AuroraMiddleware/v8
1
0
Fork 0
Code Issues 2 Pull Requests Releases Wiki Activity

[turbofan] Re-wire greedy.

Browse Source 
We completely un-wired the greedy allocator to focus on the
stackchecks in loops (splintering) work. This change re-wires greedy,
still behind its flag. For now, enabling the greedy allocator disables
the stackchecks in loops feature (and range splintering), so that we are
at the baseline we left it at.

The main contribution in this change is adapting the codebase after
the live range model refactoring, whereby RegisterAllocationData's
live_ranges() contains just top-level ranges, and children are accessed
via their parents.

BUG=

Review URL: https://codereview.chromium.org/1320363002

Cr-Commit-Position: refs/heads/master@{#30492}
This commit is contained in:
mtrofin 2015-08-31 21:42:35 -07:00 committed by Commit bot
parent 09bb401b80
commit 71779375ad
3 changed files with 17 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/compiler/greedy-allocator.cc
View File

@ -164,8 +164,12 @@ void GreedyAllocator::PreallocateFixedRanges() {
void GreedyAllocator::ScheduleAllocationCandidates() {
for (auto range : data()->live_ranges()) {
if (CanProcessRange(range) && !range->spilled()) {
scheduler().Schedule(range);
if (CanProcessRange(range)) {
for (LiveRange* child = range; child != nullptr; child = child->next()) {
if (!child->spilled()) {
scheduler().Schedule(child);
}
}
}
}
}

10
src/compiler/pipeline.cc
View File

@ -1354,9 +1354,13 @@ void Pipeline::AllocateRegisters(const RegisterConfiguration* config,
Run<SplinterLiveRangesPhase>();
}
// TODO(mtrofin): re-enable greedy once we have bots for range preprocessing.
Run<AllocateGeneralRegistersPhase<LinearScanAllocator>>();
Run<AllocateDoubleRegistersPhase<LinearScanAllocator>>();
if (FLAG_turbo_greedy_regalloc) {
Run<AllocateGeneralRegistersPhase<GreedyAllocator>>();
Run<AllocateDoubleRegistersPhase<GreedyAllocator>>();
} else {
Run<AllocateGeneralRegistersPhase<LinearScanAllocator>>();
Run<AllocateDoubleRegistersPhase<LinearScanAllocator>>();
}
if (FLAG_turbo_preprocess_ranges) {
Run<MergeSplintersPhase>();

8
src/flag-definitions.h
View File

@ -409,10 +409,10 @@ DEFINE_BOOL(turbo_greedy_regalloc, false, "use the greedy register allocator")
DEFINE_BOOL(turbo_preprocess_ranges, true,
"run pre-register allocation heuristics")
DEFINE_BOOL(turbo_loop_stackcheck, true, "enable stack checks in loops")
// TODO(mtrofin): remove these implications, as they are here just for trybot
// purposes.
DEFINE_IMPLICATION(turbo_greedy_regalloc, turbo_preprocess_ranges)
DEFINE_IMPLICATION(turbo_greedy_regalloc, turbo_loop_stackcheck)
// TODO(mtrofin): remove the 2 implications.
DEFINE_NEG_IMPLICATION(turbo_greedy_regalloc, turbo_preprocess_ranges)
DEFINE_NEG_IMPLICATION(turbo_greedy_regalloc, turbo_loop_stackcheck)
DEFINE_IMPLICATION(turbo, turbo_asm_deoptimization)
DEFINE_STRING(turbo_filter, "~~", "optimization filter for TurboFan compiler")