diff --git a/lib/Analysis/InstructionSimplify.cpp b/lib/Analysis/InstructionSimplify.cpp index 11e24e5e3e9..097b99eb0a5 100644 --- a/lib/Analysis/InstructionSimplify.cpp +++ b/lib/Analysis/InstructionSimplify.cpp @@ -3556,14 +3556,53 @@ static bool IsIdempotent(Intrinsic::ID ID) { } template -static Value *SimplifyIntrinsic(Intrinsic::ID IID, IterTy ArgBegin, IterTy ArgEnd, +static Value *SimplifyIntrinsic(Function *F, IterTy ArgBegin, IterTy ArgEnd, const Query &Q, unsigned MaxRecurse) { + Intrinsic::ID IID = F->getIntrinsicID(); + unsigned NumOperands = std::distance(ArgBegin, ArgEnd); + Type *ReturnType = F->getReturnType(); + + // Binary Ops + if (NumOperands == 2) { + Value *LHS = *ArgBegin; + Value *RHS = *(ArgBegin + 1); + if (IID == Intrinsic::usub_with_overflow || + IID == Intrinsic::ssub_with_overflow) { + // X - X -> { 0, false } + if (LHS == RHS) + return Constant::getNullValue(ReturnType); + + // X - undef -> undef + // undef - X -> undef + if (isa(LHS) || isa(RHS)) + return UndefValue::get(ReturnType); + } + + if (IID == Intrinsic::uadd_with_overflow || + IID == Intrinsic::sadd_with_overflow) { + // X + undef -> undef + if (isa(RHS)) + return UndefValue::get(ReturnType); + } + + if (IID == Intrinsic::umul_with_overflow || + IID == Intrinsic::smul_with_overflow) { + // X * 0 -> { 0, false } + if (match(RHS, m_Zero())) + return Constant::getNullValue(ReturnType); + + // X * undef -> { 0, false } + if (match(RHS, m_Undef())) + return Constant::getNullValue(ReturnType); + } + } + // Perform idempotent optimizations if (!IsIdempotent(IID)) return nullptr; // Unary Ops - if (std::distance(ArgBegin, ArgEnd) == 1) + if (NumOperands == 1) if (IntrinsicInst *II = dyn_cast(*ArgBegin)) if (II->getIntrinsicID() == IID) return II; @@ -3587,9 +3626,8 @@ static Value *SimplifyCall(Value *V, IterTy ArgBegin, IterTy ArgEnd, if (!F) return nullptr; - if (Intrinsic::ID IID = F->getIntrinsicID()) - if (Value *Ret = - SimplifyIntrinsic(IID, ArgBegin, ArgEnd, Q, MaxRecurse)) + if (F->isIntrinsic()) + if (Value *Ret = SimplifyIntrinsic(F, ArgBegin, ArgEnd, Q, MaxRecurse)) return Ret; if (!canConstantFoldCallTo(F)) diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp index 3b925eda16b..e83b9dd36ae 100644 --- a/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -13,6 +13,7 @@ #include "InstCombineInternal.h" #include "llvm/ADT/Statistic.h" +#include "llvm/Analysis/InstructionSimplify.h" #include "llvm/Analysis/MemoryBuiltins.h" #include "llvm/IR/CallSite.h" #include "llvm/IR/Dominators.h" @@ -323,6 +324,11 @@ static Value *SimplifyX86vperm2(const IntrinsicInst &II, /// the heavy lifting. /// Instruction *InstCombiner::visitCallInst(CallInst &CI) { + auto Args = CI.arg_operands(); + if (Value *V = SimplifyCall(CI.getCalledValue(), Args.begin(), Args.end(), DL, + TLI, DT, AC)) + return ReplaceInstUsesWith(CI, V); + if (isFreeCall(&CI, TLI)) return visitFree(CI); diff --git a/lib/Transforms/InstCombine/InstCombineCompares.cpp b/lib/Transforms/InstCombine/InstCombineCompares.cpp index 09ab16511e7..2dafa58d305 100644 --- a/lib/Transforms/InstCombine/InstCombineCompares.cpp +++ b/lib/Transforms/InstCombine/InstCombineCompares.cpp @@ -2139,10 +2139,6 @@ bool InstCombiner::OptimizeOverflowCheck(OverflowCheckFlavor OCF, Value *LHS, } // FALL THROUGH uadd into sadd case OCF_SIGNED_ADD: { - // X + undef -> undef - if (isa(RHS)) - return SetResult(RHS, UndefValue::get(Builder->getInt1Ty()), false); - // X + 0 -> {X, false} if (match(RHS, m_Zero())) return SetResult(LHS, Builder->getFalse(), false); @@ -2157,14 +2153,6 @@ bool InstCombiner::OptimizeOverflowCheck(OverflowCheckFlavor OCF, Value *LHS, case OCF_UNSIGNED_SUB: case OCF_SIGNED_SUB: { - // undef - X -> undef - if (isa(LHS)) - return SetResult(LHS, UndefValue::get(Builder->getInt1Ty()), false); - - // X - undef -> undef - if (isa(RHS)) - return SetResult(RHS, UndefValue::get(Builder->getInt1Ty()), false); - // X - 0 -> {X, false} if (match(RHS, m_Zero())) return SetResult(LHS, Builder->getFalse(), false); diff --git a/test/Transforms/InstCombine/intrinsics.ll b/test/Transforms/InstCombine/intrinsics.ll index 974ee4395b8..9767704c85c 100644 --- a/test/Transforms/InstCombine/intrinsics.ll +++ b/test/Transforms/InstCombine/intrinsics.ll @@ -418,10 +418,10 @@ define %ov.result.32 @ssubtest_reorder(i8 %a) { ; CHECK-NEXT: ret %ov.result.32 %1 } -define %ov.result.32 @never_overflows_ssub(i32 %a) { +define %ov.result.32 @never_overflows_ssub_test0(i32 %a) { %x = call %ov.result.32 @llvm.ssub.with.overflow.i32(i32 %a, i32 0) ret %ov.result.32 %x -; CHECK-LABEL: @never_overflows_ssub +; CHECK-LABEL: @never_overflows_ssub_test0 ; CHECK-NEXT: %[[x:.*]] = insertvalue %ov.result.32 { i32 undef, i1 false }, i32 %a, 0 ; CHECK-NEXT: ret %ov.result.32 %[[x]] } diff --git a/test/Transforms/InstSimplify/call.ll b/test/Transforms/InstSimplify/call.ll index ac94e75f489..465529aef66 100644 --- a/test/Transforms/InstSimplify/call.ll +++ b/test/Transforms/InstSimplify/call.ll @@ -1,6 +1,9 @@ ; RUN: opt < %s -instsimplify -S | FileCheck %s declare {i8, i1} @llvm.uadd.with.overflow.i8(i8 %a, i8 %b) +declare {i8, i1} @llvm.usub.with.overflow.i8(i8 %a, i8 %b) +declare {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a, i8 %b) +declare {i8, i1} @llvm.umul.with.overflow.i8(i8 %a, i8 %b) define i1 @test_uadd1() { ; CHECK-LABEL: @test_uadd1( @@ -18,6 +21,27 @@ define i8 @test_uadd2() { ; CHECK-NEXT: ret i8 42 } +define {i8, i1} @test_usub1(i8 %V) { +; CHECK-LABEL: @test_usub1( + %x = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %V, i8 %V) + ret {i8, i1} %x +; CHECK-NEXT: ret { i8, i1 } zeroinitializer +} + +define {i8, i1} @test_ssub1(i8 %V) { +; CHECK-LABEL: @test_ssub1( + %x = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %V, i8 %V) + ret {i8, i1} %x +; CHECK-NEXT: ret { i8, i1 } zeroinitializer +} + +define {i8, i1} @test_umul1(i8 %V) { +; CHECK-LABEL: @test_umul1( + %x = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %V, i8 0) + ret {i8, i1} %x +; CHECK-NEXT: ret { i8, i1 } zeroinitializer +} + declare i256 @llvm.cttz.i256(i256 %src, i1 %is_zero_undef) define i256 @test_cttz() {