https://webassembly.studio/ allows inspection of WebAssembly (WASM) files and the corresponding SpiderMonkey-generated x86 code. I'd like to similarly inspect instructions generated by V8's WASM compilers (Liftoff and TurboFan).
I'm entirely unfamiliar with V8's codebase/API (I compiled & linked it and followed some tutorials, though). There seems to be a v8::CompiledWasmModule class available, but it does not seem to expose access to generated x86/x64 instructions by either Liftoff or TurboFan.
WebAssembly - adding a new opcode describes the process of adding a WASM opcode to V8. Seemingly appropriate functions for WASM compilation/execution are available in the mentioned classes. Though, these seem rather deeply layered within the V8 codebase and would be difficult to access were I to link V8 as a library. Also, I'm unsure if this corresponds to Liftoff or TurboFan.
Could anybody familiar with the V8 codebase give me some pointers as to how I can access Liftoff and/or TurboFan's WebAssembly compilation module, as to obtain x86/x64 code?
To inspect generated code, you can run the d8 shell with the --print-wasm-code flag. You'll need either a debug build, or a release build with the v8_enable_disassembler = true GN arg.
There's no existing way to retrieve generated code via V8's API; so if that's what you want, then you'd have to add it. Keep in mind that V8 is not designed to be a standalone compiler, which means generated code assumes that it's going to run "inside V8", so if you wanted to use it for anything else, you'd have to make significant modifications.
Related
I am trying to compile C++ code to wasm and then embed it in other C++ code with wee8 (v8's wasm-api). Currently I'm getting a Segfault on instantiating the module:
auto instance = wasm::Instance::make(store, module.get(), imports);
Note that I have no problem embedding code that I write as .wat and convert to .wasm, so the problem is specifically with embedding code compiled with emcc.
I am guessing that what I'm missing is WASI support in wee8? Does it exist? How can I enable it? Alternatively: can I ask emcc not to generate any WASI calls?
Here is a minimal example which results in:
Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)
In cpp42.cpp:
int main() {
return 42;
}
Compiling this to wasm with:
emcc -O3 cpp42.cpp -o cpp42.wasm
Inspecting the compiled wasm module with wabt's wasm2wat shows that it contains the following import
(import "wasi_snapshot_preview1" "proc_exit" (func (;0;) (type 0)))
Which I suspect to be the cause of the problem.
Then embedding with wee8 like in the examples in the repo and like I do with other wasm files causes the segfault mentioned above.
Just as another check: running
wasmer cpp42.wasm
echo $?
> 42
Works without a problem.
I can answer part of your question:
I'm missing is WASI support in wee8? Does it exist?
No, wee8 does not implement WASI. Adding such support is theoretically possible, but not currently scheduled to get done.
You can implement it yourself in your wee8 embedder, and make it available to loaded modules via imports. Most (or all?) of it could probably be a reusable (among many engine implementations) library, potentially offered and maintained by the WASI project itself. (I don't know whether such a library exists already.)
You didn't say what imports object you're currently passing; it needs to be an array of wasm::Extern* pointers that's at least as long as the imports of the module, and ordered equivalently (i.e. imports[i] will be the module's ith import).
(I agree that the Wasm C/C++ API is very barebones currently. Unless/until that is changed, you'll have to build any convenience mechanisms yourself. It's all possible with the information that's available, it's just clearly less convenient than instantiating Wasm modules from JavaScript.)
VPP provides the I/S for developing custom plugins that can be hooked into a graph of nodes. I've only seen examples for such plugins written in the C language, and was wondering whether other language, Go for instance, can also be used to write such plugins.
I have no idea what "VPP" is but nonetheless the answer is: "maybe"; here's why:
Go code is able to interface with C libraries via its facility known as cgo.
cgo is a multiple-faceted thing: it allows you to "export" certain Go functions in a certain way so that they can be called from the C side, and it allows you to call functions from the C side. It also allows you to write bits of inline C code to provide glue for the C side, when necessary.
Since some time Go building toolset (at least its "reference" implementation) provides for compiling Go code into a static or dynamic library with C-compatible API.
See this.
With these things in mind, in theory, it should be possible to do what you're after.
Note some possible obstacles:
Most of the time, if a "platform" allows you to write a "plugin" in C, it presupposes your plugin will make extensive use of the platform's own API.
This usually means your plugin is supposed to include certain header files provided by the platform.
The platform might also require your plugin to link against some platform-provided library (usually shared), or libraries.
cgo can do all of the above, but you will need to scrutinize the API provided by the platform and maybe write Go helpers to make its usage more natural for the Go code.
Building/linking issues (usually the locations of the header files and the libs) may also be a thing to solve.
I'm trying to build an old application that depends on fftw. It was written against fftw2 and I am currently on fftw3 (specifically 3.3.8). It fails to link because of undefined reference to fftw_create_plan and fftw_one. Indeed, my libfftw does not have those functions anymore; the following returns nothing:
readelf -s /usr/lib/libfftw3.so | grep 'fftw_create_plan\|fftw_one'
It looks like the api has changed significantly since the code was written. Is there a compatibility layer I can use or should I just go learn fftw3's new interface?
You cannot link an FFTW2 code against FFTW3 libraries, as you are coming to realize yourself. There is also no complete interface between the two, cause the apis are really not compatible.
Having said that, you may of course link your code against FFTW2 libraries. You still can obtain them. Why is that not an option?
I've been writing a Python extension use the Python/C API to read data out of a .ROOT file and store it in a list of custom objects. The extension itself works just fine, however when I tried to use it on a different machine I ran into some problems.
The code depends upon several libraries written for the ROOT data manipulation program. The compiler is linking these libraries dynamically, which means I cannot use my extension on a machine that does not have ROOT installed.
Is there a set of flags that I can add to my compilation commands to make these libraries statically linked? Obviously this would make the file size much larger but that isn't much of an issue providing that the code runs at the same speed.
I did think about collating all of the ROOT libraries that I need into an 'archive' file. I'm not too familiar with this so I don't know if that's a good idea or not.
Any advice would be great, I've never really dealt with the static/dynamic library issue before.
Thanks, Sean.
I understand it is somehow making a connection so that a compiler when envokes connects a source code to whatever libraries that it needs to.
But what is going on a more technical level, or better put what do I need to know in order to confidentally compile code.
I'm working with C++ and MinGW, and have started to look into build files and stuff for Sublime Text 2 (Have learned mostly under unix, or Java + eclipse so far). But what I don't understand what is adding a compiler to your path do for you?
Do I need to add it for every folder I want to compile from? Or is it system wide? I'm really learning this stuff for the first time, we we're never showed how to set up development environments or even deploy code on other systems.
You probably mean include paths and library paths in the compiler:
include paths: where the compiler will look for headers; and
library paths: where the linker, invoked by the compiler, will look for binary libraries to finish building your project.
If that is the case, look here for a gentle explanation.
Basically, what is happening is that the compiler looks in certain places for symbols defined by the operating system and other libraries installed system-wide.
In addition to those paths, you need to tell the compiler where to find the symbols defined in your own project.
You may also mean something related to installing the compiler itself or configuring the editor to use it.
In that case, what is happening is that you need to tell the build system where to find the executable for the compiler.
Basically, what is probably happening is that your editor wants to know where the compiler is so that it can provide real time feedback on your code. Adding the compiler to the system path will usually, but not always, solve your problem.
In more detail:
A C++ build is a rather complex tool chain, involving determining dependencies, preprocessing, compiling, and linking. There are tools that automate that tool chain, and those tools are in turn wrapped into the functionality of modern IDEs like Eclipse, Visual C++, or Sublime Text 2. You many need to tell your editor where to find the tools it uses to provide you with those services.