Measure the peak achievable performance of GPU compute devices
A synthetic micro-benchmark that measures the peak achievable performance of GPU compute devices. It exercises tight vector / MAD / MMA loops and vendor-SDK GEMM libraries (cuBLASLt on NVIDIA, MPS on Apple) to expose what the hardware is capable of — from raw ALU peaks to near-vendor-advertised matrix throughput. clpeak began as an OpenCL-only tool. It now ships four interchangeable backends — OpenCL, Vulkan, CUDA, and Metal — running back-to-back on the same hardware, so cross-stack differences (driver lowering, instruction scheduling, extension exposure) become visible alongside the raw peak numbers.
| Release | Stable | Testing |
|---|---|---|
| Fedora Rawhide | 2.0.17-1.fc45 | - |
| Fedora 44 | 2.0.7-1.fc44 | 1.1.6-1.fc44 |
| Fedora 43 | 1.1.5-2.fc43 | - |
You can contact the maintainers of this package via email at
clpeak dash maintainers at fedoraproject dot org.