Cranelift code generation comes to Rust
1–10 of 117 posts
Re: Cranelift code generation comes to Rust
#2Any fresh compilation time benchmarks and comparisons to LLVM?
Re: Cranelift code generation comes to Rust
#3Any fresh compilation time benchmarks and comparisons to LLVM?
From the article:
> A full debug build of Cranelift itself using the Cranelift backend took 29.6 seconds on my computer, compared to 37.5 with LLVM (a reduction in wall-clock time of 20%)
That seems much smaller difference than what I would have expected
Re: Cranelift code generation comes to Rust
#4Any fresh compilation time benchmarks and comparisons to LLVM?
The very last paragraph:
> A full debug build of Cranelift itself using the Cranelift backend took 29.6 seconds on my computer, compared to 37.5 with LLVM (a reduction in wall-clock time of 20%). Those wall-clock times don't tell the full story, however, because of parallelism in the build system. Compiling with Cranelift took 125 CPU-seconds, whereas LLVM took 211 CPU-seconds, a difference of 40%. Incremental builds — rebuilding only Cranelift itself, and none of its dependencies — were faster with both backends. 66ms of CPU time compared to 90ms.
Re: Cranelift code generation comes to Rust
#5This article provides an excellent overview of the latest in speed of optimizer vs quality of optimization.
In particular, copy-and-patch compilation is still the fastest approach because it uses pre-compiled code, though leaves little room for optimization.
Cranelift uses e-graphs to represent equivalence on the IR. This allows for more optimizations than the copy-and-patch approach.
Of course, the most optimized output is going to come from a more traditional compiler toolchain like LLVM or GCC. But for users who want to get "fast enough" output as quickly as possible, newer compiler techniques provide a promising alternative.
Re: Cranelift code generation comes to Rust
#6Any fresh compilation time benchmarks and comparisons to LLVM?
Not quite fresh, but
> A paper from 2020 [0] showed that Cranelift was an order of magnitude faster than LLVM, while producing code that was approximately twice as slow on some benchmarks.
Re: Cranelift code generation comes to Rust
#7This article provides an excellent overview of the latest in speed of optimizer vs quality of optimization . In particular, copy-and-patch compilation is still the fastest approach because it uses pre-compiled code, though leaves little room for optimization. Cranelift uses e-graphs to represent equivalence on the IR. This allows for more optimizations than the copy-and-patch approach. Of course, the most optimized o…
From what I understand, the big advantage of the e-graphs approach is, that the quality of the output is (within limits) a function the time and memory given.
The more memory, the more nodes can be generated in the e-graph and the more time for search, the better the selected node.
It might never be as fast as copy-and-patch or as good as LLVM or GCC, but this flexibility is a value in itself.
Re: Cranelift code generation comes to Rust
#8This article provides an excellent overview of the latest in speed of optimizer vs quality of optimization . In particular, copy-and-patch compilation is still the fastest approach because it uses pre-compiled code, though leaves little room for optimization. Cranelift uses e-graphs to represent equivalence on the IR. This allows for more optimizations than the copy-and-patch approach. Of course, the most optimized o…
Agree, especially when developing, I'd assume speed of optimizer matters more than quality of optimization. I wonder though if LLVM spent less time on that "phase ordering" problem, what is the tradeoff between these two factors?
Re: Cranelift code generation comes to Rust
#9Any fresh compilation time benchmarks and comparisons to LLVM?
From the article: > A full debug build of Cranelift itself using the Cranelift backend took 29.6 seconds on my computer, compared to 37.5 with LLVM (a reduction in wall-clock time of 20%) That seems much smaller difference than what I would have expected
Those numbers are the entire compile time, parsing/typecheck/MIR+MIR optimization/linking
Don't have numbers handy, so hard to say how much faster Cranelift is making the codegen portion, but gets into Amdahl's Law
Re: Cranelift code generation comes to Rust
#10Finally, looking forward for wider adoption.