Outline to GPU launch before further lowering

plaidml / tpp-mlir

TPP experimentation on MLIR for linear algebra

https://arxiv.org/abs/2404.15204

Other

111 stars 29 forks source link

Open adam-smnk opened 2 weeks ago

adam-smnk commented 2 weeks ago

Reorders GPU conversion to allow easier identification and post-processing of operations targeting GPU.

Tweaks XeGPU lowering to accomodate for the change.

adam-smnk commented 17 hours ago

Bump