Releases: tud-zih-energy/FIRESTARTER
Release 2.2
Changelog for Version 2.2
- added support for AMD GPUs
- support for thread binding on all supported platforms
- bug fixes related to thread binding with cgroups and disabled cpus
- added report for estimated GPU FLOPS
With this release we provide an Alpha version for ARMv8-A with NEON support.
There are still major issues with the Alpha, e.g., a segfault when execution is stopped. We appreciate any feedback and merge requests. The source code can be found on the aarch64_support branch.
v2.1.1
Release 2.1
Changelog for Version 2.1
- added Intel OneAPI support for Intel GPUs
- added error detection
- various fixes
- removed CUDA_ONLY
Release 2.0
Changelog for Version 2.0
- added AMD Rome (Zen 2) support
- added macOS support
- CUDA support on Windows
- FIRESTARTER_CUDA initialization speed improved
- automatic tuning option to better accommodate variations in core count, frequency, and memory bandwidth between different SKUs of one generation
- automatic recording of metrics
- tweaked payloads to prevent vector registers from converging to denormal values
Release 1.7.4
FIRESTARTER has now support for AMD Zen and Zen+ CPUs. The range of supported CUDA runtimes has been extended.
Release 1.7.3
FIRESTARTER will now check automatically the supported data types on your CUDA graphics card. The range of supported CUDA runtimes has been extended. Additionally, there are some bug fixes.
Bugfix release 1.7.2
We fixed several bugs, mostly in the infrastructure and in the handling of CUDA devices. Also, the version is now correctly reported.