Skip to content

[NFC] cache repeated tree walks to avoid O(N^2) in optimizeTerminatingTails in CodeFolding #8602

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Changqing-JING wants to merge 1 commit into
base: main
Choose a base branch
from
+38 −5
Open

[NFC] cache repeated tree walks to avoid O(N^2) in optimizeTerminatingTails in CodeFolding #8602

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 38 additions & 5 deletions src/passes/CodeFolding.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -299,6 +299,7 @@ struct CodeFolding
returnTails.clear();
unoptimizables.clear();
modifieds.clear();
bodyCachePopulated = false;
if (needEHFixups) {
EHUtils::handleBlockNestedPops(func, *getModule());
}
Expand All @@ -311,6 +312,41 @@ struct CodeFolding
// inside that item
bool canMove(const std::vector<Expression*>& items, Expression* outOf) {
auto allTargets = BranchUtils::getBranchTargets(outOf);
bool hasTry = false;
bool hasTryTable = false;
if (getModule()->features.hasExceptionHandling()) {
hasTry = FindAll<Try>(outOf).has();
hasTryTable = FindAll<TryTable>(outOf).has();
}
return canMoveImpl(items, allTargets, hasTry, hasTryTable);
}

// Cached data for the function body, computed on demand to avoid repeated
// O(N) tree walks in optimizeTerminatingTails.
BranchUtils::NameSet bodyBranchTargets;
bool bodyHasTry = false;
bool bodyHasTryTable = false;
bool bodyCachePopulated = false;

// Like canMove, but uses precomputed branch targets and Try/TryTable
// presence. This avoids repeated O(N) tree walks when outOf is the
// function body and canMove is called multiple times.
bool canMoveWithCachedBodyInfo(const std::vector<Expression*>& items) {
if (!bodyCachePopulated) {
bodyBranchTargets = BranchUtils::getBranchTargets(getFunction()->body);
if (getModule()->features.hasExceptionHandling()) {
bodyHasTry = FindAll<Try>(getFunction()->body).has();
bodyHasTryTable = FindAll<TryTable>(getFunction()->body).has();
}
bodyCachePopulated = true;
}
return canMoveImpl(items, bodyBranchTargets, bodyHasTry, bodyHasTryTable);
}

bool canMoveImpl(const std::vector<Expression*>& items,
const BranchUtils::NameSet& allTargets,
bool hasTry,
bool hasTryTable) {
for (auto* item : items) {
auto exiting = BranchUtils::getExitingBranches(item);
std::vector<Name> intersection;
Expand Down Expand Up @@ -341,9 +377,7 @@ struct CodeFolding
// conservative approximation because there can be cases that
// 'try'/'try_table' is within the expression that may throw so it is
// safe to take the expression out.
// TODO: optimize this check to avoid two FindAlls.
if (effects.throws() &&
(FindAll<Try>(outOf).has() || FindAll<TryTable>(outOf).has())) {
if (effects.throws() && (hasTry || hasTryTable)) {
return false;
}
}
Expand Down Expand Up @@ -610,8 +644,7 @@ struct CodeFolding
cost += WORTH_ADDING_BLOCK_TO_REMOVE_THIS_MUCH;
// if we cannot merge to the end, then we definitely need 2 blocks,
// and a branch
// TODO: efficiency, entire body
if (!canMove(items, getFunction()->body)) {
if (!canMoveWithCachedBodyInfo(items)) {
cost += 1 + WORTH_ADDING_BLOCK_TO_REMOVE_THIS_MUCH;
// TODO: to do this, we need to maintain a map of element=>parent,
// so that we can insert the new blocks in the right place
Expand Down
Loading