Compute.CublasLtPlanCache Module Reference

Classes
class	Mila::Dnn::Compute::Cuda::CublasLtPlanCache< TPlan >
	Generic plan cache keyed on batch size bucket. More...

Functions
std::vector< int >	Mila::Dnn::Compute::Cuda::computeArchitectureBuckets (int max_batch_size)
	Computes optimal bucket boundaries for cuBLASLt plan caching based on CUDA device architecture.
int	Mila::Dnn::Compute::Cuda::getBucket (const std::vector< int > &buckets, int batch_size)
	Fast O(log N) bucket lookup.

Files
file	/__w/Mila/Mila/Mila/Src/Dnn/Compute/Devices/Cuda/Operations/Common/CublasLtPlanCache.ixx