Five New IDA Processor Modules — What Obscure Architecture Should Be Next?

Over the past few months, I’ve been exploring some of the less-traveled corners of processor architecture, developing IDA processor modules for architectures that aren’t natively supported.

So far, I’ve implemented five:

  • XMOS xCORE — Event-driven multicore architecture used in embedded audio, USB, and real-time applications.

  • TI C7x / C7000 — Advanced DSP architecture featuring parallel execution, vector instructions, and specialized processing units.

  • TI PRU / PRU-ICSS — Real-time processors found in TI SoCs, particularly interesting for industrial communication and control.

  • AMD/Xilinx MicroBlaze — Configurable soft-core processor used in FPGA-based systems.

  • NVIDIA SASS — Native GPU machine-code architecture with architecture-specific instruction encodings and execution models.

Each presented different challenges, from unusual instruction encodings and parallel execution to instruction-set variations, control-flow analysis, and binary formats.

What started as a technical challenge has gradually become a collection of processor modules. Now I’m looking for the next interesting addition.

I’m particularly interested in obscure or under-supported architectures, especially those encountered in commercial firmware, embedded devices, DSPs, and specialized processors.

A question for fellow reverse engineers

Have you encountered a processor architecture that IDA doesn’t support, but you wish it did?

Ideally, something with an available instruction-set reference, a compiler or reference disassembler, and real-world firmware or executable binaries to analyze.

I’d be interested in hearing about architectures you’ve encountered, particularly those that have made a reverse-engineering project unnecessarily difficult.

I’m also happy to hear feedback from anyone working with the modules above.

What’s the most obscure processor you’ve encountered — and should it be the next one?

Impressive lineup. Tackling VLIW execution like C7x alongside native SASS encoding is no small feat.

One piece of feedback on distribution: if these modules remain closed-source, providing precompiled binaries for Linux and macOS (x86_64 and arm64) is really a must, as many embedded workflows aren’t on Windows.

That said, I’d strongly consider open-sourcing them. From an OPSEC perspective, loading unverified third-party binaries into IDA is a tough sell for most security researchers. Plus, reversing obscure targets often requires tweaking instruction semantics or fixing edge-case decoding bugs on the fly, which isn’t possible with a binary blob. Publishing the source would definitely drive much higher adoption.

As the only IDA LICENSE I have is the Windows 8.4 unfortunately I can’t do Neither Linux nor macOS. Not to menssion newer versions of IDA. For now I will keep it close source. I understand why ppl will have concerns with only binaries (I would certainly have …). The only suggestion for now is you can run it sandboxed under some virtual environment …

Needles to say it is benign…

Thank you for the swift response

Please have a look at Hex-Rays' Contributor Program

I think another useful thing would be a small set of test binaries for each architecture, along with the expected disassembly.

I’m not entirely sure how practical that is for all five modules, especially with some of the more unusual instruction formats.

But it might make it easier for people to report decoding problems without needing access to the source code. Someone could point to a specific instruction or binary offset rather than just saying the output looks wrong.

And it could be useful when you eventually move to newer IDA versions too. At least you’d have something consistent to compare against.

That’s great but I wish Hex-Rays would provide a comprehensive API to craft your own decompiler lifter/frontend as IDA supports more than 70 unique processor families and the official decompiler only supports a dozen of those.