Back to feed

b7920

Feb 3, 2026
Meta/llama.cppCLIvb7920

cuda : revert CUDA_SCALE_LAUNCH_QUEUES override until investigated (#19227)

Hangs were reported on Jetson Orin AGX if we set CUDA_SCALE_LAUNCH_QUEUES=4x. Reverting the previous PR (#19042) and updating the document to consider setting CUDA_SCALE_LAUNCH_QUEUES=4x for faster throughput on multi-GPU systems.

macOS/iOS:

Linux:

Windows:

openEuler: