aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--llvm/test/CodeGen/LoongArch/lasx/and-not-combine.ll183
-rw-r--r--llvm/test/CodeGen/LoongArch/lsx/and-not-combine.ll183
2 files changed, 362 insertions, 4 deletions
diff --git a/llvm/test/CodeGen/LoongArch/lasx/and-not-combine.ll b/llvm/test/CodeGen/LoongArch/lasx/and-not-combine.ll
index 6754959..225afecc 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/and-not-combine.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/and-not-combine.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
-; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s
-; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s --check-prefixes=CHECK,LA32
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s --check-prefixes=CHECK,LA64
define void @and_not_combine_v32i8(ptr %res, ptr %a0, ptr %a1, ptr %a2) nounwind {
; CHECK-LABEL: and_not_combine_v32i8:
@@ -85,3 +85,182 @@ entry:
store <4 x i64> %and, ptr %res
ret void
}
+
+define void @pre_not_and_not_combine_v32i8(ptr %res, ptr %a, i8 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v32i8:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: xvreplgr2vr.b $xr1, $a1
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <32 x i8>, ptr %a
+ %b.not = xor i8 %b, -1
+ %b.not.ele = insertelement <32 x i8> poison, i8 %b.not, i64 0
+ %v1.not = shufflevector <32 x i8> %b.not.ele, <32 x i8> poison, <32 x i32> zeroinitializer
+ %v0.not = xor <32 x i8> %v0, splat (i8 -1)
+ %and = and <32 x i8> %v0.not, %v1.not
+ store <32 x i8> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v32i8(ptr %res, ptr %a, i8 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v32i8:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: xvreplgr2vr.b $xr1, $a2
+; CHECK-NEXT: xvxori.b $xr1, $xr1, 255
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <32 x i8>, ptr %a
+ %b.ele = insertelement <32 x i8> poison, i8 %b, i64 0
+ %v1 = shufflevector <32 x i8> %b.ele, <32 x i8> poison, <32 x i32> zeroinitializer
+ %v0.not = xor <32 x i8> %v0, splat (i8 -1)
+ %v1.not = xor <32 x i8> %v1, splat (i8 -1)
+ %and = and <32 x i8> %v0.not, %v1.not
+ store <32 x i8> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v16i16(ptr %res, ptr %a, i16 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v16i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: xvreplgr2vr.h $xr1, $a1
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <16 x i16>, ptr %a
+ %b.not = xor i16 %b, -1
+ %b.not.ele = insertelement <16 x i16> poison, i16 %b.not, i64 0
+ %v1.not = shufflevector <16 x i16> %b.not.ele, <16 x i16> poison, <16 x i32> zeroinitializer
+ %v0.not = xor <16 x i16> %v0, splat (i16 -1)
+ %and = and <16 x i16> %v0.not, %v1.not
+ store <16 x i16> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v16i16(ptr %res, ptr %a, i16 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v16i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: xvreplgr2vr.h $xr1, $a2
+; CHECK-NEXT: xvrepli.b $xr2, -1
+; CHECK-NEXT: xvxor.v $xr1, $xr1, $xr2
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <16 x i16>, ptr %a
+ %b.ele = insertelement <16 x i16> poison, i16 %b, i64 0
+ %v1 = shufflevector <16 x i16> %b.ele, <16 x i16> poison, <16 x i32> zeroinitializer
+ %v0.not = xor <16 x i16> %v0, splat (i16 -1)
+ %v1.not = xor <16 x i16> %v1, splat (i16 -1)
+ %and = and <16 x i16> %v0.not, %v1.not
+ store <16 x i16> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v8i32(ptr %res, ptr %a, i32 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v8i32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: xvreplgr2vr.w $xr1, $a1
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <8 x i32>, ptr %a
+ %b.not = xor i32 %b, -1
+ %b.not.ele = insertelement <8 x i32> poison, i32 %b.not, i64 0
+ %v1.not = shufflevector <8 x i32> %b.not.ele, <8 x i32> poison, <8 x i32> zeroinitializer
+ %v0.not = xor <8 x i32> %v0, splat (i32 -1)
+ %and = and <8 x i32> %v0.not, %v1.not
+ store <8 x i32> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v8i32(ptr %res, ptr %a, i32 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v8i32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xvld $xr0, $a1, 0
+; CHECK-NEXT: xvreplgr2vr.w $xr1, $a2
+; CHECK-NEXT: xvrepli.b $xr2, -1
+; CHECK-NEXT: xvxor.v $xr1, $xr1, $xr2
+; CHECK-NEXT: xvandn.v $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <8 x i32>, ptr %a
+ %b.ele = insertelement <8 x i32> poison, i32 %b, i64 0
+ %v1 = shufflevector <8 x i32> %b.ele, <8 x i32> poison, <8 x i32> zeroinitializer
+ %v0.not = xor <8 x i32> %v0, splat (i32 -1)
+ %v1.not = xor <8 x i32> %v1, splat (i32 -1)
+ %and = and <8 x i32> %v0.not, %v1.not
+ store <8 x i32> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v4i64(ptr %res, ptr %a, i64 %b) nounwind {
+; LA32-LABEL: pre_not_and_not_combine_v4i64:
+; LA32: # %bb.0:
+; LA32-NEXT: xvld $xr0, $a1, 0
+; LA32-NEXT: nor $a1, $a3, $zero
+; LA32-NEXT: nor $a2, $a2, $zero
+; LA32-NEXT: vinsgr2vr.w $vr1, $a2, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a1, 1
+; LA32-NEXT: xvreplve0.d $xr1, $xr1
+; LA32-NEXT: xvandn.v $xr0, $xr0, $xr1
+; LA32-NEXT: xvst $xr0, $a0, 0
+; LA32-NEXT: ret
+;
+; LA64-LABEL: pre_not_and_not_combine_v4i64:
+; LA64: # %bb.0:
+; LA64-NEXT: xvld $xr0, $a1, 0
+; LA64-NEXT: nor $a1, $a2, $zero
+; LA64-NEXT: xvreplgr2vr.d $xr1, $a1
+; LA64-NEXT: xvandn.v $xr0, $xr0, $xr1
+; LA64-NEXT: xvst $xr0, $a0, 0
+; LA64-NEXT: ret
+ %v0 = load <4 x i64>, ptr %a
+ %b.not = xor i64 %b, -1
+ %b.not.ele = insertelement <4 x i64> poison, i64 %b.not, i64 0
+ %v1.not = shufflevector <4 x i64> %b.not.ele, <4 x i64> poison, <4 x i32> zeroinitializer
+ %v0.not = xor <4 x i64> %v0, splat (i64 -1)
+ %and = and <4 x i64> %v0.not, %v1.not
+ store <4 x i64> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v4i64(ptr %res, ptr %a, i64 %b) nounwind {
+; LA32-LABEL: post_not_and_not_combine_v4i64:
+; LA32: # %bb.0:
+; LA32-NEXT: xvld $xr0, $a1, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a2, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a3, 1
+; LA32-NEXT: xvreplve0.d $xr1, $xr1
+; LA32-NEXT: xvrepli.b $xr2, -1
+; LA32-NEXT: xvxor.v $xr1, $xr1, $xr2
+; LA32-NEXT: xvandn.v $xr0, $xr0, $xr1
+; LA32-NEXT: xvst $xr0, $a0, 0
+; LA32-NEXT: ret
+;
+; LA64-LABEL: post_not_and_not_combine_v4i64:
+; LA64: # %bb.0:
+; LA64-NEXT: xvld $xr0, $a1, 0
+; LA64-NEXT: xvreplgr2vr.d $xr1, $a2
+; LA64-NEXT: xvrepli.b $xr2, -1
+; LA64-NEXT: xvxor.v $xr1, $xr1, $xr2
+; LA64-NEXT: xvandn.v $xr0, $xr0, $xr1
+; LA64-NEXT: xvst $xr0, $a0, 0
+; LA64-NEXT: ret
+ %v0 = load <4 x i64>, ptr %a
+ %b.ele = insertelement <4 x i64> poison, i64 %b, i64 0
+ %v1 = shufflevector <4 x i64> %b.ele, <4 x i64> poison, <4 x i32> zeroinitializer
+ %v0.not = xor <4 x i64> %v0, splat (i64 -1)
+ %v1.not = xor <4 x i64> %v1, splat (i64 -1)
+ %and = and <4 x i64> %v0.not, %v1.not
+ store <4 x i64> %and, ptr %res
+ ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lsx/and-not-combine.ll b/llvm/test/CodeGen/LoongArch/lsx/and-not-combine.ll
index 3c6d345..098ee3a 100644
--- a/llvm/test/CodeGen/LoongArch/lsx/and-not-combine.ll
+++ b/llvm/test/CodeGen/LoongArch/lsx/and-not-combine.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
-; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s
-; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA32
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA64
define void @and_not_combine_v16i8(ptr %res, ptr %a0, ptr %a1, ptr %a2) nounwind {
; CHECK-LABEL: and_not_combine_v16i8:
@@ -85,3 +85,182 @@ entry:
store <2 x i64> %and, ptr %res
ret void
}
+
+define void @pre_not_and_not_combine_v16i8(ptr %res, ptr %a, i8 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v16i8:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: vreplgr2vr.b $vr1, $a1
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <16 x i8>, ptr %a
+ %b.not = xor i8 %b, -1
+ %b.not.ele = insertelement <16 x i8> poison, i8 %b.not, i64 0
+ %v1.not = shufflevector <16 x i8> %b.not.ele, <16 x i8> poison, <16 x i32> zeroinitializer
+ %v0.not = xor <16 x i8> %v0, splat (i8 -1)
+ %and = and <16 x i8> %v0.not, %v1.not
+ store <16 x i8> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v16i8(ptr %res, ptr %a, i8 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v16i8:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: vreplgr2vr.b $vr1, $a2
+; CHECK-NEXT: vxori.b $vr1, $vr1, 255
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <16 x i8>, ptr %a
+ %b.ele = insertelement <16 x i8> poison, i8 %b, i64 0
+ %v1 = shufflevector <16 x i8> %b.ele, <16 x i8> poison, <16 x i32> zeroinitializer
+ %v0.not = xor <16 x i8> %v0, splat (i8 -1)
+ %v1.not = xor <16 x i8> %v1, splat (i8 -1)
+ %and = and <16 x i8> %v0.not, %v1.not
+ store <16 x i8> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v8i16(ptr %res, ptr %a, i16 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: vreplgr2vr.h $vr1, $a1
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <8 x i16>, ptr %a
+ %b.not = xor i16 %b, -1
+ %b.not.ele = insertelement <8 x i16> poison, i16 %b.not, i64 0
+ %v1.not = shufflevector <8 x i16> %b.not.ele, <8 x i16> poison, <8 x i32> zeroinitializer
+ %v0.not = xor <8 x i16> %v0, splat (i16 -1)
+ %and = and <8 x i16> %v0.not, %v1.not
+ store <8 x i16> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v8i16(ptr %res, ptr %a, i16 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: vreplgr2vr.h $vr1, $a2
+; CHECK-NEXT: vrepli.b $vr2, -1
+; CHECK-NEXT: vxor.v $vr1, $vr1, $vr2
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <8 x i16>, ptr %a
+ %b.ele = insertelement <8 x i16> poison, i16 %b, i64 0
+ %v1 = shufflevector <8 x i16> %b.ele, <8 x i16> poison, <8 x i32> zeroinitializer
+ %v0.not = xor <8 x i16> %v0, splat (i16 -1)
+ %v1.not = xor <8 x i16> %v1, splat (i16 -1)
+ %and = and <8 x i16> %v0.not, %v1.not
+ store <8 x i16> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v4i32(ptr %res, ptr %a, i32 %b) nounwind {
+; CHECK-LABEL: pre_not_and_not_combine_v4i32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: nor $a1, $a2, $zero
+; CHECK-NEXT: vreplgr2vr.w $vr1, $a1
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <4 x i32>, ptr %a
+ %b.not = xor i32 %b, -1
+ %b.not.ele = insertelement <4 x i32> poison, i32 %b.not, i64 0
+ %v1.not = shufflevector <4 x i32> %b.not.ele, <4 x i32> poison, <4 x i32> zeroinitializer
+ %v0.not = xor <4 x i32> %v0, splat (i32 -1)
+ %and = and <4 x i32> %v0.not, %v1.not
+ store <4 x i32> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v4i32(ptr %res, ptr %a, i32 %b) nounwind {
+; CHECK-LABEL: post_not_and_not_combine_v4i32:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: vreplgr2vr.w $vr1, $a2
+; CHECK-NEXT: vrepli.b $vr2, -1
+; CHECK-NEXT: vxor.v $vr1, $vr1, $vr2
+; CHECK-NEXT: vandn.v $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: ret
+ %v0 = load <4 x i32>, ptr %a
+ %b.ele = insertelement <4 x i32> poison, i32 %b, i64 0
+ %v1 = shufflevector <4 x i32> %b.ele, <4 x i32> poison, <4 x i32> zeroinitializer
+ %v0.not = xor <4 x i32> %v0, splat (i32 -1)
+ %v1.not = xor <4 x i32> %v1, splat (i32 -1)
+ %and = and <4 x i32> %v0.not, %v1.not
+ store <4 x i32> %and, ptr %res
+ ret void
+}
+
+define void @pre_not_and_not_combine_v2i64(ptr %res, ptr %a, i64 %b) nounwind {
+; LA32-LABEL: pre_not_and_not_combine_v2i64:
+; LA32: # %bb.0:
+; LA32-NEXT: vld $vr0, $a1, 0
+; LA32-NEXT: nor $a1, $a3, $zero
+; LA32-NEXT: nor $a2, $a2, $zero
+; LA32-NEXT: vinsgr2vr.w $vr1, $a2, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a1, 1
+; LA32-NEXT: vreplvei.d $vr1, $vr1, 0
+; LA32-NEXT: vandn.v $vr0, $vr0, $vr1
+; LA32-NEXT: vst $vr0, $a0, 0
+; LA32-NEXT: ret
+;
+; LA64-LABEL: pre_not_and_not_combine_v2i64:
+; LA64: # %bb.0:
+; LA64-NEXT: vld $vr0, $a1, 0
+; LA64-NEXT: nor $a1, $a2, $zero
+; LA64-NEXT: vreplgr2vr.d $vr1, $a1
+; LA64-NEXT: vandn.v $vr0, $vr0, $vr1
+; LA64-NEXT: vst $vr0, $a0, 0
+; LA64-NEXT: ret
+ %v0 = load <2 x i64>, ptr %a
+ %b.not = xor i64 %b, -1
+ %b.not.ele = insertelement <2 x i64> poison, i64 %b.not, i64 0
+ %v1.not = shufflevector <2 x i64> %b.not.ele, <2 x i64> poison, <2 x i32> zeroinitializer
+ %v0.not = xor <2 x i64> %v0, splat (i64 -1)
+ %and = and <2 x i64> %v0.not, %v1.not
+ store <2 x i64> %and, ptr %res
+ ret void
+}
+
+define void @post_not_and_not_combine_v2i64(ptr %res, ptr %a, i64 %b) nounwind {
+; LA32-LABEL: post_not_and_not_combine_v2i64:
+; LA32: # %bb.0:
+; LA32-NEXT: vld $vr0, $a1, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a2, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a3, 1
+; LA32-NEXT: vreplvei.d $vr1, $vr1, 0
+; LA32-NEXT: vrepli.b $vr2, -1
+; LA32-NEXT: vxor.v $vr1, $vr1, $vr2
+; LA32-NEXT: vandn.v $vr0, $vr0, $vr1
+; LA32-NEXT: vst $vr0, $a0, 0
+; LA32-NEXT: ret
+;
+; LA64-LABEL: post_not_and_not_combine_v2i64:
+; LA64: # %bb.0:
+; LA64-NEXT: vld $vr0, $a1, 0
+; LA64-NEXT: vreplgr2vr.d $vr1, $a2
+; LA64-NEXT: vrepli.b $vr2, -1
+; LA64-NEXT: vxor.v $vr1, $vr1, $vr2
+; LA64-NEXT: vandn.v $vr0, $vr0, $vr1
+; LA64-NEXT: vst $vr0, $a0, 0
+; LA64-NEXT: ret
+ %v0 = load <2 x i64>, ptr %a
+ %b.ele = insertelement <2 x i64> poison, i64 %b, i64 0
+ %v1 = shufflevector <2 x i64> %b.ele, <2 x i64> poison, <2 x i32> zeroinitializer
+ %v0.not = xor <2 x i64> %v0, splat (i64 -1)
+ %v1.not = xor <2 x i64> %v1, splat (i64 -1)
+ %and = and <2 x i64> %v0.not, %v1.not
+ store <2 x i64> %and, ptr %res
+ ret void
+}
generated by cgit v1.1 at 2025-09-27 10:58:13 +0000