Andrew Waterman
Andrew Waterman
They don't fit in the 16 KiB scratchpad. They could be rewritten to fit, but that would eliminate the microarchitectural test of AMOs that miss in the cache.
Dropping them isn't the right answer, either... I'll figure out a solution.
I actually think the answer is to rely on torture to catch the microarchitectural bugs, and change these to be simple functional tests (and so they will work on TinyConfig).
Directed microarchitectural tests are a good idea, but should probably be separate from the ISA functionality tests.
Agreed in principle, but since this change is virtually guaranteed to expose bugs in existing designs, it should be an optional pass.
Just a thought... for the existing Boolean case, excluding 0 makes sense for obvious reasons. But it's not clear that excluding 0 for the generalized UInt case is the right...
> I was planning for this to be a low-level firrtl API Derp. Not sure why I responded in the chisel context, given the repo we're on. > I now...
As to your question about Not/Cat: could you write an example of what you're trying to do? (Also, provided we agree, it should be saved for a separate PR)
That optimization is best left for synthesis, because it will usually reduce simulation performance, and the synthesis tool will always get it right.
You can put this code in a separate .v file, which also 'includes the .v chisel emits. On Thursday, March 10, 2016, Martoni [email protected] wrote: > I'm using cocotb https://github.com/potentialventures/cocotb...