Skip to content

Fix vector_reduce simplification bug. #9031

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
alexreinking merged 1 commit into from
Mar 15, 2026
+21 −2
Merged

Fix vector_reduce simplification bug. #9031

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
5 changes: 3 additions & 2 deletions src/Simplify_Exprs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,7 @@ Expr Simplify::visit(const VectorReduce *op, ExprInfo *info) {
return value;
}

if (info && op->type.is_int()) {
if (info && op->type.is_int_or_uint()) {
Copy link
Member

@abadams abadams Mar 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can't this also be broken for float? Why is there a type check at all?

Copy link
Contributor Author

@mcourteaux mcourteaux Mar 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hah. Potentially yes. Good point.

switch (op->op) {
case VectorReduce::Add:
// Alignment of result is the alignment of the arg. Bounds
Expand Down Expand Up @@ -123,7 +123,8 @@ Expr Simplify::visit(const VectorReduce *op, ExprInfo *info) {
case VectorReduce::Add: {
auto rewrite = IRMatcher::rewriter(IRMatcher::h_add(value, lanes), op->type);
if (rewrite(h_add(x * broadcast(y, arg_lanes), lanes), h_add(x, lanes) * broadcast(y, lanes)) ||
rewrite(h_add(broadcast(x, arg_lanes) * y, lanes), h_add(y, lanes) * broadcast(x, lanes))) {
rewrite(h_add(broadcast(x, arg_lanes) * y, lanes), h_add(y, lanes) * broadcast(x, lanes)) ||
rewrite(h_add(broadcast(x, arg_lanes), lanes), broadcast(x * factor, lanes))) {
return mutate(rewrite.result, info);
}
break;
Expand Down
18 changes: 18 additions & 0 deletions test/correctness/simplify.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -810,6 +810,24 @@ void check_vectors() {
int_vector);
check(VectorReduce::make(VectorReduce::Max, Broadcast::make(int_vector, 4), 8),
VectorReduce::make(VectorReduce::Max, Broadcast::make(int_vector, 4), 8));

{
// h_add(broadcast(x, 8), 4) should simplify to broadcast(x * 2, 4)
check(VectorReduce::make(VectorReduce::Add, broadcast(x, 8), 4),
broadcast(x * 2, 4));
}

{
Expr const_u8 = cast(UInt(8), 3);
check(VectorReduce::make(VectorReduce::Add, broadcast(const_u8, 9), 3), broadcast(cast(UInt(8), 9), 3));
}

{
// Test VectorReduce::Add on a variable of unsigned type to ensure the multiplied factor
// keeps the correct type and avoids type-mismatch assertion failures.
Expr u8_x = Variable::make(UInt(8), "u8_x");
check(VectorReduce::make(VectorReduce::Add, broadcast(u8_x, 9), 3), broadcast(u8_x * cast(UInt(8), 3), 3));
}
}

void check_bounds() {
Expand Down
Loading