"src/targets/vscode:/vscode.git/clone" did not exist on "64696bb1bf62d42914b20d8ef3dd05d214419c13"
Commit d3401374 authored by Khalique Ahmed's avatar Khalique Ahmed
Browse files

update fuse_ops contiguous_pointwise to check gemm input

parent 76871340
......@@ -778,6 +778,15 @@ struct find_contiguous_pointwise
{
auto ins = r.result;
auto pw = ins->inputs().front();
for(auto output : ins->outputs())
{
if(output->name() != "gpu::gemm")
continue;
auto pw_strides = pw->get_shape().strides();
// ensure pointwise shape is still valid for gemm call
if(std::none_of(pw_strides.end() - 2, pw_strides.end(), [&](auto i) { return i == 1; }))
return;
}
auto alloc = ins->inputs().back();
auto args = pw->inputs();
args.back() = alloc;
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment