On Thu, Aug 17, 2017 at 2:09 AM, Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> wrote: > > I just took a quick look at this one. It's interesting. > The IR seems correct, the locking is correct too. > The problem seems that the 'correct' presence of phi-nodes make > that some BB that were merged are not anymore. Nothing is wrong, I assume you need the phi-node relocation similar to the one used in "dead code elimination using ssa". That will get merge the block with phi node in it. > just different code. But the context checking is quite limited and see now > a path where the context could differ. I practice they won't differ and with > previous code the context checking saw that because this patch was > unexisting due to more aggressive merge of BB. Interesting.Context checking is very sensitive and give a lot of false positives. > One way to see at it is this is a false warning that appears because > of a lack of BB optimizations. > I've a vague idea of what can be done here. I'll look at it later, this evening. Thanks. I will take a look at your patches series. Chris -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
