Of all the facets of HN’s title autodestroy, I think removing “How” from titles is the worst one. I believe OP can edit it back in though.
(I passed over this article thinking it was a “look how mysteriously smart the mysteriously smart compiler is” acticle, not a “here’s how the smarts in a compiler work” one.)
I think it'd fine having it be however broken it currently is as long as the correction was checked by whoever is submitting the entry before it gets published.
The escape analysis piece is what makes this particularly interesting. ZJIT can prove an object doesn't escape a method's scope and then eliminate the heap allocation entirely — the object lives on the stack (or in registers) and the load/store optimizations follow naturally once there's no indirection.
YJIT's profile-guided approach is powerful but pays a cost every time a hot path diverges from the expected type. The BBV approach in ZJIT bakes type assumptions directly into the compiled code, so you get the same specialization without the deopt overhead on the happy path. The tradeoff is code size — more type combinations means more compiled variants — but for server-side Rails apps where the method profile is fairly stable, that's usually fine.
Curious whether they're planning to share any of the escape analysis machinery upstream to YJIT, or if the JIT designs are diverging permanently.
I'm working on an interpreter right now, and I'm considering adding JIT support in the future. Are there other blog posts like this, or deep dives that talk about how to implement and tune a JIT?
My main high-level advice would be to have an extensive set of behavioral tests (lots of scripts with assertions on the output). This helps ensure correctness when you flip on your JIT compilation.
You'll eventually run into hard to diagnose bugs - so be able to conditionally JIT parts of your code (per-function control - or even better, per-basic block) to help narrow down problem areas.
The other debugging trick I did was spit out the full state of the runtime after every instruction, and ensure that the same state is seen after every instruction even w/ JIT enabled.
You got your reply already. To add: YJIT is the one that does "basic block versioning" (Which was Maxime's thesis) while ZJIT is a more traditional design.
I am confident in that description but don't actually know what it means in practice (yes I've seen papers and talks, but I kinda need not-compiler-engineer to explain it to me.)
As I understand it BBV still holds promise, but the sheer volume of knowledge of more traditional methods might mean it gets better outcomes (also IIRC ZJIT is still lagging YJIT).
(I passed over this article thinking it was a “look how mysteriously smart the mysteriously smart compiler is” acticle, not a “here’s how the smarts in a compiler work” one.)
YJIT's profile-guided approach is powerful but pays a cost every time a hot path diverges from the expected type. The BBV approach in ZJIT bakes type assumptions directly into the compiled code, so you get the same specialization without the deopt overhead on the happy path. The tradeoff is code size — more type combinations means more compiled variants — but for server-side Rails apps where the method profile is fairly stable, that's usually fine.
Curious whether they're planning to share any of the escape analysis machinery upstream to YJIT, or if the JIT designs are diverging permanently.
Maybe you'll find the resources I link to in the documentation for my project helpful.
https://github.com/ZQuestClassic/ZQuestClassic/blob/main/doc...
Or perhaps you'd find reviewing my usage of asmjit helpful:
https://github.com/ZQuestClassic/ZQuestClassic/blob/main/src...
My main high-level advice would be to have an extensive set of behavioral tests (lots of scripts with assertions on the output). This helps ensure correctness when you flip on your JIT compilation.
You'll eventually run into hard to diagnose bugs - so be able to conditionally JIT parts of your code (per-function control - or even better, per-basic block) to help narrow down problem areas.
The other debugging trick I did was spit out the full state of the runtime after every instruction, and ensure that the same state is seen after every instruction even w/ JIT enabled.
Good luck!
I am confident in that description but don't actually know what it means in practice (yes I've seen papers and talks, but I kinda need not-compiler-engineer to explain it to me.)
As I understand it BBV still holds promise, but the sheer volume of knowledge of more traditional methods might mean it gets better outcomes (also IIRC ZJIT is still lagging YJIT).
It would be nice to have ZJIT on speed.ruby-lang.org!