ARM 'vuzp.32 Dd, Dm' is a pseudo-instruction.
While there is an encoding for it in VUZP, the result of that is undefined,
so we should avoid it. Define the instruction as a pseudo for VTRN.32
instead, as the ARM ARM indicates.
rdar://11222366
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154511 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/ARM/ARMISelDAGToDAG.cpp b/lib/Target/ARM/ARMISelDAGToDAG.cpp
index 435d6a9..1eafbbc 100644
--- a/lib/Target/ARM/ARMISelDAGToDAG.cpp
+++ b/lib/Target/ARM/ARMISelDAGToDAG.cpp
@@ -2845,7 +2845,8 @@
case MVT::v8i8: Opc = ARM::VUZPd8; break;
case MVT::v4i16: Opc = ARM::VUZPd16; break;
case MVT::v2f32:
- case MVT::v2i32: Opc = ARM::VUZPd32; break;
+ // vuzp.32 Dd, Dm is a pseudo-instruction expanded to vtrn.32 Dd, Dm.
+ case MVT::v2i32: Opc = ARM::VTRNd32; break;
case MVT::v16i8: Opc = ARM::VUZPq8; break;
case MVT::v8i16: Opc = ARM::VUZPq16; break;
case MVT::v4f32:
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 231e31a..c7219a6 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -5378,7 +5378,9 @@
def VUZPd8 : N2VDShuffle<0b00, 0b00010, "vuzp", "8">;
def VUZPd16 : N2VDShuffle<0b01, 0b00010, "vuzp", "16">;
-def VUZPd32 : N2VDShuffle<0b10, 0b00010, "vuzp", "32">;
+// vuzp.32 Dd, Dm is a pseudo-instruction expanded to vtrn.32 Dd, Dm.
+def : NEONInstAlias<"vuzp${p}.32 $Dd, $Dm",
+ (VTRNd32 DPR:$Dd, DPR:$Dm, pred:$p)>;
def VUZPq8 : N2VQShuffle<0b00, 0b00010, IIC_VPERMQ3, "vuzp", "8">;
def VUZPq16 : N2VQShuffle<0b01, 0b00010, IIC_VPERMQ3, "vuzp", "16">;
diff --git a/test/MC/ARM/neon-shuffle-encoding.s b/test/MC/ARM/neon-shuffle-encoding.s
index e4d6077..0f07d9f 100644
--- a/test/MC/ARM/neon-shuffle-encoding.s
+++ b/test/MC/ARM/neon-shuffle-encoding.s
@@ -60,6 +60,7 @@
vzip.16 q9, q8
vzip.32 q9, q8
vzip.32 d2, d3
+ vuzp.32 d2, d3
@ CHECK: vuzp.8 d17, d16 @ encoding: [0x20,0x11,0xf2,0xf3]
@ CHECK: vuzp.16 d17, d16 @ encoding: [0x20,0x11,0xf6,0xf3]
@@ -72,6 +73,7 @@
@ CHECK: vzip.16 q9, q8 @ encoding: [0xe0,0x21,0xf6,0xf3]
@ CHECK: vzip.32 q9, q8 @ encoding: [0xe0,0x21,0xfa,0xf3]
@ CHECK: vtrn.32 d2, d3 @ encoding: [0x83,0x20,0xba,0xf3]
+@ CHECK: vtrn.32 d2, d3 @ encoding: [0x83,0x20,0xba,0xf3]
@ VTRN alternate size suffices