intel / intel-xpu-backend-for-triton

OpenAI Triton backend for Intel® GPUs
MIT License
129 stars 38 forks source link

[RAISE-BP] Add support for tt.broadcast increasing tensor rank #1503

Closed mfrancepillois closed 3 months ago

mfrancepillois commented 3 months ago

Extend the broadcast support to handle tt.broadcast operations when the rank of the destination tensor is greater than the rank of the source tensor.

Closes Issue: #1429

mfrancepillois commented 3 months ago

Can we have a test with:

%1 = tt.broadcast %cst : tensor<128xi32> -> tensor<256x2x128xi32>

? Incrementing the rank in 2

A test has been added.

etiotto commented 3 months ago

@victor-eds @whitneywhtsang are you OK with the PR?