Jakob Botsch Nielsen

> Nice! I see you're hitting similar size regressions that I did too, e.g.: > > ```diff > fmov s16, #1.0000 > - stp s16, s16, [x0, #0x08] > +...

/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress

The failures look known. cc @dotnet/jit-contrib PTAL @EgorBo @kunalspathak [Diffs](https://dev.azure.com/dnceng-public/public/_build/results?buildId=673380&view=ms.vss-build-web.run-extensions-tab). I wonder if there are any worries with `stp` similar to the `ldp` issue on Apple Silicon. If so I...

> do you mean "before" the data node of the second indirection? Can you share an example or a dump to visualize this? No, the first indirection needs to happen...

@kunalspathak Any other feedback or are you ok with merging this as is?

> Ah, sorry, I thought I already approved. The JitStress* failures are existing issues I assume? Yep, they look known.

Sorry for getting to this so late. If you still would like to add this, then I wouldn't have a problem accepting this change. I think the variables should just...

Since `endRef` is the reference to the polygon that contains `endPos`, it is guaranteed that any path to `endPos` ends with this polygon. However, due to the the A\* search...

Yeah sure, I would love to see that.

> the search needs to expand with 1 more 'goal' node, that once chosen as the cheapest from the open list, is truely the best path. But how do you...