Displaying 2 results from an estimated 2 matches for "instcomine".
Did you mean:
instcombine
2015 Jul 23
0
[LLVMdev] some superoptimizer results
...ut otherwise.
- At least for me, the ability to reason about non-nullness is more
important than the value of particular bits. If I were to start this
from scratch, I'd probably start by making isKnownNonNull dominating
condition aware. Given that starting place, a separate pass (i.e.
not-instcomine) might make far more sense and help to decouple the code
more.
- I suspect that many of the cases found by this approach could be
implemented as special cases. To use nullness as an example,
recognizing the pattern { if (!p) p = malloc(X) } is much easier than
handling general control flow. T...
2015 Jul 23
4
[LLVMdev] some superoptimizer results
> I just noticed: most of the results in this batch seem to be about exploiting `[zs]ext i1` having cost 1
> in order to replace a select of cost 3.
> Could you do a run where select has cost 1 and [zs]ext i1 (and trunc to i1) has cost 2 or 3?
I tried this (or something quite similar) earlier and it caused Souper to
introduce *a lot* of selects. So the problem is that Souper's