We've just opened up a large fraction of the tools we've developed as part of the Rigel project[1] at UIUC. The goal of the project is to investigate several aspects of efficient many-core processors, including microarchitecture, architecture, compiler optimization, and programming models. To this end, we've built an execution-driven simulator, LLVM-based cross-compiler toolchain, libraries, and parallel applications all targeting a common baseline architecture, Rigel.
With these tools, you should be able to simulate cross-compiled parallel programs running on 1024 cores by the end of the quick start guide. We are working hard to release the rest of what we have internally, including more benchmarks and an RTL model of the core.
When we started the project in 2007, there really wasn't much tooling available to do the kind of cross-cutting studies we wanted to do. While many existing projects have improved since then, and many new interesting tools have come about, we've found it very useful to have an integrated set of tools targeting an ISA we define. It has allowed us to easily evaluate, for example, the effect of instruction set extensions on hardware efficiency, performance, compiler optimizations, and the application programming experience.
I'll stick around and answer any questions folks might have; we're excited to work with people to improve these tools, and are actively looking for users and contributors.
2 comments
[ 2.8 ms ] story [ 14.4 ms ] threadWe've just opened up a large fraction of the tools we've developed as part of the Rigel project[1] at UIUC. The goal of the project is to investigate several aspects of efficient many-core processors, including microarchitecture, architecture, compiler optimization, and programming models. To this end, we've built an execution-driven simulator, LLVM-based cross-compiler toolchain, libraries, and parallel applications all targeting a common baseline architecture, Rigel.
With these tools, you should be able to simulate cross-compiled parallel programs running on 1024 cores by the end of the quick start guide. We are working hard to release the rest of what we have internally, including more benchmarks and an RTL model of the core.
When we started the project in 2007, there really wasn't much tooling available to do the kind of cross-cutting studies we wanted to do. While many existing projects have improved since then, and many new interesting tools have come about, we've found it very useful to have an integrated set of tools targeting an ISA we define. It has allowed us to easily evaluate, for example, the effect of instruction set extensions on hardware efficiency, performance, compiler optimizations, and the application programming experience.
I'll stick around and answer any questions folks might have; we're excited to work with people to improve these tools, and are actively looking for users and contributors.
[1] https://rigel.crhc.illinois.edu/