Kewen asked me to have a look at this since Richard was reluctant
to approve it (given that it was his idea).
TBH I don't know much about the pass machinery, so I'm not sure I'm
really best placed. There again, perhaps I know just enough to realise
that this is indeed the hack that it's billed
Hi!
>> Can you repeat the compile-time measurement there? I also wonder
>> whether we should worry about compile-time at -O[12] when SLP is not run.
>> Thus, probably rename the cleanup pass to pre_slp_scalar_cleanup and
>> gate it on && flag_slp_vectorize
>
> Good idea, will evaluate it.
>