ml-compiler-opt VS capstone

I did a bit of work on this last summer on (much) smaller models [1] and it was briefly discussed towards the end of last year's MLGO panel [2]. For heuristic replacements specifically, you might be able to glean some things (or just use interpretable models like decision trees), but something like a neural network works fundamentally differently than the existing heuristics, so you probably wouldn't see most of the performance gains. For just tuning heuristics, the usual practice is to make most of the parameters configurable and then use something like bayesian optimization to try and find an optimal set, and this is sometimes done as a baseline in pieces of ML-in-compiler research.

1. https://github.com/google/ml-compiler-opt/pull/109

If you're using Clang/LLVM you can also enable ML inlining[1] (assuming you build from source) which can save up to around 7% if all goes well.

There are also talks of work on just brute forcing the inlining for size problem for embedded releases for smallish applications. It's definitely feasible if the problem is important enough to you to throw some compute at it [2].

1. https://github.com/google/ml-compiler-opt

2. https://doi.org/10.1145/3503222.3507744

LLVM's inlining-for-size and register-allocation-for-performance optimizations are both implemented using machine learning models trained by Google.

Continue reading | Checkout the paper, github, demo and ref article.

Looks like they do have a pretrained model:

https://github.com/google/ml-compiler-opt/releases/download/...

The code will by default auto-download it during the build process. It's about 800 kbytes, which seems very reasonable for something that will reduce the generated code size by gigabytes for a large codebase.

It just updated to the nearly latest LLVM, so that argument is void: https://github.com/capstone-engine/capstone/blob/next/docs/c...

capstone

Maybe then they can help us with the Capstone[1][2] disassembly engine auto-sync (automatic synchronization from the LLVM TableGen files) effort[3]. ARMv7, ARMv8/9, PowerPC are nearly finished, and MIPS in in near-term plans. Nobody stepped in for RISC-V yet.

[1] http://www.capstone-engine.org/

[2] https://github.com/capstone-engine/capstone

[3] https://github.com/capstone-engine/capstone/issues/2015

Bloaty is a nice tool.

When I worked on Matter a couple years ago, we had the problem that its backend http://www.capstone-engine.org/ did not support Xtensa, and produced some Python tools that could take output from bloaty or similar data from readelf or elftools, and produce several kinds of report.

https://github.com/project-chip/connectedhomeip/blob/master/...

$ git clone -b next https://github.com/capstone-engine/capstone $ cd capstone $ ./make.sh $ sudo ./make.sh install $ cd ..

People already closing in: [Implement an eBPF decompiler/disassembler · Issue #838 · capstone-engine/capstone · GitHub](https://github.com/capstone-engine/capstone/issues/838)