CUDA 10.1 Released With Performance Improvements, Lightweight GEMM Library
As the first point release since last year's CUDA 10.0 release, CUDA 10.1 is now available with a new GEMM library and various performance optimizations.
The new library to CUDA 10.1 is cuBLASLt as a lightweight GEMM library. Some of the other work and focuses for better performance include:
- New selective eigensolvers SYEVDX and SYGVDX in cuSOLVER, and performance improvements of up to 1.5X for full spectrum eigensolvers
- New encoding and batched decoding functionalities in nvJPEG
- Up to 4X faster performance for broad set of random number generators in cuRAND
- Improved performance and support for fork/join kernels in CUDA Graphs APIs
On the Linux side, CUDA 10.1 is noteworthy as now there is official Ubuntu 18.10 support alongside 18.04/16.04/14.04, among other Linux distributions.
More details on CUDA 10.1 can be found via developer.nvidia.com.
The new library to CUDA 10.1 is cuBLASLt as a lightweight GEMM library. Some of the other work and focuses for better performance include:
- New selective eigensolvers SYEVDX and SYGVDX in cuSOLVER, and performance improvements of up to 1.5X for full spectrum eigensolvers
- New encoding and batched decoding functionalities in nvJPEG
- Up to 4X faster performance for broad set of random number generators in cuRAND
- Improved performance and support for fork/join kernels in CUDA Graphs APIs
On the Linux side, CUDA 10.1 is noteworthy as now there is official Ubuntu 18.10 support alongside 18.04/16.04/14.04, among other Linux distributions.
More details on CUDA 10.1 can be found via developer.nvidia.com.
3 Comments