[llvm] 6f6e9a8 - [BasicTTIImpl][LoopUnroll] getUnrollingPreferences(): emit ORE remark when advising against unrolling due to a call in a loop
Roman Lebedev via llvm-commits
llvm-commits at lists.llvm.org
Mon Aug 2 15:04:39 PDT 2021
Author: Roman Lebedev
Date: 2021-08-03T00:57:26+03:00
New Revision: 6f6e9a867f2ace8c8b99eb8008e17dd63116bcde
URL: https://github.com/llvm/llvm-project/commit/6f6e9a867f2ace8c8b99eb8008e17dd63116bcde
DIFF: https://github.com/llvm/llvm-project/commit/6f6e9a867f2ace8c8b99eb8008e17dd63116bcde.diff
LOG: [BasicTTIImpl][LoopUnroll] getUnrollingPreferences(): emit ORE remark when advising against unrolling due to a call in a loop
I'm not sure this is the best way to approach this,
but the situation is rather not very detectable unless we explicitly call it out when refusing to advise to unroll.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D107271
Added:
llvm/test/Transforms/LoopUnroll/X86/call-remark.ll
Modified:
llvm/include/llvm/Analysis/TargetTransformInfo.h
llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
llvm/include/llvm/CodeGen/BasicTTIImpl.h
llvm/include/llvm/Transforms/Utils/UnrollLoop.h
llvm/lib/Analysis/TargetTransformInfo.cpp
llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
llvm/lib/Target/ARM/ARMTargetTransformInfo.h
llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h
llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp
llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h
llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h
llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
Removed:
################################################################################
diff --git a/llvm/include/llvm/Analysis/TargetTransformInfo.h b/llvm/include/llvm/Analysis/TargetTransformInfo.h
index 628058142e487..dd5a75fa5cac2 100644
--- a/llvm/include/llvm/Analysis/TargetTransformInfo.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfo.h
@@ -47,6 +47,7 @@ class ExtractElementInst;
class Function;
class GlobalValue;
class InstCombiner;
+class OptimizationRemarkEmitter;
class IntrinsicInst;
class LoadInst;
class LoopAccessInfo;
@@ -506,7 +507,8 @@ class TargetTransformInfo {
/// transformation. The caller will initialize UP with the current
/// target-independent defaults.
void getUnrollingPreferences(Loop *L, ScalarEvolution &,
- UnrollingPreferences &UP) const;
+ UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) const;
/// Query the target whether it would be profitable to convert the given loop
/// into a hardware loop.
@@ -1458,7 +1460,8 @@ class TargetTransformInfo::Concept {
Value *NewV) const = 0;
virtual bool isLoweredToCall(const Function *F) = 0;
virtual void getUnrollingPreferences(Loop *L, ScalarEvolution &,
- UnrollingPreferences &UP) = 0;
+ UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) = 0;
virtual void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
PeelingPreferences &PP) = 0;
virtual bool isHardwareLoopProfitable(Loop *L, ScalarEvolution &SE,
@@ -1788,8 +1791,9 @@ class TargetTransformInfo::Model final : public TargetTransformInfo::Concept {
return Impl.isLoweredToCall(F);
}
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- UnrollingPreferences &UP) override {
- return Impl.getUnrollingPreferences(L, SE, UP);
+ UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) override {
+ return Impl.getUnrollingPreferences(L, SE, UP, ORE);
}
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
PeelingPreferences &PP) override {
diff --git a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
index 449eb5513399a..4151cb1797aa8 100644
--- a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -187,7 +187,8 @@ class TargetTransformInfoImplBase {
}
void getUnrollingPreferences(Loop *, ScalarEvolution &,
- TTI::UnrollingPreferences &) const {}
+ TTI::UnrollingPreferences &,
+ OptimizationRemarkEmitter *) const {}
void getPeelingPreferences(Loop *, ScalarEvolution &,
TTI::PeelingPreferences &) const {}
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index e3b834ec42c3c..2056803484207 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -22,6 +22,7 @@
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/Analysis/LoopInfo.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/TargetTransformInfoImpl.h"
#include "llvm/CodeGen/ISDOpcodes.h"
@@ -484,7 +485,8 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
int getInlinerVectorBonusPercent() { return 150; }
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
// This unrolling functionality is target independent, but to provide some
// motivation for its intended use, for x86:
@@ -526,6 +528,15 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
continue;
}
+ if (ORE) {
+ ORE->emit([&]() {
+ return OptimizationRemark("TTI", "DontUnroll", L->getStartLoc(),
+ L->getHeader())
+ << "advising against unrolling the loop because it "
+ "contains a "
+ << ore::NV("Call", &I);
+ });
+ }
return;
}
}
diff --git a/llvm/include/llvm/Transforms/Utils/UnrollLoop.h b/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
index d95ead2def3d4..320c36b369240 100644
--- a/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
+++ b/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
@@ -117,7 +117,8 @@ MDNode *GetUnrollMetadata(MDNode *LoopID, StringRef Name);
TargetTransformInfo::UnrollingPreferences gatherUnrollingPreferences(
Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI,
- BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI, int OptLevel,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI,
+ llvm::OptimizationRemarkEmitter &ORE, int OptLevel,
Optional<unsigned> UserThreshold, Optional<unsigned> UserCount,
Optional<bool> UserAllowPartial, Optional<bool> UserRuntime,
Optional<bool> UserUpperBound, Optional<unsigned> UserFullUnrollMaxCount);
diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
index 304d24fe8e4a4..951f7d3783b97 100644
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -317,8 +317,9 @@ Optional<Value *> TargetTransformInfo::simplifyDemandedVectorEltsIntrinsic(
}
void TargetTransformInfo::getUnrollingPreferences(
- Loop *L, ScalarEvolution &SE, UnrollingPreferences &UP) const {
- return TTIImpl->getUnrollingPreferences(L, SE, UP);
+ Loop *L, ScalarEvolution &SE, UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) const {
+ return TTIImpl->getUnrollingPreferences(L, SE, UP, ORE);
}
void TargetTransformInfo::getPeelingPreferences(Loop *L, ScalarEvolution &SE,
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 8e0f06c24e2fc..6c3dd53819f94 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -1701,9 +1701,10 @@ getFalkorUnrollingPreferences(Loop *L, ScalarEvolution &SE,
}
void AArch64TTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
// Enable partial unrolling and runtime unrolling.
- BaseT::getUnrollingPreferences(L, SE, UP);
+ BaseT::getUnrollingPreferences(L, SE, UP, ORE);
// For inner loop, it is more likely to be a hot one, and the runtime check
// can be promoted out from LICM pass, so the overhead is less, let's try
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
index d55fd5b4f8158..5369eb7b7e828 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
@@ -209,7 +209,8 @@ class AArch64TTIImpl : public BasicTTIImplBase<AArch64TTIImpl> {
InstructionCost getCostOfKeepingLiveOverCall(ArrayRef<Type *> Tys);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 63f449f7a7267..d2898e149e82a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -101,7 +101,8 @@ AMDGPUTTIImpl::AMDGPUTTIImpl(const AMDGPUTargetMachine *TM, const Function &F)
TLI(ST->getTargetLowering()) {}
void AMDGPUTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
const Function &F = *L->getHeader()->getParent();
UP.Threshold = AMDGPU::getIntegerAttribute(F, "amdgpu-unroll-threshold", 300);
UP.MaxCount = std::numeric_limits<unsigned>::max();
@@ -1224,8 +1225,9 @@ unsigned GCNTTIImpl::adjustInliningThreshold(const CallBase *CB) const {
}
void GCNTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
- CommonTTI.getUnrollingPreferences(L, SE, UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
+ CommonTTI.getUnrollingPreferences(L, SE, UP, ORE);
}
void GCNTTIImpl::getPeelingPreferences(Loop *L, ScalarEvolution &SE,
@@ -1350,8 +1352,9 @@ InstructionCost R600TTIImpl::getVectorInstrCost(unsigned Opcode, Type *ValTy,
}
void R600TTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
- CommonTTI.getUnrollingPreferences(L, SE, UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
+ CommonTTI.getUnrollingPreferences(L, SE, UP, ORE);
}
void R600TTIImpl::getPeelingPreferences(Loop *L, ScalarEvolution &SE,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
index 37c0756eb7a8e..51cd42c428f75 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
@@ -53,7 +53,8 @@ class AMDGPUTTIImpl final : public BasicTTIImplBase<AMDGPUTTIImpl> {
explicit AMDGPUTTIImpl(const AMDGPUTargetMachine *TM, const Function &F);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
@@ -108,7 +109,8 @@ class GCNTTIImpl final : public BasicTTIImplBase<GCNTTIImpl> {
bool useGPUDivergenceAnalysis() const;
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
@@ -239,7 +241,8 @@ class R600TTIImpl final : public BasicTTIImplBase<R600TTIImpl> {
const AMDGPUTargetLowering *getTLI() const { return TLI; }
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
unsigned getHardwareNumberOfRegisters(bool Vec) const;
diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
index 4f3567e1390f6..d27348255e377 100644
--- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -2146,14 +2146,15 @@ bool ARMTTIImpl::emitGetActiveLaneMask() const {
return true;
}
void ARMTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
// Enable Upper bound unrolling universally, not dependant upon the conditions
// below.
UP.UpperBound = true;
// Only currently enable these preferences for M-Class cores.
if (!ST->isMClass())
- return BasicTTIImplBase::getUnrollingPreferences(L, SE, UP);
+ return BasicTTIImplBase::getUnrollingPreferences(L, SE, UP, ORE);
// Disable loop unrolling for Oz and Os.
UP.OptSizeThreshold = 0;
diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.h b/llvm/lib/Target/ARM/ARMTargetTransformInfo.h
index 889940534ce55..868c8c718277f 100644
--- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.h
+++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.h
@@ -279,7 +279,8 @@ class ARMTTIImpl : public BasicTTIImplBase<ARMTTIImpl> {
DominatorTree *DT,
const LoopAccessInfo *LAI);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
bool emitGetActiveLaneMask() const;
diff --git a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
index 108027d797546..1bdd8c3c513a1 100644
--- a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
@@ -64,7 +64,8 @@ HexagonTTIImpl::getPopcntSupport(unsigned IntTyWidthInBit) const {
// The Hexagon target can unroll loops with run-time trip counts.
void HexagonTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
UP.Runtime = UP.Partial = true;
}
diff --git a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h
index 2144fb27eb678..30470264b031a 100644
--- a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h
+++ b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h
@@ -61,7 +61,8 @@ class HexagonTTIImpl : public BasicTTIImplBase<HexagonTTIImpl> {
// The Hexagon target can unroll loops with run-time trip counts.
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp b/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp
index 20bd227b4b16e..8d1f96e4407bf 100644
--- a/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp
@@ -402,8 +402,9 @@ InstructionCost NVPTXTTIImpl::getArithmeticInstrCost(
}
void NVPTXTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
- BaseT::getUnrollingPreferences(L, SE, UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
+ BaseT::getUnrollingPreferences(L, SE, UP, ORE);
// Enable partial unrolling and runtime unrolling, but reduce the
// threshold. This partially unrolls small loops which are often
diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h b/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h
index d5a52d42ca008..2cb1405dccb53 100644
--- a/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h
+++ b/llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h
@@ -99,7 +99,8 @@ class NVPTXTTIImpl : public BasicTTIImplBase<NVPTXTTIImpl> {
const Instruction *CxtI = nullptr);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
diff --git a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
index d5a7873bd056e..3875408699e9e 100644
--- a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
@@ -787,7 +787,8 @@ bool PPCTTIImpl::isHardwareLoopProfitable(Loop *L, ScalarEvolution &SE,
}
void PPCTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
if (ST->getCPUDirective() == PPC::DIR_A2) {
// The A2 is in-order with a deep pipeline, and concatenation unrolling
// helps expose latency-hiding opportunities to the instruction scheduler.
@@ -798,7 +799,7 @@ void PPCTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
UP.AllowExpensiveTripCount = true;
}
- BaseT::getUnrollingPreferences(L, SE, UP);
+ BaseT::getUnrollingPreferences(L, SE, UP, ORE);
}
void PPCTTIImpl::getPeelingPreferences(Loop *L, ScalarEvolution &SE,
diff --git a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
index 8ac3038d51d60..8c82ee07e18f8 100644
--- a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
+++ b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
@@ -72,7 +72,8 @@ class PPCTTIImpl : public BasicTTIImplBase<PPCTTIImpl> {
TargetLibraryInfo *LibInfo);
bool getTgtMemIntrinsic(IntrinsicInst *Inst, MemIntrinsicInfo &Info);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
bool isLSRCostLess(TargetTransformInfo::LSRCost &C1,
diff --git a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
index 03c4da8495ab2..4310e4a20cfe1 100644
--- a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
@@ -243,7 +243,8 @@ SystemZTTIImpl::getPopcntSupport(unsigned TyWidth) {
}
void SystemZTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) {
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) {
// Find out if L contains a call, what the machine instruction count
// estimate is, and how many stores there are.
bool HasCall = false;
diff --git a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
index 51cf557ae99b7..77f9f0f6ac388 100644
--- a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
+++ b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
@@ -52,7 +52,8 @@ class SystemZTTIImpl : public BasicTTIImplBase<SystemZTTIImpl> {
TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth);
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP);
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE);
void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
TTI::PeelingPreferences &PP);
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
index d9bc7c6d2c3fb..f1ebcbc6fc513 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
@@ -114,7 +114,8 @@ bool WebAssemblyTTIImpl::areInlineCompatible(const Function *Caller,
}
void WebAssemblyTTIImpl::getUnrollingPreferences(
- Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) const {
+ Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) const {
// Scan the loop: don't unroll loops with calls. This is a standard approach
// for most (all?) targets.
for (BasicBlock *BB : L->blocks())
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h
index 1a33bd20d0275..670f5d7dcb1eb 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h
@@ -49,7 +49,8 @@ class WebAssemblyTTIImpl final : public BasicTTIImplBase<WebAssemblyTTIImpl> {
TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth) const;
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
- TTI::UnrollingPreferences &UP) const;
+ TTI::UnrollingPreferences &UP,
+ OptimizationRemarkEmitter *ORE) const;
/// @}
diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
index 71eb393fcdd7c..1ecbb86724e1d 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollAndJamPass.cpp
@@ -286,8 +286,8 @@ tryToUnrollAndJamLoop(Loop *L, DominatorTree &DT, LoopInfo *LI,
AssumptionCache &AC, DependenceInfo &DI,
OptimizationRemarkEmitter &ORE, int OptLevel) {
TargetTransformInfo::UnrollingPreferences UP =
- gatherUnrollingPreferences(L, SE, TTI, nullptr, nullptr, OptLevel, None,
- None, None, None, None, None);
+ gatherUnrollingPreferences(L, SE, TTI, nullptr, nullptr, ORE, OptLevel,
+ None, None, None, None, None, None);
TargetTransformInfo::PeelingPreferences PP =
gatherPeelingPreferences(L, SE, TTI, None, None);
diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
index 49501f324a49f..34dad2ac213bb 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -184,7 +184,8 @@ static const unsigned NoThreshold = std::numeric_limits<unsigned>::max();
/// flags, TTI overrides and user specified parameters.
TargetTransformInfo::UnrollingPreferences llvm::gatherUnrollingPreferences(
Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI,
- BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI, int OptLevel,
+ BlockFrequencyInfo *BFI, ProfileSummaryInfo *PSI,
+ OptimizationRemarkEmitter &ORE, int OptLevel,
Optional<unsigned> UserThreshold, Optional<unsigned> UserCount,
Optional<bool> UserAllowPartial, Optional<bool> UserRuntime,
Optional<bool> UserUpperBound, Optional<unsigned> UserFullUnrollMaxCount) {
@@ -214,7 +215,7 @@ TargetTransformInfo::UnrollingPreferences llvm::gatherUnrollingPreferences(
UP.MaxIterationsCountToAnalyze = UnrollMaxIterationsCountToAnalyze;
// Override with any target specific settings
- TTI.getUnrollingPreferences(L, SE, UP);
+ TTI.getUnrollingPreferences(L, SE, UP, &ORE);
// Apply size attributes
bool OptForSize = L->getHeader()->getParent()->hasOptSize() ||
@@ -1079,7 +1080,7 @@ static LoopUnrollResult tryToUnrollLoop(
bool NotDuplicatable;
bool Convergent;
TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences(
- L, SE, TTI, BFI, PSI, OptLevel, ProvidedThreshold, ProvidedCount,
+ L, SE, TTI, BFI, PSI, ORE, OptLevel, ProvidedThreshold, ProvidedCount,
ProvidedAllowPartial, ProvidedRuntime, ProvidedUpperBound,
ProvidedFullUnrollMaxCount);
TargetTransformInfo::PeelingPreferences PP = gatherPeelingPreferences(
diff --git a/llvm/test/Transforms/LoopUnroll/X86/call-remark.ll b/llvm/test/Transforms/LoopUnroll/X86/call-remark.ll
new file mode 100644
index 0000000000000..6b559637c206b
--- /dev/null
+++ b/llvm/test/Transforms/LoopUnroll/X86/call-remark.ll
@@ -0,0 +1,45 @@
+; RUN: opt -debugify -loop-unroll -mcpu=znver3 -pass-remarks=loop-unroll -pass-remarks-analysis=loop-unroll < %s -S 2>&1 | FileCheck --check-prefixes=ALL,UNROLL %s
+; RUN: opt -debugify -loop-unroll -mcpu=znver3 -pass-remarks=TTI -pass-remarks-analysis=TTI < %s -S 2>&1 | FileCheck --check-prefixes=ALL,TTI %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; TTI: remark: <stdin>:7:1: advising against unrolling the loop because it contains a call
+; UNROLL: remark: <stdin>:14:1: unrolled loop by a factor of 8 with run-time trip count
+
+define void @contains_external_call(i32 %count) {
+; ALL-LABEL: @contains_external_call(
+; ALL-NOT: unroll
+entry:
+ %cmp.not3 = icmp eq i32 %count, 0
+ br i1 %cmp.not3, label %for.cond.cleanup, label %for.body
+
+for.body:
+ %i.04 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
+ tail call void @sideeffect()
+ %inc = add nuw nsw i32 %i.04, 1
+ %cmp.not = icmp eq i32 %inc, %count
+ br i1 %cmp.not, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+ ret void
+}
+
+declare void @sideeffect()
+
+define i32 @no_external_calls(i32 %count) {
+; ALL-LABEL: @no_external_calls(
+; ALL: unroll
+entry:
+ %cmp.not5 = icmp eq i32 %count, 0
+ br i1 %cmp.not5, label %for.end, label %for.body
+
+for.body:
+ %i.06 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
+ %inc = add nuw nsw i32 %i.06, 1
+ %cmp.not = icmp eq i32 %inc, %count
+ br i1 %cmp.not, label %for.end, label %for.body
+
+for.end:
+ ret i32 %count
+}
More information about the llvm-commits
mailing list