For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

zero_scale_global_kernel

def zero_scale_global_kernel(scale_global: Pointer[Float32, MutAnyOrigin, _safe=False])

Zeros the global FP8 scale factor in a single thread.

Writes 0 to scale_global[0] directly rather than using enqueue_fill, which can deadlock under CUDA graph replay when the GPU is already spinning inside a collectives kernel.

Args:

• ​scale_global (Pointer[Float32, MutAnyOrigin, _safe=False]): Pointer to the single FP32 global scale to zero.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit