summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorRui Ueyama <ruiu@google.com>2015-10-26 16:20:00 +0000
committerRui Ueyama <ruiu@google.com>2015-10-26 16:20:00 +0000
commitdf985afa149aa083eb6ab23872681fa4af15dbca (patch)
treedd7577c4f9bd1dbb103db720c6490a08c4db5577
parent83553d0cac09085ac9405ee171414a5a040e8e0f (diff)
downloadbcm5719-llvm-df985afa149aa083eb6ab23872681fa4af15dbca.tar.gz
bcm5719-llvm-df985afa149aa083eb6ab23872681fa4af15dbca.zip
COFF: De-parallelize ICF for now.
There was a threading issue in the ICF code for COFF. That seems like a venign bug in the sense that it doesn't produce an incorrect output, but it oftentimes misses reducible sections. As a result, mergeable sections could remain in outputs, which makes the output nondeterministic. Basically the algorithm we are using for ICF is this: We group sections so that identical sections will eventually be in the same group. Initially, all sections are in one group. We split the group by relocation targets until we get a convergence (if relocation targets are in different gruops, the sections are different). Once a group is split, they will never be merged. Each section has a group ID. That variable itself is atomic, so there's no threading issue at the level that we can use thread sanitizer. The point is, when we split a group, we re-assign new group IDs to group of sections. That are multiple separate writes to atomic varaibles. Thus, splitting a group is not an atomic operation, and there's a small chance that the other thread observes inconsistent group IDs. Over-splitting is always "safe", so it will never create incorrect output. I suspect that the nondeterminism stems from that point. However, I cannot prove or fix that at this moment, so I'm going to avoid using threads here. llvm-svn: 251300
-rw-r--r--lld/COFF/ICF.cpp57
1 files changed, 21 insertions, 36 deletions
diff --git a/lld/COFF/ICF.cpp b/lld/COFF/ICF.cpp
index 243dfa01b09..3dab8a79ed6 100644
--- a/lld/COFF/ICF.cpp
+++ b/lld/COFF/ICF.cpp
@@ -58,7 +58,6 @@ using namespace llvm;
namespace lld {
namespace coff {
-static const size_t NJOBS = 256;
typedef std::vector<SectionChunk *>::iterator ChunkIterator;
typedef bool (*Comparator)(const SectionChunk *, const SectionChunk *);
@@ -191,11 +190,11 @@ void ICF::run(const std::vector<Chunk *> &Vec) {
SC->GroupID = getHash(SC) | (uint64_t(1) << 63);
}
});
- std::vector<std::vector<SectionChunk *>> VChunks(NJOBS);
+ std::vector<SectionChunk *> Chunks;
for (Chunk *C : Vec) {
if (auto *SC = dyn_cast<SectionChunk>(C)) {
if (SC->GroupID) {
- VChunks[SC->GroupID % NJOBS].push_back(SC);
+ Chunks.push_back(SC);
} else {
SC->GroupID = NextID++;
}
@@ -204,29 +203,17 @@ void ICF::run(const std::vector<Chunk *> &Vec) {
// From now on, sections in Chunks are ordered so that sections in
// the same group are consecutive in the vector.
- parallel_for_each(VChunks.begin(), VChunks.end(),
- [&](std::vector<SectionChunk *> &Chunks) {
- std::sort(Chunks.begin(), Chunks.end(),
- [](SectionChunk *A, SectionChunk *B) {
- return A->GroupID < B->GroupID;
- });
- });
+ std::sort(Chunks.begin(), Chunks.end(),
+ [](SectionChunk *A, SectionChunk *B) {
+ return A->GroupID < B->GroupID;
+ });
// Split groups until we get a convergence.
int Cnt = 1;
- parallel_for_each(VChunks.begin(), VChunks.end(),
- [&](std::vector<SectionChunk *> &Chunks) {
- forEachGroup(Chunks, equalsConstant);
- });
+ forEachGroup(Chunks, equalsConstant);
for (;;) {
- std::atomic<bool> Redo(false);
- parallel_for_each(VChunks.begin(), VChunks.end(),
- [&](std::vector<SectionChunk *> &Chunks) {
- if (forEachGroup(Chunks, equalsVariable))
- Redo = true;
- });
- if (!Redo)
+ if (!forEachGroup(Chunks, equalsVariable))
break;
++Cnt;
}
@@ -234,22 +221,20 @@ void ICF::run(const std::vector<Chunk *> &Vec) {
llvm::outs() << "\nICF needed " << Cnt << " iterations.\n";
// Merge sections in the same group.
- for (std::vector<SectionChunk *> &Chunks : VChunks) {
- for (auto It = Chunks.begin(), End = Chunks.end(); It != End;) {
- SectionChunk *Head = *It++;
- auto Bound = std::find_if(It, End, [&](SectionChunk *SC) {
- return Head->GroupID != SC->GroupID;
- });
- if (It == Bound)
- continue;
+ for (auto It = Chunks.begin(), End = Chunks.end(); It != End;) {
+ SectionChunk *Head = *It++;
+ auto Bound = std::find_if(It, End, [&](SectionChunk *SC) {
+ return Head->GroupID != SC->GroupID;
+ });
+ if (It == Bound)
+ continue;
+ if (Config->Verbose)
+ llvm::outs() << "Selected " << Head->getDebugName() << "\n";
+ while (It != Bound) {
+ SectionChunk *SC = *It++;
if (Config->Verbose)
- llvm::outs() << "Selected " << Head->getDebugName() << "\n";
- while (It != Bound) {
- SectionChunk *SC = *It++;
- if (Config->Verbose)
- llvm::outs() << " Removed " << SC->getDebugName() << "\n";
- Head->replace(SC);
- }
+ llvm::outs() << " Removed " << SC->getDebugName() << "\n";
+ Head->replace(SC);
}
}
}
OpenPOWER on IntegriCloud