[LLVMdev] some superoptimizer results

[John Regehr]

> Your timing for mentioning this is slightly ironic.  I'm in the process of 
> deciding whether I should just delete all the code in question.  It's been in 
> tree for a while now and I haven't been able to justify the time to make it 
> ready for production.

Ouch, I'll try to get the next batch of Souper results done before you 
remove the code!  Also if I have time I'll do runs with and without your 
code and then we'll have another data point about its efficacy.

> - Enabling this seems to tickle a lot of unrelated bugs.  Taking the time to 
> triage those might be good purely from a robustness standpoint since in 
> *theory* this is just returning a result the known bits could have figured 
> out otherwise.

I would like to hear more. Do you have any suspicions about which pass 
you're tickling problems in? If you think it'll be helpful I can crank up 
Csmith with your option turned on and go looking for trouble.

Love to see the lessons learned, people should always do that.

John