This is the Quda benchmarking program, aiming to be equivalent to Benchmark_Grid.
It does include:
running over a fixed range of (local) lattice sizes
memory benchmark, i.e., axpy
wilson fermions
domain-wall fermions
Id does NOT include:
explicit communication benchmarks
Also: the build script is not portable outside of Tursa because I have not figured out how to link against libnvidia-ml.so without specifying the (very machine-dependent) full path.
This is the Quda benchmarking program, aiming to be equivalent to `Benchmark_Grid`.
It does include:
- running over a fixed range of (local) lattice sizes
- memory benchmark, i.e., `axpy`
- wilson fermions
- domain-wall fermions
Id does NOT include:
- explicit communication benchmarks
Also: the build script is not portable outside of Tursa because I have not figured out how to link against `libnvidia-ml.so` without specifying the (very machine-dependent) full path.
Benchmark_Quda now produces json output, format is generally compatible with Benchmark_Grid, but contains somewhat different information
added some logic to choose the number of benchmarking iterations automatically as in "run this benchmark for X seconds".
added start/stop_time to the json output. This can be used for example to line up benchmarking results with simultaneously taken energy measurements.
from my side, this is ready to be merged.
update:
* `Benchmark_Quda` now produces json output, format is generally compatible with `Benchmark_Grid`, but contains somewhat different information
* added some logic to choose the number of benchmarking iterations automatically as in "run this benchmark for X seconds".
* added `start/stop_time` to the json output. This can be used for example to line up benchmarking results with simultaneously taken energy measurements.
from my side, this is ready to be merged.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
This is the Quda benchmarking program, aiming to be equivalent to
Benchmark_Grid.It does include:
axpyId does NOT include:
Also: the build script is not portable outside of Tursa because I have not figured out how to link against
libnvidia-ml.sowithout specifying the (very machine-dependent) full path.update:
Benchmark_Qudanow produces json output, format is generally compatible withBenchmark_Grid, but contains somewhat different informationstart/stop_timeto the json output. This can be used for example to line up benchmarking results with simultaneously taken energy measurements.from my side, this is ready to be merged.
2351edf3f7tofb43d16830