<p><span class="h-card" translate="no"><a href="https://photog.social/@darktable" class="u-url mention">@<span>darktable</span></a></span> or just follow the tag :-) Or just pin it.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> Big Memory Leak</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@chrisvest" class="u-url mention">@<span>chrisvest</span></a></span> oh no! I&#39;m so sorry</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> but I mean the question isn&#39;t what bits the optimization looks at, it&#39;s whether the optimization is driven by a partial undef. I don&#39;t think we&#39;ve run into any of them in LLVM, but Nuno would know better</p>

<p><span class="h-card" translate="no"><a href="https://donotsta.re/users/mwk" class="u-url mention">@<span>mwk</span></a></span> I don&#39;t doubt that, but it seems whatever ABC does gets the best results as far as yosys is concerned. It&#39;s twice as good as flowmap on e.g. Sentinel. That&#39;s why I say &quot;if the AIG/whatever stuff has to still exist b/c it&#39;s the best we can do thus far, at least make it accessible&quot;.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> Russia</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> I was referencing that in my immigration petition, so it&#39;s uh, rather inconvenient for it to disappear.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@regehr" class="u-url mention">@<span>regehr</span></a></span> optimizations that look at the low bit of the mul only, actually!</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@regehr" class="u-url mention">@<span>regehr</span></a></span> <span class="h-card" translate="no"><a href="https://types.pl/@lenary" class="u-url mention">@<span>lenary</span></a></span> yeah, this is true! I have no idea what we&#39;ll end up with in the end</p><p>current plan: start with very basic unit tests to exercise all rules at least, investigate quickcheck style input generation, then do an SMT2 backend and see what we can do with that</p><p>I&#39;ve fed this compiler large real-world netlists and we had like 2 bugs in this ruleset since writing it which is very encouraging but also not very motivating to build complex validation infra :D</p>