jhlee525 opened a new pull request, #18627: URL: https://github.com/apache/tvm/pull/18627
This PR contains some fix for metal codegen & runtime, especially enabling adds bfloat16 support for Metal(Metal3.1+). - add bf16 Intrinsics for metals - Registers BF16-specific simdgroup matrix intrinsics (8×8×8) - Cast immediate values to the correct dtype in make_filled_simdgroup_matrix - Explicit `bfloat()` casts for bf16 immediates in codegen (Metal does not support an h suffix for bf16 literals) - Uses Metal 3.1 language version in `metal_module.mm` when it's available -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
