On Thu, Mar 02, 2017 at 05:41:59AM +0000, Dibyendu Majumdar wrote: > On 2 March 2017 at 05:21, Luc Van Oostenryck > <luc.vanoostenryck@xxxxxxxxx> wrote: > >> Anyway have hit a bunch of other issues with sparse-llvm ... :-( > > > > Each day its problem (and happily its solution too!). > > > > I will submit test cases for the new problems when I get some time. OK. Thanks. > But I am beginning to think that there is quite a bit of work needed > to fix the issues, and in any case it might be better to create an > LLVM backend from the parse tree rather than the linearized version. Possible, but this will certainly need some work too. > The current approach is very low level - for example struct member > access bypasses the natural LLVM way of doing it. Yes, that's true. > This approach will > have the consequence that LLVM will generate poor quality code as it > will not have enough information to optimise properly. The current > approach is more suited to backends that directly emit machine code I > think. As far as I understood, the idea behind the linearized code and the optimization made on it was to have it's own backend. Sparse-llvm only came much later. Regards, Luc -- 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
