Skip to content

Commit

Permalink
Triton kernel requires TMA
Browse files Browse the repository at this point in the history
Summary: Fix a minor bug in fp8_gemm_rowwise where TMA is not available on AMD/HIP

Reviewed By: nmacchioni

Differential Revision: D65915379

fbshipit-source-id: de6effd2f7e830115eadaf17a4c1add2101ef78e
  • Loading branch information
xuzhao9 authored and facebook-github-bot committed Nov 14, 2024
1 parent c2273e5 commit f48ad1e
Showing 1 changed file with 4 additions and 1 deletion.
5 changes: 4 additions & 1 deletion tritonbench/operators/fp8_gemm_rowwise/operator.py
Original file line number Diff line number Diff line change
Expand Up @@ -43,8 +43,11 @@ def parse_args(args: List[str]) -> argparse.Namespace:
matmul_fp8_row as triton_fp8_row,
)

assert hasattr(
triton.runtime.driver.active.utils, "fill_1d_tma_descriptor"
), "TMA is required by the Triton kernel."
HAS_TRITON = True
except ImportError:
except (ImportError, AssertionError):
HAS_TRITON = False


Expand Down

0 comments on commit f48ad1e

Please sign in to comment.