Our current result, based on 5.0 (so year old bits) is
Last Aug 2020 or so I took a screenshot, and we were here
I'm including Java here just to demonstrate the hardware hasn't changed. Our benchmark implementation is essentially the same as the official one, but made single threaded. Note that didn't change since at least 2017, the only variable is the .NET version.
Did we really regress 3.1->5.0? And did this regression persist to 6.0?
Was there a regression that only shows up in the parallelized version? Is that stable enough to be able to include in our perf repo alongside the single threaded one?
Again using Java as a benchmark, we get the same result on Fasta and are leading on all other scenarios. binary-trees the curious outlier.
Our current result, based on 5.0 (so year old bits) is
Last Aug 2020 or so I took a screenshot, and we were here
I'm including Java here just to demonstrate the hardware hasn't changed. Our benchmark implementation is essentially the same as the official one, but made single threaded. Note that didn't change since at least 2017, the only variable is the .NET version.
Did we really regress 3.1->5.0? And did this regression persist to 6.0?
Was there a regression that only shows up in the parallelized version? Is that stable enough to be able to include in our perf repo alongside the single threaded one?
Again using Java as a benchmark, we get the same result on Fasta and are leading on all other scenarios. binary-trees the curious outlier.