Clemens Lutz

icon indicating a website link www.clemenslutz.com icon indicating an email [email protected]

icon company NVIDIA icon location Santa Clara, CA icon location CUDA DevTech Engineer at NVIDIA. Interested in systems, networks, and the large-scale data processing these enable. Mostly coding C/C++, Rust, OpenCL & CUDA.

Results 2 issues of Clemens Lutz

Add cooperative kernel launch API

Pascal and newer devices support cooperative groups. These groups enable kernels to, e.g., globally synchronize the grid without terminating the kernel. A special function `cuLaunchCooperativeKernel` was added to CUDA 9.0...

Add tuning option for shared memory size

Pascal and newer devices support a shared memory size larger than 48 KiB per thread group. This is an opt-in feature that was introduced in CUDA 9.0 by specifying the...