Pihlaja

So further complication, even with constants: if we merge two ops with immediates and then try to reassoc, we can end up moving the actually useful operands into ops that...

opt_reassoc is now a separate pass and constant reassocs are enabled as these don't require any dominance checks. For non-constant cases (a+b)+c -> (a+c)+b needs to check that c dominates...

Also the (a+b)+(c+d) case... if we handle it as (a+b)+x, we end up with (a+(c+d))+b and we'd rather prefer to do the reverse... so simplistic rewriting might not be a...

With [2f220a3](https://github.com/signaldust/bunny-jit/commit/2f220a3a3f2cb045ddfc6d7fd858e678e9fe3045) commutative ops should now get reliably reassociated to promote LICM. Since [40eaa54](https://github.com/signaldust/bunny-jit/commit/40eaa54e669f73f2fd1792817e2281804720dd19) we can also handle some cases of add/sub (and float mul/div). Unlike commutative ops, this isn't...

Looking at Chrome's dev-tools, there doesn't seem to be anything useful in console (ie. no errors); the only issue it complains is about component edit box form field elements not...

Interesting, I just tried on another system running Windows and on an Android phone.. and I'm getting the same (broken) results on both. Tried clearing all side data too, no...

Ok, I figured out what is going on. There is both a "square wave" and "pulse" in the waveform drop-down. Choosing "square wave" works perfectly fine and allows changing the...

Just to sanity check, this is happening with optimization level set to at least 1? The lcu+callp is a reasonable thing to do. Most likely what is going is an...

bacc03e forbids RA from moving constants to satisfy input constraints, forcing a rematerialization instead, which seems sufficient for bin/test_callp to directly load function address into x9 on ARM (the same...

I could use a repro test-case for the second issue with the excess pushes/pops... Since the only optimization that runs after RA is the DCE (dead-code elimination) this must be...