Skip to content
Merged
Changes from all commits
Commits
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
12 changes: 8 additions & 4 deletions src/transform/atomicadd_vectorize.cc
Original file line number Diff line number Diff line change
Expand Up @@ -231,21 +231,25 @@ class AtomicAddVectorizeRewriter : public StmtExprMutator {
// Ref: src/tl_templates/cuda/atomic.h::AtomicAdd
const IntImm memory_order =
node->args.size() >= 3 ? Downcast<IntImm>(node->args[2]) : IntImm(0);

Array<PrimExpr> new_args;
Call address_of_dst =
Call(DataType::Handle(), builtin::address_of(), {dst_node});
Call address_of_value =
Call(DataType::Handle(), builtin::address_of(), {value_node});
Array<PrimExpr> new_args;
if (vector_size_ == 4) {
new_args.push_back(StringImm("AtomicAddx4"));
new_args.push_back(address_of_dst);
new_args.push_back(address_of_value);
} else if (vector_size_ == 2) {
new_args.push_back(StringImm("AtomicAddx2"));
new_args.push_back(address_of_dst);
new_args.push_back(address_of_value);
} else {
new_args.push_back(StringImm("AtomicAdd"));
new_args.push_back(dst_node);
new_args.push_back(value_node);
}
new_args.push_back(address_of_dst);
new_args.push_back(address_of_value);

new_args.push_back(memory_order);

Call new_call =
Expand Down
Loading