Add NEON VLD1-dup instructions (load 1 element to all lanes).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120194 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 43 insertions, 1 deletions

diff --git a/test/CodeGen/ARM/2009-11-02-NegativeLane.ll b/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
index 89c9037..ca5ae8b 100644
--- a/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
+++ b/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
@@ -1,4 +1,4 @@
-; RUN: llc -mcpu=cortex-a8 < %s | grep vdup.16
+; RUN: llc -mcpu=cortex-a8 < %s | FileCheck %s
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64"
 target triple = "armv7-eabi"
 
@@ -7,6 +7,7 @@ entry:
   br i1 undef, label %return, label %bb
 
 bb:                                               ; preds = %bb, %entry
+; CHECK: vld1.16 {d16[], d17[]}
   %0 = load i16* undef, align 2
   %1 = insertelement <8 x i16> undef, i16 %0, i32 2
   %2 = insertelement <8 x i16> %1, i16 undef, i32 3
diff --git a/test/CodeGen/ARM/vlddup.ll b/test/CodeGen/ARM/vlddup.ll
new file mode 100644
index 0000000..f172ed5
--- /dev/null
+++ b/test/CodeGen/ARM/vlddup.ll
@@ -0,0 +1,41 @@
+; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
+
+define <8 x i8> @vld1dupi8(i8* %A) nounwind {
+;CHECK: vld1dupi8:
+;Check the (default) alignment value.
+;CHECK: vld1.8 {d16[]}, [r0]
+	%tmp1 = load i8* %A, align 8
+	%tmp2 = insertelement <8 x i8> undef, i8 %tmp1, i32 0
+	%tmp3 = shufflevector <8 x i8> %tmp2, <8 x i8> undef, <8 x i32> zeroinitializer
+        ret <8 x i8> %tmp3
+}
+
+define <4 x i16> @vld1dupi16(i16* %A) nounwind {
+;CHECK: vld1dupi16:
+;Check the alignment value.  Max for this instruction is 16 bits:
+;CHECK: vld1.16 {d16[]}, [r0, :16]
+	%tmp1 = load i16* %A, align 8
+	%tmp2 = insertelement <4 x i16> undef, i16 %tmp1, i32 0
+	%tmp3 = shufflevector <4 x i16> %tmp2, <4 x i16> undef, <4 x i32> zeroinitializer
+        ret <4 x i16> %tmp3
+}
+
+define <2 x i32> @vld1dupi32(i32* %A) nounwind {
+;CHECK: vld1dupi32:
+;Check the alignment value.  Max for this instruction is 32 bits:
+;CHECK: vld1.32 {d16[]}, [r0, :32]
+	%tmp1 = load i32* %A, align 8
+	%tmp2 = insertelement <2 x i32> undef, i32 %tmp1, i32 0
+	%tmp3 = shufflevector <2 x i32> %tmp2, <2 x i32> undef, <2 x i32> zeroinitializer
+        ret <2 x i32> %tmp3
+}
+
+define <16 x i8> @vld1dupQi8(i8* %A) nounwind {
+;CHECK: vld1dupQi8:
+;Check the (default) alignment value.
+;CHECK: vld1.8 {d16[], d17[]}, [r0]
+	%tmp1 = load i8* %A, align 8
+	%tmp2 = insertelement <16 x i8> undef, i8 %tmp1, i32 0
+	%tmp3 = shufflevector <16 x i8> %tmp2, <16 x i8> undef, <16 x i32> zeroinitializer
+        ret <16 x i8> %tmp3
+}