From: Owen Anderson <resistor@mac.com>
Date: Wed, 27 Oct 2010 17:40:08 +0000 (+0000)
Subject: Add correct NEON encodings for vsli and vsri.
X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=0745c389d903bf9d8a8705ff49bea818a6be6c52;p=oota-llvm.git

Add correct NEON encodings for vsli and vsri.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@117459 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index cbc49bab5cf..0e2473650ed 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -1747,16 +1747,16 @@ class N2VQShAdd<bit op24, bit op23, bits<4> op11_8, bit op7, bit op4,
 // both double- and quad-register.
 class N2VDShIns<bit op24, bit op23, bits<4> op11_8, bit op7, bit op4,
                 Format f, 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), f, IIC_VSHLiD,
-           OpcodeStr, Dt, "$dst, $src2, $SIMM", "$src1 = $dst",
-           [(set DPR:$dst, (Ty (ShOp DPR:$src1, DPR:$src2, (i32 imm:$SIMM))))]>;
+  : N2VImm<op24, op23, op11_8, op7, 0, op4, (outs DPR:$Vd),
+           (ins DPR:$src1, DPR:$Vm, i32imm:$SIMM), f, IIC_VSHLiD,
+           OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "$src1 = $Vd",
+           [(set DPR:$Vd, (Ty (ShOp DPR:$src1, DPR:$Vm, (i32 imm:$SIMM))))]>;
 class N2VQShIns<bit op24, bit op23, bits<4> op11_8, bit op7, bit op4,
                 Format f, string OpcodeStr, string Dt, ValueType Ty,SDNode ShOp>
-  : N2VImm<op24, op23, op11_8, op7, 1, op4, (outs QPR:$dst),
-           (ins QPR:$src1, QPR:$src2, i32imm:$SIMM), f, IIC_VSHLiQ,
-           OpcodeStr, Dt, "$dst, $src2, $SIMM", "$src1 = $dst",
-           [(set QPR:$dst, (Ty (ShOp QPR:$src1, QPR:$src2, (i32 imm:$SIMM))))]>;
+  : N2VImm<op24, op23, op11_8, op7, 1, op4, (outs QPR:$Vd),
+           (ins QPR:$src1, QPR:$Vm, i32imm:$SIMM), f, IIC_VSHLiQ,
+           OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "$src1 = $Vd",
+           [(set QPR:$Vd, (Ty (ShOp QPR:$src1, QPR:$Vm, (i32 imm:$SIMM))))]>;
 
 // Convert, with fractional bits immediate,
 // both double- and quad-register.
diff --git a/test/MC/ARM/neon-shiftaccum-encoding.ll b/test/MC/ARM/neon-shiftaccum-encoding.ll
index f9692412bf3..e2e71092b7b 100644
--- a/test/MC/ARM/neon-shiftaccum-encoding.ll
+++ b/test/MC/ARM/neon-shiftaccum-encoding.ll
@@ -307,3 +307,141 @@ declare <16 x i8> @llvm.arm.neon.vrshiftu.v16i8(<16 x i8>, <16 x i8>) nounwind r
 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
+
+define <8 x i8> @vsli8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+	%tmp1 = load <8 x i8>* %A
+	%tmp2 = load <8 x i8>* %B
+; CHECK: vsli.8	d17, d16, #7            @ encoding: [0x30,0x15,0xcf,0xf3]
+	%tmp3 = call <8 x i8> @llvm.arm.neon.vshiftins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
+	ret <8 x i8> %tmp3
+}
+
+define <4 x i16> @vsli16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+	%tmp1 = load <4 x i16>* %A
+	%tmp2 = load <4 x i16>* %B
+; CHECK: vsli.16	d17, d16, #15           @ encoding: [0x30,0x15,0xdf,0xf3]
+	%tmp3 = call <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >)
+	ret <4 x i16> %tmp3
+}
+
+define <2 x i32> @vsli32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+	%tmp1 = load <2 x i32>* %A
+	%tmp2 = load <2 x i32>* %B
+; CHECK: vsli.32	d17, d16, #31           @ encoding: [0x30,0x15,0xff,0xf3]
+	%tmp3 = call <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2, <2 x i32> < i32 31, i32 31 >)
+	ret <2 x i32> %tmp3
+}
+
+define <1 x i64> @vsli64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+	%tmp1 = load <1 x i64>* %A
+	%tmp2 = load <1 x i64>* %B
+; CHECK: vsli.64	d17, d16, #63           @ encoding: [0xb0,0x15,0xff,0xf3]
+	%tmp3 = call <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2, <1 x i64> < i64 63 >)
+	ret <1 x i64> %tmp3
+}
+
+define <16 x i8> @vsliQ8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+	%tmp1 = load <16 x i8>* %A
+	%tmp2 = load <16 x i8>* %B
+; CHECK: vsli.8	q8, q9, #7              @ encoding: [0x72,0x05,0xcf,0xf3]
+	%tmp3 = call <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2, <16 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
+	ret <16 x i8> %tmp3
+}
+
+define <8 x i16> @vsliQ16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+	%tmp1 = load <8 x i16>* %A
+	%tmp2 = load <8 x i16>* %B
+; CHECK: vsli.16	q8, q9, #15             @ encoding: [0x72,0x05,0xdf,0xf3]
+	%tmp3 = call <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >)
+	ret <8 x i16> %tmp3
+}
+
+define <4 x i32> @vsliQ32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+	%tmp1 = load <4 x i32>* %A
+	%tmp2 = load <4 x i32>* %B
+; CHECK: vsli.32	q8, q9, #31             @ encoding: [0x72,0x05,0xff,0xf3]
+	%tmp3 = call <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >)
+	ret <4 x i32> %tmp3
+}
+
+define <2 x i64> @vsliQ64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+	%tmp1 = load <2 x i64>* %A
+	%tmp2 = load <2 x i64>* %B
+; CHECK: vsli.64	q8, q9, #63             @ encoding: [0xf2,0x05,0xff,0xf3]
+	%tmp3 = call <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2, <2 x i64> < i64 63, i64 63 >)
+	ret <2 x i64> %tmp3
+}
+
+define <8 x i8> @vsri8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+	%tmp1 = load <8 x i8>* %A
+	%tmp2 = load <8 x i8>* %B
+; CHECK: vsri.8	d17, d16, #8            @ encoding: [0x30,0x14,0xc8,0xf3]
+	%tmp3 = call <8 x i8> @llvm.arm.neon.vshiftins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
+	ret <8 x i8> %tmp3
+}
+
+define <4 x i16> @vsri16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+	%tmp1 = load <4 x i16>* %A
+	%tmp2 = load <4 x i16>* %B
+; CHECK: vsri.16	d17, d16, #16           @ encoding: [0x30,0x14,0xd0,0xf3
+	%tmp3 = call <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
+	ret <4 x i16> %tmp3
+}
+
+define <2 x i32> @vsri32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+	%tmp1 = load <2 x i32>* %A
+	%tmp2 = load <2 x i32>* %B
+; CHECK: vsri.32	d17, d16, #32           @ encoding: [0x30,0x14,0xe0,0xf3]
+	%tmp3 = call <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2, <2 x i32> < i32 -32, i32 -32 >)
+	ret <2 x i32> %tmp3
+}
+
+define <1 x i64> @vsri64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+	%tmp1 = load <1 x i64>* %A
+	%tmp2 = load <1 x i64>* %B
+; CHECK: vsri.64	d17, d16, #64           @ encoding: [0xb0,0x14,0xc0,0xf3]
+	%tmp3 = call <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2, <1 x i64> < i64 -64 >)
+	ret <1 x i64> %tmp3
+}
+
+define <16 x i8> @vsriQ8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+	%tmp1 = load <16 x i8>* %A
+	%tmp2 = load <16 x i8>* %B
+; CHECK: vsri.8	q8, q9, #8              @ encoding: [0x72,0x04,0xc8,0xf3]
+	%tmp3 = call <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8> %tmp1, <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 >)
+	ret <16 x i8> %tmp3
+}
+
+define <8 x i16> @vsriQ16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+	%tmp1 = load <8 x i16>* %A
+	%tmp2 = load <8 x i16>* %B
+; CHECK: vsri.16	q8, q9, #16             @ encoding: [0x72,0x04,0xd0,0xf3]
+	%tmp3 = call <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
+	ret <8 x i16> %tmp3
+}
+
+define <4 x i32> @vsriQ32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+	%tmp1 = load <4 x i32>* %A
+	%tmp2 = load <4 x i32>* %B
+; CHECK: vsri.32	q8, q9, #32             @ encoding: [0x72,0x04,0xe0,0xf3]
+	%tmp3 = call <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
+	ret <4 x i32> %tmp3
+}
+
+define <2 x i64> @vsriQ64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+	%tmp1 = load <2 x i64>* %A
+	%tmp2 = load <2 x i64>* %B
+; CHECK: vsri.64	q8, q9, #64             @ encoding: [0xf2,0x04,0xc0,0xf3]
+	%tmp3 = call <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2, <2 x i64> < i64 -64, i64 -64 >)
+	ret <2 x i64> %tmp3
+}
+
+declare <8 x i8>  @llvm.arm.neon.vshiftins.v8i8(<8 x i8>, <8 x i8>, <8 x i8>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16>, <4 x i16>, <4 x i16>) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64>, <1 x i64>, <1 x i64>) nounwind readnone
+
+declare <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
+declare <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16>, <8 x i16>, <8 x i16>) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64>, <2 x i64>, <2 x i64>) nounwind readnone