Alan Hayward
Alan Hayward
The first commit in this PR is a copy of https://github.com/dotnet/runtime/pull/71705. Once that PR is merged I will remove that commit from this patch. This was just so I could...
PR is now rebased on top of HEAD, and is good for a review. Taking this out of draft. I do have the one failure being hit a few times...
>The diffs look like size wise regressions, why is that? I would expect this pattern should give smaller code in most cases. For all the cases where the code is...
As a side point - when rebasing is it best to pull/rebase/forcepush or is a merge commit preferred?
> > Assuming the branch taken/not taken is fairly random, then the new code will be faster. > > I'm not sure this is a fair assumption in practice. Can...
> What does it look like for always taken/always not taken branches? Not so great.... For both of these it's mod2 with inputs of all zeros. SingleSeqAlways - Always passes...
> > As I understood it, it wouldn't be safe to move that? > > It would be fine to do the transformation regardless of side effects on the condition....
> > > As I understood it, it wouldn't be safe to move that? > > > > > > It would be fine to do the transformation regardless of...
>What is the diff of the code? This is my full set of benchmark changes: https://github.com/a74nh/performance/commit/b1c00bafdcb19edcc12640b8e3e2f16752277e80 (Delay sharing was just internal process on my end that we need to go...
New version allows for side effects in the `JTRUE` condition.