From: Ahmed Bougacha <ahmed.bougacha@gmail.com>
Date: Thu, 16 Apr 2015 23:57:07 +0000 (+0000)
Subject: [AArch64] Don't assert on f16 in DUP PerfectShuffle generator.
X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=7e3c3ae7c14b3846bd8ad80ca5ef16a82d1dd318;p=oota-llvm.git

[AArch64] Don't assert on f16 in DUP PerfectShuffle generator.

Found by code inspection, but breaking i16 at least breaks other tests.
They aren't checking this in particular though, so also add some
explicit tests for the already working types.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235148 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/AArch64/AArch64ISelLowering.cpp b/lib/Target/AArch64/AArch64ISelLowering.cpp
index 9a45ca711bf..90a5e5ef6b7 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -5046,7 +5046,7 @@ static SDValue GeneratePerfectShuffle(unsigned PFEntry, SDValue LHS,
     unsigned Opcode;
     if (EltTy == MVT::i8)
       Opcode = AArch64ISD::DUPLANE8;
-    else if (EltTy == MVT::i16)
+    else if (EltTy == MVT::i16 || EltTy == MVT::f16)
       Opcode = AArch64ISD::DUPLANE16;
     else if (EltTy == MVT::i32 || EltTy == MVT::f32)
       Opcode = AArch64ISD::DUPLANE32;
diff --git a/test/CodeGen/AArch64/arm64-dup.ll b/test/CodeGen/AArch64/arm64-dup.ll
index 849e227116f..c6b7de366d2 100644
--- a/test/CodeGen/AArch64/arm64-dup.ll
+++ b/test/CodeGen/AArch64/arm64-dup.ll
@@ -321,3 +321,40 @@ entry:
   %sub = sub <4 x i16> %a, %mul
   ret <4 x i16> %sub
 }
+
+; Also test the DUP path in the PerfectShuffle generator.
+
+; CHECK-LABEL: test_perfectshuffle_dupext_v4i16:
+; CHECK-NEXT: dup.4h v0, v0[0]
+; CHECK-NEXT: ext.8b v0, v0, v1, #4
+define <4 x i16> @test_perfectshuffle_dupext_v4i16(<4 x i16> %a, <4 x i16> %b) nounwind {
+  %r = shufflevector <4 x i16> %a, <4 x i16> %b, <4 x i32> <i32 0, i32 0, i32 4, i32 5>
+  ret <4 x i16> %r
+}
+
+; CHECK-LABEL: test_perfectshuffle_dupext_v4f16:
+; CHECK-NEXT: dup.4h v0, v0[0]
+; CHECK-NEXT: ext.8b v0, v0, v1, #4
+; CHECK-NEXT: ret
+define <4 x half> @test_perfectshuffle_dupext_v4f16(<4 x half> %a, <4 x half> %b) nounwind {
+  %r = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 0, i32 0, i32 4, i32 5>
+  ret <4 x half> %r
+}
+
+; CHECK-LABEL: test_perfectshuffle_dupext_v4i32:
+; CHECK-NEXT: dup.4s v0, v0[0]
+; CHECK-NEXT: ext.16b v0, v0, v1, #8
+; CHECK-NEXT: ret
+define <4 x i32> @test_perfectshuffle_dupext_v4i32(<4 x i32> %a, <4 x i32> %b) nounwind {
+  %r = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 0, i32 4, i32 5>
+  ret <4 x i32> %r
+}
+
+; CHECK-LABEL: test_perfectshuffle_dupext_v4f32:
+; CHECK-NEXT: dup.4s v0, v0[0]
+; CHECK-NEXT: ext.16b v0, v0, v1, #8
+; CHECK-NEXT: ret
+define <4 x float> @test_perfectshuffle_dupext_v4f32(<4 x float> %a, <4 x float> %b) nounwind {
+  %r = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 4, i32 5>
+  ret <4 x float> %r
+}