Ah, I didn't remember that we didn't have type refinement there. Yes, if we implemented that, I would think it should help very significantly.
- Alon On Wed, Dec 10, 2014 at 1:24 PM, Nicholas Wilson < [email protected]> wrote: > Hi, > > One of the big limitations of ASYNCIFY I found was the fact that indirect > calls (virtual methods) are pessimistically treated as async. At the stage > when the Asyncify transformation happens, there's just about enough type > information left in the IR to refine this. I had the code to do that in a > branch, but I never got it quite working (I'm an LLVM newbie!) The code > size decrease is considerable though if you could get that optimisation to > work. I might be able to come back to it sometime, but not in employed time > unfortunately. > > The heuristics of classifying functions as async/possibly async/sync are > good otherwise though, and if carried through to indirect calls might well > be useful for the emterpreter approach too. > > Nick > > -- > You received this message because you are subscribed to the Google Groups > "emscripten-discuss" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "emscripten-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
