Closed dcrc2 closed 3 years ago
Looks good, has it been run on sqrl benchmark?
Ah, good thing you reminded me about this: compilation succeeds but the results are reported as being incorrect.
compilation succeeds but the results are reported as being incorrect.
So there are two issues in sqrl:
Nice find!
I'm fine with reducing the largest size. I've been wondering about how best to do the correctness checks on PR, but at least for now the overnight is our real benchmark! Perhaps we need to make the benchmark tests individually fail-able. In any case, I'm happy for it to be reduced for now and we'll figure out elsewhere.
Update pytest benchmarking to use
tsmod2ksmod
, matching the corresponding change inrun-bench.py
:https://github.com/microsoft/knossos-ksc/pull/838/files#diff-19d15b57f2c48a8fac44562c3dfd51f7d9be1f31236e51dd4bd574deabd44cf7
AB#19336