Lunderberg commented on PR #17283: URL: https://github.com/apache/tvm/pull/17283#issuecomment-2302896806
Hmm. I'm seeing local `__truncsfhf2` and `__extendhfsf2` functions, which are defined [here](https://github.com/apache/tvm/blob/main/src/target/llvm/codegen_llvm.cc#L1136) along with a comment saying that they were based on the functions in `buildin_fp16.cc`. However, I don't see any calls to these local functions in the LLVM IR. It looks like the generated LLVM IR instead uses `fpext` ([here](https://github.com/apache/tvm/actions/runs/10490857078/job/29058556506?pr=17283#step:5:658)) to convert from float16 to float32. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
