VLaDOS Dev Blog: Intro (Also Titled: Do More Research Before Committing to a Project Like This)
VLaDOS is now defunct and its (empty) repository was removed. It had a very short life after I learned about already-existing tools to do meet its goals, such as LLHD (and, to an extent, Verilator). It makes far more sense to contribute to this project than to start something from scratch. I researched tools like this earlier this year and couldn’t find anything, but I guess LLHD released their paper earlier this year and I missed it.
Thanks to everyone who responded on Twitter with great feedback on this subject. I’m really glad there are a lot of people who work with HDLs who want to make the tool ecosystem better.
Today, I’ve finally begun work on the SystemVerilog compiler and simulator, VLaDOS. This project has been a long time coming, and I’ve had to learn a lot to even get to the point of being able to start writing it. This is probably the most ambitious project I have ever taken up, and is my first major open-source project. Because of that, I’m going to do development blogs for each alpha release of the tool that has major features – starting with this one.
What is VLaDOS?
VLaDOS aims to be a free and open-source SystemVerilog compiler and simulator. The project has a few goals:
- Obviously, compile code that adheres to the SystemVerilog LRM into a simulator executable.
- As much as is reasonable and possible, act as a drop-in replacement for most other EDA vendor compilers and simulators. Semiconductor companies should have minimal pain switching to this tool – otherwise they may never do so.
- Remain free and open-source to allow anyone to simulate SystemVerilog anywhere.
What’s with the name “VLaDOS?”
VLaDOS is an acronym for (System)Verilog Language: a Definitively Open Simulator. It is also a play-on-words of GLaDOS, the AI antagonist of the popular Portal video game series.
Why are you working on this?
I’ve discussed this in length in my “Modern” Hardware Design post, but the gist of the issue is that EDA vendors generally provide sub-par tools at extreme prices per license. Not only are these licenses are a non-trivial development cost for semiconductor companies, but any individual who isn’t able to pay for a license has no way of simulating SystemVerilog code on their local machine.
In this case, the only feasible option is to use EDAPlayground, but there are obvious downsides to this, namely that it requires an account, and that you have to run the code in a web browser.
I like software development, I work in the semiconductor industry, and I see this as a major pain point for the industry at large. Given the circumstances, I almost feel like I have an obligation to work on things such as this.
Basically all the reasons that anyone would choose Rust over C++ for an application like this. It’s more modern. It has attractive features. Its not tied to linking with other C++ code (aside from LLVM, which there are Rust bindings for). I can think of several more reasons off the top of my head, but you get the idea.
Maybe a few months down the line I’ll realize that Rust was a terrible choice and I’ll revert to using C++, but I highly doubt it.
This sounds great! How can I help?
Compilers are huge undertakings, and I know that I won’t be able to complete this project by myself. If you know Rust and want to work on an interesting domain-specific language that was used to code your CPU/GPU, feel free to chat on the project’s Gitter community. Eventually, descriptive issues will be filed, and you should look there instead.