Index: llvm/lib/Analysis/MemoryLocation.cpp =================================================================== --- llvm/lib/Analysis/MemoryLocation.cpp +++ llvm/lib/Analysis/MemoryLocation.cpp @@ -222,9 +222,10 @@ return MemoryLocation(Arg, LocationSize::precise(LenCI->getZExtValue()), AATags); break; + case LibFunc_bcmp: case LibFunc_memcmp: assert((ArgIdx == 0 || ArgIdx == 1) && - "Invalid argument index for memcmp"); + "Invalid argument index for memcmp/bcmp"); if (const ConstantInt *LenCI = dyn_cast(Call->getArgOperand(2))) return MemoryLocation(Arg, LocationSize::precise(LenCI->getZExtValue()), Index: llvm/test/Analysis/BasicAA/libfuncs.ll =================================================================== --- llvm/test/Analysis/BasicAA/libfuncs.ll +++ llvm/test/Analysis/BasicAA/libfuncs.ll @@ -1,7 +1,5 @@ -; RUN: opt -inferattrs -basic-aa -aa-eval -print-all-alias-modref-info -disable-output 2>&1 %s | FileCheck %s - -target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64" -target triple = "x86_64-apple-macosx10.7" +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; RUN: opt -mtriple=i386-pc-linux-gnu -inferattrs -basic-aa -aa-eval -print-all-alias-modref-info -disable-output 2>&1 %s | FileCheck %s ; CHECK-LABEL: Function: test_memcmp_const_size ; CHECK: Just Ref: Ptr: i8* %a <-> %res = tail call i32 @memcmp(i8* %a, i8* %b, i64 4) @@ -45,5 +43,48 @@ ret i32 %res } - declare i32 @memcmp(i8*, i8*, i64) +declare i32 @bcmp(i8*, i8*, i64) + +; CHECK-LABEL: Function: test_bcmp_const_size +; CHECK: Just Ref: Ptr: i8* %a <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +; CHECK-NEXT: Just Ref: Ptr: i8* %b <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +; CHECK-NEXT: Just Ref: Ptr: i8* %a.gep.1 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +; CHECK-NEXT: NoModRef: Ptr: i8* %a.gep.5 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +; CHECK-NEXT: Just Ref: Ptr: i8* %b.gep.1 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +; CHECK-NEXT: NoModRef: Ptr: i8* %b.gep.5 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) +define i32 @test_bcmp_const_size(i8* noalias %a, i8* noalias %b) { +entry: + %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 4) + %a.gep.1 = getelementptr i8, i8* %a, i32 1 + store i8 0, i8* %a.gep.1 + %a.gep.5 = getelementptr i8, i8* %a, i32 5 + store i8 1, i8* %a.gep.5 + %b.gep.1 = getelementptr i8, i8* %b, i32 1 + store i8 2, i8* %b.gep.1 + %b.gep.5 = getelementptr i8, i8* %b, i32 5 + store i8 3, i8* %b.gep.5 + ret i32 %res +} + +; CHECK-LABEL: Function: test_bcmp_variable_size +; CHECK: Just Ref: Ptr: i8* %a <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +; CHECK-NEXT: Just Ref: Ptr: i8* %b <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +; CHECK-NEXT: Just Ref: Ptr: i8* %a.gep.1 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +; CHECK-NEXT: Just Ref: Ptr: i8* %a.gep.5 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +; CHECK-NEXT: Just Ref: Ptr: i8* %b.gep.1 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +; CHECK-NEXT: Just Ref: Ptr: i8* %b.gep.5 <-> %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) +define i32 @test_bcmp_variable_size(i8* noalias %a, i8* noalias %b, i64 %n) { +entry: + %res = tail call i32 @bcmp(i8* %a, i8* %b, i64 %n) + %a.gep.1 = getelementptr i8, i8* %a, i32 1 + store i8 0, i8* %a.gep.1 + %a.gep.5 = getelementptr i8, i8* %a, i32 5 + store i8 1, i8* %a.gep.5 + %b.gep.1 = getelementptr i8, i8* %b, i32 1 + store i8 2, i8* %b.gep.1 + %b.gep.5 = getelementptr i8, i8* %b, i32 5 + store i8 3, i8* %b.gep.5 + ret i32 %res +} + Index: llvm/test/Transforms/DeadStoreElimination/MSSA/libcalls.ll =================================================================== --- llvm/test/Transforms/DeadStoreElimination/MSSA/libcalls.ll +++ llvm/test/Transforms/DeadStoreElimination/MSSA/libcalls.ll @@ -130,3 +130,64 @@ %res = call i32 @memcmp(i8* nonnull %foo, i8* nonnull %stack.ptr, i64 %n) ret i32 %res } + + +declare i32 @bcmp(i8*, i8*, i64) + +define i1 @test_bcmp_const_size(i8* noalias %foo) { +; CHECK-LABEL: @test_bcmp_const_size( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[STACK:%.*]] = alloca [10 x i8], align 1 +; CHECK-NEXT: [[STACK_PTR:%.*]] = bitcast [10 x i8]* [[STACK]] to i8* +; CHECK-NEXT: store i8 49, i8* [[STACK_PTR]], align 1 +; CHECK-NEXT: [[GEP_1:%.*]] = getelementptr i8, i8* [[STACK_PTR]], i64 1 +; CHECK-NEXT: store i8 50, i8* [[GEP_1]], align 1 +; CHECK-NEXT: [[CALL:%.*]] = call i32 @bcmp(i8* nonnull dereferenceable(2) [[FOO:%.*]], i8* nonnull dereferenceable(2) [[STACK_PTR]], i64 2) +; CHECK-NEXT: [[RES:%.*]] = icmp eq i32 [[CALL]], 0 +; CHECK-NEXT: ret i1 [[RES]] +; +entry: + %stack = alloca [10 x i8] + %stack.ptr = bitcast [10 x i8]* %stack to i8* + store i8 49, i8* %stack.ptr, align 1 + %gep.1 = getelementptr i8, i8* %stack.ptr, i64 1 + store i8 50, i8* %gep.1, align 1 + %gep.2 = getelementptr i8, i8* %stack.ptr, i64 2 + store i8 51, i8* %gep.2, align 1 + %gep.3 = getelementptr i8, i8* %stack.ptr, i64 3 + store i8 52, i8* %gep.3, align 1 + %call = call i32 @bcmp(i8* nonnull dereferenceable(2) %foo, i8* nonnull dereferenceable(2) %stack.ptr, i64 2) + %res = icmp eq i32 %call, 0 + ret i1 %res +} + +define i1 @test_bcmp_variable_size(i8* noalias %foo, i64 %n) { +; CHECK-LABEL: @test_bcmp_variable_size( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[STACK:%.*]] = alloca [10 x i8], align 1 +; CHECK-NEXT: [[STACK_PTR:%.*]] = bitcast [10 x i8]* [[STACK]] to i8* +; CHECK-NEXT: store i8 49, i8* [[STACK_PTR]], align 1 +; CHECK-NEXT: [[GEP_1:%.*]] = getelementptr i8, i8* [[STACK_PTR]], i64 1 +; CHECK-NEXT: store i8 50, i8* [[GEP_1]], align 1 +; CHECK-NEXT: [[GEP_2:%.*]] = getelementptr i8, i8* [[STACK_PTR]], i64 2 +; CHECK-NEXT: store i8 51, i8* [[GEP_2]], align 1 +; CHECK-NEXT: [[GEP_3:%.*]] = getelementptr i8, i8* [[STACK_PTR]], i64 3 +; CHECK-NEXT: store i8 52, i8* [[GEP_3]], align 1 +; CHECK-NEXT: [[CALL:%.*]] = call i32 @bcmp(i8* nonnull [[FOO:%.*]], i8* nonnull [[STACK_PTR]], i64 [[N:%.*]]) +; CHECK-NEXT: [[RES:%.*]] = icmp eq i32 [[CALL]], 0 +; CHECK-NEXT: ret i1 [[RES]] +; +entry: + %stack = alloca [10 x i8] + %stack.ptr = bitcast [10 x i8]* %stack to i8* + store i8 49, i8* %stack.ptr, align 1 + %gep.1 = getelementptr i8, i8* %stack.ptr, i64 1 + store i8 50, i8* %gep.1, align 1 + %gep.2 = getelementptr i8, i8* %stack.ptr, i64 2 + store i8 51, i8* %gep.2, align 1 + %gep.3 = getelementptr i8, i8* %stack.ptr, i64 3 + store i8 52, i8* %gep.3, align 1 + %call = call i32 @bcmp(i8* nonnull %foo, i8* nonnull %stack.ptr, i64 %n) + %res = icmp eq i32 %call, 0 + ret i1 %res +}