From 5d782754939c748dd0cde9dbb9755f35bdfac79b Mon Sep 17 00:00:00 2001 From: Bob Wilson Date: Thu, 8 Oct 2009 22:33:53 +0000 Subject: [PATCH] Convert more NEON tests to use FileCheck. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83587 91177308-0d34-0410-b5e6-96231b3b80d8 --- test/CodeGen/ARM/vqabs.ll | 17 ++++-- test/CodeGen/ARM/vqadd.ll | 42 +++++++++++---- test/CodeGen/ARM/vqdmlal.ll | 8 +-- test/CodeGen/ARM/vqdmlsl.ll | 8 +-- test/CodeGen/ARM/vqdmulh.ll | 22 ++++++-- test/CodeGen/ARM/vqdmull.ll | 8 +-- test/CodeGen/ARM/vqmovn.ll | 29 ++++++---- test/CodeGen/ARM/vqneg.ll | 17 ++++-- test/CodeGen/ARM/vqrshl.ll | 42 +++++++++++---- test/CodeGen/ARM/vqrshrn.ll | 29 ++++++---- test/CodeGen/ARM/vqshl.ll | 102 ++++++++++++++++++++++++++++-------- test/CodeGen/ARM/vqshrn.ll | 29 ++++++---- test/CodeGen/ARM/vqsub.ll | 42 +++++++++++---- 13 files changed, 295 insertions(+), 100 deletions(-) diff --git a/test/CodeGen/ARM/vqabs.ll b/test/CodeGen/ARM/vqabs.ll index 84e59387742..d923557345e 100644 --- a/test/CodeGen/ARM/vqabs.ll +++ b/test/CodeGen/ARM/vqabs.ll @@ -1,39 +1,48 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqabs\\.s8} %t | count 2 -; RUN: grep {vqabs\\.s16} %t | count 2 -; RUN: grep {vqabs\\.s32} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqabss8(<8 x i8>* %A) nounwind { +;CHECK: vqabss8: +;CHECK: vqabs.s8 %tmp1 = load <8 x i8>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqabs.v8i8(<8 x i8> %tmp1) ret <8 x i8> %tmp2 } define <4 x i16> @vqabss16(<4 x i16>* %A) nounwind { +;CHECK: vqabss16: +;CHECK: vqabs.s16 %tmp1 = load <4 x i16>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqabs.v4i16(<4 x i16> %tmp1) ret <4 x i16> %tmp2 } define <2 x i32> @vqabss32(<2 x i32>* %A) nounwind { +;CHECK: vqabss32: +;CHECK: vqabs.s32 %tmp1 = load <2 x i32>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqabs.v2i32(<2 x i32> %tmp1) ret <2 x i32> %tmp2 } define <16 x i8> @vqabsQs8(<16 x i8>* %A) nounwind { +;CHECK: vqabsQs8: +;CHECK: vqabs.s8 %tmp1 = load <16 x i8>* %A %tmp2 = call <16 x i8> @llvm.arm.neon.vqabs.v16i8(<16 x i8> %tmp1) ret <16 x i8> %tmp2 } define <8 x i16> @vqabsQs16(<8 x i16>* %A) nounwind { +;CHECK: vqabsQs16: +;CHECK: vqabs.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i16> @llvm.arm.neon.vqabs.v8i16(<8 x i16> %tmp1) ret <8 x i16> %tmp2 } define <4 x i32> @vqabsQs32(<4 x i32>* %A) nounwind { +;CHECK: vqabsQs32: +;CHECK: vqabs.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i32> @llvm.arm.neon.vqabs.v4i32(<4 x i32> %tmp1) ret <4 x i32> %tmp2 diff --git a/test/CodeGen/ARM/vqadd.ll b/test/CodeGen/ARM/vqadd.ll index bce677a93ef..a1669b60ab5 100644 --- a/test/CodeGen/ARM/vqadd.ll +++ b/test/CodeGen/ARM/vqadd.ll @@ -1,14 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqadd\\.s8} %t | count 2 -; RUN: grep {vqadd\\.s16} %t | count 2 -; RUN: grep {vqadd\\.s32} %t | count 2 -; RUN: grep {vqadd\\.s64} %t | count 2 -; RUN: grep {vqadd\\.u8} %t | count 2 -; RUN: grep {vqadd\\.u16} %t | count 2 -; RUN: grep {vqadd\\.u32} %t | count 2 -; RUN: grep {vqadd\\.u64} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqadds8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqadds8: +;CHECK: vqadd.s8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqadds.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -16,6 +10,8 @@ define <8 x i8> @vqadds8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqadds16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqadds16: +;CHECK: vqadd.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqadds.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -23,6 +19,8 @@ define <4 x i16> @vqadds16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqadds32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqadds32: +;CHECK: vqadd.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqadds.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -30,6 +28,8 @@ define <2 x i32> @vqadds32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqadds64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqadds64: +;CHECK: vqadd.s64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqadds.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -37,6 +37,8 @@ define <1 x i64> @vqadds64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <8 x i8> @vqaddu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqaddu8: +;CHECK: vqadd.u8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqaddu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -44,6 +46,8 @@ define <8 x i8> @vqaddu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqaddu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqaddu16: +;CHECK: vqadd.u16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqaddu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -51,6 +55,8 @@ define <4 x i16> @vqaddu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqaddu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqaddu32: +;CHECK: vqadd.u32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqaddu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -58,6 +64,8 @@ define <2 x i32> @vqaddu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqaddu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqaddu64: +;CHECK: vqadd.u64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqaddu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -65,6 +73,8 @@ define <1 x i64> @vqaddu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <16 x i8> @vqaddQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqaddQs8: +;CHECK: vqadd.s8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqadds.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -72,6 +82,8 @@ define <16 x i8> @vqaddQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqaddQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqaddQs16: +;CHECK: vqadd.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqadds.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -79,6 +91,8 @@ define <8 x i16> @vqaddQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqaddQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqaddQs32: +;CHECK: vqadd.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqadds.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -86,6 +100,8 @@ define <4 x i32> @vqaddQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqaddQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqaddQs64: +;CHECK: vqadd.s64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqadds.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) @@ -93,6 +109,8 @@ define <2 x i64> @vqaddQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { } define <16 x i8> @vqaddQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqaddQu8: +;CHECK: vqadd.u8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqaddu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -100,6 +118,8 @@ define <16 x i8> @vqaddQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqaddQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqaddQu16: +;CHECK: vqadd.u16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqaddu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -107,6 +127,8 @@ define <8 x i16> @vqaddQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqaddQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqaddQu32: +;CHECK: vqadd.u32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqaddu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -114,6 +136,8 @@ define <4 x i32> @vqaddQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqaddQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqaddQu64: +;CHECK: vqadd.u64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqaddu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) diff --git a/test/CodeGen/ARM/vqdmlal.ll b/test/CodeGen/ARM/vqdmlal.ll index 3f9cde2d017..3e4b7f55509 100644 --- a/test/CodeGen/ARM/vqdmlal.ll +++ b/test/CodeGen/ARM/vqdmlal.ll @@ -1,8 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqdmlal\\.s16} %t | count 1 -; RUN: grep {vqdmlal\\.s32} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <4 x i32> @vqdmlals16(<4 x i32>* %A, <4 x i16>* %B, <4 x i16>* %C) nounwind { +;CHECK: vqdmlals16: +;CHECK: vqdmlal.s16 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i16>* %B %tmp3 = load <4 x i16>* %C @@ -11,6 +11,8 @@ define <4 x i32> @vqdmlals16(<4 x i32>* %A, <4 x i16>* %B, <4 x i16>* %C) nounwi } define <2 x i64> @vqdmlals32(<2 x i64>* %A, <2 x i32>* %B, <2 x i32>* %C) nounwind { +;CHECK: vqdmlals32: +;CHECK: vqdmlal.s32 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i32>* %B %tmp3 = load <2 x i32>* %C diff --git a/test/CodeGen/ARM/vqdmlsl.ll b/test/CodeGen/ARM/vqdmlsl.ll index 280291670dd..ab3c51f8d20 100644 --- a/test/CodeGen/ARM/vqdmlsl.ll +++ b/test/CodeGen/ARM/vqdmlsl.ll @@ -1,8 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqdmlsl\\.s16} %t | count 1 -; RUN: grep {vqdmlsl\\.s32} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <4 x i32> @vqdmlsls16(<4 x i32>* %A, <4 x i16>* %B, <4 x i16>* %C) nounwind { +;CHECK: vqdmlsls16: +;CHECK: vqdmlsl.s16 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i16>* %B %tmp3 = load <4 x i16>* %C @@ -11,6 +11,8 @@ define <4 x i32> @vqdmlsls16(<4 x i32>* %A, <4 x i16>* %B, <4 x i16>* %C) nounwi } define <2 x i64> @vqdmlsls32(<2 x i64>* %A, <2 x i32>* %B, <2 x i32>* %C) nounwind { +;CHECK: vqdmlsls32: +;CHECK: vqdmlsl.s32 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i32>* %B %tmp3 = load <2 x i32>* %C diff --git a/test/CodeGen/ARM/vqdmulh.ll b/test/CodeGen/ARM/vqdmulh.ll index 1600dc5d61e..c9e6dd9c9e7 100644 --- a/test/CodeGen/ARM/vqdmulh.ll +++ b/test/CodeGen/ARM/vqdmulh.ll @@ -1,10 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqdmulh\\.s16} %t | count 2 -; RUN: grep {vqdmulh\\.s32} %t | count 2 -; RUN: grep {vqrdmulh\\.s16} %t | count 2 -; RUN: grep {vqrdmulh\\.s32} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <4 x i16> @vqdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqdmulhs16: +;CHECK: vqdmulh.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqdmulh.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -12,6 +10,8 @@ define <4 x i16> @vqdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqdmulhs32: +;CHECK: vqdmulh.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqdmulh.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -19,6 +19,8 @@ define <2 x i32> @vqdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <8 x i16> @vqdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqdmulhQs16: +;CHECK: vqdmulh.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqdmulh.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -26,6 +28,8 @@ define <8 x i16> @vqdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqdmulhQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqdmulhQs32: +;CHECK: vqdmulh.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqdmulh.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -39,6 +43,8 @@ declare <8 x i16> @llvm.arm.neon.vqdmulh.v8i16(<8 x i16>, <8 x i16>) nounwind re declare <4 x i32> @llvm.arm.neon.vqdmulh.v4i32(<4 x i32>, <4 x i32>) nounwind readnone define <4 x i16> @vqrdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqrdmulhs16: +;CHECK: vqrdmulh.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -46,6 +52,8 @@ define <4 x i16> @vqrdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqrdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqrdmulhs32: +;CHECK: vqrdmulh.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -53,6 +61,8 @@ define <2 x i32> @vqrdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <8 x i16> @vqrdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqrdmulhQs16: +;CHECK: vqrdmulh.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -60,6 +70,8 @@ define <8 x i16> @vqrdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqrdmulhQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqrdmulhQs32: +;CHECK: vqrdmulh.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) diff --git a/test/CodeGen/ARM/vqdmull.ll b/test/CodeGen/ARM/vqdmull.ll index 8ddd4d62edf..9062536392d 100644 --- a/test/CodeGen/ARM/vqdmull.ll +++ b/test/CodeGen/ARM/vqdmull.ll @@ -1,8 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqdmull\\.s16} %t | count 1 -; RUN: grep {vqdmull\\.s32} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <4 x i32> @vqdmulls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqdmulls16: +;CHECK: vqdmull.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqdmull.v4i32(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -10,6 +10,8 @@ define <4 x i32> @vqdmulls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i64> @vqdmulls32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqdmulls32: +;CHECK: vqdmull.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqdmull.v2i64(<2 x i32> %tmp1, <2 x i32> %tmp2) diff --git a/test/CodeGen/ARM/vqmovn.ll b/test/CodeGen/ARM/vqmovn.ll index 06e5f1edada..96c84d809a8 100644 --- a/test/CodeGen/ARM/vqmovn.ll +++ b/test/CodeGen/ARM/vqmovn.ll @@ -1,63 +1,72 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqmovn\\.s16} %t | count 1 -; RUN: grep {vqmovn\\.s32} %t | count 1 -; RUN: grep {vqmovn\\.s64} %t | count 1 -; RUN: grep {vqmovn\\.u16} %t | count 1 -; RUN: grep {vqmovn\\.u32} %t | count 1 -; RUN: grep {vqmovn\\.u64} %t | count 1 -; RUN: grep {vqmovun\\.s16} %t | count 1 -; RUN: grep {vqmovun\\.s32} %t | count 1 -; RUN: grep {vqmovun\\.s64} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqmovns16(<8 x i16>* %A) nounwind { +;CHECK: vqmovns16: +;CHECK: vqmovn.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqmovns.v8i8(<8 x i16> %tmp1) ret <8 x i8> %tmp2 } define <4 x i16> @vqmovns32(<4 x i32>* %A) nounwind { +;CHECK: vqmovns32: +;CHECK: vqmovn.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqmovns.v4i16(<4 x i32> %tmp1) ret <4 x i16> %tmp2 } define <2 x i32> @vqmovns64(<2 x i64>* %A) nounwind { +;CHECK: vqmovns64: +;CHECK: vqmovn.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqmovns.v2i32(<2 x i64> %tmp1) ret <2 x i32> %tmp2 } define <8 x i8> @vqmovnu16(<8 x i16>* %A) nounwind { +;CHECK: vqmovnu16: +;CHECK: vqmovn.u16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqmovnu.v8i8(<8 x i16> %tmp1) ret <8 x i8> %tmp2 } define <4 x i16> @vqmovnu32(<4 x i32>* %A) nounwind { +;CHECK: vqmovnu32: +;CHECK: vqmovn.u32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqmovnu.v4i16(<4 x i32> %tmp1) ret <4 x i16> %tmp2 } define <2 x i32> @vqmovnu64(<2 x i64>* %A) nounwind { +;CHECK: vqmovnu64: +;CHECK: vqmovn.u64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqmovnu.v2i32(<2 x i64> %tmp1) ret <2 x i32> %tmp2 } define <8 x i8> @vqmovuns16(<8 x i16>* %A) nounwind { +;CHECK: vqmovuns16: +;CHECK: vqmovun.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqmovnsu.v8i8(<8 x i16> %tmp1) ret <8 x i8> %tmp2 } define <4 x i16> @vqmovuns32(<4 x i32>* %A) nounwind { +;CHECK: vqmovuns32: +;CHECK: vqmovun.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqmovnsu.v4i16(<4 x i32> %tmp1) ret <4 x i16> %tmp2 } define <2 x i32> @vqmovuns64(<2 x i64>* %A) nounwind { +;CHECK: vqmovuns64: +;CHECK: vqmovun.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqmovnsu.v2i32(<2 x i64> %tmp1) ret <2 x i32> %tmp2 diff --git a/test/CodeGen/ARM/vqneg.ll b/test/CodeGen/ARM/vqneg.ll index 36265596f93..f55a132c178 100644 --- a/test/CodeGen/ARM/vqneg.ll +++ b/test/CodeGen/ARM/vqneg.ll @@ -1,39 +1,48 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqneg\\.s8} %t | count 2 -; RUN: grep {vqneg\\.s16} %t | count 2 -; RUN: grep {vqneg\\.s32} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqnegs8(<8 x i8>* %A) nounwind { +;CHECK: vqnegs8: +;CHECK: vqneg.s8 %tmp1 = load <8 x i8>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqneg.v8i8(<8 x i8> %tmp1) ret <8 x i8> %tmp2 } define <4 x i16> @vqnegs16(<4 x i16>* %A) nounwind { +;CHECK: vqnegs16: +;CHECK: vqneg.s16 %tmp1 = load <4 x i16>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqneg.v4i16(<4 x i16> %tmp1) ret <4 x i16> %tmp2 } define <2 x i32> @vqnegs32(<2 x i32>* %A) nounwind { +;CHECK: vqnegs32: +;CHECK: vqneg.s32 %tmp1 = load <2 x i32>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqneg.v2i32(<2 x i32> %tmp1) ret <2 x i32> %tmp2 } define <16 x i8> @vqnegQs8(<16 x i8>* %A) nounwind { +;CHECK: vqnegQs8: +;CHECK: vqneg.s8 %tmp1 = load <16 x i8>* %A %tmp2 = call <16 x i8> @llvm.arm.neon.vqneg.v16i8(<16 x i8> %tmp1) ret <16 x i8> %tmp2 } define <8 x i16> @vqnegQs16(<8 x i16>* %A) nounwind { +;CHECK: vqnegQs16: +;CHECK: vqneg.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i16> @llvm.arm.neon.vqneg.v8i16(<8 x i16> %tmp1) ret <8 x i16> %tmp2 } define <4 x i32> @vqnegQs32(<4 x i32>* %A) nounwind { +;CHECK: vqnegQs32: +;CHECK: vqneg.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i32> @llvm.arm.neon.vqneg.v4i32(<4 x i32> %tmp1) ret <4 x i32> %tmp2 diff --git a/test/CodeGen/ARM/vqrshl.ll b/test/CodeGen/ARM/vqrshl.ll index e680f93acc9..f94e92fd0f5 100644 --- a/test/CodeGen/ARM/vqrshl.ll +++ b/test/CodeGen/ARM/vqrshl.ll @@ -1,14 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqrshl\\.s8} %t | count 2 -; RUN: grep {vqrshl\\.s16} %t | count 2 -; RUN: grep {vqrshl\\.s32} %t | count 2 -; RUN: grep {vqrshl\\.s64} %t | count 2 -; RUN: grep {vqrshl\\.u8} %t | count 2 -; RUN: grep {vqrshl\\.u16} %t | count 2 -; RUN: grep {vqrshl\\.u32} %t | count 2 -; RUN: grep {vqrshl\\.u64} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqrshls8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqrshls8: +;CHECK: vqrshl.s8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqrshifts.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -16,6 +10,8 @@ define <8 x i8> @vqrshls8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqrshls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqrshls16: +;CHECK: vqrshl.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqrshifts.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -23,6 +19,8 @@ define <4 x i16> @vqrshls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqrshls32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqrshls32: +;CHECK: vqrshl.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqrshifts.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -30,6 +28,8 @@ define <2 x i32> @vqrshls32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqrshls64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqrshls64: +;CHECK: vqrshl.s64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqrshifts.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -37,6 +37,8 @@ define <1 x i64> @vqrshls64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <8 x i8> @vqrshlu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqrshlu8: +;CHECK: vqrshl.u8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqrshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -44,6 +46,8 @@ define <8 x i8> @vqrshlu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqrshlu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqrshlu16: +;CHECK: vqrshl.u16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqrshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -51,6 +55,8 @@ define <4 x i16> @vqrshlu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqrshlu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqrshlu32: +;CHECK: vqrshl.u32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqrshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -58,6 +64,8 @@ define <2 x i32> @vqrshlu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqrshlu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqrshlu64: +;CHECK: vqrshl.u64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqrshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -65,6 +73,8 @@ define <1 x i64> @vqrshlu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <16 x i8> @vqrshlQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqrshlQs8: +;CHECK: vqrshl.s8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqrshifts.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -72,6 +82,8 @@ define <16 x i8> @vqrshlQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqrshlQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqrshlQs16: +;CHECK: vqrshl.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqrshifts.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -79,6 +91,8 @@ define <8 x i16> @vqrshlQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqrshlQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqrshlQs32: +;CHECK: vqrshl.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqrshifts.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -86,6 +100,8 @@ define <4 x i32> @vqrshlQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqrshlQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqrshlQs64: +;CHECK: vqrshl.s64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqrshifts.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) @@ -93,6 +109,8 @@ define <2 x i64> @vqrshlQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { } define <16 x i8> @vqrshlQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqrshlQu8: +;CHECK: vqrshl.u8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqrshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -100,6 +118,8 @@ define <16 x i8> @vqrshlQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqrshlQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqrshlQu16: +;CHECK: vqrshl.u16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqrshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -107,6 +127,8 @@ define <8 x i16> @vqrshlQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqrshlQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqrshlQu32: +;CHECK: vqrshl.u32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqrshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -114,6 +136,8 @@ define <4 x i32> @vqrshlQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqrshlQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqrshlQu64: +;CHECK: vqrshl.u64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqrshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) diff --git a/test/CodeGen/ARM/vqrshrn.ll b/test/CodeGen/ARM/vqrshrn.ll index bb046fa1d7d..41c2e7cc18f 100644 --- a/test/CodeGen/ARM/vqrshrn.ll +++ b/test/CodeGen/ARM/vqrshrn.ll @@ -1,63 +1,72 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqrshrn\\.s16} %t | count 1 -; RUN: grep {vqrshrn\\.s32} %t | count 1 -; RUN: grep {vqrshrn\\.s64} %t | count 1 -; RUN: grep {vqrshrn\\.u16} %t | count 1 -; RUN: grep {vqrshrn\\.u32} %t | count 1 -; RUN: grep {vqrshrn\\.u64} %t | count 1 -; RUN: grep {vqrshrun\\.s16} %t | count 1 -; RUN: grep {vqrshrun\\.s32} %t | count 1 -; RUN: grep {vqrshrun\\.s64} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqrshrns8(<8 x i16>* %A) nounwind { +;CHECK: vqrshrns8: +;CHECK: vqrshrn.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftns.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqrshrns16(<4 x i32>* %A) nounwind { +;CHECK: vqrshrns16: +;CHECK: vqrshrn.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftns.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqrshrns32(<2 x i64>* %A) nounwind { +;CHECK: vqrshrns32: +;CHECK: vqrshrn.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftns.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 } define <8 x i8> @vqrshrnu8(<8 x i16>* %A) nounwind { +;CHECK: vqrshrnu8: +;CHECK: vqrshrn.u16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftnu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqrshrnu16(<4 x i32>* %A) nounwind { +;CHECK: vqrshrnu16: +;CHECK: vqrshrn.u32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftnu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqrshrnu32(<2 x i64>* %A) nounwind { +;CHECK: vqrshrnu32: +;CHECK: vqrshrn.u64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftnu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 } define <8 x i8> @vqrshruns8(<8 x i16>* %A) nounwind { +;CHECK: vqrshruns8: +;CHECK: vqrshrun.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftnsu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqrshruns16(<4 x i32>* %A) nounwind { +;CHECK: vqrshruns16: +;CHECK: vqrshrun.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftnsu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqrshruns32(<2 x i64>* %A) nounwind { +;CHECK: vqrshruns32: +;CHECK: vqrshrun.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftnsu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 diff --git a/test/CodeGen/ARM/vqshl.ll b/test/CodeGen/ARM/vqshl.ll index bfc4e88437d..e8fdcd3cb58 100644 --- a/test/CodeGen/ARM/vqshl.ll +++ b/test/CodeGen/ARM/vqshl.ll @@ -1,26 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqshl\\.s8} %t | count 4 -; RUN: grep {vqshl\\.s16} %t | count 4 -; RUN: grep {vqshl\\.s32} %t | count 4 -; RUN: grep {vqshl\\.s64} %t | count 4 -; RUN: grep {vqshl\\.u8} %t | count 4 -; RUN: grep {vqshl\\.u16} %t | count 4 -; RUN: grep {vqshl\\.u32} %t | count 4 -; RUN: grep {vqshl\\.u64} %t | count 4 -; RUN: grep {vqshl\\.s8.*#7} %t | count 2 -; RUN: grep {vqshl\\.s16.*#15} %t | count 2 -; RUN: grep {vqshl\\.s32.*#31} %t | count 2 -; RUN: grep {vqshl\\.s64.*#63} %t | count 2 -; RUN: grep {vqshl\\.u8.*#7} %t | count 2 -; RUN: grep {vqshl\\.u16.*#15} %t | count 2 -; RUN: grep {vqshl\\.u32.*#31} %t | count 2 -; RUN: grep {vqshl\\.u64.*#63} %t | count 2 -; RUN: grep {vqshlu\\.s8} %t | count 2 -; RUN: grep {vqshlu\\.s16} %t | count 2 -; RUN: grep {vqshlu\\.s32} %t | count 2 -; RUN: grep {vqshlu\\.s64} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqshls8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqshls8: +;CHECK: vqshl.s8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqshifts.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -28,6 +10,8 @@ define <8 x i8> @vqshls8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqshls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqshls16: +;CHECK: vqshl.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqshifts.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -35,6 +19,8 @@ define <4 x i16> @vqshls16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqshls32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqshls32: +;CHECK: vqshl.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqshifts.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -42,6 +28,8 @@ define <2 x i32> @vqshls32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqshls64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqshls64: +;CHECK: vqshl.s64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqshifts.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -49,6 +37,8 @@ define <1 x i64> @vqshls64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <8 x i8> @vqshlu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqshlu8: +;CHECK: vqshl.u8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -56,6 +46,8 @@ define <8 x i8> @vqshlu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqshlu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqshlu16: +;CHECK: vqshl.u16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -63,6 +55,8 @@ define <4 x i16> @vqshlu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqshlu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqshlu32: +;CHECK: vqshl.u32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -70,6 +64,8 @@ define <2 x i32> @vqshlu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqshlu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqshlu64: +;CHECK: vqshl.u64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -77,6 +73,8 @@ define <1 x i64> @vqshlu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <16 x i8> @vqshlQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqshlQs8: +;CHECK: vqshl.s8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqshifts.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -84,6 +82,8 @@ define <16 x i8> @vqshlQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqshlQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqshlQs16: +;CHECK: vqshl.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqshifts.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -91,6 +91,8 @@ define <8 x i16> @vqshlQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqshlQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqshlQs32: +;CHECK: vqshl.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqshifts.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -98,6 +100,8 @@ define <4 x i32> @vqshlQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqshlQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqshlQs64: +;CHECK: vqshl.s64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqshifts.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) @@ -105,6 +109,8 @@ define <2 x i64> @vqshlQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { } define <16 x i8> @vqshlQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqshlQu8: +;CHECK: vqshl.u8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqshiftu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -112,6 +118,8 @@ define <16 x i8> @vqshlQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqshlQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqshlQu16: +;CHECK: vqshl.u16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -119,6 +127,8 @@ define <8 x i16> @vqshlQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqshlQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqshlQu32: +;CHECK: vqshl.u32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -126,6 +136,8 @@ define <4 x i32> @vqshlQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqshlQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqshlQu64: +;CHECK: vqshl.u64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) @@ -133,144 +145,192 @@ define <2 x i64> @vqshlQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { } define <8 x i8> @vqshls_n8(<8 x i8>* %A) nounwind { +;CHECK: vqshls_n8: +;CHECK: vqshl.s8{{.*#7}} %tmp1 = load <8 x i8>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshifts.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshls_n16(<4 x i16>* %A) nounwind { +;CHECK: vqshls_n16: +;CHECK: vqshl.s16{{.*#15}} %tmp1 = load <4 x i16>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshifts.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshls_n32(<2 x i32>* %A) nounwind { +;CHECK: vqshls_n32: +;CHECK: vqshl.s32{{.*#31}} %tmp1 = load <2 x i32>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshifts.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >) ret <2 x i32> %tmp2 } define <1 x i64> @vqshls_n64(<1 x i64>* %A) nounwind { +;CHECK: vqshls_n64: +;CHECK: vqshl.s64{{.*#63}} %tmp1 = load <1 x i64>* %A %tmp2 = call <1 x i64> @llvm.arm.neon.vqshifts.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 63 >) ret <1 x i64> %tmp2 } define <8 x i8> @vqshlu_n8(<8 x i8>* %A) nounwind { +;CHECK: vqshlu_n8: +;CHECK: vqshl.u8{{.*#7}} %tmp1 = load <8 x i8>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshiftu.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshlu_n16(<4 x i16>* %A) nounwind { +;CHECK: vqshlu_n16: +;CHECK: vqshl.u16{{.*#15}} %tmp1 = load <4 x i16>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshiftu.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshlu_n32(<2 x i32>* %A) nounwind { +;CHECK: vqshlu_n32: +;CHECK: vqshl.u32{{.*#31}} %tmp1 = load <2 x i32>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshiftu.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >) ret <2 x i32> %tmp2 } define <1 x i64> @vqshlu_n64(<1 x i64>* %A) nounwind { +;CHECK: vqshlu_n64: +;CHECK: vqshl.u64{{.*#63}} %tmp1 = load <1 x i64>* %A %tmp2 = call <1 x i64> @llvm.arm.neon.vqshiftu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 63 >) ret <1 x i64> %tmp2 } define <8 x i8> @vqshlsu_n8(<8 x i8>* %A) nounwind { +;CHECK: vqshlsu_n8: +;CHECK: vqshlu.s8 %tmp1 = load <8 x i8>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshiftsu.v8i8(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshlsu_n16(<4 x i16>* %A) nounwind { +;CHECK: vqshlsu_n16: +;CHECK: vqshlu.s16 %tmp1 = load <4 x i16>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshiftsu.v4i16(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshlsu_n32(<2 x i32>* %A) nounwind { +;CHECK: vqshlsu_n32: +;CHECK: vqshlu.s32 %tmp1 = load <2 x i32>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshiftsu.v2i32(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >) ret <2 x i32> %tmp2 } define <1 x i64> @vqshlsu_n64(<1 x i64>* %A) nounwind { +;CHECK: vqshlsu_n64: +;CHECK: vqshlu.s64 %tmp1 = load <1 x i64>* %A %tmp2 = call <1 x i64> @llvm.arm.neon.vqshiftsu.v1i64(<1 x i64> %tmp1, <1 x i64> < i64 63 >) ret <1 x i64> %tmp2 } define <16 x i8> @vqshlQs_n8(<16 x i8>* %A) nounwind { +;CHECK: vqshlQs_n8: +;CHECK: vqshl.s8{{.*#7}} %tmp1 = load <16 x i8>* %A %tmp2 = call <16 x i8> @llvm.arm.neon.vqshifts.v16i8(<16 x i8> %tmp1, <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> %tmp2 } define <8 x i16> @vqshlQs_n16(<8 x i16>* %A) nounwind { +;CHECK: vqshlQs_n16: +;CHECK: vqshl.s16{{.*#15}} %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i16> @llvm.arm.neon.vqshifts.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >) ret <8 x i16> %tmp2 } define <4 x i32> @vqshlQs_n32(<4 x i32>* %A) nounwind { +;CHECK: vqshlQs_n32: +;CHECK: vqshl.s32{{.*#31}} %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i32> @llvm.arm.neon.vqshifts.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >) ret <4 x i32> %tmp2 } define <2 x i64> @vqshlQs_n64(<2 x i64>* %A) nounwind { +;CHECK: vqshlQs_n64: +;CHECK: vqshl.s64{{.*#63}} %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i64> @llvm.arm.neon.vqshifts.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 63, i64 63 >) ret <2 x i64> %tmp2 } define <16 x i8> @vqshlQu_n8(<16 x i8>* %A) nounwind { +;CHECK: vqshlQu_n8: +;CHECK: vqshl.u8{{.*#7}} %tmp1 = load <16 x i8>* %A %tmp2 = call <16 x i8> @llvm.arm.neon.vqshiftu.v16i8(<16 x i8> %tmp1, <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> %tmp2 } define <8 x i16> @vqshlQu_n16(<8 x i16>* %A) nounwind { +;CHECK: vqshlQu_n16: +;CHECK: vqshl.u16{{.*#15}} %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i16> @llvm.arm.neon.vqshiftu.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >) ret <8 x i16> %tmp2 } define <4 x i32> @vqshlQu_n32(<4 x i32>* %A) nounwind { +;CHECK: vqshlQu_n32: +;CHECK: vqshl.u32{{.*#31}} %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i32> @llvm.arm.neon.vqshiftu.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >) ret <4 x i32> %tmp2 } define <2 x i64> @vqshlQu_n64(<2 x i64>* %A) nounwind { +;CHECK: vqshlQu_n64: +;CHECK: vqshl.u64{{.*#63}} %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i64> @llvm.arm.neon.vqshiftu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 63, i64 63 >) ret <2 x i64> %tmp2 } define <16 x i8> @vqshlQsu_n8(<16 x i8>* %A) nounwind { +;CHECK: vqshlQsu_n8: +;CHECK: vqshlu.s8 %tmp1 = load <16 x i8>* %A %tmp2 = call <16 x i8> @llvm.arm.neon.vqshiftsu.v16i8(<16 x i8> %tmp1, <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> %tmp2 } define <8 x i16> @vqshlQsu_n16(<8 x i16>* %A) nounwind { +;CHECK: vqshlQsu_n16: +;CHECK: vqshlu.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i16> @llvm.arm.neon.vqshiftsu.v8i16(<8 x i16> %tmp1, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >) ret <8 x i16> %tmp2 } define <4 x i32> @vqshlQsu_n32(<4 x i32>* %A) nounwind { +;CHECK: vqshlQsu_n32: +;CHECK: vqshlu.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i32> @llvm.arm.neon.vqshiftsu.v4i32(<4 x i32> %tmp1, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >) ret <4 x i32> %tmp2 } define <2 x i64> @vqshlQsu_n64(<2 x i64>* %A) nounwind { +;CHECK: vqshlQsu_n64: +;CHECK: vqshlu.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i64> @llvm.arm.neon.vqshiftsu.v2i64(<2 x i64> %tmp1, <2 x i64> < i64 63, i64 63 >) ret <2 x i64> %tmp2 diff --git a/test/CodeGen/ARM/vqshrn.ll b/test/CodeGen/ARM/vqshrn.ll index fb53c36eeef..0e00c863a52 100644 --- a/test/CodeGen/ARM/vqshrn.ll +++ b/test/CodeGen/ARM/vqshrn.ll @@ -1,63 +1,72 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqshrn\\.s16} %t | count 1 -; RUN: grep {vqshrn\\.s32} %t | count 1 -; RUN: grep {vqshrn\\.s64} %t | count 1 -; RUN: grep {vqshrn\\.u16} %t | count 1 -; RUN: grep {vqshrn\\.u32} %t | count 1 -; RUN: grep {vqshrn\\.u64} %t | count 1 -; RUN: grep {vqshrun\\.s16} %t | count 1 -; RUN: grep {vqshrun\\.s32} %t | count 1 -; RUN: grep {vqshrun\\.s64} %t | count 1 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqshrns8(<8 x i16>* %A) nounwind { +;CHECK: vqshrns8: +;CHECK: vqshrn.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshiftns.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshrns16(<4 x i32>* %A) nounwind { +;CHECK: vqshrns16: +;CHECK: vqshrn.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshiftns.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshrns32(<2 x i64>* %A) nounwind { +;CHECK: vqshrns32: +;CHECK: vqshrn.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshiftns.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 } define <8 x i8> @vqshrnu8(<8 x i16>* %A) nounwind { +;CHECK: vqshrnu8: +;CHECK: vqshrn.u16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshiftnu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshrnu16(<4 x i32>* %A) nounwind { +;CHECK: vqshrnu16: +;CHECK: vqshrn.u32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshiftnu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshrnu32(<2 x i64>* %A) nounwind { +;CHECK: vqshrnu32: +;CHECK: vqshrn.u64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshiftnu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 } define <8 x i8> @vqshruns8(<8 x i16>* %A) nounwind { +;CHECK: vqshruns8: +;CHECK: vqshrun.s16 %tmp1 = load <8 x i16>* %A %tmp2 = call <8 x i8> @llvm.arm.neon.vqshiftnsu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >) ret <8 x i8> %tmp2 } define <4 x i16> @vqshruns16(<4 x i32>* %A) nounwind { +;CHECK: vqshruns16: +;CHECK: vqshrun.s32 %tmp1 = load <4 x i32>* %A %tmp2 = call <4 x i16> @llvm.arm.neon.vqshiftnsu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >) ret <4 x i16> %tmp2 } define <2 x i32> @vqshruns32(<2 x i64>* %A) nounwind { +;CHECK: vqshruns32: +;CHECK: vqshrun.s64 %tmp1 = load <2 x i64>* %A %tmp2 = call <2 x i32> @llvm.arm.neon.vqshiftnsu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >) ret <2 x i32> %tmp2 diff --git a/test/CodeGen/ARM/vqsub.ll b/test/CodeGen/ARM/vqsub.ll index bae4ebe9be6..4231fca37e3 100644 --- a/test/CodeGen/ARM/vqsub.ll +++ b/test/CodeGen/ARM/vqsub.ll @@ -1,14 +1,8 @@ -; RUN: llc < %s -march=arm -mattr=+neon > %t -; RUN: grep {vqsub\\.s8} %t | count 2 -; RUN: grep {vqsub\\.s16} %t | count 2 -; RUN: grep {vqsub\\.s32} %t | count 2 -; RUN: grep {vqsub\\.s64} %t | count 2 -; RUN: grep {vqsub\\.u8} %t | count 2 -; RUN: grep {vqsub\\.u16} %t | count 2 -; RUN: grep {vqsub\\.u32} %t | count 2 -; RUN: grep {vqsub\\.u64} %t | count 2 +; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s define <8 x i8> @vqsubs8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqsubs8: +;CHECK: vqsub.s8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqsubs.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -16,6 +10,8 @@ define <8 x i8> @vqsubs8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqsubs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqsubs16: +;CHECK: vqsub.s16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -23,6 +19,8 @@ define <4 x i16> @vqsubs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqsubs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqsubs32: +;CHECK: vqsub.s32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -30,6 +28,8 @@ define <2 x i32> @vqsubs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqsubs64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqsubs64: +;CHECK: vqsub.s64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -37,6 +37,8 @@ define <1 x i64> @vqsubs64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <8 x i8> @vqsubu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { +;CHECK: vqsubu8: +;CHECK: vqsub.u8 %tmp1 = load <8 x i8>* %A %tmp2 = load <8 x i8>* %B %tmp3 = call <8 x i8> @llvm.arm.neon.vqsubu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) @@ -44,6 +46,8 @@ define <8 x i8> @vqsubu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { } define <4 x i16> @vqsubu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { +;CHECK: vqsubu16: +;CHECK: vqsub.u16 %tmp1 = load <4 x i16>* %A %tmp2 = load <4 x i16>* %B %tmp3 = call <4 x i16> @llvm.arm.neon.vqsubu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) @@ -51,6 +55,8 @@ define <4 x i16> @vqsubu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { } define <2 x i32> @vqsubu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { +;CHECK: vqsubu32: +;CHECK: vqsub.u32 %tmp1 = load <2 x i32>* %A %tmp2 = load <2 x i32>* %B %tmp3 = call <2 x i32> @llvm.arm.neon.vqsubu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) @@ -58,6 +64,8 @@ define <2 x i32> @vqsubu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { } define <1 x i64> @vqsubu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { +;CHECK: vqsubu64: +;CHECK: vqsub.u64 %tmp1 = load <1 x i64>* %A %tmp2 = load <1 x i64>* %B %tmp3 = call <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2) @@ -65,6 +73,8 @@ define <1 x i64> @vqsubu64(<1 x i64>* %A, <1 x i64>* %B) nounwind { } define <16 x i8> @vqsubQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqsubQs8: +;CHECK: vqsub.s8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqsubs.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -72,6 +82,8 @@ define <16 x i8> @vqsubQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqsubQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqsubQs16: +;CHECK: vqsub.s16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -79,6 +91,8 @@ define <8 x i16> @vqsubQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqsubQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqsubQs32: +;CHECK: vqsub.s32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -86,6 +100,8 @@ define <4 x i32> @vqsubQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqsubQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqsubQs64: +;CHECK: vqsub.s64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqsubs.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2) @@ -93,6 +109,8 @@ define <2 x i64> @vqsubQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { } define <16 x i8> @vqsubQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { +;CHECK: vqsubQu8: +;CHECK: vqsub.u8 %tmp1 = load <16 x i8>* %A %tmp2 = load <16 x i8>* %B %tmp3 = call <16 x i8> @llvm.arm.neon.vqsubu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) @@ -100,6 +118,8 @@ define <16 x i8> @vqsubQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { } define <8 x i16> @vqsubQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { +;CHECK: vqsubQu16: +;CHECK: vqsub.u16 %tmp1 = load <8 x i16>* %A %tmp2 = load <8 x i16>* %B %tmp3 = call <8 x i16> @llvm.arm.neon.vqsubu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) @@ -107,6 +127,8 @@ define <8 x i16> @vqsubQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { } define <4 x i32> @vqsubQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { +;CHECK: vqsubQu32: +;CHECK: vqsub.u32 %tmp1 = load <4 x i32>* %A %tmp2 = load <4 x i32>* %B %tmp3 = call <4 x i32> @llvm.arm.neon.vqsubu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) @@ -114,6 +136,8 @@ define <4 x i32> @vqsubQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { } define <2 x i64> @vqsubQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { +;CHECK: vqsubQu64: +;CHECK: vqsub.u64 %tmp1 = load <2 x i64>* %A %tmp2 = load <2 x i64>* %B %tmp3 = call <2 x i64> @llvm.arm.neon.vqsubu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)