Chris Leary

I'd think we need a construct on the top level proc in the DSL and it could just be extracted as part of IR conversion? Is there a natural place...

I think @meheff's work on JIT scaling really helped the situation here. I wrote a hacky shell script and left it running a few minutes and wrote a hacky plot...

Marking as "long term enhancement" because I believe there's not a pressing need to have this association today.

Snippets from my docker run output: ``` //xls/ir:function_test PASSED in 1.9s //xls/ir:node_test PASSED in 2.8s //xls/fuzzer:run_fuzz_test PASSED in 149.3s ``` So something definitely seems odd in your run. I was...

@hongted may have thoughts on this -- it took a little work for us to get the caching to a point where it wouldn't overflow.

@meheff has a change (that looks through aggregate/index opacity using the generic dataflow analysis) that we think should address this.

Closing due to staleness though appreciate all the discussion, notably we've somewhat refined our "continuously working" vs "best effort" policies around environments, continuous working state is reflected in our per-commit...

Note that we're also nearly exporting release binaries which should be fully statically linked I believe and work on systems with a sufficient libc (cc @proppy in case I'm mistaken...

Note there's active progress on this now, e.g. https://github.com/google/xls/commit/e5f3d0de01d2863d5b648754abb1ac09cf34ba9b is a prelude to a github action that can host release binaries. These will be downloaded/used as canonical URLs by e.g....

We had some discussion about how best to support proofs in the build/test system. I think if the proofs are noted to be something like `psan_prove!(predicate)` in the DSLX code...