Regarding the specific questions: * Memory initialization: we actually just moved to doing this a few days ago :) On emscripten incoming and binaryen master, asm2wasm will import the memory initialization into the wasm module.
* establishStackSpace() receives parameters, while start() does not, so we can't use start(). Overall I think having an option for more standalone wasm output could be useful. But some things currently need JS support: * Our libc does syscalls out into JS, so without the JS side you'd need to implement those yourself. * The libc also reuses web platform things like data/time support, by calling out (in a non-syscall manner) to JS. * If you want integration support for input/output, WebGL, Web Audio, etc. then you definitely need JS. However, for code that just does pure computation, with little or no libc/OS support, it should be possible to emit a standalone wasm module. But we'd need to figure out things like establishStackSpace, and I suspect there is more I can't think of now, experimenting in this direction is probably the best way to find out. On Sat, Sep 10, 2016 at 12:29 PM, Joshua Warner <allpowerfu...@gmail.com> wrote: > Hi, > > I've been experimenting with implementing my own WebAssembly interpreter ( > https://github.com/joshuawarner32/rust-wasm), and I'm starting to throw > some more complicated examples at it. An emscripten/binaryen-compiled `int > main() { return 42; }` works just fine, but once I graduate to `int main() > { printf("Hello"); }`, I start running into some problems with > initial-conditions that the WebAssembly module expects. > > Taking a step back, it looks like emscripten is generating a large > javascript wrapper around the wasm module, and the wasm module is highly > dependent on this external helper. I'd like to avoid shipping a javascript > engine next to my wasm interpreter, which means I'm left translating that > wrapper code by hand. > > Some of the relevant things that the wrapper seems to contain: > * Memory initialization - why can't this be done with standard `(memory > ... (segment ...))` stuff, directly in the wasm module? > * Calling some initialization routines - possibly including > `establishStackSpace` - again, why can't this be done in the `start` > function of the wasm module? > * (Some other things that I'm clearly missing, since I have yet to get > HelloWorld working on _any_ wasm interpreter (including v8), without the > wrapping javascript) > > In short, I'd like to get emscripten to generate a wasm module that, to > the extent possible, doesn't depend on any javascript wrapper. > > How small can I get the logic in that wrapper? > > Any pointers on places in the emscripten code that I can poke at to push > things in this direction? I'd love to upstream some improvements if > possible. > > Thanks, > Joshua > > -- > 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 emscripten-discuss+unsubscr...@googlegroups.com. > 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 emscripten-discuss+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
