diff --git a/llvm/lib/Analysis/LoopUnrollAnalyzer.cpp b/llvm/lib/Analysis/LoopUnrollAnalyzer.cpp index d8fe92ab14cc..4b8f5d7543f7 100644 --- a/llvm/lib/Analysis/LoopUnrollAnalyzer.cpp +++ b/llvm/lib/Analysis/LoopUnrollAnalyzer.cpp @@ -13,6 +13,7 @@ //===----------------------------------------------------------------------===// #include "llvm/Analysis/LoopUnrollAnalyzer.h" +#include "llvm/Analysis/ConstantFolding.h" #include "llvm/Analysis/InstructionSimplify.h" #include "llvm/Analysis/LoopInfo.h" #include "llvm/Analysis/ScalarEvolutionExpressions.h" @@ -105,7 +106,6 @@ bool UnrolledInstAnalyzer::visitLoad(LoadInst &I) { auto AddressIt = SimplifiedAddresses.find(AddrOp); if (AddressIt == SimplifiedAddresses.end()) return false; - const APInt &SimplifiedAddrOp = AddressIt->second.Offset; auto *GV = dyn_cast(AddressIt->second.Base); // We're only interested in loads that can be completely folded to a @@ -113,37 +113,13 @@ bool UnrolledInstAnalyzer::visitLoad(LoadInst &I) { if (!GV || !GV->hasDefinitiveInitializer() || !GV->isConstant()) return false; - ConstantDataSequential *CDS = - dyn_cast(GV->getInitializer()); - if (!CDS) + Constant *Res = + ConstantFoldLoadFromConst(GV->getInitializer(), I.getType(), + AddressIt->second.Offset, I.getDataLayout()); + if (!Res) return false; - // We might have a vector load from an array. FIXME: for now we just bail - // out in this case, but we should be able to resolve and simplify such - // loads. - if (CDS->getElementType() != I.getType()) - return false; - - unsigned ElemSize = CDS->getElementType()->getPrimitiveSizeInBits() / 8U; - if (SimplifiedAddrOp.getActiveBits() > 64) - return false; - int64_t SimplifiedAddrOpV = SimplifiedAddrOp.getSExtValue(); - if (SimplifiedAddrOpV < 0) { - // FIXME: For now we conservatively ignore out of bound accesses, but - // we're allowed to perform the optimization in this case. - return false; - } - uint64_t Index = static_cast(SimplifiedAddrOpV) / ElemSize; - if (Index >= CDS->getNumElements()) { - // FIXME: For now we conservatively ignore out of bound accesses, but - // we're allowed to perform the optimization in this case. - return false; - } - - Constant *CV = CDS->getElementAsConstant(Index); - assert(CV && "Constant expected."); - SimplifiedValues[&I] = CV; - + SimplifiedValues[&I] = Res; return true; } diff --git a/llvm/test/Transforms/LoopUnroll/full-unroll-heuristics-load-const.ll b/llvm/test/Transforms/LoopUnroll/full-unroll-heuristics-load-const.ll new file mode 100644 index 000000000000..241ed2824b8f --- /dev/null +++ b/llvm/test/Transforms/LoopUnroll/full-unroll-heuristics-load-const.ll @@ -0,0 +1,39 @@ +; RUN: opt < %s -S -passes=loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=10 -unroll-max-percent-threshold-boost=100 | FileCheck %s -check-prefix=TEST1 +; RUN: opt < %s -S -passes=loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=20 -unroll-max-percent-threshold-boost=200 | FileCheck %s -check-prefix=TEST2 +; RUN: opt < %s -S -passes=loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=20 -unroll-max-percent-threshold-boost=100 | FileCheck %s -check-prefix=TEST3 + +; This test is a copy of full-unroll-heuristics.ll but with the constant +; wrapped in an extra struct. This should not hinder the analysis. + +; If the absolute threshold is too low, we should not unroll: +; TEST1: %array_const_idx = getelementptr inbounds { [9 x i32] }, ptr @known_constant, i64 0, i32 0, i64 %iv + +; Otherwise, we should: +; TEST2-NOT: %array_const_idx = getelementptr inbounds { [9 x i32] }, ptr @known_constant, i64 0, i32 0, i64 %iv + +; If we do not boost threshold, the unroll will not happen: +; TEST3: %array_const_idx = getelementptr inbounds { [9 x i32] }, ptr @known_constant, i64 0, i32 0, i64 %iv + +@known_constant = internal unnamed_addr constant { [9 x i32] } { [9 x i32] [i32 0, i32 -1, i32 0, i32 -1, i32 5, i32 -1, i32 0, i32 -1, i32 0] }, align 16 + +define i32 @foo(ptr noalias nocapture readonly %src) { +entry: + br label %loop + +loop: ; preds = %loop, %entry + %iv = phi i64 [ 0, %entry ], [ %inc, %loop ] + %r = phi i32 [ 0, %entry ], [ %add, %loop ] + %arrayidx = getelementptr inbounds i32, ptr %src, i64 %iv + %src_element = load i32, ptr %arrayidx, align 4 + %array_const_idx = getelementptr inbounds { [9 x i32] }, ptr @known_constant, i64 0, i32 0, i64 %iv + %const_array_element = load i32, ptr %array_const_idx, align 4 + %mul = mul nsw i32 %src_element, %const_array_element + %add = add nsw i32 %mul, %r + %inc = add nuw nsw i64 %iv, 1 + %exitcond86.i = icmp eq i64 %inc, 9 + br i1 %exitcond86.i, label %loop.end, label %loop + +loop.end: ; preds = %loop + %r.lcssa = phi i32 [ %r, %loop ] + ret i32 %r.lcssa +}