Skip to content

Conversation

@mondus
Copy link
Contributor

@mondus mondus commented Nov 5, 2025

Issue discovered in FLAME GPU2 PR #1334
and replicated with Jitify 2 where SimpleTest fails on Google Colab and Sheffield HPC with older drivers.

Jitify uses JITIFY_USE_CONTEXT_INDEPENDENT_LOADING to define jitify2::CudaFunction as either a CUkernel or a CUfunction however ConfiguredKernel::configure_1d_max_occupancy unconditionally casts this as a CUFunction (see L2566).

Fix is to conditionally obtain the function form the kernel using cuKernelGetFunction where JITIFY_USE_CONTEXT_INDEPENDENT_LOADING is defined.

JITIFY_USE_CONTEXT_INDEPENDENT_LOADING update for occupancy API
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant