I'm just starting to get acquainted with LLVM and I have been looking for information regarding the actual state of LLVM's JIT API and implentation. A lot of examples I had found, show usage either of ExecutionEngine class or MCJit/ORC API, but never show them in conjunction.
Therefore my question is: what exactly is the relation between ExecutionEngine class and MCJit/ORC API? Currently, my understanding is that ExecutionEngine is a kind of API class for both MCJit and ORC, hiding their implementation details. Is this correct? Can it be assumed that one should always use ExecutionEngine class instead of MCJit/ORC API? If so, why official LLVM's tutorial uses ORC API directly and doesn't even mention the existence of ExecutionEngine class? Is it just outdated?
Related
I'm working on my project for completion undergraduate courses, consisting of an experimental analysis of registers allocation algorithms. For this task, I am using the set of tools from the LLVM project.
However, I have read the documentation of the LLVM project and not yet found a way to put the pieces of the puzzle together. So far I know:
As passes work as engage them to LLVM and know I must implement MachineFunctionPass pass.
I follow some suggestions to look at the Basic Allocator code, but could not understand much.
The allocator I intend to implement is based on graph coloring heuristics, as the theory of operation of such methods I’m well aware.
So I look for is some sort of "How to", a defined set of steps to implement such allocator. It sounds like carelessness, but I have to deliver the work in about six months and I'm a little confused.
If anyone can give me some guidance or reference to any supporting material (besides the own documentation), I would be grateful.
My English may be a little wrong, I am a Brazilian student.
Well, after a while I have finished my project and have written an tutorial of How to write an LLVM register allocator using the RegAllocBase interface.
The tutorial have been written in reStructuredText and it's available on GitHub: https://github.com/nael8r/How-To-Write-An-LLVM-Register-Allocator
How can in retrieve the Operating system context for my application programmatically?
As others have posted this site answers the question: http://msdn.microsoft.com/en-us/library/windows/desktop/dn424972(v=vs.85).aspx
Use the Version Helper API
To know whether this will actually suit your needs though -Why- you want to do this needs to be examined. If this is related to feature support testing, then you may want to read the suggestions here: http://msdn.microsoft.com/en-us/library/windows/desktop/ms724832(v=vs.85).aspx. The article discusses some various ways to perform feature detection such as:
A combination of LoadLibrary and GetProcAddress
GetSystemMetrics e.g. GetSystemMetrics(SM_CMONITORS)
Of course, this is more relevant if you are directly coding outside of the .NET world.
In ruby there is a method on the Kernel object called set_trace_function. It will execute a given proc object on occasion, when a method is called or returns, or when an exception is raised. You can use this to make your own loggers and learn interesting things about your program &c...
Is there a similar facility in C++ or maybe in boost? Also, what is this facility called in general?
Thanks!
z.
Like Seth said this sort of stuff isn't directly available by the language. In order to do something similar in C++ you have to inject this sort of "monitoring code" into your own code during compilation.
In C or C++ this is often referred to as instrumentation or profiling.
To learn more I suggest you google it.
If what you are looking for instead is a logging or debugging helper framework, then you might want to read this (which is actually the second answer when you google instrumentation):
Instrumentation (diagnostic) library for C++
LLVM is very modular and allows you to fairly easily define new backends. However most of the documentation/tutorials on creating an LLVM backend focus on adding a new processor instruction set and registers. I'm wondering what it would take to create a VHDL backend for LLVM? Are there examples of using LLVM to go from one higher level language to another?
Just to clarify: are there examples of translating LLVM IR to a higher level language instead of to an assembly language? For example: you could read in C with Clang, use LLVM to do some optimization and then write out code in another language like Java or maybe Fortran.
Yes !
There are many LLVM back-end targeting VHDL/Verilog around :
(open source) Legup paper
(commercial) Xilinx HLS
(online) C-to-verilog
And I know there are many others...
The interesting thing about such low-level representations as LLVM or GIMPLE (also called RTL by the the way) is that they expose static-single assignments (SSA) forms : this can be translated to hardware quite directly, as SSA can be seen as a tree of multiplexers...
There's nothing really special about the LLVM IR. It's a standard DAG with variable arity. Decompiling LLVM IR is a lot like decompiling machine language.
You might be able to leverage some frontend optimizations such as constant folding, but that sounds pretty minor compared to the whole task.
My only experience with LLVM was writing a binary translator for a class project, from a toy CISC to a custom RISC.
I'd say, since it's the closest thing to a standard IR (well, GCC GIMPLE is a close second), see if it fits with your algorithms and style and evaluate it as one alternative.
Note that GCC also started out prioritizing portability above all, and has also accomplished a lot.
I'm not sure I follow how parts of your question relate one to another.
To target LLVM into a high-level language like C is very possible and you seem to have found one reference point.
VHDL is a whole other business however. Do you consider VHDL a high-level language? It may be, but but describing hardware/logic. Sure VHDL has some constructs that you can employ to actually program in it, but it's hardly a fruitful endeavor. VHDL describes hardware and thus makes translating LLVM IR into it a very hard problem, unless of course you design a CPU with a custom instruction set in VHDL and translate LLVM IR into your instructions.
This thread was one of the first things I found while looking for the same thing.
I found a project that's rather far along that cleanly builds under/with llvm 3.5. It's pretty darn cool. It spits out HDL and does various other cool FPGA related things. While it's designed to work with TTAs and generate images for FPGA (or simulate them), it can probably also be made to do some trivial HDL generation from c functions.
It was perfect for my purposes because I wanted to upload to an Altera FPGA, and the fpga_stdout example even spits out Quartus build scripts and project files.
TTA-Based Co-design Environment
I also tried the things listed in the accepted answer and a couple others and found that they weren't going to work for me or weren't very high quality (usually both). TCE is professional feeling, but purely academic I believe. Very nice all the way around.
It seems the question was partially answered, so I’d like to give it a shot:
What it would take to create a VHDL backend for LLVM?
What it would take to translate LLVM IR to a higher level language (presumably with the intention of converting between high-level langs)?
I will give you some background on 2. And expand at a later date on 1.
If you want to convert LLVM IR to a high-level language such as C or Java:
You would have to take the LLVM instructions, and abstract that out into its equivalent C code. Then you need to take the remaining features that LLVM does not have an equivalent for (like classes and abstractions for C++) and write a routine that would find those patterns in the LLVM (like reused blocks) and write C. For the basic stuff, its pretty straightforward. But, just follow the train of thought and you quickly find yourself realizing the true difficultly of the problem, after all not everyone writes simple C. To compound the difficulty further, you may not get the same LLVM IR when compiling the generated C! (Consider the resulting feedback loop)
As for Java, you are in for an even harder battle going direct from LLVM IR, and in either case still have the problem you likely won't get the same code compiling to LLVM IR, if one even can do that. Rather, you would translate LLVM IR to JVM Bytecode. Then you could use a reverse compiler to get your Java.
A group of Chinese students was apparently able to do this, but they wondered why such little interest in their research. I would say its bc they don't fully understand just what the LLVM guys have done, and how it is better than the JVM. (In fact, LLVM arguably makes the JVM obsolete ;)
Even though this seems useful in that one can use LLVM as an intermediary between C and Java to convert bidirectionally, this solution is actually of little use because we are asking the wrong question. See, the entire reason you would want that for practical purposes is to have a common code base and increase performance.
But the real problem is that we need a language that has abstracted the common features of modern languages, and that gives you a central language that you can build from. http://julialang.org/ has answered the question 😉
Looks like the best place to start is with the CBackend in the LLVM source:
llvm/lib/Target/CBackend/CBackend.cpp
tl,dr: I don't think LLVM is the right tool
What your are looking for is way to translate LLVM code to a higher language that's what emscripten do for Javascript.
But it looks like you miss a bit the point of LLVM as it's meant to generate static code in order to achieve that they use a specific intermediate language build for that purpose.
As you can see the way emscripten works is by implementing a stack, but without using javascript as a human would have done it.
They are several project that try to achieve what you original question was, like MyHDL that turns python to VHDL or Verilog.
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 3 years ago.
Improve this question
What are your tips on implementing a plugin style system?
In C (and I think C++ too although I haven't done it myself), this is most typically done using dynamically loaded modules. The API:s for such are platform-dependent.
On POSIX (Linux), you use the dlopen() family of functions. Basically you build your plugin separately, then load it at run-time, look up its symbols by name, and can then call them.
For Win32, there is LoadLibrary() which does something very similar, you build your code into a DLL.
For a handy wrapper that makes all of these easy and transparent, check out GLib's GModule API.
In the '92/'93 time frame I worked on a plugin architecture for Aldus PageMaker, which was coded in C++. PageMaker was built on a C++ OOP framework called VAMP, which assisted its portability between Mac OS and Windows.
So we tried to use the features of C++ to build a plugin architecture. This proved to be very problematic for C++ classes due to the so-called brittle base class problem. I proceeded to write a paper that was published in journals and that I presented at OOPSLA '93 in a reflection workshop. I also made contact with Bjarne Stroustrup at a Usenix conference in Portland and proceeded to dialog with him for several months, where he championed the issue of dealing with the brittle base class problem on my behalf. (Alas, other issues were deemed more important at that time.)
Microsoft introduced the COM/DCOM system and for that platform that was looked on as a viable solution to the problem. C++ could be used as an implementation language for COM via abstract classes used to define COM interfaces.
However, these days developers shun away from COM/DCOM.
In contrast, NeXT devised a plugin architecture using Objective C in the early 90s in the NeXT Step framework. Today that lives on vibrantly in Mac OS X on Apple's computers and important platforms such as the iPhone.
I submit Objective C enabled solving the plugin problem in a superior manner.
I personally regard the brittle base class problem of C++ to be it's most fatal flaw.
If were building a plugin architecture with the C-based family of languages, would do so using Objective C.
The best platform and language neutral advice I can give is this:
Design your entire app around the plugin SDK.
IMO, a plugin SDK should not be an afterthought. If you design your app to basically be an empty shell which loads plugins, then the core features are implemented in your own SDK, you get the following benefits:
High modularity of components, and clear separation of purpose (it kind of forces your architecture to be good)
It forces your SDK to be really good
It allows other third party developers to make extremely powerful, core-level features as well
New developers/hires can easily start work on a major new feature without having to touch the main app - they can do all their work in a plugin (which prevents them screwing up anything else)
In C/C++, you probably use dynamic link libraries and either function pointers (C) or interfaces (classes solely consisting of pure virtual methods, for C++). However even if you use Javascript, I'd still recommend the above architecture.
Qt provides QPluginLoader:
http://qt-project.org/doc/qt-4.8/qpluginloader.html
If you need/want more fine grained control, Qt also provides a means to load libraries on the fly with QLibrary:
http://qt-project.org/doc/qt-4.8/qlibrary.html
Even better, these are portable across platforms.
This may not be what you're looking for, but you could embed a scripting language in your application, such as Lua. Lua was designed to be embedded in other programs and used as a scripting language for writing plugins. I believe it's fairly easy to add the Lua interpreter to your program, though I don't know Lua so I can't vouch for how effective of a solution this would be. Others with more experience with Lua, please add comments about your experience with embedding Lua in another application.
This would, of course, mean that your plugins need to be written in Lua. If you don't like Lua then the de-facto standard Perl, Python and Ruby interpreters are all written in C, and can be embedded in a C program. I know of a number of programs that use these languages as scripting language extensions.
However, I don't know what you're looking for, as your question is a little vague. Perhaps more information about what you want people to be able to do with said plugins would be appropriate. For some tasks, a full-blown scripting language may be a bit overkill.
I have written an article about how to implement a plugin system using Dynamic Linking Libraries. The article is written from the point-of-view of a Windows programmer but the technique can be applied to a Linux/Unix type environment.
The article can be found here: http://3dgep.com/?p=1759
The main point is, you should create a "common" DLL that is implicitly linked by both the main application (the core application) and by the plugin implementations. The plugins can then be explicitly linked and loaded dynamically at run-time by the core application.
The article also shows how you can safely share static (singleton) instance of a class across multiple DLLs by using the "common" DLL.
The article also shows how you can export a "C" function or variables from a DLL and use the exported functions in the application at run-time.
It's best to use a framework like ACE (http://www.cs.wustl.edu/~schmidt/ACE.html) that shields you (as good as possible) from platform specific coding.
ACE contains a plugin framework that is based on shared libraries that you can use to create dynamically assembled applications.
For a higher level abstraction check out CIAO (http://www.cs.wustl.edu/~schmidt/CIAO.html) the Open Source C++ implementation of the CORBA Component Model.
Look at Poco Class Loader, it can be interesting for you.
I have written a plugin library Pugg that loads C++ classes from dll files and here is the logic I used:
User exports a c function from dll that has a unique name. This name has to be unique enough as functions cannot be distinguished using their arguments while loading from dlls.
C function registers one or several factory classes called "Driver". Every Driver class is associated with a string. When the main application wants to create a class, it gathers the related factory class using the associated string. I also implemented a version checking system to not load old plugins.
Dll loading is accomplished using the LoadLibraryA and GetProcAddress functions (Pugg currently works on windows).
One thing worth mentioning is that main application and dlls should be compiled using the same compiler and using the same compilation options (release/debug modes, optimization settings, stl versions etc...). Otherwise there might be issues with mapping of classes.
I have had some success using a fairly naive system:
Create API Specification for plug-ins
Use a singleton plug-in manager
Use LoadLibrary/GetProcAddress based run time dynamic linking
Implement Inversion of control based event handling for notifying the plug-ins
This podcast on plugin architectures might also be interesting.