Avoid HEXL benchmark overhead
Fixes #326 (closed)
There's a fair amount of variance, but you'll see from the second command that it takes ~28s, so this seems like a good accuracy/runtime tradeoff. I'm happy to adjust as desired.
Fixes #326 (closed)
There's a fair amount of variance, but you'll see from the second command that it takes ~28s, so this seems like a good accuracy/runtime tradeoff. I'm happy to adjust as desired.