Marek Materzok
Marek Materzok
Do we need this anywhere now? Isn't a function returning a module something weird?
> If we had quasi-register, which store high till the end of the cycle, we could implement iterable algorithm in one cycle, but this will make critical path very long......
Did you check if there is even any benefit of trying to optimize them in this way? If some method is called in only one place, there is absolutely no...
> each unit test is parameterized using two values important from running test case: seed and test_number. You mean, unit tests which use random test generation. We have a number...
Sanity check: did you assign the `ModuleConnector` to some attribute on an `Elaboratable`?
This would be kind of weird, as: * In silicon, binary multiplexers are basically just one-hot multiplexers with binary decoders. * In FPGA, the two stages might just get glued...
That was a fun little experiment, and the results (https://github.com/kuznia-rdzeni/coreblocks/actions/runs/5446254887) are: * IPC is only minimally improved, which probably means inter-instruction dependencies aren't a bottleneck right now. * Fmax is...
Superseded by #792.
As this is optimization-related, I don't think this is a priority; also, this would probably increase our resource usage a lot if run in CI.
Fixed in #394, should probably be tested better.