search

intel / llvm

Intel staging area for llvm.org contribution. Home for Intel LLVM-based projects.
Other
1.25k stars 738 forks source link

BF16 in CUDA builtins.hpp #15418

Open jinz2014 opened 1 month ago

jinz2014 commented 1 month ago

Describe the bug

BF16 is not added in https://github.com/intel/llvm/blob/sycl/sycl/include/sycl/ext/oneapi/experimental/cuda/builtins.hpp

To reproduce

The error message is "no matching function for call to 'ldg'"

Environment

No response

Additional context

No response

JackAKirk commented 1 month ago

Thanks, do you have an example cuda application that uses __ldg for bf16?

jinz2014 commented 1 month ago

Please see https://github.com/zjin-lcf/HeCBench/blob/master/src/addBiasResidualLayerNorm-sycl/

JackAKirk commented 1 month ago

cc @npmiller