RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2024-12-13 23:18:51 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
550eae18a9
llvm/unittests/Analysis/ScalarEvolutionTest.cpp
Chandler Carruth bfe1f1c5a3 [PM] Port ScalarEvolution to the new pass manager. 
This change makes ScalarEvolution a stand-alone object and just produces
one from a pass as needed. Making this work well requires making the
object movable, using references instead of overwritten pointers in
a number of places, and other refactorings.

I've also wired it up to the new pass manager and added a RUN line to
a test to exercise it under the new pass manager. This includes basic
printing support much like with other analyses.

But there is a big and somewhat scary change here. Prior to this patch
ScalarEvolution was never *actually* invalidated!!! Re-running the pass
just re-wired up the various other analyses and didn't remove any of the
existing entries in the SCEV caches or clear out anything at all. This
might seem OK as everything in SCEV that can uses ValueHandles to track
updates to the values that serve as SCEV keys. However, this still means
that as we ran SCEV over each function in the module, we kept
accumulating more and more SCEVs into the cache. At the end, we would
have a SCEV cache with every value that we ever needed a SCEV for in the
entire module!!! Yowzers. The releaseMemory routine would dump all of
this, but that isn't realy called during normal runs of the pipeline as
far as I can see.

To make matters worse, there *is* actually a key that we don't update
with value handles -- there is a map keyed off of Loop*s. Because
LoopInfo *does* release its memory from run to run, it is entirely
possible to run SCEV over one function, then over another function, and
then lookup a Loop* from the second function but find an entry inserted
for the first function! Ouch.

To make matters still worse, there are plenty of updates that *don't*
trip a value handle. It seems incredibly unlikely that today GVN or
another pass that invalidates SCEV can update values in *just* such
a way that a subsequent run of SCEV will incorrectly find lookups in
a cache, but it is theoretically possible and would be a nightmare to
debug.

With this refactoring, I've fixed all this by actually destroying and
recreating the ScalarEvolution object from run to run. Technically, this
could increase the amount of malloc traffic we see, but then again it is
also technically correct. ;] I don't actually think we're suffering from
tons of malloc traffic from SCEV because if we were, the fact that we
never clear the memory would seem more likely to have come up as an
actual problem before now. So, I've made the simple fix here. If in fact
there are serious issues with too much allocation and deallocation,
I can work on a clever fix that preserves the allocations (while
clearing the data) between each run, but I'd prefer to do that kind of
optimization with a test case / benchmark that shows why we need such
cleverness (and that can test that we actually make it faster). It's
possible that this will make some things faster by making the SCEV
caches have higher locality (due to being significantly smaller) so
until there is a clear benchmark, I think the simple change is best.

Differential Revision: http://reviews.llvm.org/D12063

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@245193 91177308-0d34-0410-b5e6-96231b3b80d8
2015-08-17 02:08:17 +00:00

242 lines
9.7 KiB
C++
Raw Blame History

//===- ScalarEvolutionsTest.cpp - ScalarEvolution unit tests --------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/LegacyPassManager.h"
#include "gtest/gtest.h"
namespace llvm {
namespace {
// We use this fixture to ensure that we clean up ScalarEvolution before
// deleting the PassManager.
class ScalarEvolutionsTest : public testing::Test {
protected:
LLVMContext Context;
Module M;
TargetLibraryInfoImpl TLII;
TargetLibraryInfo TLI;
std::unique_ptr<AssumptionCache> AC;
std::unique_ptr<DominatorTree> DT;
std::unique_ptr<LoopInfo> LI;
ScalarEvolutionsTest() : M("", Context), TLII(), TLI(TLII) {}
ScalarEvolution buildSE(Function &F) {
AC.reset(new AssumptionCache(F));
DT.reset(new DominatorTree(F));
LI.reset(new LoopInfo(*DT));
return ScalarEvolution(F, TLI, *AC, *DT, *LI);
}
};
TEST_F(ScalarEvolutionsTest, SCEVUnknownRAUW) {
FunctionType *FTy = FunctionType::get(Type::getVoidTy(Context),
std::vector<Type *>(), false);
Function *F = cast<Function>(M.getOrInsertFunction("f", FTy));
BasicBlock *BB = BasicBlock::Create(Context, "entry", F);
ReturnInst::Create(Context, nullptr, BB);
Type *Ty = Type::getInt1Ty(Context);
Constant *Init = Constant::getNullValue(Ty);
Value *V0 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V0");
Value *V1 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V1");
Value *V2 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V2");
ScalarEvolution SE = buildSE(*F);
const SCEV *S0 = SE.getSCEV(V0);
const SCEV *S1 = SE.getSCEV(V1);
const SCEV *S2 = SE.getSCEV(V2);
const SCEV *P0 = SE.getAddExpr(S0, S0);
const SCEV *P1 = SE.getAddExpr(S1, S1);
const SCEV *P2 = SE.getAddExpr(S2, S2);
const SCEVMulExpr *M0 = cast<SCEVMulExpr>(P0);
const SCEVMulExpr *M1 = cast<SCEVMulExpr>(P1);
const SCEVMulExpr *M2 = cast<SCEVMulExpr>(P2);
EXPECT_EQ(cast<SCEVConstant>(M0->getOperand(0))->getValue()->getZExtValue(),
2u);
EXPECT_EQ(cast<SCEVConstant>(M1->getOperand(0))->getValue()->getZExtValue(),
2u);
EXPECT_EQ(cast<SCEVConstant>(M2->getOperand(0))->getValue()->getZExtValue(),
2u);
// Before the RAUWs, these are all pointing to separate values.
EXPECT_EQ(cast<SCEVUnknown>(M0->getOperand(1))->getValue(), V0);
EXPECT_EQ(cast<SCEVUnknown>(M1->getOperand(1))->getValue(), V1);
EXPECT_EQ(cast<SCEVUnknown>(M2->getOperand(1))->getValue(), V2);
// Do some RAUWs.
V2->replaceAllUsesWith(V1);
V1->replaceAllUsesWith(V0);
// After the RAUWs, these should all be pointing to V0.
EXPECT_EQ(cast<SCEVUnknown>(M0->getOperand(1))->getValue(), V0);
EXPECT_EQ(cast<SCEVUnknown>(M1->getOperand(1))->getValue(), V0);
EXPECT_EQ(cast<SCEVUnknown>(M2->getOperand(1))->getValue(), V0);
}
TEST_F(ScalarEvolutionsTest, SCEVMultiplyAddRecs) {
Type *Ty = Type::getInt32Ty(Context);
SmallVector<Type *, 10> Types;
Types.append(10, Ty);
FunctionType *FTy = FunctionType::get(Type::getVoidTy(Context), Types, false);
Function *F = cast<Function>(M.getOrInsertFunction("f", FTy));
BasicBlock *BB = BasicBlock::Create(Context, "entry", F);
ReturnInst::Create(Context, nullptr, BB);
ScalarEvolution SE = buildSE(*F);
// It's possible to produce an empty loop through the default constructor,
// but you can't add any blocks to it without a LoopInfo pass.
Loop L;
const_cast<std::vector<BasicBlock*>&>(L.getBlocks()).push_back(BB);
Function::arg_iterator AI = F->arg_begin();
SmallVector<const SCEV *, 5> A;
A.push_back(SE.getSCEV(&*AI++));
A.push_back(SE.getSCEV(&*AI++));
A.push_back(SE.getSCEV(&*AI++));
A.push_back(SE.getSCEV(&*AI++));
A.push_back(SE.getSCEV(&*AI++));
const SCEV *A_rec = SE.getAddRecExpr(A, &L, SCEV::FlagAnyWrap);
SmallVector<const SCEV *, 5> B;
B.push_back(SE.getSCEV(&*AI++));
B.push_back(SE.getSCEV(&*AI++));
B.push_back(SE.getSCEV(&*AI++));
B.push_back(SE.getSCEV(&*AI++));
B.push_back(SE.getSCEV(&*AI++));
const SCEV *B_rec = SE.getAddRecExpr(B, &L, SCEV::FlagAnyWrap);
/* Spot check that we perform this transformation:
{A0,+,A1,+,A2,+,A3,+,A4} * {B0,+,B1,+,B2,+,B3,+,B4} =
{A0*B0,+,
A1*B0 + A0*B1 + A1*B1,+,
A2*B0 + 2A1*B1 + A0*B2 + 2A2*B1 + 2A1*B2 + A2*B2,+,
A3*B0 + 3A2*B1 + 3A1*B2 + A0*B3 + 3A3*B1 + 6A2*B2 + 3A1*B3 + 3A3*B2 +
3A2*B3 + A3*B3,+,
A4*B0 + 4A3*B1 + 6A2*B2 + 4A1*B3 + A0*B4 + 4A4*B1 + 12A3*B2 + 12A2*B3 +
4A1*B4 + 6A4*B2 + 12A3*B3 + 6A2*B4 + 4A4*B3 + 4A3*B4 + A4*B4,+,
5A4*B1 + 10A3*B2 + 10A2*B3 + 5A1*B4 + 20A4*B2 + 30A3*B3 + 20A2*B4 +
30A4*B3 + 30A3*B4 + 20A4*B4,+,
15A4*B2 + 20A3*B3 + 15A2*B4 + 60A4*B3 + 60A3*B4 + 90A4*B4,+,
35A4*B3 + 35A3*B4 + 140A4*B4,+,
70A4*B4}
*/
const SCEVAddRecExpr *Product =
dyn_cast<SCEVAddRecExpr>(SE.getMulExpr(A_rec, B_rec));
ASSERT_TRUE(Product);
ASSERT_EQ(Product->getNumOperands(), 9u);
SmallVector<const SCEV *, 16> Sum;
Sum.push_back(SE.getMulExpr(A[0], B[0]));
EXPECT_EQ(Product->getOperand(0), SE.getAddExpr(Sum));
Sum.clear();
// SCEV produces different an equal but different expression for these.
// Re-enable when PR11052 is fixed.
#if 0
Sum.push_back(SE.getMulExpr(A[1], B[0]));
Sum.push_back(SE.getMulExpr(A[0], B[1]));
Sum.push_back(SE.getMulExpr(A[1], B[1]));
EXPECT_EQ(Product->getOperand(1), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(A[2], B[0]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 2), A[1], B[1]));
Sum.push_back(SE.getMulExpr(A[0], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 2), A[2], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 2), A[1], B[2]));
Sum.push_back(SE.getMulExpr(A[2], B[2]));
EXPECT_EQ(Product->getOperand(2), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(A[3], B[0]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[2], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[1], B[2]));
Sum.push_back(SE.getMulExpr(A[0], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[3], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 6), A[2], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[1], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[3], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 3), A[2], B[3]));
Sum.push_back(SE.getMulExpr(A[3], B[3]));
EXPECT_EQ(Product->getOperand(3), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(A[4], B[0]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[3], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 6), A[2], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[1], B[3]));
Sum.push_back(SE.getMulExpr(A[0], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[4], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 12), A[3], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 12), A[2], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[1], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 6), A[4], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 12), A[3], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 6), A[2], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[4], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 4), A[3], B[4]));
Sum.push_back(SE.getMulExpr(A[4], B[4]));
EXPECT_EQ(Product->getOperand(4), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 5), A[4], B[1]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 10), A[3], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 10), A[2], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 5), A[1], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 20), A[4], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 30), A[3], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 20), A[2], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 30), A[4], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 30), A[3], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 20), A[4], B[4]));
EXPECT_EQ(Product->getOperand(5), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 15), A[4], B[2]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 20), A[3], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 15), A[2], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 60), A[4], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 60), A[3], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 90), A[4], B[4]));
EXPECT_EQ(Product->getOperand(6), SE.getAddExpr(Sum));
Sum.clear();
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 35), A[4], B[3]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 35), A[3], B[4]));
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 140), A[4], B[4]));
EXPECT_EQ(Product->getOperand(7), SE.getAddExpr(Sum));
Sum.clear();
#endif
Sum.push_back(SE.getMulExpr(SE.getConstant(Ty, 70), A[4], B[4]));
EXPECT_EQ(Product->getOperand(8), SE.getAddExpr(Sum));
}
} // end anonymous namespace
} // end namespace llvm
Reference in New Issue View Git Blame Copy Permalink