NLnet User-operated Grant Request for 3mdeb Power ISA Simulator

  • 2021-08-071

Project name

Libre-SOC 3mdeb Cavatools: Power ISA Simulator

Website / wiki

Please be short and to the point in your answers; focus primarily on the what and how, not so much on the why. Add longer descriptions as attachments (see below). If English isn't your first language, don't worry - our reviewers don't care about spelling errors, only about great ideas. We apologise for the inconvenience of having to submit in English. On the up side, you can be as technical as you need to be (but you don't have to). Do stay concrete. Use plain text in your reply only, if you need any HTML to make your point please include this as attachment.

Abstract: Can you explain the whole project and its expected outcome(s).

Cavatools is currently a high performance user-operated simulator of the RISC-V ISA. The primary objective of the project is to extend it to implement the scalar Power ISA and the Libre-SOC Draft SVP64 Extensions. This will allow rapid prototyping of Extensions to the Power ISA long before they reach silicon (which is very costly). In turn this helps Libre-SOC to deliver on its commitment to provide user-trustable processors for use in Internet routers, desktop, smartphone and other user-operated devices where security and transparency is expected.

Have you been involved with projects or organisations relevant to this project before? And if so, can you tell us a bit about your contributions?

3mdeb is currently helping Libre-SOC with the (horribly slow, easy-to-read, easy-to-use) Libre-SOC Power ISA Simulator which is 20,000 times slower than cavatools. 3mdeb is also helping with ISA level unit tests in Libre-SOC's code base that will be used to cross-validate a huge range of Power ISA simulators and actual silicon implementations.

Requested Amount

EUR $50,000.

Explain what the requested budget will be used for?

  • To create a compiler which takes Libre-SOC Machine-readable Power ISA specification files and generate c code
  • To extend cavatools to include support for the Scalar parts of the Power ISA
  • To then add support for Libre-SOC's Draft SVP64 Extensions
  • To enhance it to include gdb "remote" machine interface support
  • To add Power ISA RADIX MMU emulation
  • To extend cavatools to run a very basic linux initramfs in-memory with basic serial console access
  • To demonstrate running first a single core linux kernel and later a SMP one, with busybox
  • To use the exact same Specification c compiler to create an "illegal instruction trap" emulator, integrated into the linux kernel, for emulating Power ISA SIMD instructions (extending the existing trap-and-emulate code already present in ppc64 linux kernel source code)

Does the project have other funding sources, both past and present?

Although there is NLnet funding for the Libre-SOC Simulator (written in python) and associayed unit tests, cavatools, which is written in c by Peter Hsu, does not have funding for the Power ISA aditions. cavatools itself is a very new project.

Compare your own project with existing or historical efforts.

Although there are quite a few Power ISA simulators, none of them are up-to-date or are suited to high performance, like cavatools. cavatools is multi-process and extremely fast, using relatively little memory, where power-gem5, which has a different focus and has huge flexibility and usefulness for research, uses vast amounts of memory and is much slower. cavatools also has hardware-level cycle-accurate emulation which is extremely useful and important for analysing experimental instructions, which is a feature that no other Power ISA Simulator has. DolphinPC and pearpc are over 15 years old and were targetted at 32 bit emulation of much older Power ISA processors. Libre-SOC's python-based simulator only achieves aroubd 2,000 instructions per second on high performance hardware whereas cavatools achieves 200,000 instructions per second per processor on modest hardware. IBM's own Power ISA simulator is proprietary and, because it contains confidential experimentation internal to IBM, may not be made public.

What are significant technical challenges you expect to solve during the project, if any?

This is at its heart a compiler project, which can be a challenging area. However the language being implemented is quite small and limited, so the project is relatively straightforward.

Describe the ecosystem of the project, and how you will engage with relevant actors and promote the outcomes?

The project will be developed entirely publicly and transparently, using Libre-SOC Project Resources which are already set up for trustable auditability and transparency. The mailing lists therefore are always publicly available.

Online conferences and talks will be given as progress is made, as well as working with Libre-SOC to send out development reports and progress.

Extra info to be submitted