llama.cpp

41e3f026 - cuda : revert CUDA_SCALE_LAUNCH_QUEUES override until investigated (#19227)

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.