developed by Marcel Köster
A modern, lightweight & fast GPU compiler for high-performance .Net programs
ILGPU is a new JIT (just-in-time) compiler for high-performance GPU programs (also known as kernels) written in .Net-based languages.
It combines the convenience of C++ AMP with the high performance of CUDA.
Functions in the scope of kernels do not have to be annotated (e.g. default C# functions) and are allowed to work on value types.
All kernels (including all hardware features like shared memory, atomics and warp shuffles) can be executed and debugged on the CPU using the integrated multi-threaded CPU accelerator.
And the best feature: it's free! ILGPU is released under the University of Illinois/NCSA Open Source License.
ILGPU is a free and non-sponsored project. It is being developed by a professional and passionate compiler, GPU and computer graphics developer. Support the project with contributions or some small donations in order to speed up the development process and to keep the project alive.
Thank you for your feedback on version 0.4-beta. Based on hundreds of emails and several bug reports I will release a much improved version of the compiler in December.
A new public Discord server has been created.
I've just created a new Discord server to simplify communication: https://discord.gg/KTvqYZP. For general questions or feature requests, I recommend using the new Discord server.
Instead of publishing v0.5, I decided to first release a public beta version (called v0.4.0-beta) of the new compiler to get additional feedback from the community. Note that ILGPU is now completely written in C# without any native dependencies.
A new public Google group has been created.
Due to the large number of emails I usually receive, I have decided to set up a public discussion group. For general questions or feature requests, I recommend using the public discussion forum.
The roadmap has been updated
Version v0.5 will not longer use LLVM for code generation. Instead, ILGPU will offer a novel Intermediate Representation (IR), a set of relevant code transformation phases and a new custom backend. The implementation is based on .Net Standard, which ensures that ILGPU programs can run on all .Net implementations without the need to cross-compile and maintain native library dependencies.
A new major version has been released
The new ILGPU version includes several important enhancements: dotnetcore support, convenient kernel loading and caching. Please refer to the updated ILGPU samples or the upgrade guide for more information. Note that all libraries contain required backwards compatibility that is marked with the Obsolete attribute.
High performance kernel compilation, dispatch and execution times. Furthermore, type-safe kernel delegates avoid boxing.
Use the power of C# or VB.Net to write high-level kernels and execute them on the GPU. No need to program C++, Cuda or OpenCL.
Single- or multi-threaded execution of kernels on the CPU. This is also useful for debugging or emulation of specific target platforms.
High-level kernel debugging using your favorite .Net debugger. Furthermore, the single-threaded execution feature allows to focus on the algorithm instead of the parallelism.
Functions do not have to be annotated in order to use them in the scope of kernels.
Compile your applications for any cpu. ILGPU will automatically adjust everything else for X86 or X64 platforms.
Focus on the algorithm and not on the details. Implicitly grouped kernels let you implement high-level kernels without paying attention to low-level index computations or tiling.
Multi-dimensional index types simplify address computations and kernel writing.
No pointer arithmetic and dramatically simplified index computations due to views to memory regions.
Support for shared (scratch-pad) memory in kernels via array views. Static or dynamic allocation of shared memory is supported.
Easy access to atomic functions and low-level-intrinsics like warp shuffles. All functions are supported during CPU debugging.
Default math functions and operations are mapped to high-performance math functions. Furthermore, there is support for fast math and forced 32bit math to avoid doubles.
Exceptions require support for exception handlers and a limited support for reference types. Changes of the "intended" control flow (which can be caused by exceptions) are currently not supported. However, there might be a conversion phase in the future that converts several exceptions into debug assertions.
Debug assertions are supported on all accelerators. Note that debug assertions are not available in Release mode.
Reference types are currently not supported. However, a limited support for reference types will be added in the future. This will also allow the implementation of delegates.
Lambda functions (or delegates in general) are currently not supported since they require a limited support for reference types and custom code-transformation passes. Support for lambda functions will be added in the future.
There is currently no support for harware-based kernel debugging. This will be added in the future. However, CPU-based kernel debugging is recommended in all cases due to the advanced debugging and testing capabilities.
Version 0.3 supports .Net Standard 2.0. Consequently, ILGPU can also run in the scope of the .Net Framework or .Net Core.
.Net Standard support is already available, which allows writing portable .Net applications with .Net Core. However, the platform-dependent parts of ILGPU must be ported to Linux and Mac. Linux library support has been tested with Ubuntu and Fedora. General support for Linux and Mac will be available in one of the next releases. This allows you to compile your application (including GPU code) only once.