rustc-perf VS nanoserde

> what do people use to run benchmarks on CI?

Typically, you purchase/rent a server that does nothing but sequentially run queued benchmarks (and the size/performance of this server doesn't really matter, as long as the performance is consistent), then sends the report somewhere for hosting and processing. Of course, this could be triggered by something running in CI, and the CI job could wait for the results, if benchmarking is an important part of your workflow.

But CI and benchmarks really shouldn't be run on the same host.

> What does the rust project use?

It's not clear exactly where the Rust benchmark "perf-runner" is hosted, but here are the specifications of the machine at least: https://github.com/rust-lang/rustc-perf/blob/414230abc695bd7...

> What do other projects use?

Essentially what I described above, a dedicated machine that runs benchmarks. The Rust project seems to do it via GitHub comments (as I understand https://github.com/rust-lang/rustc-perf/tree/master/collecto...), others have API servers that respond to HTTP requests done from CI/chat, others have remote GUIs that triggers the runs. I don't think there is a single solution that everyone/most are using.

hmm really really hard to answer :'), it's tradeoffs I think, no matter what you think Rust (cmiiw, I'm not qualified to say this) has (and probably in the future will adds more with guards on compiler metrics https://perf.rust-lang.org/) several phases that given the diffs to other language, might not available to any language compiler out there, if it's available I think rustc already did their best in here (some already being parallized etc etc, might be wrong since I can't refs any reference MRs, but it does exists though labels regarding this)

About a year ago I was looking for a tool like Rust perf for my application code. I did some research and found a lot of prior art. However, nothing checked all the boxes I was looking for, so I built Bencher!

https://www.pingcap.com/blog/rust-compilation-model-calamity... is a good overview. In general it varies depending on the crate but we track the performance at https://perf.rust-lang.org/ - if you look at cargo, for example, over 60% of the time is spent in codegen through LLVM: https://perf.rust-lang.org/detailed-query.html?commit=222d1f...

Something like https://perf.rust-lang.org/?

The performance full-report link is dead: https://github.com/rust-lang/rustc-perf/blob/master/triage/2022-10-04.md

Looking for better arguments, rustc might generate better IR because the compiler is more mature, has much more manpower, and has its performance [closely monitored[(https://perf.rust-lang.org/) and triaged. Zigc might generatte better IR because the language might be easier to generate IR for.

At the end of the day most people care about wall clock time. It's a real physical value that's easy to understand and easy to compare between systems. Plus, if two functions execute say, 1 billion instructions each, but one spends extra time stalled waiting for data fetches from RAM or doing IO, you definitely want to account for that in normal benchmarking.

Instruction counting is more of a specialized tool but I like to use it whenever I can because it has low variance and makes comparing changes a lot easier. Compare how bumpy these graphs are for instruction count (first link) and wall clock time (second link):

https://perf.rust-lang.org/

https://perf.rust-lang.org/?start=&end=&kind=raw&stat=wall-t...

what do you think of the parser in nanoserde? https://github.com/not-fl3/nanoserde/blob/master/derive/src/parse.rs

Yeah proc macros have a massive hole in their user experience. Nobody actually wants to use them without syn + quote functionality, so the majority of people just pay the compile time cost of syn/quote/procmacro2. But then a subset of people can't afford to pay the compile time cost, so they either: 1. Don't use them. And remove all of their deps that use them ... which is generally a lot. 2. Hand-manage the TokenStream (which is very much not fun). Macroquad doesn't want syn + quote in their tree because they care _deeply about compile times, so they opted to re-implement serde. Serde is the crowned jewel of the rust ecosystem. Someone feeling unable to use it is a massive failure of the system.