summaryrefslogtreecommitdiff
path: root/test/CodeGen/ARM/vsub.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/ARM/vsub.ll')
-rw-r--r--test/CodeGen/ARM/vsub.ll96
1 files changed, 46 insertions, 50 deletions
diff --git a/test/CodeGen/ARM/vsub.ll b/test/CodeGen/ARM/vsub.ll
index df77bb31fc8b..6b95b97378e0 100644
--- a/test/CodeGen/ARM/vsub.ll
+++ b/test/CodeGen/ARM/vsub.ll
@@ -1,7 +1,7 @@
; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
define <8 x i8> @vsubi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
-;CHECK: vsubi8:
+;CHECK-LABEL: vsubi8:
;CHECK: vsub.i8
%tmp1 = load <8 x i8>* %A
%tmp2 = load <8 x i8>* %B
@@ -10,7 +10,7 @@ define <8 x i8> @vsubi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
}
define <4 x i16> @vsubi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
-;CHECK: vsubi16:
+;CHECK-LABEL: vsubi16:
;CHECK: vsub.i16
%tmp1 = load <4 x i16>* %A
%tmp2 = load <4 x i16>* %B
@@ -19,7 +19,7 @@ define <4 x i16> @vsubi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
}
define <2 x i32> @vsubi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
-;CHECK: vsubi32:
+;CHECK-LABEL: vsubi32:
;CHECK: vsub.i32
%tmp1 = load <2 x i32>* %A
%tmp2 = load <2 x i32>* %B
@@ -28,7 +28,7 @@ define <2 x i32> @vsubi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
}
define <1 x i64> @vsubi64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
-;CHECK: vsubi64:
+;CHECK-LABEL: vsubi64:
;CHECK: vsub.i64
%tmp1 = load <1 x i64>* %A
%tmp2 = load <1 x i64>* %B
@@ -37,7 +37,7 @@ define <1 x i64> @vsubi64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
}
define <2 x float> @vsubf32(<2 x float>* %A, <2 x float>* %B) nounwind {
-;CHECK: vsubf32:
+;CHECK-LABEL: vsubf32:
;CHECK: vsub.f32
%tmp1 = load <2 x float>* %A
%tmp2 = load <2 x float>* %B
@@ -46,7 +46,7 @@ define <2 x float> @vsubf32(<2 x float>* %A, <2 x float>* %B) nounwind {
}
define <16 x i8> @vsubQi8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
-;CHECK: vsubQi8:
+;CHECK-LABEL: vsubQi8:
;CHECK: vsub.i8
%tmp1 = load <16 x i8>* %A
%tmp2 = load <16 x i8>* %B
@@ -55,7 +55,7 @@ define <16 x i8> @vsubQi8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
}
define <8 x i16> @vsubQi16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
-;CHECK: vsubQi16:
+;CHECK-LABEL: vsubQi16:
;CHECK: vsub.i16
%tmp1 = load <8 x i16>* %A
%tmp2 = load <8 x i16>* %B
@@ -64,7 +64,7 @@ define <8 x i16> @vsubQi16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
}
define <4 x i32> @vsubQi32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
-;CHECK: vsubQi32:
+;CHECK-LABEL: vsubQi32:
;CHECK: vsub.i32
%tmp1 = load <4 x i32>* %A
%tmp2 = load <4 x i32>* %B
@@ -73,7 +73,7 @@ define <4 x i32> @vsubQi32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
}
define <2 x i64> @vsubQi64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
-;CHECK: vsubQi64:
+;CHECK-LABEL: vsubQi64:
;CHECK: vsub.i64
%tmp1 = load <2 x i64>* %A
%tmp2 = load <2 x i64>* %B
@@ -82,7 +82,7 @@ define <2 x i64> @vsubQi64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
}
define <4 x float> @vsubQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
-;CHECK: vsubQf32:
+;CHECK-LABEL: vsubQf32:
;CHECK: vsub.f32
%tmp1 = load <4 x float>* %A
%tmp2 = load <4 x float>* %B
@@ -90,39 +90,35 @@ define <4 x float> @vsubQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
ret <4 x float> %tmp3
}
-define <8 x i8> @vsubhni16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
-;CHECK: vsubhni16:
-;CHECK: vsubhn.i16
- %tmp1 = load <8 x i16>* %A
- %tmp2 = load <8 x i16>* %B
- %tmp3 = call <8 x i8> @llvm.arm.neon.vsubhn.v8i8(<8 x i16> %tmp1, <8 x i16> %tmp2)
- ret <8 x i8> %tmp3
+define <8 x i8> @vsubhni16_natural(<8 x i16> %A, <8 x i16> %B) nounwind {
+; CHECK-LABEL: vsubhni16_natural:
+; CHECK: vsubhn.i16
+ %sum = sub <8 x i16> %A, %B
+ %shift = lshr <8 x i16> %sum, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+ %trunc = trunc <8 x i16> %shift to <8 x i8>
+ ret <8 x i8> %trunc
}
-define <4 x i16> @vsubhni32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
-;CHECK: vsubhni32:
-;CHECK: vsubhn.i32
- %tmp1 = load <4 x i32>* %A
- %tmp2 = load <4 x i32>* %B
- %tmp3 = call <4 x i16> @llvm.arm.neon.vsubhn.v4i16(<4 x i32> %tmp1, <4 x i32> %tmp2)
- ret <4 x i16> %tmp3
+define <4 x i16> @vsubhni32_natural(<4 x i32> %A, <4 x i32> %B) nounwind {
+; CHECK-LABEL: vsubhni32_natural:
+; CHECK: vsubhn.i32
+ %sum = sub <4 x i32> %A, %B
+ %shift = lshr <4 x i32> %sum, <i32 16, i32 16, i32 16, i32 16>
+ %trunc = trunc <4 x i32> %shift to <4 x i16>
+ ret <4 x i16> %trunc
}
-define <2 x i32> @vsubhni64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
-;CHECK: vsubhni64:
-;CHECK: vsubhn.i64
- %tmp1 = load <2 x i64>* %A
- %tmp2 = load <2 x i64>* %B
- %tmp3 = call <2 x i32> @llvm.arm.neon.vsubhn.v2i32(<2 x i64> %tmp1, <2 x i64> %tmp2)
- ret <2 x i32> %tmp3
+define <2 x i32> @vsubhni64_natural(<2 x i64> %A, <2 x i64> %B) nounwind {
+; CHECK-LABEL: vsubhni64_natural:
+; CHECK: vsubhn.i64
+ %sum = sub <2 x i64> %A, %B
+ %shift = lshr <2 x i64> %sum, <i64 32, i64 32>
+ %trunc = trunc <2 x i64> %shift to <2 x i32>
+ ret <2 x i32> %trunc
}
-declare <8 x i8> @llvm.arm.neon.vsubhn.v8i8(<8 x i16>, <8 x i16>) nounwind readnone
-declare <4 x i16> @llvm.arm.neon.vsubhn.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
-declare <2 x i32> @llvm.arm.neon.vsubhn.v2i32(<2 x i64>, <2 x i64>) nounwind readnone
-
define <8 x i8> @vrsubhni16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
-;CHECK: vrsubhni16:
+;CHECK-LABEL: vrsubhni16:
;CHECK: vrsubhn.i16
%tmp1 = load <8 x i16>* %A
%tmp2 = load <8 x i16>* %B
@@ -131,7 +127,7 @@ define <8 x i8> @vrsubhni16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
}
define <4 x i16> @vrsubhni32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
-;CHECK: vrsubhni32:
+;CHECK-LABEL: vrsubhni32:
;CHECK: vrsubhn.i32
%tmp1 = load <4 x i32>* %A
%tmp2 = load <4 x i32>* %B
@@ -140,7 +136,7 @@ define <4 x i16> @vrsubhni32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
}
define <2 x i32> @vrsubhni64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
-;CHECK: vrsubhni64:
+;CHECK-LABEL: vrsubhni64:
;CHECK: vrsubhn.i64
%tmp1 = load <2 x i64>* %A
%tmp2 = load <2 x i64>* %B
@@ -153,7 +149,7 @@ declare <4 x i16> @llvm.arm.neon.vrsubhn.v4i16(<4 x i32>, <4 x i32>) nounwind re
declare <2 x i32> @llvm.arm.neon.vrsubhn.v2i32(<2 x i64>, <2 x i64>) nounwind readnone
define <8 x i16> @vsubls8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
-;CHECK: vsubls8:
+;CHECK-LABEL: vsubls8:
;CHECK: vsubl.s8
%tmp1 = load <8 x i8>* %A
%tmp2 = load <8 x i8>* %B
@@ -164,7 +160,7 @@ define <8 x i16> @vsubls8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
}
define <4 x i32> @vsubls16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
-;CHECK: vsubls16:
+;CHECK-LABEL: vsubls16:
;CHECK: vsubl.s16
%tmp1 = load <4 x i16>* %A
%tmp2 = load <4 x i16>* %B
@@ -175,7 +171,7 @@ define <4 x i32> @vsubls16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
}
define <2 x i64> @vsubls32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
-;CHECK: vsubls32:
+;CHECK-LABEL: vsubls32:
;CHECK: vsubl.s32
%tmp1 = load <2 x i32>* %A
%tmp2 = load <2 x i32>* %B
@@ -186,7 +182,7 @@ define <2 x i64> @vsubls32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
}
define <8 x i16> @vsublu8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
-;CHECK: vsublu8:
+;CHECK-LABEL: vsublu8:
;CHECK: vsubl.u8
%tmp1 = load <8 x i8>* %A
%tmp2 = load <8 x i8>* %B
@@ -197,7 +193,7 @@ define <8 x i16> @vsublu8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
}
define <4 x i32> @vsublu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
-;CHECK: vsublu16:
+;CHECK-LABEL: vsublu16:
;CHECK: vsubl.u16
%tmp1 = load <4 x i16>* %A
%tmp2 = load <4 x i16>* %B
@@ -208,7 +204,7 @@ define <4 x i32> @vsublu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
}
define <2 x i64> @vsublu32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
-;CHECK: vsublu32:
+;CHECK-LABEL: vsublu32:
;CHECK: vsubl.u32
%tmp1 = load <2 x i32>* %A
%tmp2 = load <2 x i32>* %B
@@ -219,7 +215,7 @@ define <2 x i64> @vsublu32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
}
define <8 x i16> @vsubws8(<8 x i16>* %A, <8 x i8>* %B) nounwind {
-;CHECK: vsubws8:
+;CHECK-LABEL: vsubws8:
;CHECK: vsubw.s8
%tmp1 = load <8 x i16>* %A
%tmp2 = load <8 x i8>* %B
@@ -229,7 +225,7 @@ define <8 x i16> @vsubws8(<8 x i16>* %A, <8 x i8>* %B) nounwind {
}
define <4 x i32> @vsubws16(<4 x i32>* %A, <4 x i16>* %B) nounwind {
-;CHECK: vsubws16:
+;CHECK-LABEL: vsubws16:
;CHECK: vsubw.s16
%tmp1 = load <4 x i32>* %A
%tmp2 = load <4 x i16>* %B
@@ -239,7 +235,7 @@ define <4 x i32> @vsubws16(<4 x i32>* %A, <4 x i16>* %B) nounwind {
}
define <2 x i64> @vsubws32(<2 x i64>* %A, <2 x i32>* %B) nounwind {
-;CHECK: vsubws32:
+;CHECK-LABEL: vsubws32:
;CHECK: vsubw.s32
%tmp1 = load <2 x i64>* %A
%tmp2 = load <2 x i32>* %B
@@ -249,7 +245,7 @@ define <2 x i64> @vsubws32(<2 x i64>* %A, <2 x i32>* %B) nounwind {
}
define <8 x i16> @vsubwu8(<8 x i16>* %A, <8 x i8>* %B) nounwind {
-;CHECK: vsubwu8:
+;CHECK-LABEL: vsubwu8:
;CHECK: vsubw.u8
%tmp1 = load <8 x i16>* %A
%tmp2 = load <8 x i8>* %B
@@ -259,7 +255,7 @@ define <8 x i16> @vsubwu8(<8 x i16>* %A, <8 x i8>* %B) nounwind {
}
define <4 x i32> @vsubwu16(<4 x i32>* %A, <4 x i16>* %B) nounwind {
-;CHECK: vsubwu16:
+;CHECK-LABEL: vsubwu16:
;CHECK: vsubw.u16
%tmp1 = load <4 x i32>* %A
%tmp2 = load <4 x i16>* %B
@@ -269,7 +265,7 @@ define <4 x i32> @vsubwu16(<4 x i32>* %A, <4 x i16>* %B) nounwind {
}
define <2 x i64> @vsubwu32(<2 x i64>* %A, <2 x i32>* %B) nounwind {
-;CHECK: vsubwu32:
+;CHECK-LABEL: vsubwu32:
;CHECK: vsubw.u32
%tmp1 = load <2 x i64>* %A
%tmp2 = load <2 x i32>* %B
generated by cgit v1.3 (git 2.53.0) at 2026-07-28 00:15:01 +0000