cuda.bindings._v2.driver.CooperativeLaunchMultiDeviceFlags

class cuda.bindings._v2.driver.CooperativeLaunchMultiDeviceFlags(value: int)

Flags for launch_cooperative_kernel_multi_device (CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_*).

NO_PRE_LAUNCH_SYNC: 1

If set, each kernel launched as part of cuLaunchCooperativeKernelMultiDevice only waits for prior work in the stream corresponding to that GPU to complete before the kernel begins execution.

NO_POST_LAUNCH_SYNC: 2

If set, any subsequent work pushed in a stream that participated in a call to cuLaunchCooperativeKernelMultiDevice will only wait for the kernel launched on the GPU corresponding to that stream to complete before it begins execution.