Skip to content
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

Jump to bottom

[Flang][OpenMP] Erase trip count for generic kernels #125

Merged
merged 2 commits into from
Jul 31, 2024
Merged

[Flang][OpenMP] Erase trip count for generic kernels #125

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
78604ec
[Flang][OpenMP] Erase trip count for generic kernels
DominikAdamski Jul 23, 2024
0e8e02e
Apply Sergio's remark
DominikAdamski Jul 30, 2024
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
34 changes: 20 additions & 14 deletions flang/lib/Lower/OpenMP/OpenMP.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1678,21 +1678,27 @@ genLoopNestOp(lower::AbstractConverter &converter, lower::SymMap &symTable,
firOpBuilder.getModule().getOperation());
auto targetOp = loopNestOp->getParentOfType<mlir::omp::TargetOp>();

if (offloadMod && targetOp && !offloadMod.getIsTargetDevice() &&
targetOp.isTargetSPMDLoop()) {
// Lower loop bounds and step, and process collapsing again, putting lowered
// values outside of omp.target this time. This enables calculating and
// accessing the trip count in the host, which is needed when lowering to
// LLVM IR via the OMPIRBuilder.
HostClausesInsertionGuard guard(firOpBuilder);
mlir::omp::CollapseClauseOps collapseClauseOps;
llvm::SmallVector<const semantics::Symbol *> iv;
ClauseProcessor cp(converter, semaCtx, item->clauses);
cp.processCollapse(loc, eval, collapseClauseOps, iv);
targetOp.getTripCountMutable().assign(calculateTripCount(
converter.getFirOpBuilder(), loc, collapseClauseOps));
if (offloadMod && targetOp && !offloadMod.getIsTargetDevice()) {
if (targetOp.isTargetSPMDLoop()) {
// Lower loop bounds and step, and process collapsing again, putting
// lowered values outside of omp.target this time. This enables
// calculating and accessing the trip count in the host, which is needed
// when lowering to LLVM IR via the OMPIRBuilder.
HostClausesInsertionGuard guard(firOpBuilder);
mlir::omp::CollapseClauseOps collapseClauseOps;
llvm::SmallVector<const semantics::Symbol *> iv;
ClauseProcessor cp(converter, semaCtx, item->clauses);
cp.processCollapse(loc, eval, collapseClauseOps, iv);
targetOp.getTripCountMutable().assign(calculateTripCount(
converter.getFirOpBuilder(), loc, collapseClauseOps));
} else {
DominikAdamski marked this conversation as resolved.
Show resolved Hide resolved
// The MLIR target operation was updated during PFT lowering,
// and it is no longer an SPMD kernel. Erase the trip count because
// as it is now invalid.
if (targetOp.getTripCountMutable().size())
targetOp.getTripCountMutable().erase(0);
}
}

return loopNestOp;
}

Expand Down