Well, once you have the LLVM backend compiler running, build times will be reduced, which would help our team quite a bit. I would be sad if we were stuck on the JavaScript compiler just because we require a resizable heap. :)
On Thu, Dec 19, 2013 at 9:53 PM, Alon Zakai <[email protected]> wrote: > I don't think it would buy us anything over what you are currently doing, > though. It would look superfically more similar, but the underlying issues > would remain (heap can change, so js engines optimize less, and we must > limit our eliminator as well). > > - Alon > > > > On Thu, Dec 19, 2013 at 9:36 PM, Chad Austin <[email protected]> wrote: > >> Great. Even if it's not as fast as validated asm.js, that would be fine. >> >> >> On Thu, Dec 19, 2013 at 9:14 PM, Alon Zakai <[email protected]>wrote: >> >>> We could emit something close to asm but with heap resizing, but it >>> would not be as optimizable because the heap would not be constant. Without >>> proper browser support, our options are limited I'm afraid. >>> >>> - Alon >>> >>> >>> >>> On Thu, Dec 19, 2013 at 8:55 PM, Chad Austin <[email protected]> wrote: >>> >>>> On Thu, Dec 19, 2013 at 5:53 PM, Alon Zakai <[email protected]>wrote: >>>> >>>>> 1) faster compilation speed, no need to process IR in JS, can use LLVM >>>>> IR in C++ directly >>>>> >>>> >>>> I'm super excited about this! In particular, it means we can avoid >>>> invoking llvm-dis on the build and avoid multiply parsing the LLVM IR >>>> files. >>>> >>>> >>>>> In time we can support all those things, although there are some >>>>> features we never will - the new compiler will stay streamlined by >>>>> focusing >>>>> on one mode of codegen, optimized and relooped asm.js, as opposed to the >>>>> old compiler which supported several other modes (non-asmjs typed arrays, >>>>> and no typed arrays). Of course the old compiler will remain viable for >>>>> things that need those codegen modes. Otherwise, things like C++ >>>>> exceptions >>>>> etc. should certainly be supported in the new compiler and are just a >>>>> matter of time and how much people need them. >>>>> >>>> >>>> I agree 100% with dropping support for TA0 and TA1 codegen modes. I >>>> doubt many people depended on those, especially now that even IE supports >>>> typed arrays. >>>> >>>> Will there be a way in the new compiler to emit asm.js-like code but >>>> with runtime heap resizing? We can't use asm.js until at least Chrome and >>>> Firefox support resizable typed arrays. :/ >>>> >>>> Again, very excited! >>>> >>>> Thanks, >>>> Chad >>>> >>>> -- >>>> 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/groups/opt_out. >>>> >>> >>> -- >>> 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/groups/opt_out. >>> >> >> >> >> -- >> Chad Austin >> Technical Director, IMVU >> http://www.imvu.com/members/Chad/ >> >> -- >> 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/groups/opt_out. >> > > -- > 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/groups/opt_out. > -- Chad Austin Technical Director, IMVU http://www.imvu.com/members/Chad/ -- 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/groups/opt_out.
