I am working on a JIT that uses LLVM. The language has a small run-time written in C++ which I compile down to LLVM IR using clang
clang++ runtime.cu --cuda-gpu-arch=sm_50 -c -emit-llvm
and then load the *.bc files, generate additional IR, and execute on the fly. The reason for the CUDA stuff is that I want to add some GPU acceleration to the runtime. However, this introduces CUDA specific external functions which gives errors such as:
LLVM ERROR: Program used external function 'cudaSetupArgument' which could not be resolved!
As discussed here, this is usually solved by including the appropriate libraries when compiling the program:
g++ main.c cmal.o -L/usr/local/cuda/lib64 -lcudart
However, I am not sure how to include libraries in JITed modules using LLVM. I found this question which suggested that is used to be possible to add libraries to modules in the JIT like this:
[your module]->addLibrary("m");
Unfortunately, this has been deprecated. Can anyone tell me the best way to do this now? Let me know if I need to provide more information!
Furthermore, I am not really sure if this is the best way to be incorporating GPU offloading into my JIT, so if anyone can point me to a better method then please do :)
Thanks!
EDIT: I am using LLVM 5.0 and the JIT engine I am using is from llvm/ExecutionEngine/ExecutionEngine.h, more specifically I create it like this:
EngineBuilder EB(std::move(module));
ExecutionEngine *EE = EB.create(targetMachine);
You need to teach your JIT engine about other symbols explicitly.
If they are in a dynamic library (dylib, so, dll) then you can just call
sys::DynamicLibrary::LoadLibraryPermanently("path_to_some.dylib")
with a path to the dynamic library.
If the symbols are in an object file or an archive, then it requires a bit more work: you would need to load them into memory and add to the ExecutionEngine using its APIs.
Here is an example for an object file:
std::string objectFileName("some_object_file.o");
ErrorOr<std::unique_ptr<MemoryBuffer>> buffer =
MemoryBuffer::getFile(objectFileName.c_str());
if (!buffer) {
// handle error
}
Expected<std::unique_ptr<ObjectFile>> objectOrError =
ObjectFile::createObjectFile(buffer.get()->getMemBufferRef());
if (!objectOrError) {
// handle error
}
std::unique_ptr<ObjectFile> objectFile(std::move(objectOrError.get()));
auto owningObject = OwningBinary<ObjectFile>(std::move(objectFile),
std::move(buffer.get()));
executionEngine.addObjectFile(std::move(owningObject));
For archives replace template types ObjectFile with Archive, and call
executionEngine.addArchive(std::move(owningArchive));
at the end.
Related
I've found this How do I use a C library in a Rust library compiled to WebAssembly?, but this relies on wasm-merge, which has been discontinued. My problem is the following, I have some C++ code that I would like to call from Rust in order to have the option to compile the resulting package either to native code for use in mobile apps or to Webassembly for use in Node.js. At the moment, I have the following setup:
libTest.cpp
extern "C"{
int test_function(int i){
return i;
}
}
lib.rs
use wasm_bindgen::prelude::*;
#[link(name = "Test")]
extern "C"{
pub fn test_function(i: i32) -> i32 ;
}
#[wasm_bindgen]
pub fn test_function_js(i : i32) -> i32{
let res = unsafe{test_function(i)};
res
}
build.rs
fn main() {
cc::Build::new()
.cpp(true)
.file("libTest.cpp")
.compile("libTest.a");
}
This compiles and works when compiling to native code using a simple cargo build, but does not work for building to wasm, for which I'm doing cargo build --target wasm32-unknown-unknown. There I get the two errors
= note: rust-lld: error: /[path to my project]/target/wasm32-unknown-unknown/debug/build/rustCpp-cc5e129d4ee03598/out/libTest.a: archive has no index; run ranlib to add one
rust-lld: error: unable to find library -lstdc++
Is this the right way to go about this and if yes, how do I resolve the above error? If not, how do I best go about calling C++ from Rust and compiling it to wasm?
(This is not really a full answer, but too long for a comment.)
I can compile your example with
cc::Build::new()
.archiver("llvm-ar") // Takes care of "archive has no index" - emar might be an alternative
.cpp_link_stdlib(None) // Takes care of "unable to find library -lstdc++"
… // rest of your flags
but I'm not sure whether the resulting binary is useful to you. Especially, it contains WASI imports when compiled in debug mode, and you'll probably get linker errors if you start using any interesting functions (e.g. sin).
You could in theory give the C++ compiler a full stdlib to work with through .flag("--sysroot=/usr/share/wasi-sysroot/") (if you have wasi-sdk or wasi-libc++ installed), but
I'm unsure how to best account for differences of where this folder is usually installed (maybe like this)
I think you have to also pass this flag at link time, but I don't know how (it seems to work without, though)
that would target wasi, and may not be useful for whatever bindgen-based environment you have in mind.
Compilers such as GCC and Clang allow to compile C++ programs without the C++ standard library, e.g. using the -nostdlib command line flag. It seems that such often fail to link thou, for example:
void f() noexcept { throw 42; }
int main() { f(); }
Usually fails to link due to undefined symbols like __cxa_allocate_exception, typeinfo for int, __cxa_throw, __gxx_personality_v0, __clang_call_terminate, __cxa_begin_catch, std::terminate() etc.
Even a simple
int main() {}
Fails to link with
ld: warning: cannot find entry symbol _start; defaulting to 0000000000400120
and is killed by the OS upon execution. Using -c the compiler still runs the linker which blatantly fails with:
ld: error in mytest(.eh_frame); no .eh_frame_hdr table will be created.
Is it a realistic goal to program and compile C++ applications or libraries without using and linking to the standard library? How can I compile my code using GCC or Clang on Linux? What core language features would one be unable to use without the standard library?
You will basically find all of your questions answered at osdev.org, but I'll give a brief summary anyway.
When you give GCC -nostdlib, you are saying "no startup or library files". This includes:
crti.o, crtbegin.o, crtend.o and crtn.o. Generally kernel developers only care about implementing crti.o and crtend.o and let GCC supply crtbegin.o and crtend.o by passing -print-file-name= to the linker. Generally these are just stubs that consist of .init and .fini respectively, leaving room for GCC to shove the contents of crtbegin.o and crtend.o respectively. These files are necessary for calling global constructors/destructors.
You can't avoid linking libgcc (the "low-level runtime library" (-lgcc) because even if you pass -nostdlib GCC will emit calls to its functions whenever you use it, leading to inexplicable linking errors for seemingly no reason. This is the case even when you're implementing/porting a C library.
You don't "need" libstdc++ no, but typically kernel developers want it. Porting a C library then implementing the C++ standard library from scratch is an extremely difficult task.
Since you only want to get rid of the "standard library", but keeping libc (on a Linux system) you're essentially programming C++ with just a C library. Of course, there's nothing wrong with this and you do you, but ultimately I don't see the point unless you plan on developing a kernel.
Required reading:
OSDev's C++ page - If you really care about RTTI/exception support, it's more annoying to implement than it sounds. Typically people just pass -fno-rtti or -fno-exceptions and then worry about it down the line or not at all.
"Standard" is a misnomer. In this context it doesn't mean "the library (set of functions, classes etc) as defined by the C++ standard" but "the usual set of libraries and objects (compiled files in a certain format) gcc links with by default". Some of those are necessary for most or even all programs to function.
If you use this flag, it's your responsibility to provide any missing functionality. There are several ways to do so:
Cherry-pick libraries and objects that your program really needs out of the default set. (Makes little sense as the result will most probably be exactly the same as with the default link flags).
Provide your own implementation of missing functionality.
Explicitly disable, through compiler flags, language features your program isn't using. I know of two such features: exceptions and RTTI. This is needed because the compiler needs to generate exceptions-related code and RTTI info even if these features are not explicitly used in this module.
I want to use LEAP Motion in D.
Therefore It doesn't have C library and It has only C++ library.
I tried SWIG 2.0.9 below command.
swig -c++ -d -d2 leap.i
This command output Leap.d, Leap_im.d, Leap_wrap.cxx, Leap_wrap.h.
However, I don't know how to use to wrapper in D and I can't find how to use the wrapper.
Link error displays to use it intact.
How use these wrapper in D2?
And can I use without Leap.cpp (source of Leap.dll)?
Update:
Thanks two answers. and sorry for reply late because of busy.
Say first conclusion I could build Leap sample code on Win64 by following the steps below.
Output wrappers by above command.
Create x64 DLL with VC2010 from Leap_wrap.cxx, Leap_wrap.h, and import Leap.lib(x64).
Compile Leap.d and Leap_im.d with dmd -c.
Build LeapTest.d with Leap.obj and Leap_im.obj
all command is below.
swig -c++ -d -d2 leap.i
dmd -c Leap.d Leap_im.d -m64
dmd LeapTest.d Leap.obj Leap_im.obj -m64
execute LeapTest.exe (require x64 Leap.dll and Leap_wrap.dll)
I could run Leap Program.
But program crach onFrame event callback.
I'll try again on x86 and investigate the causes.
Few helpful links (some information may be outdated):
http://klickverbot.at/blog/2010/11/announcing-d-support-in-swig/
http://www.swig.org/Doc2.0/D.html
http://www.swig.org/tutorial.html
I have never used SWIG personally but my guess based on general knowledge about SWIG:
Leap_wrap.cxx is C++ source file that wraps calls to C++ functions from target library in extern(C) calls
Leap_wrap.h is header file with all extern(C) wrappers listed
Leap_im.d is D module based on Leap_wrap.h with same extern(C) function listed
Leap.d is D module that uses Leap_im.d as an implementation and reproduces API similar to original C++ one.
So in your D code you want to import Leap.d module. Than compile Leap_wrap.cxx to an object file with your C++ compiler and provide D object files, Leap_wrap.o and target library at linking stage. That should do the trick.
P.S. Leap.cpp source should not be needed. All stuff links directly from Leap_wrap.cxx to target library binary.
Go to IRC, either FreeNode, or OFTC, channel #D. In order to help you, we have to see what is in those files. My first guess is that you have to compile both D files, and the C++ file into object files, and link them together. I suppose SWIG is going to flatten the C++ API into bunch of C functions, and that is probably what Leap_wrap.cxx does.
If the LEAP API is not complex (ie. just bunch of simple C++ classes), it may be possible to directly interface with it. Read more about it here: http://dlang.org/cpp_interface.html .
I'm currently working on a compiler project using llvm. I have followed various tutorials to the point where I have a parser to create a syntax tree and then the tree is converted into an llvm Module using the provided IRBuilder.
My goal is to create an executable, and I am confused as what to do next. All the tutorials I've found just create the llvm module and print out the assembly using Module.dump(). Additionally, the only documentation I can find is for llvm developers, and not end users of the project.
If I want to generate machine code, what are the next steps? The llvm-mc project looks like it may do what I want, but I can't find any sort of documentation on it.
Perhaps I'm expecting llvm to do something that it doesn't. My expectation is that I can build a Module, then there would be an API that I can call with the Module and a target triple and an object file will be produced. I have found documentation and examples on producing a JIT, and I am not interested in that. I am looking for how to produce compiled binaries.
I am working on OS X, if that has any impact.
Use llc -filetype=obj to emit a linkable object file from your IR. You can look at the code of llc to see the LLVM API calls it makes to emit such code. At least for Mac OS X and Linux, the objects emitted in such a manner should be pretty good (i.e. this is not a "alpha quality" option by now).
LLVM does not contain a linker (yet!), however. So to actually link this object file into some executable or shared library, you will need to use the system linker. Note that even if you have an executable consisting of a single object file, the latter has to be linked anyway. Developers in the LLVM community are working on a real linker for LLVM, called lld. You can visit its page or search the mailing list archives to follow its progress.
As you can read on the llc guide, it is indeed intended to just generate the assembly, and then "The assembly language output can then be passed through a native assembler and linker to generate a native executable" - e.g. the gnu assembler (as) and linker (ld).
So the main answer here is to use native tools for assembling and linking.
However, there's experimental support for generating the native object directly from an IR file, via llc:
-filetype - Choose a file type (not all types are supported by all targets):
=asm - Emit an assembly ('.s') file
=obj - Emit a native object ('.o') file [experimental]
Or you can use llvm-mc to assemble it from the .s file:
-filetype - Choose an output file type:
=asm - Emit an assembly ('.s') file
=null - Don't emit anything (for timing purposes)
=obj - Emit a native object ('.o') file
I don't know about linkers, though.
In addition, I recommend checking out the tools/bugpoint/ToolRunner.h file, which exposes a wrapper combining llc and the platform's native C toolchain for generating machine code. From its header comment:
This file exposes an abstraction around a platform C compiler, used to compile C and assembly code.
Check out these functions in llvm-c/TargetMachine.h:
/** Emits an asm or object file for the given module to the filename. This
wraps several c++ only classes (among them a file stream). Returns any
error in ErrorMessage. Use LLVMDisposeMessage to dispose the message. */
LLVMBool LLVMTargetMachineEmitToFile(LLVMTargetMachineRef T, LLVMModuleRef M,
char *Filename, LLVMCodeGenFileType codegen, char **ErrorMessage);
/** Compile the LLVM IR stored in \p M and store the result in \p OutMemBuf. */
LLVMBool LLVMTargetMachineEmitToMemoryBuffer(LLVMTargetMachineRef T, LLVMModuleRef M,
LLVMCodeGenFileType codegen, char** ErrorMessage, LLVMMemoryBufferRef *OutMemBuf);
To run the example BrainF program, compile it and run:
echo ,. > test.bf
./BrainF test.bf -o test.bc
llc -filetype=obj test.bc
gcc test.o -o a.out
./a.out
then type a single letter and press Enter. It should echo that letter back to you. (That's what ,. does.)
The above was tested with LLVM version 3.5.0.
I have been running into trouble recently trying to symbolicate a crash log of an iOS app. For some reason the UUID of the dSYM was not indexed in Spotlight. After some manual search and a healthy dose of command line incantations, I managed to symbolicate partially the crash log.
At first I thought the dSYM might be incomplete or something like that, but then I realized that the method calls missing were the ones occurring in C++ code: this project is an Objective-C app that calls into C++ libraries (via Objective-C++) which call back to Objective-C code (again, via Objective-C++ code). The calls that I'm missing are, specifically, the ones that happen in C++ land.
So, my question is: is there some way that the symbolication process can resolve the function calls of C++ code? Which special options do I need to set, if any?
One useful program that comes with the apple sdk is atos (address to symbol). Basically, here's what you want to do:
atos -o myExecutable -arch armv7 0x(address here)
It should print out the name of the symbol at that address.
I'm not well versed in Objective-C, but I'd make sure that the C++ code is being compiled with symbols. Particularly, did you make sure to include -rdynamic and/or -g when compiling the C++ code?
try
dwarfdump --lookup=0xYOUR_ADRESS YOUR_DSYM_FILE
you will have to look up each adress manually ( or write a script to do this ) but if the symbols are ok ( your dSym file is bigger than say 20MB) this will do the job .