[llvm] 2f97ff8 - [SLP] Add additional memory versioning tests.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Sep 16 05:31:34 PDT 2021
Author: Florian Hahn
Date: 2021-09-16T13:31:14+01:00
New Revision: 2f97ff8e7bcf35595682b53fce8ecb7b4aabee6d
URL: https://github.com/llvm/llvm-project/commit/2f97ff8e7bcf35595682b53fce8ecb7b4aabee6d
DIFF: https://github.com/llvm/llvm-project/commit/2f97ff8e7bcf35595682b53fce8ecb7b4aabee6d.diff
LOG: [SLP] Add additional memory versioning tests.
Added:
llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks-in-loops.ll
Modified:
llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll
Removed:
################################################################################
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks-in-loops.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks-in-loops.ll
new file mode 100644
index 000000000000..6beb1103ceaa
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks-in-loops.ll
@@ -0,0 +1,176 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -slp-vectorizer -mtriple=arm64-apple-ios -S %s | FileCheck %s
+
+define void @loop1(i32* %A, i32* %B, i64 %N) {
+; CHECK-LABEL: @loop1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT: [[B_GEP_0:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[IV]]
+; CHECK-NEXT: [[B_0:%.*]] = load i32, i32* [[B_GEP_0]], align 4
+; CHECK-NEXT: [[A_GEP_0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[IV]]
+; CHECK-NEXT: [[A_0:%.*]] = load i32, i32* [[A_GEP_0]], align 4
+; CHECK-NEXT: [[ADD_0:%.*]] = add i32 [[A_0]], 20
+; CHECK-NEXT: [[XOR_0:%.*]] = xor i32 [[ADD_0]], [[B_0]]
+; CHECK-NEXT: store i32 [[XOR_0]], i32* [[A_GEP_0]], align 4
+; CHECK-NEXT: [[IV_1:%.*]] = or i64 [[IV]], 1
+; CHECK-NEXT: [[B_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_1]]
+; CHECK-NEXT: [[B_1:%.*]] = load i32, i32* [[B_GEP_1]], align 4
+; CHECK-NEXT: [[A_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_1]]
+; CHECK-NEXT: [[A_1:%.*]] = load i32, i32* [[A_GEP_1]], align 4
+; CHECK-NEXT: [[ADD_1:%.*]] = add i32 [[A_1]], 20
+; CHECK-NEXT: [[XOR_1:%.*]] = xor i32 [[ADD_1]], [[B_1]]
+; CHECK-NEXT: store i32 [[XOR_1]], i32* [[A_GEP_1]], align 4
+; CHECK-NEXT: [[IV_2:%.*]] = or i64 [[IV]], 2
+; CHECK-NEXT: [[B_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_2]]
+; CHECK-NEXT: [[B_2:%.*]] = load i32, i32* [[B_GEP_2]], align 4
+; CHECK-NEXT: [[A_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_2]]
+; CHECK-NEXT: [[A_2:%.*]] = load i32, i32* [[A_GEP_2]], align 4
+; CHECK-NEXT: [[ADD_2:%.*]] = add i32 [[A_2]], 20
+; CHECK-NEXT: [[XOR_2:%.*]] = xor i32 [[ADD_2]], [[B_2]]
+; CHECK-NEXT: store i32 [[XOR_2]], i32* [[A_GEP_2]], align 4
+; CHECK-NEXT: [[IV_3:%.*]] = or i64 [[IV]], 3
+; CHECK-NEXT: [[B_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_3]]
+; CHECK-NEXT: [[B_3:%.*]] = load i32, i32* [[B_GEP_3]], align 4
+; CHECK-NEXT: [[A_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_3]]
+; CHECK-NEXT: [[A_3:%.*]] = load i32, i32* [[A_GEP_3]], align 4
+; CHECK-NEXT: [[ADD_3:%.*]] = add i32 [[A_3]], 20
+; CHECK-NEXT: [[XOR_3:%.*]] = xor i32 [[ADD_3]], [[B_3]]
+; CHECK-NEXT: store i32 [[XOR_3]], i32* [[A_GEP_3]], align 4
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 16
+; CHECK-NEXT: [[COND:%.*]] = icmp ult i64 [[IV_NEXT]], [[N:%.*]]
+; CHECK-NEXT: br i1 [[COND]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK: exit:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+ %B.gep.0 = getelementptr inbounds i32, i32* %B, i64 %iv
+ %B.0 = load i32, i32* %B.gep.0, align 4
+ %A.gep.0 = getelementptr inbounds i32, i32* %A, i64 %iv
+ %A.0 = load i32, i32* %A.gep.0, align 4
+ %add.0 = add i32 %A.0, 20
+ %xor.0 = xor i32 %add.0, %B.0
+ store i32 %xor.0, i32* %A.gep.0, align 4
+ %iv.1 = or i64 %iv, 1
+ %B.gep.1 = getelementptr inbounds i32, i32* %B, i64 %iv.1
+ %B.1 = load i32, i32* %B.gep.1, align 4
+ %A.gep.1 = getelementptr inbounds i32, i32* %A, i64 %iv.1
+ %A.1 = load i32, i32* %A.gep.1, align 4
+ %add.1 = add i32 %A.1, 20
+ %xor.1 = xor i32 %add.1, %B.1
+ store i32 %xor.1, i32* %A.gep.1, align 4
+ %iv.2 = or i64 %iv, 2
+ %B.gep.2 = getelementptr inbounds i32, i32* %B, i64 %iv.2
+ %B.2 = load i32, i32* %B.gep.2, align 4
+ %A.gep.2 = getelementptr inbounds i32, i32* %A, i64 %iv.2
+ %A.2 = load i32, i32* %A.gep.2, align 4
+ %add.2 = add i32 %A.2, 20
+ %xor.2 = xor i32 %add.2, %B.2
+ store i32 %xor.2, i32* %A.gep.2, align 4
+ %iv.3 = or i64 %iv, 3
+ %B.gep.3 = getelementptr inbounds i32, i32* %B, i64 %iv.3
+ %B.3 = load i32, i32* %B.gep.3, align 4
+ %A.gep.3 = getelementptr inbounds i32, i32* %A, i64 %iv.3
+ %A.3 = load i32, i32* %A.gep.3, align 4
+ %add.3 = add i32 %A.3, 20
+ %xor.3 = xor i32 %add.3, %B.3
+ store i32 %xor.3, i32* %A.gep.3, align 4
+ %iv.next = add nuw nsw i64 %iv, 16
+ %cond = icmp ult i64 %iv.next, %N
+ br i1 %cond, label %loop, label %exit
+
+exit:
+ ret void
+}
+
+define void @loop_iv_update_at_start(float* %src, float* %dst) #0 {
+; CHECK-LABEL: @loop_iv_update_at_start(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
+; CHECK-NEXT: [[COND:%.*]] = icmp ult i32 [[IV]], 2000
+; CHECK-NEXT: [[SRC_GEP_0:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 0
+; CHECK-NEXT: [[SRC_0:%.*]] = load float, float* [[SRC_GEP_0]], align 8
+; CHECK-NEXT: [[ADD_0:%.*]] = fadd float [[SRC_0]], 1.000000e+00
+; CHECK-NEXT: [[MUL_0:%.*]] = fmul float [[ADD_0]], [[SRC_0]]
+; CHECK-NEXT: [[DST_GEP_0:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 0
+; CHECK-NEXT: store float [[MUL_0]], float* [[DST_GEP_0]], align 8
+; CHECK-NEXT: [[SRC_GEP_1:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 1
+; CHECK-NEXT: [[SRC_1:%.*]] = load float, float* [[SRC_GEP_1]], align 8
+; CHECK-NEXT: [[ADD_1:%.*]] = fadd float [[SRC_1]], 1.000000e+00
+; CHECK-NEXT: [[MUL_1:%.*]] = fmul float [[ADD_1]], [[SRC_1]]
+; CHECK-NEXT: [[DST_GEP_1:%.*]] = getelementptr inbounds float, float* [[DST]], i64 1
+; CHECK-NEXT: store float [[MUL_1]], float* [[DST_GEP_1]], align 8
+; CHECK-NEXT: [[SRC_GEP_2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
+; CHECK-NEXT: [[SRC_2:%.*]] = load float, float* [[SRC_GEP_2]], align 8
+; CHECK-NEXT: [[ADD_2:%.*]] = fadd float [[SRC_2]], 1.000000e+00
+; CHECK-NEXT: [[MUL_2:%.*]] = fmul float [[ADD_2]], [[SRC_2]]
+; CHECK-NEXT: [[DST_GEP_2:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
+; CHECK-NEXT: store float [[MUL_2]], float* [[DST_GEP_2]], align 8
+; CHECK-NEXT: [[SRC_GEP_3:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
+; CHECK-NEXT: [[SRC_3:%.*]] = load float, float* [[SRC_GEP_3]], align 8
+; CHECK-NEXT: [[ADD_3:%.*]] = fadd float [[SRC_3]], 1.000000e+00
+; CHECK-NEXT: [[MUL_3:%.*]] = fmul float [[ADD_3]], [[SRC_3]]
+; CHECK-NEXT: [[DST_GEP_3:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
+; CHECK-NEXT: store float [[MUL_3]], float* [[DST_GEP_3]], align 8
+; CHECK-NEXT: [[SRC_GEP_4:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 4
+; CHECK-NEXT: [[SRC_4:%.*]] = load float, float* [[SRC_GEP_4]], align 8
+; CHECK-NEXT: [[ADD_4:%.*]] = fadd float [[SRC_4]], 1.000000e+00
+; CHECK-NEXT: [[MUL_4:%.*]] = fmul float [[ADD_4]], [[SRC_4]]
+; CHECK-NEXT: [[DST_GEP_4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 4
+; CHECK-NEXT: store float [[MUL_4]], float* [[DST_GEP_4]], align 8
+; CHECK-NEXT: br i1 [[COND]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK: exit:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
+ %iv.next = add i32 %iv, 1
+ %cond = icmp ult i32 %iv, 2000
+
+ %src.gep.0 = getelementptr inbounds float, float* %src, i64 0
+ %src.0 = load float, float* %src.gep.0, align 8
+ %add.0 = fadd float %src.0, 1.0
+ %mul.0 = fmul float %add.0, %src.0
+ %dst.gep.0 = getelementptr inbounds float, float* %dst, i64 0
+ store float %mul.0, float* %dst.gep.0, align 8
+
+ %src.gep.1 = getelementptr inbounds float, float* %src, i64 1
+ %src.1 = load float, float* %src.gep.1, align 8
+ %add.1 = fadd float %src.1, 1.0
+ %mul.1 = fmul float %add.1, %src.1
+ %dst.gep.1 = getelementptr inbounds float, float* %dst, i64 1
+ store float %mul.1, float* %dst.gep.1, align 8
+ %src.gep.2 = getelementptr inbounds float, float* %src, i64 2
+ %src.2 = load float, float* %src.gep.2, align 8
+ %add.2 = fadd float %src.2, 1.0
+ %mul.2 = fmul float %add.2, %src.2
+ %dst.gep.2 = getelementptr inbounds float, float* %dst, i64 2
+ store float %mul.2, float* %dst.gep.2, align 8
+ %src.gep.3 = getelementptr inbounds float, float* %src, i64 3
+ %src.3 = load float, float* %src.gep.3, align 8
+ %add.3 = fadd float %src.3, 1.0
+ %mul.3 = fmul float %add.3, %src.3
+ %dst.gep.3 = getelementptr inbounds float, float* %dst, i64 3
+ store float %mul.3, float* %dst.gep.3, align 8
+ %src.gep.4 = getelementptr inbounds float, float* %src, i64 4
+ %src.4 = load float, float* %src.gep.4, align 8
+ %add.4 = fadd float %src.4, 1.0
+ %mul.4 = fmul float %add.4, %src.4
+ %dst.gep.4 = getelementptr inbounds float, float* %dst, i64 4
+ store float %mul.4, float* %dst.gep.4, align 8
+ br i1 %cond, label %loop, label %exit
+
+exit:
+ ret void
+}
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll
index dc7e2a18cffa..5d77d5c2a009 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll
@@ -72,6 +72,254 @@ entry:
ret void
}
+define void @needs_versioning_profitable_2_sources(i32* %dst, i32* %A, i32* %B) {
+; CHECK-LABEL: @needs_versioning_profitable_2_sources(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A_0:%.*]] = load i32, i32* [[A:%.*]], align 4
+; CHECK-NEXT: [[B_0:%.*]] = load i32, i32* [[B:%.*]], align 4
+; CHECK-NEXT: [[R_0:%.*]] = add i32 [[A_0]], [[B_0]]
+; CHECK-NEXT: [[MUL_0:%.*]] = mul i32 [[R_0]], 2
+; CHECK-NEXT: store i32 [[MUL_0]], i32* [[DST:%.*]], align 4
+; CHECK-NEXT: [[A_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 1
+; CHECK-NEXT: [[A_1:%.*]] = load i32, i32* [[A_GEP_1]], align 4
+; CHECK-NEXT: [[B_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 1
+; CHECK-NEXT: [[B_1:%.*]] = load i32, i32* [[B_GEP_1]], align 4
+; CHECK-NEXT: [[R_1:%.*]] = add i32 [[A_1]], [[B_1]]
+; CHECK-NEXT: [[MUL_1:%.*]] = mul i32 [[R_1]], 2
+; CHECK-NEXT: [[DST_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 1
+; CHECK-NEXT: store i32 [[MUL_1]], i32* [[DST_GEP_1]], align 4
+; CHECK-NEXT: [[A_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 2
+; CHECK-NEXT: [[A_2:%.*]] = load i32, i32* [[A_GEP_2]], align 4
+; CHECK-NEXT: [[B_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2
+; CHECK-NEXT: [[B_2:%.*]] = load i32, i32* [[B_GEP_2]], align 4
+; CHECK-NEXT: [[R_2:%.*]] = add i32 [[A_2]], [[B_2]]
+; CHECK-NEXT: [[MUL_2:%.*]] = mul i32 [[R_2]], 2
+; CHECK-NEXT: [[DST_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
+; CHECK-NEXT: store i32 [[MUL_2]], i32* [[DST_GEP_2]], align 4
+; CHECK-NEXT: [[A_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 3
+; CHECK-NEXT: [[A_3:%.*]] = load i32, i32* [[A_GEP_3]], align 4
+; CHECK-NEXT: [[B_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3
+; CHECK-NEXT: [[B_3:%.*]] = load i32, i32* [[B_GEP_3]], align 4
+; CHECK-NEXT: [[R_3:%.*]] = add i32 [[A_3]], [[B_3]]
+; CHECK-NEXT: [[MUL_3:%.*]] = mul i32 [[R_3]], 2
+; CHECK-NEXT: [[DST_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
+; CHECK-NEXT: store i32 [[MUL_3]], i32* [[DST_GEP_3]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %A.0 = load i32, i32* %A, align 4
+ %B.0 = load i32, i32* %B, align 4
+ %r.0 = add i32 %A.0, %B.0
+ %mul.0 = mul i32 %r.0, 2
+ store i32 %mul.0, i32* %dst, align 4
+ %A.gep.1 = getelementptr inbounds i32, i32* %A, i64 1
+ %A.1 = load i32, i32* %A.gep.1, align 4
+ %B.gep.1 = getelementptr inbounds i32, i32* %B, i64 1
+ %B.1 = load i32, i32* %B.gep.1, align 4
+ %r.1 = add i32 %A.1, %B.1
+ %mul.1 = mul i32 %r.1, 2
+ %dst.gep.1 = getelementptr inbounds i32, i32* %dst, i64 1
+ store i32 %mul.1, i32* %dst.gep.1, align 4
+ %A.gep.2 = getelementptr inbounds i32, i32* %A, i64 2
+ %A.2 = load i32, i32* %A.gep.2, align 4
+ %B.gep.2 = getelementptr inbounds i32, i32* %B, i64 2
+ %B.2 = load i32, i32* %B.gep.2, align 4
+ %r.2 = add i32 %A.2, %B.2
+ %mul.2 = mul i32 %r.2, 2
+ %dst.gep.2 = getelementptr inbounds i32, i32* %dst, i64 2
+ store i32 %mul.2, i32* %dst.gep.2, align 4
+ %A.gep.3 = getelementptr inbounds i32, i32* %A, i64 3
+ %A.3 = load i32, i32* %A.gep.3, align 4
+ %B.gep.3 = getelementptr inbounds i32, i32* %B, i64 3
+ %B.3 = load i32, i32* %B.gep.3, align 4
+ %r.3 = add i32 %A.3, %B.3
+ %mul.3 = mul i32 %r.3, 2
+ %dst.gep.3 = getelementptr inbounds i32, i32* %dst, i64 3
+ store i32 %mul.3, i32* %dst.gep.3, align 4
+
+ ret void
+}
+
+declare void @use(i32)
+
+declare void @bar()
+
+define void @needs_versioning_profitable_split_points(i32* %dst, i32* %src) {
+; CHECK-LABEL: @needs_versioning_profitable_split_points(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: call void @bar()
+; CHECK-NEXT: call void @bar()
+; CHECK-NEXT: call void @bar()
+; CHECK-NEXT: [[SRC_0:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; CHECK-NEXT: [[R_0:%.*]] = ashr i32 [[SRC_0]], 16
+; CHECK-NEXT: store i32 [[R_0]], i32* [[DST:%.*]], align 4
+; CHECK-NEXT: [[SRC_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 1
+; CHECK-NEXT: [[SRC_1:%.*]] = load i32, i32* [[SRC_GEP_1]], align 4
+; CHECK-NEXT: [[R_1:%.*]] = ashr i32 [[SRC_1]], 16
+; CHECK-NEXT: [[DST_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 1
+; CHECK-NEXT: store i32 [[R_1]], i32* [[DST_GEP_1]], align 4
+; CHECK-NEXT: [[SRC_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
+; CHECK-NEXT: [[SRC_2:%.*]] = load i32, i32* [[SRC_GEP_2]], align 4
+; CHECK-NEXT: [[R_2:%.*]] = ashr i32 [[SRC_2]], 16
+; CHECK-NEXT: [[DST_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
+; CHECK-NEXT: store i32 [[R_2]], i32* [[DST_GEP_2]], align 4
+; CHECK-NEXT: [[SRC_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
+; CHECK-NEXT: [[SRC_3:%.*]] = load i32, i32* [[SRC_GEP_3]], align 4
+; CHECK-NEXT: [[R_3:%.*]] = ashr i32 [[SRC_3]], 16
+; CHECK-NEXT: [[DST_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
+; CHECK-NEXT: store i32 [[R_3]], i32* [[DST_GEP_3]], align 4
+; CHECK-NEXT: call void @bar()
+; CHECK-NEXT: ret void
+;
+entry:
+ call void @bar()
+ call void @bar()
+ call void @bar()
+
+ %src.0 = load i32, i32* %src, align 4
+ %r.0 = ashr i32 %src.0, 16
+ store i32 %r.0, i32* %dst, align 4
+ %src.gep.1 = getelementptr inbounds i32, i32* %src, i64 1
+ %src.1 = load i32, i32* %src.gep.1, align 4
+ %r.1 = ashr i32 %src.1, 16
+ %dst.gep.1 = getelementptr inbounds i32, i32* %dst, i64 1
+ store i32 %r.1, i32* %dst.gep.1, align 4
+ %src.gep.2 = getelementptr inbounds i32, i32* %src, i64 2
+ %src.2 = load i32, i32* %src.gep.2, align 4
+ %r.2 = ashr i32 %src.2, 16
+ %dst.gep.2 = getelementptr inbounds i32, i32* %dst, i64 2
+ store i32 %r.2, i32* %dst.gep.2, align 4
+ %src.gep.3 = getelementptr inbounds i32, i32* %src, i64 3
+ %src.3 = load i32, i32* %src.gep.3, align 4
+ %r.3 = ashr i32 %src.3, 16
+ %dst.gep.3 = getelementptr inbounds i32, i32* %dst, i64 3
+ store i32 %r.3, i32* %dst.gep.3, align 4
+
+ call void @bar()
+ ret void
+}
+
+define void @needs_versioning_profitable_load_used_outside_region1(i32* %dst, i32* %src, i1 %c) {
+; CHECK-LABEL: @needs_versioning_profitable_load_used_outside_region1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 [[C:%.*]], label [[THEN:%.*]], label [[EXIT:%.*]]
+; CHECK: then:
+; CHECK-NEXT: [[SRC_0:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; CHECK-NEXT: [[R_0:%.*]] = ashr i32 [[SRC_0]], 16
+; CHECK-NEXT: store i32 [[R_0]], i32* [[DST:%.*]], align 4
+; CHECK-NEXT: [[SRC_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 1
+; CHECK-NEXT: [[SRC_1:%.*]] = load i32, i32* [[SRC_GEP_1]], align 4
+; CHECK-NEXT: [[R_1:%.*]] = ashr i32 [[SRC_1]], 16
+; CHECK-NEXT: [[DST_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 1
+; CHECK-NEXT: store i32 [[R_1]], i32* [[DST_GEP_1]], align 4
+; CHECK-NEXT: [[SRC_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
+; CHECK-NEXT: [[SRC_2:%.*]] = load i32, i32* [[SRC_GEP_2]], align 4
+; CHECK-NEXT: [[R_2:%.*]] = ashr i32 [[SRC_2]], 16
+; CHECK-NEXT: [[DST_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
+; CHECK-NEXT: store i32 [[R_2]], i32* [[DST_GEP_2]], align 4
+; CHECK-NEXT: [[SRC_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
+; CHECK-NEXT: [[SRC_3:%.*]] = load i32, i32* [[SRC_GEP_3]], align 4
+; CHECK-NEXT: [[R_3:%.*]] = ashr i32 [[SRC_3]], 16
+; CHECK-NEXT: [[DST_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
+; CHECK-NEXT: store i32 [[R_3]], i32* [[DST_GEP_3]], align 4
+; CHECK-NEXT: [[SRC_GEP_5:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 5
+; CHECK-NEXT: [[L:%.*]] = load i32, i32* [[SRC_GEP_5]], align 4
+; CHECK-NEXT: call void @use(i32 [[L]])
+; CHECK-NEXT: br label [[EXIT]]
+; CHECK: exit:
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 %c, label %then, label %exit
+
+then:
+ %src.0 = load i32, i32* %src, align 4
+ %r.0 = ashr i32 %src.0, 16
+ store i32 %r.0, i32* %dst, align 4
+ %src.gep.1 = getelementptr inbounds i32, i32* %src, i64 1
+ %src.1 = load i32, i32* %src.gep.1, align 4
+ %r.1 = ashr i32 %src.1, 16
+ %dst.gep.1 = getelementptr inbounds i32, i32* %dst, i64 1
+ store i32 %r.1, i32* %dst.gep.1, align 4
+ %src.gep.2 = getelementptr inbounds i32, i32* %src, i64 2
+ %src.2 = load i32, i32* %src.gep.2, align 4
+ %r.2 = ashr i32 %src.2, 16
+ %dst.gep.2 = getelementptr inbounds i32, i32* %dst, i64 2
+ store i32 %r.2, i32* %dst.gep.2, align 4
+ %src.gep.3 = getelementptr inbounds i32, i32* %src, i64 3
+ %src.3 = load i32, i32* %src.gep.3, align 4
+ %r.3 = ashr i32 %src.3, 16
+ %dst.gep.3 = getelementptr inbounds i32, i32* %dst, i64 3
+ store i32 %r.3, i32* %dst.gep.3, align 4
+ %src.gep.5 = getelementptr inbounds i32, i32* %src, i64 5
+ %l = load i32, i32* %src.gep.5
+ call void @use(i32 %l)
+ br label %exit
+
+exit:
+ ret void
+}
+
+define void @needs_versioning_profitable_load_used_outside_region2(i32* %dst, i32* %src, i1 %c) {
+; CHECK-LABEL: @needs_versioning_profitable_load_used_outside_region2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 [[C:%.*]], label [[THEN:%.*]], label [[EXIT:%.*]]
+; CHECK: then:
+; CHECK-NEXT: [[SRC_0:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; CHECK-NEXT: [[R_0:%.*]] = ashr i32 [[SRC_0]], 16
+; CHECK-NEXT: store i32 [[R_0]], i32* [[DST:%.*]], align 4
+; CHECK-NEXT: [[SRC_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 1
+; CHECK-NEXT: [[SRC_1:%.*]] = load i32, i32* [[SRC_GEP_1]], align 4
+; CHECK-NEXT: [[R_1:%.*]] = ashr i32 [[SRC_1]], 16
+; CHECK-NEXT: [[DST_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 1
+; CHECK-NEXT: store i32 [[R_1]], i32* [[DST_GEP_1]], align 4
+; CHECK-NEXT: [[SRC_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
+; CHECK-NEXT: [[SRC_2:%.*]] = load i32, i32* [[SRC_GEP_2]], align 4
+; CHECK-NEXT: [[SRC_GEP_5:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 5
+; CHECK-NEXT: [[L:%.*]] = load i32, i32* [[SRC_GEP_5]], align 4
+; CHECK-NEXT: [[R_2:%.*]] = ashr i32 [[SRC_2]], 16
+; CHECK-NEXT: [[DST_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
+; CHECK-NEXT: store i32 [[R_2]], i32* [[DST_GEP_2]], align 4
+; CHECK-NEXT: [[SRC_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
+; CHECK-NEXT: [[SRC_3:%.*]] = load i32, i32* [[SRC_GEP_3]], align 4
+; CHECK-NEXT: [[R_3:%.*]] = ashr i32 [[SRC_3]], 16
+; CHECK-NEXT: [[DST_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
+; CHECK-NEXT: store i32 [[R_3]], i32* [[DST_GEP_3]], align 4
+; CHECK-NEXT: call void @use(i32 [[L]])
+; CHECK-NEXT: br label [[EXIT]]
+; CHECK: exit:
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 %c, label %then, label %exit
+
+then:
+ %src.0 = load i32, i32* %src, align 4
+ %r.0 = ashr i32 %src.0, 16
+ store i32 %r.0, i32* %dst, align 4
+ %src.gep.1 = getelementptr inbounds i32, i32* %src, i64 1
+ %src.1 = load i32, i32* %src.gep.1, align 4
+ %r.1 = ashr i32 %src.1, 16
+ %dst.gep.1 = getelementptr inbounds i32, i32* %dst, i64 1
+ store i32 %r.1, i32* %dst.gep.1, align 4
+ %src.gep.2 = getelementptr inbounds i32, i32* %src, i64 2
+ %src.2 = load i32, i32* %src.gep.2, align 4
+ %src.gep.5 = getelementptr inbounds i32, i32* %src, i64 5
+ %l = load i32, i32* %src.gep.5
+ %r.2 = ashr i32 %src.2, 16
+ %dst.gep.2 = getelementptr inbounds i32, i32* %dst, i64 2
+ store i32 %r.2, i32* %dst.gep.2, align 4
+ %src.gep.3 = getelementptr inbounds i32, i32* %src, i64 3
+ %src.3 = load i32, i32* %src.gep.3, align 4
+ %r.3 = ashr i32 %src.3, 16
+ %dst.gep.3 = getelementptr inbounds i32, i32* %dst, i64 3
+ store i32 %r.3, i32* %dst.gep.3, align 4
+ call void @use(i32 %l)
+ br label %exit
+
+exit:
+ ret void
+}
define void @no_version(i32* nocapture %dst, i32* nocapture readonly %src) {
; CHECK-LABEL: @no_version(
@@ -452,3 +700,686 @@ bb15:
store double %tmp17, double* %tmp14, align 8
ret void
}
+
+%struct.2 = type { [4 x float] }
+
+; Make sure we do not crash when we encounter a SCEVCouldNotCompute.
+define void @no_lcssa_phi(%struct.2* %A, float* %B, i1 %c) {
+; CHECK-LABEL: @no_lcssa_phi(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[PTR_PHI:%.*]] = phi %struct.2* [ [[A:%.*]], [[BB:%.*]] ], [ null, [[LOOP]] ]
+; CHECK-NEXT: br i1 [[C:%.*]], label [[EXIT:%.*]], label [[LOOP]]
+; CHECK: exit:
+; CHECK-NEXT: [[B_GEP_0:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 0
+; CHECK-NEXT: [[L_0:%.*]] = load float, float* [[B_GEP_0]], align 8
+; CHECK-NEXT: [[ADD_0:%.*]] = fadd float [[L_0]], 1.000000e+01
+; CHECK-NEXT: [[MUL_0:%.*]] = fmul float [[ADD_0]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_0:%.*]] = getelementptr inbounds [[STRUCT_2:%.*]], %struct.2* [[PTR_PHI]], i64 0, i32 0, i32 0
+; CHECK-NEXT: store float [[MUL_0]], float* [[A_GEP_0]], align 8
+; CHECK-NEXT: [[B_GEP_1:%.*]] = getelementptr inbounds float, float* [[B]], i64 1
+; CHECK-NEXT: [[L_1:%.*]] = load float, float* [[B_GEP_1]], align 8
+; CHECK-NEXT: [[ADD_1:%.*]] = fadd float [[L_1]], 1.000000e+01
+; CHECK-NEXT: [[MUL_1:%.*]] = fmul float [[ADD_1]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_1:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI]], i64 0, i32 0, i32 1
+; CHECK-NEXT: store float [[MUL_1]], float* [[A_GEP_1]], align 8
+; CHECK-NEXT: [[B_GEP_2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2
+; CHECK-NEXT: [[L_2:%.*]] = load float, float* [[B_GEP_2]], align 8
+; CHECK-NEXT: [[ADD_2:%.*]] = fadd float [[L_2]], 1.000000e+01
+; CHECK-NEXT: [[MUL_2:%.*]] = fmul float [[ADD_2]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_2:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI]], i64 0, i32 0, i32 2
+; CHECK-NEXT: store float [[MUL_2]], float* [[A_GEP_2]], align 8
+; CHECK-NEXT: [[B_GEP_3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3
+; CHECK-NEXT: [[L_3:%.*]] = load float, float* [[B_GEP_3]], align 8
+; CHECK-NEXT: [[ADD_3:%.*]] = fadd float [[L_3]], 1.000000e+01
+; CHECK-NEXT: [[MUL_3:%.*]] = fmul float [[ADD_3]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_3:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI]], i64 0, i32 0, i32 3
+; CHECK-NEXT: store float [[MUL_3]], float* [[A_GEP_3]], align 8
+; CHECK-NEXT: ret void
+;
+bb:
+ br label %loop
+
+loop:
+ %ptr.phi = phi %struct.2* [ %A, %bb ], [ null, %loop ]
+ br i1 %c, label %exit, label %loop
+
+exit:
+ %B.gep.0 = getelementptr inbounds float, float* %B, i64 0
+ %l.0 = load float, float* %B.gep.0, align 8
+ %add.0 = fadd float %l.0, 10.0
+ %mul.0 = fmul float %add.0, 30.0
+ %A.gep.0 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi, i64 0, i32 0, i32 0
+ store float %mul.0, float* %A.gep.0, align 8
+ %B.gep.1 = getelementptr inbounds float, float* %B, i64 1
+ %l.1 = load float, float* %B.gep.1, align 8
+ %add.1 = fadd float %l.1, 10.0
+ %mul.1 = fmul float %add.1, 30.0
+ %A.gep.1 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi, i64 0, i32 0, i32 1
+ store float %mul.1, float* %A.gep.1, align 8
+ %B.gep.2 = getelementptr inbounds float, float* %B, i64 2
+ %l.2 = load float, float* %B.gep.2, align 8
+ %add.2 = fadd float %l.2, 10.0
+ %mul.2 = fmul float %add.2, 30.0
+ %A.gep.2 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi, i64 0, i32 0, i32 2
+ store float %mul.2, float* %A.gep.2, align 8
+ %B.gep.3 = getelementptr inbounds float, float* %B, i64 3
+ %l.3 = load float, float* %B.gep.3, align 8
+ %add.3 = fadd float %l.3, 10.0
+ %mul.3 = fmul float %add.3, 30.0
+ %A.gep.3 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi, i64 0, i32 0, i32 3
+ store float %mul.3, float* %A.gep.3, align 8
+ ret void
+}
+
+; Make sure lcssa phis as pointer bases are handled properly.
+define void @lcssa_phi(%struct.2* %A, float* %B, i1 %c) {
+; CHECK-LABEL: @lcssa_phi(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[PTR_PHI:%.*]] = phi %struct.2* [ [[A:%.*]], [[BB:%.*]] ], [ null, [[LOOP]] ]
+; CHECK-NEXT: br i1 [[C:%.*]], label [[EXIT:%.*]], label [[LOOP]]
+; CHECK: exit:
+; CHECK-NEXT: [[PTR_PHI_LCSSA:%.*]] = phi %struct.2* [ [[PTR_PHI]], [[LOOP]] ]
+; CHECK-NEXT: [[B_GEP_0:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 0
+; CHECK-NEXT: [[L_0:%.*]] = load float, float* [[B_GEP_0]], align 8
+; CHECK-NEXT: [[ADD_0:%.*]] = fadd float [[L_0]], 1.000000e+01
+; CHECK-NEXT: [[MUL_0:%.*]] = fmul float [[ADD_0]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_0:%.*]] = getelementptr inbounds [[STRUCT_2:%.*]], %struct.2* [[PTR_PHI_LCSSA]], i64 0, i32 0, i32 0
+; CHECK-NEXT: store float [[MUL_0]], float* [[A_GEP_0]], align 8
+; CHECK-NEXT: [[B_GEP_1:%.*]] = getelementptr inbounds float, float* [[B]], i64 1
+; CHECK-NEXT: [[L_1:%.*]] = load float, float* [[B_GEP_1]], align 8
+; CHECK-NEXT: [[ADD_1:%.*]] = fadd float [[L_1]], 1.000000e+01
+; CHECK-NEXT: [[MUL_1:%.*]] = fmul float [[ADD_1]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_1:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI_LCSSA]], i64 0, i32 0, i32 1
+; CHECK-NEXT: store float [[MUL_1]], float* [[A_GEP_1]], align 8
+; CHECK-NEXT: [[B_GEP_2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2
+; CHECK-NEXT: [[L_2:%.*]] = load float, float* [[B_GEP_2]], align 8
+; CHECK-NEXT: [[ADD_2:%.*]] = fadd float [[L_2]], 1.000000e+01
+; CHECK-NEXT: [[MUL_2:%.*]] = fmul float [[ADD_2]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_2:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI_LCSSA]], i64 0, i32 0, i32 2
+; CHECK-NEXT: store float [[MUL_2]], float* [[A_GEP_2]], align 8
+; CHECK-NEXT: [[B_GEP_3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3
+; CHECK-NEXT: [[L_3:%.*]] = load float, float* [[B_GEP_3]], align 8
+; CHECK-NEXT: [[ADD_3:%.*]] = fadd float [[L_3]], 1.000000e+01
+; CHECK-NEXT: [[MUL_3:%.*]] = fmul float [[ADD_3]], 3.000000e+01
+; CHECK-NEXT: [[A_GEP_3:%.*]] = getelementptr inbounds [[STRUCT_2]], %struct.2* [[PTR_PHI_LCSSA]], i64 0, i32 0, i32 3
+; CHECK-NEXT: store float [[MUL_3]], float* [[A_GEP_3]], align 8
+; CHECK-NEXT: ret void
+;
+bb:
+ br label %loop
+
+loop:
+ %ptr.phi = phi %struct.2* [ %A, %bb ], [ null, %loop ]
+ br i1 %c, label %exit, label %loop
+
+exit:
+ %ptr.phi.lcssa = phi %struct.2* [ %ptr.phi, %loop ]
+ %B.gep.0 = getelementptr inbounds float, float* %B, i64 0
+ %l.0 = load float, float* %B.gep.0, align 8
+ %add.0 = fadd float %l.0, 10.0
+ %mul.0 = fmul float %add.0, 30.0
+ %A.gep.0 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi.lcssa, i64 0, i32 0, i32 0
+ store float %mul.0, float* %A.gep.0, align 8
+ %B.gep.1 = getelementptr inbounds float, float* %B, i64 1
+ %l.1 = load float, float* %B.gep.1, align 8
+ %add.1 = fadd float %l.1, 10.0
+ %mul.1 = fmul float %add.1, 30.0
+ %A.gep.1 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi.lcssa, i64 0, i32 0, i32 1
+ store float %mul.1, float* %A.gep.1, align 8
+ %B.gep.2 = getelementptr inbounds float, float* %B, i64 2
+ %l.2 = load float, float* %B.gep.2, align 8
+ %add.2 = fadd float %l.2, 10.0
+ %mul.2 = fmul float %add.2, 30.0
+ %A.gep.2 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi.lcssa, i64 0, i32 0, i32 2
+ store float %mul.2, float* %A.gep.2, align 8
+ %B.gep.3 = getelementptr inbounds float, float* %B, i64 3
+ %l.3 = load float, float* %B.gep.3, align 8
+ %add.3 = fadd float %l.3, 10.0
+ %mul.3 = fmul float %add.3, 30.0
+ %A.gep.3 = getelementptr inbounds %struct.2, %struct.2* %ptr.phi.lcssa, i64 0, i32 0, i32 3
+ store float %mul.3, float* %A.gep.3, align 8
+ ret void
+}
+
+%struct.spam = type { [60 x i32], i32, [12 x i8] }
+
+declare void @foo(i8*)
+
+; Test case with a basic block where parts can be vectorized without versioning.
+define i32 @block_partly_vectorized_without_versioning(%struct.spam* readonly %arg, i8* nocapture readonly %arg1, i8* nocapture %arg2, i8* nocapture readonly %arg3, i8* %A, i8* %B) {
+; CHECK-LABEL: @block_partly_vectorized_without_versioning(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: [[T:%.*]] = alloca <16 x i8>, align 16
+; CHECK-NEXT: [[T4:%.*]] = getelementptr inbounds <16 x i8>, <16 x i8>* [[T]], i64 0, i64 0
+; CHECK-NEXT: [[T5:%.*]] = getelementptr inbounds i8, i8* [[ARG3:%.*]], i64 1
+; CHECK-NEXT: [[T6:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 2
+; CHECK-NEXT: [[T7:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 3
+; CHECK-NEXT: [[T8:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 4
+; CHECK-NEXT: [[T9:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 5
+; CHECK-NEXT: [[T10:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 6
+; CHECK-NEXT: [[T11:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 7
+; CHECK-NEXT: [[T12:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 8
+; CHECK-NEXT: [[T13:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 9
+; CHECK-NEXT: [[T14:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 10
+; CHECK-NEXT: [[T15:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 11
+; CHECK-NEXT: [[T16:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 12
+; CHECK-NEXT: [[T17:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 13
+; CHECK-NEXT: [[T18:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 14
+; CHECK-NEXT: [[T19:%.*]] = bitcast i8* [[ARG1:%.*]] to <16 x i8>*
+; CHECK-NEXT: [[A_GEP_0:%.*]] = getelementptr i8, i8* [[A:%.*]], i64 0
+; CHECK-NEXT: [[B_GEP_0:%.*]] = getelementptr i8, i8* [[B:%.*]], i64 0
+; CHECK-NEXT: [[A_GEP_1:%.*]] = getelementptr i8, i8* [[A]], i64 1
+; CHECK-NEXT: [[B_GEP_1:%.*]] = getelementptr i8, i8* [[B]], i64 1
+; CHECK-NEXT: [[A_GEP_2:%.*]] = getelementptr i8, i8* [[A]], i64 2
+; CHECK-NEXT: [[B_GEP_2:%.*]] = getelementptr i8, i8* [[B]], i64 2
+; CHECK-NEXT: [[A_GEP_3:%.*]] = getelementptr i8, i8* [[A]], i64 3
+; CHECK-NEXT: [[B_GEP_3:%.*]] = getelementptr i8, i8* [[B]], i64 3
+; CHECK-NEXT: [[A_GEP_4:%.*]] = getelementptr i8, i8* [[A]], i64 4
+; CHECK-NEXT: [[B_GEP_4:%.*]] = getelementptr i8, i8* [[B]], i64 4
+; CHECK-NEXT: [[A_GEP_5:%.*]] = getelementptr i8, i8* [[A]], i64 5
+; CHECK-NEXT: [[B_GEP_5:%.*]] = getelementptr i8, i8* [[B]], i64 5
+; CHECK-NEXT: [[A_GEP_6:%.*]] = getelementptr i8, i8* [[A]], i64 6
+; CHECK-NEXT: [[B_GEP_6:%.*]] = getelementptr i8, i8* [[B]], i64 6
+; CHECK-NEXT: [[A_GEP_7:%.*]] = getelementptr i8, i8* [[A]], i64 7
+; CHECK-NEXT: [[B_GEP_7:%.*]] = getelementptr i8, i8* [[B]], i64 7
+; CHECK-NEXT: [[A_GEP_8:%.*]] = getelementptr i8, i8* [[A]], i64 8
+; CHECK-NEXT: [[B_GEP_8:%.*]] = getelementptr i8, i8* [[B]], i64 8
+; CHECK-NEXT: [[A_GEP_9:%.*]] = getelementptr i8, i8* [[A]], i64 9
+; CHECK-NEXT: [[B_GEP_9:%.*]] = getelementptr i8, i8* [[B]], i64 9
+; CHECK-NEXT: [[A_GEP_10:%.*]] = getelementptr i8, i8* [[A]], i64 10
+; CHECK-NEXT: [[B_GEP_10:%.*]] = getelementptr i8, i8* [[B]], i64 10
+; CHECK-NEXT: [[A_GEP_11:%.*]] = getelementptr i8, i8* [[A]], i64 11
+; CHECK-NEXT: [[B_GEP_11:%.*]] = getelementptr i8, i8* [[B]], i64 11
+; CHECK-NEXT: [[A_GEP_12:%.*]] = getelementptr i8, i8* [[A]], i64 12
+; CHECK-NEXT: [[B_GEP_12:%.*]] = getelementptr i8, i8* [[B]], i64 12
+; CHECK-NEXT: [[A_GEP_13:%.*]] = getelementptr i8, i8* [[A]], i64 13
+; CHECK-NEXT: [[B_GEP_13:%.*]] = getelementptr i8, i8* [[B]], i64 13
+; CHECK-NEXT: [[A_GEP_14:%.*]] = getelementptr i8, i8* [[A]], i64 14
+; CHECK-NEXT: [[B_GEP_14:%.*]] = getelementptr i8, i8* [[B]], i64 14
+; CHECK-NEXT: [[A_GEP_15:%.*]] = getelementptr i8, i8* [[A]], i64 15
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i8* [[A_GEP_0]] to <16 x i8>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* [[TMP0]], align 1
+; CHECK-NEXT: [[B_GEP_15:%.*]] = getelementptr i8, i8* [[B]], i64 15
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast i8* [[B_GEP_0]] to <16 x i8>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[TMP2]], align 1
+; CHECK-NEXT: [[TMP4:%.*]] = xor <16 x i8> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[R_GEP_0:%.*]] = getelementptr i8, i8* [[ARG1]], i64 0
+; CHECK-NEXT: [[R_GEP_1:%.*]] = getelementptr i8, i8* [[ARG1]], i64 1
+; CHECK-NEXT: [[R_GEP_2:%.*]] = getelementptr i8, i8* [[ARG1]], i64 2
+; CHECK-NEXT: [[R_GEP_3:%.*]] = getelementptr i8, i8* [[ARG1]], i64 3
+; CHECK-NEXT: [[R_GEP_4:%.*]] = getelementptr i8, i8* [[ARG1]], i64 4
+; CHECK-NEXT: [[R_GEP_5:%.*]] = getelementptr i8, i8* [[ARG1]], i64 5
+; CHECK-NEXT: [[R_GEP_6:%.*]] = getelementptr i8, i8* [[ARG1]], i64 6
+; CHECK-NEXT: [[R_GEP_7:%.*]] = getelementptr i8, i8* [[ARG1]], i64 7
+; CHECK-NEXT: [[R_GEP_8:%.*]] = getelementptr i8, i8* [[ARG1]], i64 8
+; CHECK-NEXT: [[R_GEP_9:%.*]] = getelementptr i8, i8* [[ARG1]], i64 9
+; CHECK-NEXT: [[R_GEP_10:%.*]] = getelementptr i8, i8* [[ARG1]], i64 10
+; CHECK-NEXT: [[R_GEP_11:%.*]] = getelementptr i8, i8* [[ARG1]], i64 11
+; CHECK-NEXT: [[R_GEP_12:%.*]] = getelementptr i8, i8* [[ARG1]], i64 12
+; CHECK-NEXT: [[R_GEP_13:%.*]] = getelementptr i8, i8* [[ARG1]], i64 13
+; CHECK-NEXT: [[R_GEP_14:%.*]] = getelementptr i8, i8* [[ARG1]], i64 14
+; CHECK-NEXT: [[R_GEP_15:%.*]] = getelementptr i8, i8* [[ARG1]], i64 15
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast i8* [[R_GEP_0]] to <16 x i8>*
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* [[TMP5]], align 1
+; CHECK-NEXT: [[T21:%.*]] = getelementptr inbounds i8, i8* [[ARG3]], i64 15
+; CHECK-NEXT: [[T22:%.*]] = bitcast i8* [[ARG3]] to <16 x i8>*
+; CHECK-NEXT: call void @foo(i8* nonnull [[T4]])
+; CHECK-NEXT: [[T26:%.*]] = load i8, i8* [[ARG3]], align 1
+; CHECK-NEXT: [[T27:%.*]] = load i8, i8* [[ARG2:%.*]], align 1
+; CHECK-NEXT: [[T28:%.*]] = xor i8 [[T27]], [[T26]]
+; CHECK-NEXT: store i8 [[T28]], i8* [[ARG2]], align 1
+; CHECK-NEXT: [[T29:%.*]] = load i8, i8* [[T5]], align 1
+; CHECK-NEXT: [[T30:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 1
+; CHECK-NEXT: [[T31:%.*]] = load i8, i8* [[T30]], align 1
+; CHECK-NEXT: [[T32:%.*]] = xor i8 [[T31]], [[T29]]
+; CHECK-NEXT: store i8 [[T32]], i8* [[T30]], align 1
+; CHECK-NEXT: [[T33:%.*]] = load i8, i8* [[T6]], align 1
+; CHECK-NEXT: [[T34:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 2
+; CHECK-NEXT: [[T35:%.*]] = load i8, i8* [[T34]], align 1
+; CHECK-NEXT: [[T36:%.*]] = xor i8 [[T35]], [[T33]]
+; CHECK-NEXT: store i8 [[T36]], i8* [[T34]], align 1
+; CHECK-NEXT: [[T37:%.*]] = load i8, i8* [[T7]], align 1
+; CHECK-NEXT: [[T38:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 3
+; CHECK-NEXT: [[T39:%.*]] = load i8, i8* [[T38]], align 1
+; CHECK-NEXT: [[T40:%.*]] = xor i8 [[T39]], [[T37]]
+; CHECK-NEXT: store i8 [[T40]], i8* [[T38]], align 1
+; CHECK-NEXT: [[T41:%.*]] = load i8, i8* [[T8]], align 1
+; CHECK-NEXT: [[T42:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 4
+; CHECK-NEXT: [[T43:%.*]] = load i8, i8* [[T42]], align 1
+; CHECK-NEXT: [[T44:%.*]] = xor i8 [[T43]], [[T41]]
+; CHECK-NEXT: store i8 [[T44]], i8* [[T42]], align 1
+; CHECK-NEXT: [[T45:%.*]] = load i8, i8* [[T9]], align 1
+; CHECK-NEXT: [[T46:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 5
+; CHECK-NEXT: [[T47:%.*]] = load i8, i8* [[T46]], align 1
+; CHECK-NEXT: [[T48:%.*]] = xor i8 [[T47]], [[T45]]
+; CHECK-NEXT: store i8 [[T48]], i8* [[T46]], align 1
+; CHECK-NEXT: [[T49:%.*]] = load i8, i8* [[T10]], align 1
+; CHECK-NEXT: [[T50:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 6
+; CHECK-NEXT: [[T51:%.*]] = load i8, i8* [[T50]], align 1
+; CHECK-NEXT: [[T52:%.*]] = xor i8 [[T51]], [[T49]]
+; CHECK-NEXT: store i8 [[T52]], i8* [[T50]], align 1
+; CHECK-NEXT: [[T53:%.*]] = load i8, i8* [[T11]], align 1
+; CHECK-NEXT: [[T54:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 7
+; CHECK-NEXT: [[T55:%.*]] = load i8, i8* [[T54]], align 1
+; CHECK-NEXT: [[T56:%.*]] = xor i8 [[T55]], [[T53]]
+; CHECK-NEXT: store i8 [[T56]], i8* [[T54]], align 1
+; CHECK-NEXT: [[T57:%.*]] = load i8, i8* [[T12]], align 1
+; CHECK-NEXT: [[T58:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 8
+; CHECK-NEXT: [[T59:%.*]] = load i8, i8* [[T58]], align 1
+; CHECK-NEXT: [[T60:%.*]] = xor i8 [[T59]], [[T57]]
+; CHECK-NEXT: store i8 [[T60]], i8* [[T58]], align 1
+; CHECK-NEXT: [[T61:%.*]] = load i8, i8* [[T13]], align 1
+; CHECK-NEXT: [[T62:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 9
+; CHECK-NEXT: [[T63:%.*]] = load i8, i8* [[T62]], align 1
+; CHECK-NEXT: [[T64:%.*]] = xor i8 [[T63]], [[T61]]
+; CHECK-NEXT: store i8 [[T64]], i8* [[T62]], align 1
+; CHECK-NEXT: [[T65:%.*]] = load i8, i8* [[T14]], align 1
+; CHECK-NEXT: [[T66:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 10
+; CHECK-NEXT: [[T67:%.*]] = load i8, i8* [[T66]], align 1
+; CHECK-NEXT: [[T68:%.*]] = xor i8 [[T67]], [[T65]]
+; CHECK-NEXT: store i8 [[T68]], i8* [[T66]], align 1
+; CHECK-NEXT: [[T69:%.*]] = load i8, i8* [[T15]], align 1
+; CHECK-NEXT: [[T70:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 11
+; CHECK-NEXT: [[T71:%.*]] = load i8, i8* [[T70]], align 1
+; CHECK-NEXT: [[T72:%.*]] = xor i8 [[T71]], [[T69]]
+; CHECK-NEXT: store i8 [[T72]], i8* [[T70]], align 1
+; CHECK-NEXT: [[T73:%.*]] = load i8, i8* [[T16]], align 1
+; CHECK-NEXT: [[T74:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 12
+; CHECK-NEXT: [[T75:%.*]] = load i8, i8* [[T74]], align 1
+; CHECK-NEXT: [[T76:%.*]] = xor i8 [[T75]], [[T73]]
+; CHECK-NEXT: store i8 [[T76]], i8* [[T74]], align 1
+; CHECK-NEXT: [[T77:%.*]] = load i8, i8* [[T17]], align 1
+; CHECK-NEXT: [[T78:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 13
+; CHECK-NEXT: [[T79:%.*]] = load i8, i8* [[T78]], align 1
+; CHECK-NEXT: [[T80:%.*]] = xor i8 [[T79]], [[T77]]
+; CHECK-NEXT: store i8 [[T80]], i8* [[T78]], align 1
+; CHECK-NEXT: [[T81:%.*]] = load i8, i8* [[T18]], align 1
+; CHECK-NEXT: [[T82:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 14
+; CHECK-NEXT: [[T83:%.*]] = load i8, i8* [[T82]], align 1
+; CHECK-NEXT: [[T84:%.*]] = xor i8 [[T83]], [[T81]]
+; CHECK-NEXT: store i8 [[T84]], i8* [[T82]], align 1
+; CHECK-NEXT: [[T85:%.*]] = load i8, i8* [[T21]], align 1
+; CHECK-NEXT: [[T86:%.*]] = getelementptr inbounds i8, i8* [[ARG2]], i64 15
+; CHECK-NEXT: [[T87:%.*]] = load i8, i8* [[T86]], align 1
+; CHECK-NEXT: [[T88:%.*]] = xor i8 [[T87]], [[T85]]
+; CHECK-NEXT: store i8 [[T88]], i8* [[T86]], align 1
+; CHECK-NEXT: ret i32 1
+;
+bb:
+ %t = alloca <16 x i8>, align 16
+ %t4 = getelementptr inbounds <16 x i8>, <16 x i8>* %t, i64 0, i64 0
+ %t5 = getelementptr inbounds i8, i8* %arg3, i64 1
+ %t6 = getelementptr inbounds i8, i8* %arg3, i64 2
+ %t7 = getelementptr inbounds i8, i8* %arg3, i64 3
+ %t8 = getelementptr inbounds i8, i8* %arg3, i64 4
+ %t9 = getelementptr inbounds i8, i8* %arg3, i64 5
+ %t10 = getelementptr inbounds i8, i8* %arg3, i64 6
+ %t11 = getelementptr inbounds i8, i8* %arg3, i64 7
+ %t12 = getelementptr inbounds i8, i8* %arg3, i64 8
+ %t13 = getelementptr inbounds i8, i8* %arg3, i64 9
+ %t14 = getelementptr inbounds i8, i8* %arg3, i64 10
+ %t15 = getelementptr inbounds i8, i8* %arg3, i64 11
+ %t16 = getelementptr inbounds i8, i8* %arg3, i64 12
+ %t17 = getelementptr inbounds i8, i8* %arg3, i64 13
+ %t18 = getelementptr inbounds i8, i8* %arg3, i64 14
+ %t19 = bitcast i8* %arg1 to <16 x i8>*
+ %A.gep.0 = getelementptr i8, i8* %A, i64 0
+ %A.0 = load i8, i8* %A.gep.0
+ %B.gep.0 = getelementptr i8, i8* %B, i64 0
+ %B.0 = load i8, i8* %B.gep.0
+ %xor.0 = xor i8 %A.0, %B.0
+ %A.gep.1 = getelementptr i8, i8* %A, i64 1
+ %A.1 = load i8, i8* %A.gep.1
+ %B.gep.1 = getelementptr i8, i8* %B, i64 1
+ %B.1 = load i8, i8* %B.gep.1
+ %xor.1 = xor i8 %A.1, %B.1
+ %A.gep.2 = getelementptr i8, i8* %A, i64 2
+ %A.2 = load i8, i8* %A.gep.2
+ %B.gep.2 = getelementptr i8, i8* %B, i64 2
+ %B.2 = load i8, i8* %B.gep.2
+ %xor.2 = xor i8 %A.2, %B.2
+ %A.gep.3 = getelementptr i8, i8* %A, i64 3
+ %A.3 = load i8, i8* %A.gep.3
+ %B.gep.3 = getelementptr i8, i8* %B, i64 3
+ %B.3 = load i8, i8* %B.gep.3
+ %xor.3 = xor i8 %A.3, %B.3
+ %A.gep.4 = getelementptr i8, i8* %A, i64 4
+ %A.4 = load i8, i8* %A.gep.4
+ %B.gep.4 = getelementptr i8, i8* %B, i64 4
+ %B.4 = load i8, i8* %B.gep.4
+ %xor.4 = xor i8 %A.4, %B.4
+ %A.gep.5 = getelementptr i8, i8* %A, i64 5
+ %A.5 = load i8, i8* %A.gep.5
+ %B.gep.5 = getelementptr i8, i8* %B, i64 5
+ %B.5 = load i8, i8* %B.gep.5
+ %xor.5 = xor i8 %A.5, %B.5
+ %A.gep.6 = getelementptr i8, i8* %A, i64 6
+ %A.6 = load i8, i8* %A.gep.6
+ %B.gep.6 = getelementptr i8, i8* %B, i64 6
+ %B.6 = load i8, i8* %B.gep.6
+ %xor.6 = xor i8 %A.6, %B.6
+ %A.gep.7 = getelementptr i8, i8* %A, i64 7
+ %A.7 = load i8, i8* %A.gep.7
+ %B.gep.7 = getelementptr i8, i8* %B, i64 7
+ %B.7 = load i8, i8* %B.gep.7
+ %xor.7 = xor i8 %A.7, %B.7
+ %A.gep.8 = getelementptr i8, i8* %A, i64 8
+ %A.8 = load i8, i8* %A.gep.8
+ %B.gep.8 = getelementptr i8, i8* %B, i64 8
+ %B.8 = load i8, i8* %B.gep.8
+ %xor.8 = xor i8 %A.8, %B.8
+ %A.gep.9 = getelementptr i8, i8* %A, i64 9
+ %A.9 = load i8, i8* %A.gep.9
+ %B.gep.9 = getelementptr i8, i8* %B, i64 9
+ %B.9 = load i8, i8* %B.gep.9
+ %xor.9 = xor i8 %A.9, %B.9
+ %A.gep.10 = getelementptr i8, i8* %A, i64 10
+ %A.10 = load i8, i8* %A.gep.10
+ %B.gep.10 = getelementptr i8, i8* %B, i64 10
+ %B.10 = load i8, i8* %B.gep.10
+ %xor.10 = xor i8 %A.10, %B.10
+ %A.gep.11 = getelementptr i8, i8* %A, i64 11
+ %A.11 = load i8, i8* %A.gep.11
+ %B.gep.11 = getelementptr i8, i8* %B, i64 11
+ %B.11 = load i8, i8* %B.gep.11
+ %xor.11 = xor i8 %A.11, %B.11
+ %A.gep.12 = getelementptr i8, i8* %A, i64 12
+ %A.12 = load i8, i8* %A.gep.12
+ %B.gep.12 = getelementptr i8, i8* %B, i64 12
+ %B.12 = load i8, i8* %B.gep.12
+ %xor.12 = xor i8 %A.12, %B.12
+ %A.gep.13 = getelementptr i8, i8* %A, i64 13
+ %A.13 = load i8, i8* %A.gep.13
+ %B.gep.13 = getelementptr i8, i8* %B, i64 13
+ %B.13 = load i8, i8* %B.gep.13
+ %xor.13 = xor i8 %A.13, %B.13
+ %A.gep.14 = getelementptr i8, i8* %A, i64 14
+ %A.14 = load i8, i8* %A.gep.14
+ %B.gep.14 = getelementptr i8, i8* %B, i64 14
+ %B.14 = load i8, i8* %B.gep.14
+ %xor.14 = xor i8 %A.14, %B.14
+ %A.gep.15 = getelementptr i8, i8* %A, i64 15
+ %A.15 = load i8, i8* %A.gep.15
+ %B.gep.15 = getelementptr i8, i8* %B, i64 15
+ %B.15 = load i8, i8* %B.gep.15
+ %xor.15 = xor i8 %A.15, %B.15
+ %R.gep.0 = getelementptr i8, i8* %arg1, i64 0
+ store i8 %xor.0, i8* %R.gep.0
+ %R.gep.1 = getelementptr i8, i8* %arg1, i64 1
+ store i8 %xor.1, i8* %R.gep.1
+ %R.gep.2 = getelementptr i8, i8* %arg1, i64 2
+ store i8 %xor.2, i8* %R.gep.2
+ %R.gep.3 = getelementptr i8, i8* %arg1, i64 3
+ store i8 %xor.3, i8* %R.gep.3
+ %R.gep.4 = getelementptr i8, i8* %arg1, i64 4
+ store i8 %xor.4, i8* %R.gep.4
+ %R.gep.5 = getelementptr i8, i8* %arg1, i64 5
+ store i8 %xor.5, i8* %R.gep.5
+ %R.gep.6 = getelementptr i8, i8* %arg1, i64 6
+ store i8 %xor.6, i8* %R.gep.6
+ %R.gep.7 = getelementptr i8, i8* %arg1, i64 7
+ store i8 %xor.7, i8* %R.gep.7
+ %R.gep.8 = getelementptr i8, i8* %arg1, i64 8
+ store i8 %xor.8, i8* %R.gep.8
+ %R.gep.9 = getelementptr i8, i8* %arg1, i64 9
+ store i8 %xor.9, i8* %R.gep.9
+ %R.gep.10 = getelementptr i8, i8* %arg1, i64 10
+ store i8 %xor.10, i8* %R.gep.10
+ %R.gep.11 = getelementptr i8, i8* %arg1, i64 11
+ store i8 %xor.11, i8* %R.gep.11
+ %R.gep.12 = getelementptr i8, i8* %arg1, i64 12
+ store i8 %xor.12, i8* %R.gep.12
+ %R.gep.13 = getelementptr i8, i8* %arg1, i64 13
+ store i8 %xor.13, i8* %R.gep.13
+ %R.gep.14 = getelementptr i8, i8* %arg1, i64 14
+ store i8 %xor.14, i8* %R.gep.14
+ %R.gep.15 = getelementptr i8, i8* %arg1, i64 15
+ store i8 %xor.15, i8* %R.gep.15
+
+
+ %t21 = getelementptr inbounds i8, i8* %arg3, i64 15
+ %t22 = bitcast i8* %arg3 to <16 x i8>*
+
+ call void @foo(i8* nonnull %t4)
+ %t26 = load i8, i8* %arg3, align 1
+ %t27 = load i8, i8* %arg2, align 1
+ %t28 = xor i8 %t27, %t26
+ store i8 %t28, i8* %arg2, align 1
+ %t29 = load i8, i8* %t5, align 1
+ %t30 = getelementptr inbounds i8, i8* %arg2, i64 1
+ %t31 = load i8, i8* %t30, align 1
+ %t32 = xor i8 %t31, %t29
+ store i8 %t32, i8* %t30, align 1
+ %t33 = load i8, i8* %t6, align 1
+ %t34 = getelementptr inbounds i8, i8* %arg2, i64 2
+ %t35 = load i8, i8* %t34, align 1
+ %t36 = xor i8 %t35, %t33
+ store i8 %t36, i8* %t34, align 1
+ %t37 = load i8, i8* %t7, align 1
+ %t38 = getelementptr inbounds i8, i8* %arg2, i64 3
+ %t39 = load i8, i8* %t38, align 1
+ %t40 = xor i8 %t39, %t37
+ store i8 %t40, i8* %t38, align 1
+ %t41 = load i8, i8* %t8, align 1
+ %t42 = getelementptr inbounds i8, i8* %arg2, i64 4
+ %t43 = load i8, i8* %t42, align 1
+ %t44 = xor i8 %t43, %t41
+ store i8 %t44, i8* %t42, align 1
+ %t45 = load i8, i8* %t9, align 1
+ %t46 = getelementptr inbounds i8, i8* %arg2, i64 5
+ %t47 = load i8, i8* %t46, align 1
+ %t48 = xor i8 %t47, %t45
+ store i8 %t48, i8* %t46, align 1
+ %t49 = load i8, i8* %t10, align 1
+ %t50 = getelementptr inbounds i8, i8* %arg2, i64 6
+ %t51 = load i8, i8* %t50, align 1
+ %t52 = xor i8 %t51, %t49
+ store i8 %t52, i8* %t50, align 1
+ %t53 = load i8, i8* %t11, align 1
+ %t54 = getelementptr inbounds i8, i8* %arg2, i64 7
+ %t55 = load i8, i8* %t54, align 1
+ %t56 = xor i8 %t55, %t53
+ store i8 %t56, i8* %t54, align 1
+ %t57 = load i8, i8* %t12, align 1
+ %t58 = getelementptr inbounds i8, i8* %arg2, i64 8
+ %t59 = load i8, i8* %t58, align 1
+ %t60 = xor i8 %t59, %t57
+ store i8 %t60, i8* %t58, align 1
+ %t61 = load i8, i8* %t13, align 1
+ %t62 = getelementptr inbounds i8, i8* %arg2, i64 9
+ %t63 = load i8, i8* %t62, align 1
+ %t64 = xor i8 %t63, %t61
+ store i8 %t64, i8* %t62, align 1
+ %t65 = load i8, i8* %t14, align 1
+ %t66 = getelementptr inbounds i8, i8* %arg2, i64 10
+ %t67 = load i8, i8* %t66, align 1
+ %t68 = xor i8 %t67, %t65
+ store i8 %t68, i8* %t66, align 1
+ %t69 = load i8, i8* %t15, align 1
+ %t70 = getelementptr inbounds i8, i8* %arg2, i64 11
+ %t71 = load i8, i8* %t70, align 1
+ %t72 = xor i8 %t71, %t69
+ store i8 %t72, i8* %t70, align 1
+ %t73 = load i8, i8* %t16, align 1
+ %t74 = getelementptr inbounds i8, i8* %arg2, i64 12
+ %t75 = load i8, i8* %t74, align 1
+ %t76 = xor i8 %t75, %t73
+ store i8 %t76, i8* %t74, align 1
+ %t77 = load i8, i8* %t17, align 1
+ %t78 = getelementptr inbounds i8, i8* %arg2, i64 13
+ %t79 = load i8, i8* %t78, align 1
+ %t80 = xor i8 %t79, %t77
+ store i8 %t80, i8* %t78, align 1
+ %t81 = load i8, i8* %t18, align 1
+ %t82 = getelementptr inbounds i8, i8* %arg2, i64 14
+ %t83 = load i8, i8* %t82, align 1
+ %t84 = xor i8 %t83, %t81
+ store i8 %t84, i8* %t82, align 1
+ %t85 = load i8, i8* %t21, align 1
+ %t86 = getelementptr inbounds i8, i8* %arg2, i64 15
+ %t87 = load i8, i8* %t86, align 1
+ %t88 = xor i8 %t87, %t85
+ store i8 %t88, i8* %t86, align 1
+ ret i32 1
+}
+
+; A test case where instructions required to compute the pointer bounds get
+; vectorized before versioning. Make sure there is no crash.
+define void @crash_instructions_deleted(float* %t, i32* %a, i32** noalias %ptr) {
+; CHECK-LABEL: @crash_instructions_deleted(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: [[T15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 2
+; CHECK-NEXT: [[T16:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[T15]] to <2 x i32>*
+; CHECK-NEXT: store <2 x i32> <i32 0, i32 10>, <2 x i32>* [[TMP0]], align 8
+; CHECK-NEXT: [[T17:%.*]] = load i32*, i32** [[PTR:%.*]], align 8
+; CHECK-NEXT: br label [[BB18:%.*]]
+; CHECK: bb18:
+; CHECK-NEXT: [[T19:%.*]] = sext i32 0 to i64
+; CHECK-NEXT: [[T20:%.*]] = add nsw i64 1, [[T19]]
+; CHECK-NEXT: [[T21:%.*]] = getelementptr inbounds i32, i32* [[T17]], i64 [[T20]]
+; CHECK-NEXT: [[T22:%.*]] = bitcast i32* [[T21]] to i8*
+; CHECK-NEXT: [[T23:%.*]] = getelementptr inbounds i8, i8* [[T22]], i64 1
+; CHECK-NEXT: [[T24:%.*]] = getelementptr inbounds i8, i8* [[T22]], i64 2
+; CHECK-NEXT: [[T25:%.*]] = getelementptr inbounds i8, i8* [[T22]], i64 3
+; CHECK-NEXT: [[T26:%.*]] = load i8, i8* [[T22]], align 1
+; CHECK-NEXT: [[T27:%.*]] = uitofp i8 [[T26]] to float
+; CHECK-NEXT: [[T28:%.*]] = fdiv float [[T27]], 2.550000e+02
+; CHECK-NEXT: [[T29:%.*]] = getelementptr inbounds float, float* [[T:%.*]], i64 0
+; CHECK-NEXT: store float [[T28]], float* [[T29]], align 8
+; CHECK-NEXT: [[T30:%.*]] = load i8, i8* [[T23]], align 1
+; CHECK-NEXT: [[T31:%.*]] = uitofp i8 [[T30]] to float
+; CHECK-NEXT: [[T32:%.*]] = fdiv float [[T31]], 2.550000e+02
+; CHECK-NEXT: [[T33:%.*]] = getelementptr inbounds float, float* [[T]], i64 1
+; CHECK-NEXT: store float [[T32]], float* [[T33]], align 4
+; CHECK-NEXT: [[T34:%.*]] = load i8, i8* [[T24]], align 1
+; CHECK-NEXT: [[T35:%.*]] = uitofp i8 [[T34]] to float
+; CHECK-NEXT: [[T36:%.*]] = fdiv float [[T35]], 2.550000e+02
+; CHECK-NEXT: [[T37:%.*]] = getelementptr inbounds float, float* [[T]], i64 2
+; CHECK-NEXT: store float [[T36]], float* [[T37]], align 8
+; CHECK-NEXT: [[T38:%.*]] = load i8, i8* [[T25]], align 1
+; CHECK-NEXT: [[T39:%.*]] = uitofp i8 [[T38]] to float
+; CHECK-NEXT: [[T40:%.*]] = fdiv float [[T39]], 2.550000e+02
+; CHECK-NEXT: [[T41:%.*]] = getelementptr inbounds float, float* [[T]], i64 3
+; CHECK-NEXT: store float [[T40]], float* [[T41]], align 4
+; CHECK-NEXT: ret void
+;
+bb:
+ %t6 = icmp slt i32 10, 0
+ %t7 = icmp sgt i32 20, 20
+ %t9 = select i1 %t7, i32 5, i32 0
+ %t10 = select i1 %t6, i32 0, i32 %t9
+ %t11 = icmp slt i32 10, 0
+ %t12 = icmp sgt i32 20, 20
+ %t13 = select i1 %t12, i32 5, i32 10
+ %t14 = select i1 %t11, i32 0, i32 %t13
+ %t15 = getelementptr inbounds i32, i32* %a, i32 2
+ store i32 %t10, i32* %t15, align 8
+ %t16 = getelementptr inbounds i32, i32* %a, i32 3
+ store i32 %t14, i32* %t16, align 4
+ %t17 = load i32*, i32** %ptr, align 8
+ br label %bb18
+
+bb18: ; preds = %bb5
+ %t19 = sext i32 %t10 to i64
+ %t20 = add nsw i64 1, %t19
+ %t21 = getelementptr inbounds i32, i32* %t17, i64 %t20
+ %t22 = bitcast i32* %t21 to i8*
+ %t23 = getelementptr inbounds i8, i8* %t22, i64 1
+ %t24 = getelementptr inbounds i8, i8* %t22, i64 2
+ %t25 = getelementptr inbounds i8, i8* %t22, i64 3
+ %t26 = load i8, i8* %t22, align 1
+ %t27 = uitofp i8 %t26 to float
+ %t28 = fdiv float %t27, 2.550000e+02
+ %t29 = getelementptr inbounds float, float* %t, i64 0
+ store float %t28, float* %t29, align 8
+ %t30 = load i8, i8* %t23, align 1
+ %t31 = uitofp i8 %t30 to float
+ %t32 = fdiv float %t31, 2.550000e+02
+ %t33 = getelementptr inbounds float, float* %t, i64 1
+ store float %t32, float* %t33, align 4
+ %t34 = load i8, i8* %t24, align 1
+ %t35 = uitofp i8 %t34 to float
+ %t36 = fdiv float %t35, 2.550000e+02
+ %t37 = getelementptr inbounds float, float* %t, i64 2
+ store float %t36, float* %t37, align 8
+ %t38 = load i8, i8* %t25, align 1
+ %t39 = uitofp i8 %t38 to float
+ %t40 = fdiv float %t39, 2.550000e+02
+ %t41 = getelementptr inbounds float, float* %t, i64 3
+ store float %t40, float* %t41, align 4
+ ret void
+}
+
+; A test case where there are no instructions accessing a tracked object in a
+; block for which versioning was requested.
+define void @crash_no_tracked_instructions(float** %arg, float* %arg.2, float* %arg.3, i1 %c) {
+; CHECK-LABEL: @crash_no_tracked_instructions(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[T19:%.*]] = load float*, float** [[ARG:%.*]], align 8
+; CHECK-NEXT: [[T20:%.*]] = load float, float* [[ARG_3:%.*]], align 4
+; CHECK-NEXT: [[T21:%.*]] = getelementptr inbounds float, float* [[ARG_2:%.*]], i64 0
+; CHECK-NEXT: br i1 [[C:%.*]], label [[BB22:%.*]], label [[BB30:%.*]]
+; CHECK: bb22:
+; CHECK-NEXT: [[T23:%.*]] = fmul float [[T20]], 9.900000e+01
+; CHECK-NEXT: [[T24:%.*]] = fmul float [[T23]], 9.900000e+01
+; CHECK-NEXT: [[T25:%.*]] = getelementptr inbounds float, float* [[T19]], i64 2
+; CHECK-NEXT: [[T26:%.*]] = fmul float [[T23]], 1.000000e+01
+; CHECK-NEXT: store float [[T26]], float* [[T25]], align 4
+; CHECK-NEXT: [[T27:%.*]] = load float, float* [[T21]], align 8
+; CHECK-NEXT: [[T28:%.*]] = fadd float [[T24]], 2.000000e+01
+; CHECK-NEXT: [[T29:%.*]] = fadd float [[T26]], 2.000000e+01
+; CHECK-NEXT: br label [[BB30]]
+; CHECK: bb30:
+; CHECK-NEXT: [[T31:%.*]] = phi float [ [[T28]], [[BB22]] ], [ 0.000000e+00, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[T32:%.*]] = phi float [ [[T29]], [[BB22]] ], [ [[T20]], [[ENTRY]] ]
+; CHECK-NEXT: br label [[BB36:%.*]]
+; CHECK: bb36:
+; CHECK-NEXT: [[T37:%.*]] = fmul float [[T31]], 3.000000e+00
+; CHECK-NEXT: [[T38:%.*]] = getelementptr inbounds float, float* [[ARG_3]], i64 0
+; CHECK-NEXT: store float [[T37]], float* [[T38]], align 4
+; CHECK-NEXT: [[T39:%.*]] = fmul float [[T32]], 3.000000e+00
+; CHECK-NEXT: [[T40:%.*]] = getelementptr inbounds float, float* [[ARG_3]], i64 1
+; CHECK-NEXT: store float [[T39]], float* [[T40]], align 4
+; CHECK-NEXT: br label [[BB41:%.*]]
+; CHECK: bb41:
+; CHECK-NEXT: ret void
+;
+entry:
+ %t19 = load float*, float** %arg
+ %t20 = load float, float* %arg.3, align 4
+ %t21 = getelementptr inbounds float, float* %arg.2, i64 0
+ br i1 %c, label %bb22, label %bb30
+
+bb22:
+ %t23 = fmul float %t20, 99.0
+ %t24 = fmul float %t23, 99.0
+ %t25 = getelementptr inbounds float, float* %t19, i64 2
+ %t26 = fmul float %t23, 10.0
+ store float %t26, float* %t25, align 4
+ %t27 = load float, float* %t21, align 8
+ %t28 = fadd float %t24, 20.0
+ %t29 = fadd float %t26, 20.0
+ br label %bb30
+
+bb30:
+ %t31 = phi float [ %t28, %bb22 ], [ 0.0, %entry ]
+ %t32 = phi float [ %t29, %bb22 ], [ %t20, %entry ]
+ br label %bb36
+
+bb36:
+ %t37 = fmul float %t31, 3.0
+ %t38 = getelementptr inbounds float, float* %arg.3, i64 0
+ store float %t37, float* %t38, align 4
+ %t39 = fmul float %t32, 3.0
+ %t40 = getelementptr inbounds float, float* %arg.3, i64 1
+ store float %t39, float* %t40, align 4
+ br label %bb41
+
+bb41:
+ ret void
+}
More information about the llvm-commits
mailing list