Hi Luc, On 10 August 2017 at 00:15, Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> wrote: > On Thu, Aug 10, 2017 at 12:23 AM, Dibyendu Majumdar > <mobile@xxxxxxxxxxxxxxx> wrote: >> >> Thank you. I am trying out this change as I am hoping it will help >> avoid the incorrect simplifications we saw in some cases. So far my >> findings are: >> > >> But it doesn't help with the issue with pseudos defined in one branch >> of the code (https://github.com/dibyendumajumdar/dmr_c/blob/master/tests/bugs/simplifybug.c). >> >> Is there another fix / patch that you made to overcome above issue or >> would you expect both issues to be fixed by this change? > > I don't know exactly what you have as problem with this other test. > I quickly looked at the output of test-linearize and I saw no problem > with self-defined pseudos. The phi-nodes are very wrong though but that's > another problem. > Yes I suppose that is what it is. LLVM complains that 'instruction does not dominate all uses'. The unsimplified IR works fine (https://github.com/dibyendumajumdar/dmr_c/blob/master/tests/bugs/simplifybug.lin). The simplified IR has an issue (https://github.com/dibyendumajumdar/dmr_c/blob/master/tests/bugs/simplifybug_O1.lin). > You may try the new SSA construction at : > https://github.com/lucvoo/sparse/tree/sssa > It will help a lot. > Is that a standalone change I can apply? > But in both case, I saw that sparse-llvm crashes > (which is normal as none of the LLVM fixes are applied here). > > In the coming days, I'll do a branch that aggregates all the good stuff. > I think that this is a very good idea. Since the official sparse repo is so behind - it will be good if you maintain a branch in your repo that has all the things that you have tested thoroughly and are happy with - this can be the "next sparse version" for testing. I would suggest aggregating only changes that you are 100% confident about. It will allow me (and others like me) to test the changes and report back if any issues are found. > But until then, can you explain exactly what is wrong with this second test? > >> I will run my test cases with this change and report if anything breaks. > > Thanks. > Testing in other environments, with other goals, is very useful. > So far I haven't found any breaks which is good. I will continue testing tomorrow. Regards Dibyendu -- 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
