3]
+       %tmp3 = lshr <8 x i8> %tmp2, < i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8 >
+        %tmp4 = add <8 x i8> %tmp1, %tmp3
+       ret <8 x i8> %tmp4
+}
+
+define <4 x i16> @vsrau16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+       %tmp1 = load <4 x i16>* %A
+       %tmp2 = load <4 x i16>* %B
+; CHECK: vsra.u16      d16, d17, #16   @ encoding: [0x31,0x01,0xd0,0xf3]
+       %tmp3 = lshr <4 x i16> %tmp2, < i16 16, i16 16, i16 16, i16 16 >
+        %tmp4 = add <4 x i16> %tmp1, %tmp3
+       ret <4 x i16> %tmp4
+}
+
+define <2 x i32> @vsrau32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = load <2 x i32>* %B
+; CHECK: vsra.u32      d16, d17, #32   @ encoding: [0x31,0x01,0xe0,0xf3]
+       %tmp3 = lshr <2 x i32> %tmp2, < i32 32, i32 32 >
+        %tmp4 = add <2 x i32> %tmp1, %tmp3
+       ret <2 x i32> %tmp4
+}
+
+define <1 x i64> @vsrau64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+       %tmp1 = load <1 x i64>* %A
+       %tmp2 = load <1 x i64>* %B
+; CHECK: vsra.u64      d16, d17, #64   @ encoding: [0xb1,0x01,0xc0,0xf3]
+       %tmp3 = lshr <1 x i64> %tmp2, < i64 64 >
+        %tmp4 = add <1 x i64> %tmp1, %tmp3
+       ret <1 x i64> %tmp4
+}
+
+define <16 x i8> @vsraQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+       %tmp1 = load <16 x i8>* %A
+       %tmp2 = load <16 x i8>* %B
+; CHECK: vsra.u8       q9, q8, #8              @ encoding: [0x70,0x21,0xc8,0xf3]
+       %tmp3 = lshr <16 x i8> %tmp2, < i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8 >
+        %tmp4 = add <16 x i8> %tmp1, %tmp3
+       ret <16 x i8> %tmp4
+}
+
+define <8 x i16> @vsraQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+       %tmp1 = load <8 x i16>* %A
+       %tmp2 = load <8 x i16>* %B
+; CHECK: vsra.u16      q9, q8, #16     @ encoding: [0x70,0x21,0xd0,0xf3]
+       %tmp3 = lshr <8 x i16> %tmp2, < i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16 >
+        %tmp4 = add <8 x i16> %tmp1, %tmp3
+       ret <8 x i16> %tmp4
+}
+
+define <4 x i32> @vsraQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = load <4 x i32>* %B
+; CHECK: vsra.u32      q9, q8, #32     @ encoding: [0x70,0x21,0xe0,0xf3]
+       %tmp3 = lshr <4 x i32> %tmp2, < i32 32, i32 32, i32 32, i32 32 >
+        %tmp4 = add <4 x i32> %tmp1, %tmp3
+       ret <4 x i32> %tmp4
+}
+
+define <2 x i64> @vsraQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+       %tmp1 = load <2 x i64>* %A
+       %tmp2 = load <2 x i64>* %B
+; CHECK: vsra.u64      q9, q8, #64     @ encoding: [0xf0,0x21,0xc0,0xf3]
+       %tmp3 = lshr <2 x i64> %tmp2, < i64 64, i64 64 >
+        %tmp4 = add <2 x i64> %tmp1, %tmp3
+       ret <2 x i64> %tmp4
+}
+
+define <8 x i8> @vrsras8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+       %tmp1 = load <8 x i8>* %A
+       %tmp2 = load <8 x i8>* %B
+; CHECK: vrsra.s8      d16, d17, #8    @ encoding: [0x31,0x03,0xc8,0xf2]
+       %tmp3 = call <8 x i8> @llvm.arm.neon.vrshifts.v8i8(<8 x i8> %tmp2, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
+        %tmp4 = add <8 x i8> %tmp1, %tmp3
+       ret <8 x i8> %tmp4
+}
+
+define <4 x i16> @vrsras16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+       %tmp1 = load <4 x i16>* %A
+       %tmp2 = load <4 x i16>* %B
+; CHECK: vrsra.s16     d16, d17, #16   @ encoding: [0x31,0x03,0xd0,0xf2]
+       %tmp3 = call <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16> %tmp2, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
+        %tmp4 = add <4 x i16> %tmp1, %tmp3
+       ret <4 x i16> %tmp4
+}
+
+define <2 x i32> @vrsras32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = load <2 x i32>* %B
+; CHECK: vrsra.s32     d16, d17, #32   @ encoding: [0x31,0x03,0xe0,0xf2]
+       %tmp3 = call <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32> %tmp2, <2 x i32> < i32 -32, i32 -32 >)
+        %tmp4 = add <2 x i32> %tmp1, %tmp3
+       ret <2 x i32> %tmp4
+}
+
+define <1 x i64> @vrsras64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+       %tmp1 = load <1 x i64>* %A
+       %tmp2 = load <1 x i64>* %B
+; CHECK: vrsra.s64     d16, d17, #64   @ encoding: [0xb1,0x03,0xc0,0xf2]
+       %tmp3 = call <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64> %tmp2, <1 x i64> < i64 -64 >)
+        %tmp4 = add <1 x i64> %tmp1, %tmp3
+       ret <1 x i64> %tmp4
+}
+
+define <8 x i8> @vrsrau8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+       %tmp1 = load <8 x i8>* %A
+       %tmp2 = load <8 x i8>* %B
+; CHECK: vrsra.u8      d16, d17, #8    @ encoding: [0x31,0x03,0xc8,0xf3]
+       %tmp3 = call <8 x i8> @llvm.arm.neon.vrshiftu.v8i8(<8 x i8> %tmp2, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
+        %tmp4 = add <8 x i8> %tmp1, %tmp3
+       ret <8 x i8> %tmp4
+}
+
+define <4 x i16> @vrsrau16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+       %tmp1 = load <4 x i16>* %A
+       %tmp2 = load <4 x i16>* %B
+; CHECK: vrsra.u16     d16, d17, #16   @ encoding: [0x31,0x03,0xd0,0xf3]
+       %tmp3 = call <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16> %tmp2, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
+        %tmp4 = add <4 x i16> %tmp1, %tmp3
+       ret <4 x i16> %tmp4
+}
+
+define <2 x i32> @vrsrau32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = load <2 x i32>* %B
+; CHECK: vrsra.u32     d16, d17, #32   @ encoding: [0x31,0x03,0xe0,0xf3]
+       %tmp3 = call <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32> %tmp2, <2 x i32> < i32 -32, i32 -32 >)
+        %tmp4 = add <2 x i32> %tmp1, %tmp3
+       ret <2 x i32> %tmp4
+}
+
+define <1 x i64> @vrsrau64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+       %tmp1 = load <1 x i64>* %A
+       %tmp2 = load <1 x i64>* %B
+; CHECK: vrsra.u64     d16, d17, #64   @ encoding: [0xb1,0x03,0xc0,0xf3]
+       %tmp3 = call <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64> %tmp2, <1 x i64> < i64 -64 >)
+        %tmp4 = add <1 x i64> %tmp1, %tmp3
+       ret <1 x i64> %tmp4
+}
+
+define <16 x i8> @vrsraQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+       %tmp1 = load <16 x i8>* %A
+       %tmp2 = load <16 x i8>* %B
+; CHECK: vrsra.s8      q9, q8, #8      @ encoding: [0x70,0x23,0xc8,0xf2]
+       %tmp3 = call <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8> %tmp2, <16 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
+        %tmp4 = add <16 x i8> %tmp1, %tmp3
+       ret <16 x i8> %tmp4
+}
+
+define <8 x i16> @vrsraQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+       %tmp1 = load <8 x i16>* %A
+       %tmp2 = load <8 x i16>* %B
+; CHECK: vrsra.s16     q9, q8, #16     @ encoding: [0x70,0x23,0xd0,0xf2]
+       %tmp3 = call <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16> %tmp2, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
+        %tmp4 = add <8 x i16> %tmp1, %tmp3
+       ret <8 x i16> %tmp4
+}
+
+define <4 x i32> @vrsraQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = load <4 x i32>* %B
+; CHECK: vrsra.s32     q9, q8, #32     @ encoding: [0x70,0x23,0xe0,0xf2]
+       %tmp3 = call <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32> %tmp2, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
+        %tmp4 = add <4 x i32> %tmp1, %tmp3
+       ret <4 x i32> %tmp4
+}
+
+define <2 x i64> @vrsraQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+       %tmp1 = load <2 x i64>* %A
+       %tmp2 = load <2 x i64>* %B
+; CHECK: vrsra.s64     q9, q8, #64     @ encoding: [0xf0,0x23,0xc0,0xf2]
+       %tmp3 = call <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64> %tmp2, <2 x i64> < i64 -64, i64 -64 >)
+        %tmp4 = add <2 x i64> %tmp1, %tmp3
+       ret <2 x i64> %tmp4
+}
+
+define <16 x i8> @vrsraQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+       %tmp1 = load <16 x i8>* %A
+       %tmp2 = load <16 x i8>* %B
+; CHECK: vrsra.u8      q9, q8, #8      @ encoding: [0x70,0x23,0xc8,0xf3]
+       %tmp3 = call <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8> %tmp2, <16 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
+        %tmp4 = add <16 x i8> %tmp1, %tmp3
+       ret <16 x i8> %tmp4
+}
+
+define <8 x i16> @vrsraQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+       %tmp1 = load <8 x i16>* %A
+       %tmp2 = load <8 x i16>* %B
+; CHECK: vrsra.u16     q9, q8, #16     @ encoding: [0x70,0x23,0xd0,0xf3]
+       %tmp3 = call <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16> %tmp2, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
+        %tmp4 = add <8 x i16> %tmp1, %tmp3
+       ret <8 x i16> %tmp4
+}
+
+define <4 x i32> @vrsraQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = load <4 x i32>* %B
+; CHECK: vrsra.u32     q9, q8, #32     @ encoding: [0x70,0x23,0xe0,0xf3]
+       %tmp3 = call <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32> %tmp2, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
+        %tmp4 = add <4 x i32> %tmp1, %tmp3
+       ret <4 x i32> %tmp4
+}
+
+define <2 x i64> @vrsraQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+       %tmp1 = load <2 x i64>* %A
+       %tmp2 = load <2 x i64>* %B
+; CHECK: vrsra.u64     q9, q8, #64     @ encoding: [0xf0,0x23,0xc0,0xf3]
+       %tmp3 = call <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64> %tmp2, <2 x i64> < i64 -64, i64 -64 >)
+        %tmp4 = add <2 x i64> %tmp1, %tmp3
+       ret <2 x i64> %tmp4
+}
+
+declare <8 x i8>  @llvm.arm.neon.vrshifts.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vrshifts.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vrshifts.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vrshifts.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
+
+declare <8 x i8>  @llvm.arm.neon.vrshiftu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vrshiftu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vrshiftu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vrshiftu.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
+
+declare <16 x i8> @llvm.arm.neon.vrshifts.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
+declare <8 x i16> @llvm.arm.neon.vrshifts.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vrshifts.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vrshifts.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
+
+declare <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
+declare <8 x i16> @llvm.arm.neon.vrshiftu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vrshiftu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vrshiftu.v2i64(<2 x i64>, <2 x i64>) nounwind readnone