On Mon, Jun 26, 2023 at 5:05 AM René N. <wither2...@gmail.com> wrote:
> Hi Sam, > please excuse my late response. Thank you for all your answers! This > helped me so much. > > From all those information you gave me, I tried to make a diagram to > summarize the process of Emscripten while creating C->Wasm. See the picture > below. Did I got it right now? > Your picture looks about right to me, yes. I assume the first disgram is the expanded form of the second box of the second diagram? I don't know that we have any specific documentation about these internals, but I could be wrong. > [image: Emscripten C-to-Wasm Sketch.png] > > I am wondering if these information are documented somewhere. Since I need > those information for my uni paper, I have to list the sources and refer to > "official" documents. > s...@google.com schrieb am Montag, 5. Juni 2023 um 04:46:48 UTC+2: > >> On Fri, Jun 2, 2023 at 11:40 PM René N. <withe...@gmail.com> wrote: >> >>> Thank you for answering Sam! Very glad to get a feedback this quick. >>> >>> So if I understand it correctly, emcc is doing the whole C -> Wasm >>> Binary since it's the driver of the whole toolchain, right? >>> I think I am irritated because emcc is called "Emscripten Compiler >>> *Frontend*". So I thought that emcc is just a part of the compilation >>> process since a "Frontend" to me is the component of a compiler which takes >>> a source code and generates the IR while the Backend is takes the IR and >>> generates the target language. >>> >> >> You are right, it is not quite correct to call emcc a frontend, since it >> doesn't actually parse any code. clang is both a compiler driver, and a C >> frontend. >> >>> >>> Also I would like to ask: Where does the "(upstream) LLVM WebAssembly >>> Backend" comes into play? It should be in step 3 (LLVM IR -> wasm object >>> file), right? >>> >> >> Yes. >> >> >>> >>> Referring to step 4 & 5: While wasm-opt is a Binaryen-Tool, wasm-ld is a >>> LLVM tool, right? I'm asking this because Emscripten's website >>> <https://emscripten.org/docs/compiling/WebAssembly.html> says, that >>> "Emscripten's WebAssembly support depends on Binaryen". This statement >>> seems wrong, since (non-optimized) Wasm Binary is generated without >>> Binaryen. >>> >> >> Historically it has been true that binaryen was always required, but >> fairly recently we have made it possible to perform complete debug builds >> without binaryen (by moving a lot of things that binaryen used to do into >> python code in emscripten). The documentation is a little inaccurate >> there. However, to build anything for production (i.e.. any release >> build) binaryen is required, so I think we can still think of it as a >> required dependency of emscirpten. Binaryen is also requires even for >> debug builds in some cases (for example if you use asyncify there is no way >> to avoid it). >> >> >>> Referring step 6: emcc.py creates JS wrapper code if needed, does it >>> also creates the html code (optionally)? >>> >> >> Yes it can optionally create both JS and HTML. >> >> >>> >>> >>> Thanks again for helping me understand Emscripten and toolchains in >>> general! >>> >>> s...@google.com schrieb am Freitag, 2. Juni 2023 um 19:00:21 UTC+2: >>> >>>> Hi René, >>>> >>>> Great questions. It sounds like you have a pretty good understanding >>>> of the various phases. I will reiterate your list, filling in a few >>>> details for you. At the high level emcc is the compiler driver, rather >>>> than the compiler itself. gcc and clang both take this roll too, and under >>>> the hood both clang and gcc fork separate processes for the actual >>>> compiling and linking. >>>> >>>> emcc -> >>>> 1. clang.exe: C-Code -> LLVM IR >>>> 2. clang.exe: LLVM IR -> LLVM IR (optimized) (clang/llvm) ( this >>>> really happens as part of (1) when you build with optimizations enabled) >>>> 3. clang.exe: LLVM IR -> wasm object file >>>> 4. wasm-ld.exe: combine files -> Wasm Binary >>>> 5. wasm-opt.exe: Wasm Binary -> Optimized Wasm Binary (optional) >>>> 6. emcc.py: Generate JS wrapper code (optional) >>>> >>>> >>>> On Fri, Jun 2, 2023 at 9:25 AM René N. <withe...@gmail.com> wrote: >>>> >>>>> Hello guys, >>>>> >>>>> for a uni paper I need to explain how Emscripten works (on >>>>> higher-level) to emit WebAssembly-Binary in the end. >>>>> I'm new into compilers and toolchains and I'm not sure if I understand >>>>> it correctly, how Emscripten is converting compatible code (like C) to >>>>> WebAssembly Binary. >>>>> >>>>> If I got it right, for creating .wasm, these tools are used by >>>>> Emscripten: >>>>> >>>>> - emsdk as configurator for the whole toolchain >>>>> - emcc (which includes Clang+LLVM) >>>>> - the (upstream) LLVM WebAssembly Backend >>>>> - Binaryen >>>>> >>>>> So if i got it right, to create .wasm, the compilation works like this: >>>>> >>>>> 1. C-Code -> LLVM IR >>>>> 2. LLVM IR -> LLVM IR (optimized) >>>>> 3. LLVM IR (optimized) -> Wasm Binary >>>>> 4. Wasm Binary -> Wasm Binary (optimized) >>>>> 5. (Wasm Binary (optimized) -> JS) [optional] >>>>> >>>>> Did I forget something? >>>>> >>>>> What I am especially unsure of, is, which tool is doing what: >>>>> So emcc uses Clang+LLVM. Now I'm not sure if emcc emits LLVM IR only >>>>> (since Clang is creating LLVM IR), or does it also convert it to emit Wasm >>>>> Binary (which means that the upstream LLVM Wasm Backend lies in the emcc)? >>>>> What I'm quite sure of, is that step 1 is done by Clang and steps 4 & >>>>> 5 are done by Binaryen. >>>>> >>>>> Also, to me it seems like 'emcc' has 2 different meanings: 1. as >>>>> part(!!) of the compilation process and 2. as command representation for >>>>> the whole toolchain >>>>> >>>>> Also, which component/tool of Emscripten is creating the JS and HTML >>>>> Gluecode? >>>>> >>>>> I'm happy if someone can help me 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 emscripten-disc...@googlegroups.com. >>>>> To view this discussion on the web visit >>>>> https://groups.google.com/d/msgid/emscripten-discuss/3efb5151-1d5c-4ae1-89a5-b37ed3085793n%40googlegroups.com >>>>> <https://groups.google.com/d/msgid/emscripten-discuss/3efb5151-1d5c-4ae1-89a5-b37ed3085793n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>> . >>>>> >>>> -- >>> 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-disc...@googlegroups.com. >>> >> To view this discussion on the web visit >>> https://groups.google.com/d/msgid/emscripten-discuss/495eeb80-5d31-4d6a-a15d-92c9b215151fn%40googlegroups.com >>> <https://groups.google.com/d/msgid/emscripten-discuss/495eeb80-5d31-4d6a-a15d-92c9b215151fn%40googlegroups.com?utm_medium=email&utm_source=footer> >>> . >>> >> -- > 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. > To view this discussion on the web visit > https://groups.google.com/d/msgid/emscripten-discuss/8b07e3ae-db8a-45f7-ab48-ef293964405en%40googlegroups.com > <https://groups.google.com/d/msgid/emscripten-discuss/8b07e3ae-db8a-45f7-ab48-ef293964405en%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- 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. To view this discussion on the web visit https://groups.google.com/d/msgid/emscripten-discuss/CAL_va28VUfxjzGXg6za%3DgPuc9NMYkwgEACkLe-Fj-C_z-k70Xg%40mail.gmail.com.
