{r,w}_level is broken on AsyncFIFO{Buffered,}

[anuejn]

It seems like the r_level / w_level on AsyncFIFO / AsyncFIFOBufered are broken in multiple ways:

• If the FIFO is full the level indicators become 0
• The semantics of w_level of AsyncFIFOBuffered are quite unintuitive since it cannot reach depth and is quite different from r_level / w_level of AsyncFIFO.

@awygle started to work on a fix for this and I picked up his work and added a simulation-based test case (master...anuejn:fifo_bugfix).
However, it is unclear to me, how one would implement the same behavior of w_level for AsyncFIFOBuffered as for AsyncFIFO without violating CDC rules. Maybe anyone has an idea here? Am I wrong here?

[awygle]

Yeah I goofed this up pretty badly (we either need a lot more simulation test cases for AsyncFIFO or we need to implement #387...). Thanks for the test case! I'll pick this back up ASAP, just kinda swamped at the moment.

[whitequark]

However, it is unclear to me, how one would implement the same behavior of w_level for AsyncFIFOBuffered as for AsyncFIFO without violating CDC rules.

Nominating for today's meeting because of this.

[whitequark]

Per IRC discussion, @anuejn may be able to fix this bug as they already have a partial fix for r_level. The fix for w_level would involve resynchronizing the state of the output buffer (which has the same CDC latency as the Gray counters themselves) and then doing the same thing as for r_level.

[whitequark]

Partially fixed by #513.