search

intel / intel-xpu-backend-for-triton

OpenAI Triton backend for Intel® GPUs
MIT License
143 stars 44 forks source link

Fix UT test_core.py::test_poison_return failure #2685

Closed Retribution98 closed 1 week ago

Retribution98 commented 1 week ago

Closes #2482

Cuda converts tt.store to inline_asm command, and xpu converts it to llvm.store. Then the LLVM optimizer for the SROA function removes llvm.store with the poison value, but ignores the cuda assembly instruction.