Commit c2266463 authored by Austin Kerbow's avatar Austin Kerbow
Browse files

Resubmit: [DA][TTI][AMDGPU] Add option to select GPUDA with TTI

Summary:
Enable the new diveregence analysis by default for AMDGPU.

Resubmit with test updates since GPUDA was causing failures on Windows.

Reviewers: rampitec, nhaehnle, arsenm, thakis

Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D73315
parent c5bd3d07
......@@ -54,7 +54,8 @@ public:
private:
// Whether analysis should be performed by GPUDivergenceAnalysis.
bool shouldUseGPUDivergenceAnalysis(const Function &F) const;
bool shouldUseGPUDivergenceAnalysis(const Function &F,
const TargetTransformInfo &TTI) const;
// (optional) handle to new DivergenceAnalysis
std::unique_ptr<GPUDivergenceAnalysis> gpuDA;
......
......@@ -342,6 +342,10 @@ public:
/// branches.
bool hasBranchDivergence() const;
/// Return true if the target prefers to use GPU divergence analysis to
/// replace the legacy version.
bool useGPUDivergenceAnalysis() const;
/// Returns whether V is a source of divergence.
///
/// This function provides the target-dependent information for
......@@ -1198,6 +1202,7 @@ public:
virtual int
getUserCost(const User *U, ArrayRef<const Value *> Operands) = 0;
virtual bool hasBranchDivergence() = 0;
virtual bool useGPUDivergenceAnalysis() = 0;
virtual bool isSourceOfDivergence(const Value *V) = 0;
virtual bool isAlwaysUniform(const Value *V) = 0;
virtual unsigned getFlatAddressSpace() = 0;
......@@ -1452,6 +1457,7 @@ public:
return Impl.getUserCost(U, Operands);
}
bool hasBranchDivergence() override { return Impl.hasBranchDivergence(); }
bool useGPUDivergenceAnalysis() override { return Impl.useGPUDivergenceAnalysis(); }
bool isSourceOfDivergence(const Value *V) override {
return Impl.isSourceOfDivergence(V);
}
......
......@@ -152,6 +152,8 @@ public:
bool hasBranchDivergence() { return false; }
bool useGPUDivergenceAnalysis() { return false; }
bool isSourceOfDivergence(const Value *V) { return false; }
bool isAlwaysUniform(const Value *V) { return false; }
......
......@@ -207,6 +207,8 @@ public:
bool hasBranchDivergence() { return false; }
bool useGPUDivergenceAnalysis() { return false; }
bool isSourceOfDivergence(const Value *V) { return false; }
bool isAlwaysUniform(const Value *V) { return false; }
......
......@@ -301,14 +301,13 @@ FunctionPass *llvm::createLegacyDivergenceAnalysisPass() {
void LegacyDivergenceAnalysis::getAnalysisUsage(AnalysisUsage &AU) const {
AU.addRequired<DominatorTreeWrapperPass>();
AU.addRequired<PostDominatorTreeWrapperPass>();
if (UseGPUDA)
AU.addRequired<LoopInfoWrapperPass>();
AU.addRequired<LoopInfoWrapperPass>();
AU.setPreservesAll();
}
bool LegacyDivergenceAnalysis::shouldUseGPUDivergenceAnalysis(
const Function &F) const {
if (!UseGPUDA)
const Function &F, const TargetTransformInfo &TTI) const {
if (!(UseGPUDA || TTI.useGPUDivergenceAnalysis()))
return false;
// GPUDivergenceAnalysis requires a reducible CFG.
......@@ -337,7 +336,7 @@ bool LegacyDivergenceAnalysis::runOnFunction(Function &F) {
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
auto &PDT = getAnalysis<PostDominatorTreeWrapperPass>().getPostDomTree();
if (shouldUseGPUDivergenceAnalysis(F)) {
if (shouldUseGPUDivergenceAnalysis(F, TTI)) {
// run the new GPU divergence analysis
auto &LI = getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
gpuDA = std::make_unique<GPUDivergenceAnalysis>(F, DT, PDT, LI, TTI);
......
......@@ -212,6 +212,10 @@ bool TargetTransformInfo::hasBranchDivergence() const {
return TTIImpl->hasBranchDivergence();
}
bool TargetTransformInfo::useGPUDivergenceAnalysis() const {
return TTIImpl->useGPUDivergenceAnalysis();
}
bool TargetTransformInfo::isSourceOfDivergence(const Value *V) const {
return TTIImpl->isSourceOfDivergence(V);
}
......
......@@ -69,6 +69,11 @@ static cl::opt<unsigned> UnrollThresholdIf(
cl::desc("Unroll threshold increment for AMDGPU for each if statement inside loop"),
cl::init(150), cl::Hidden);
static cl::opt<bool> UseLegacyDA(
"amdgpu-use-legacy-divergence-analysis",
cl::desc("Enable legacy divergence analysis for AMDGPU"),
cl::init(false), cl::Hidden);
static bool dependsOnLocalPhi(const Loop *L, const Value *Cond,
unsigned Depth = 0) {
const Instruction *I = dyn_cast<Instruction>(Cond);
......@@ -601,6 +606,11 @@ static bool isArgPassedInSGPR(const Argument *A) {
}
}
/// \returns true if the new GPU divergence analysis is enabled.
bool GCNTTIImpl::useGPUDivergenceAnalysis() const {
return !UseLegacyDA;
}
/// \returns true if the result of the value could potentially be
/// different across workitems in a wavefront.
bool GCNTTIImpl::isSourceOfDivergence(const Value *V) const {
......
......@@ -136,6 +136,7 @@ public:
HasFP32Denormals(ST->hasFP32Denormals(F)) { }
bool hasBranchDivergence() { return true; }
bool useGPUDivergenceAnalysis() const;
void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
TTI::UnrollingPreferences &UP);
......
; RUN: opt -mtriple=amdgcn-- -analyze -divergence %s | FileCheck %s
; RUN: opt -mtriple=amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
; CHECK: DIVERGENT: %orig = atomicrmw xchg i32* %ptr, i32 %val seq_cst
define i32 @test1(i32* %ptr, i32 %val) #0 {
......
; RUN: opt -mtriple=amdgcn-- -analyze -divergence %s | FileCheck %s
; RUN: opt -mtriple=amdgcn-- -analyze -amdgpu-use-legacy-divergence-analysis -divergence %s | FileCheck %s
; CHECK: DIVERGENT: %swizzle = call i32 @llvm.amdgcn.ds.swizzle(i32 %src, i32 100) #0
define amdgpu_kernel void @ds_swizzle(i32 addrspace(1)* %out, i32 %src) #0 {
......
; RUN: opt %s -mtriple amdgcn-- -analyze -divergence | FileCheck %s
; RUN: opt %s -mtriple amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence | FileCheck %s
; CHECK-LABEL: Printing analysis 'Legacy Divergence Analysis' for function 'test_amdgpu_ps':
; CHECK: DIVERGENT:
......
;RUN: opt -mtriple=amdgcn-mesa-mesa3d -analyze -divergence %s | FileCheck %s
;RUN: opt -mtriple=amdgcn-mesa-mesa3d -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
;CHECK: DIVERGENT: %orig = call i32 @llvm.amdgcn.buffer.atomic.swap.i32(
define float @buffer_atomic_swap(<4 x i32> inreg %rsrc, i32 inreg %data) #0 {
......
;RUN: opt -mtriple=amdgcn-mesa-mesa3d -analyze -divergence %s | FileCheck %s
;RUN: opt -mtriple=amdgcn-mesa-mesa3d -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
;CHECK: DIVERGENT: %orig = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(
define float @image_atomic_swap(<8 x i32> inreg %rsrc, i32 inreg %addr, i32 inreg %data) #0 {
......
; RUN: opt -mtriple=amdgcn-- -analyze -divergence %s | FileCheck %s
; RUN: opt -mtriple=amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
; Test that we consider loads from flat and private addrspaces to be divergent.
......
; RUN: opt %s -mtriple amdgcn-- -analyze -divergence | FileCheck %s
; RUN: opt %s -mtriple amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence | FileCheck %s
; CHECK: DIVERGENT: %tmp5 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp2
; CHECK: DIVERGENT: %tmp10 = load volatile float, float addrspace(1)* %tmp5, align 4
......
; RUN: opt -mtriple=amdgcn-- -analyze -divergence %s | FileCheck %s
; RUN: opt -mtriple=amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
; CHECK-LABEL: 'test1':
; CHECK-NEXT: DIVERGENT: i32 %bound
......
; RUN: opt %s -mtriple amdgcn-- -analyze -divergence | FileCheck %s
; RUN: opt %s -mtriple amdgcn-- -amdgpu-use-legacy-divergence-analysis -analyze -divergence | FileCheck %s
; CHECK: DIVERGENT: %tmp = cmpxchg volatile
define amdgpu_kernel void @unreachable_loop(i32 %tidx) #0 {
......
; RUN: opt -mtriple amdgcn-unknown-amdhsa -analyze -divergence %s | FileCheck %s
; RUN: opt -mtriple amdgcn-unknown-amdhsa -amdgpu-use-legacy-divergence-analysis -analyze -divergence %s | FileCheck %s
declare i32 @llvm.amdgcn.workitem.id.x() #0
declare i32 @llvm.amdgcn.workitem.id.y() #0
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment