Skip to content

[Bugfix] Fix pdl related intrin handling to avoid strict annotation codegen #1650

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

Merged
LeiWang1999 merged 1 commit into from
Jan 9, 2026
Merged

[Bugfix] Fix pdl related intrin handling to avoid strict annotation codegen #1650

Show file tree
Hide file tree
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
2 changes: 1 addition & 1 deletion 3rdparty/tvm
Open in desktop
Submodule tvm updated from 001022 to e47e76
15 changes: 4 additions & 11 deletions src/transform/lower_pdl.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,17 +57,10 @@ class MarkCudaSyncCalls : public StmtExprMutator {
}

PrimExpr VisitExpr_(const tir::CallNode *op) final {
if (op && op->op.same_as(builtin::call_extern())) {
if (!op->args.empty()) {
if (const auto *str_node = op->args[0].as<tvm::tir::StringImmNode>()) {
std::string func_name = str_node->value;
if (func_name == "cudaTriggerProgrammaticLaunchCompletion") {
has_trigger_launch_ = true;
} else if (func_name == "cudaGridDependencySynchronize") {
has_grid_sync_ = true;
}
}
}
if (op->op.same_as(tl::pdl_trigger())) {
has_trigger_launch_ = true;
} else if (op->op.same_as(tl::pdl_sync())) {
has_grid_sync_ = true;
}
return StmtExprMutator::VisitExpr_(op);
}
Expand Down
13 changes: 2 additions & 11 deletions src/transform/warp_specialized_rewriter.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,17 +164,8 @@ class WarpSpecializedRoleMarker : public StmtVisitor {
if (call->op.same_as(loop_break())) {
role = Role::kBoth;
}
if (call->op.same_as(builtin::call_extern())) {
if (!call->args.empty()) {
if (const auto *str_node =
call->args[0].as<tvm::tir::StringImmNode>()) {
std::string func_name = str_node->value;
if (func_name == "cudaGridDependencySynchronize" ||
func_name == "cudaTriggerProgrammaticLaunchCompletion") {
role = Role::kBoth;
}
}
}
if (call->op.same_as(pdl_sync()) || call->op.same_as(pdl_trigger())) {
role = Role::kBoth;
}
}
SetRole(op, role);
Expand Down
2 changes: 1 addition & 1 deletion testing/python/language/test_tilelang_language_pdl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,10 +48,10 @@ def test_pdl_trigger():
def test_pdl_sync():
N = 64
program = kernels_with_pdl_sync(N)

pdl_kernel = tilelang.compile(program, target="cuda -arch=sm_90")
code = pdl_kernel.get_kernel_source()
assert "cudaGridDependencySynchronize" in code
assert "__restrict__" not in code


if __name__ == "__main__":
Expand Down
Loading