All: our poor server is smoking today* so I've had to reduce the page size of comments. There are 1500+ comments in this thread but if you want to read more than a few dozen you'll need to page through them by clicking the More link at the bottom. I apologize!
Also, if you're cool with read-only access, just log out (edit: or use an incognito tab) and all will be fast again.
* yes, HN still runs on one core, at least the part that serves logged-in requests, and yes this will all get better someday...it kills me that this isn't done yet but one day you will all see
# This match takes 10 sec. in Ruby 3.1, and 0.003 sec. in Ruby 3.2
/^a*b?a*$/ =~ "a" * 50000 + "x"
而另外一組 regexp 也可以看出類似的效果:
用一些記憶體空間換取效能,降低被 DoS 的一些機會。另外一方面,引入了 regexp timeout 的 workaround,緩解真的被打的時候的資源消耗上限:
The optimization above cannot be applied to some kind of regular expressions, such as those including advanced features (e.g., back-references or look-around), or with a huge fixed number of repetitions. As a fallback measure, a timeout feature for Regexp matches is also introduced.
Trigraph sequences were introduced in C89 as alternate spellings of some characters to allow the implementation of C in character sets which do not provide a sufficient number of non-alphabetic graphics
而且是強制要求實做:
Implementations are required to support these alternate spellings, even if the character set in use is ASCII, in order to allow transportation of code from systems which must use the trigraphs. AMD1 also added digraphs (see §6.4.6 and §MSE.4).
其中遇到的問題就是當年得決定 C 可以用的 charset,得考慮到很多不同機器 charset 相容性的問題:
The C89 Committee faced a serious problem in trying to define a character set for C. Not all of the character sets in general use have the right number of characters, nor do they support the graphical symbols that C users expect to see. For instance, many character sets for languages other than English resemble ASCII except that codes used for graphic characters in ASCII are instead used for alphabetic characters or diacritical marks. C relies upon a richer set of graphic characters than most other programming languages, so the representation of programs in character sets other than ASCII is a greater problem than for most other programming languages.
The solution is an internationally agreed-upon repertoire in terms of which an international representation of C can be defined. ISO has defined such a standard, ISO/IEC 646, which describes an invariant subset of ASCII.
The characters in the ASCII repertoire used by C and absent from the ISO/IEC 646 invariant repertoire are:
The preview supports code written in Python, Java, and JavaScript, using VS Code, IntelliJ IDEA, PyCharm, WebStorm, and AWS Cloud9. Support for the AWS Lambda Console is in the works and should be ready very soon.
至於 training 的資料集,這邊有提到的是 open source 專案與 Amazon 自家的東西:
CodeWhisperer code generation is powered by ML models trained on various data sources, including Amazon and open-source code.
開發應該需要一段時間,不知道是剛好,還是被 GitHub Copilot 轉 GA 的事件強迫推出 Preview 版...
首先是在 -O3 的情況下 (也就是作者使用的參數),可以看到類似的結果:(我桌機的 CPU 是定速,沒有跑動態調整)
$ repeat 10 ./a
[-] Took: 248830 ns.
[-] Took: 249150 ns.
[-] Took: 248760 ns.
[-] Took: 248730 ns.
[-] Took: 248770 ns.
[-] Took: 248861 ns.
[-] Took: 248760 ns.
[-] Took: 253050 ns.
[-] Took: 248640 ns.
[-] Took: 249211 ns.
$ repeat 10 ./b
[-] Took: 686660 ns.
[-] Took: 696090 ns.
[-] Took: 696310 ns.
[-] Took: 694431 ns.
[-] Took: 691971 ns.
[-] Took: 697690 ns.
[-] Took: 693241 ns.
[-] Took: 692900 ns.
[-] Took: 654751 ns.
[-] Took: 679101 ns.
從版本 A 的 objdump -d -S -M intel a 可以看到作者 screenshot 內也有看的 unroll 與 SSE2 指令集:
$ repeat 10 ./a
[-] Took: 571140 ns.
[-] Took: 570280 ns.
[-] Took: 571271 ns.
[-] Took: 573971 ns.
[-] Took: 571981 ns.
[-] Took: 569650 ns.
[-] Took: 566361 ns.
[-] Took: 571600 ns.
[-] Took: 571330 ns.
[-] Took: 571030 ns.
$ repeat 10 ./b
[-] Took: 697521 ns.
[-] Took: 696961 ns.
[-] Took: 696201 ns.
[-] Took: 694921 ns.
[-] Took: 696930 ns.
[-] Took: 695001 ns.
[-] Took: 701661 ns.
[-] Took: 698100 ns.
[-] Took: 702430 ns.
[-] Took: 702641 ns.
從 objdump 可以看到版本 A 的變化,退化成一次只處理一個,但把所有的數字都用 xmmN 存放計算:
$ repeat 10 ./a
[-] Took: 1097091 ns.
[-] Took: 1092941 ns.
[-] Took: 1092501 ns.
[-] Took: 1091991 ns.
[-] Took: 1092441 ns.
[-] Took: 1093970 ns.
[-] Took: 1091341 ns.
[-] Took: 1093931 ns.
[-] Took: 1094111 ns.
[-] Took: 1092231 ns.
$ repeat 10 ./b
[-] Took: 2703282 ns.
[-] Took: 2705933 ns.
[-] Took: 2703582 ns.
[-] Took: 2702622 ns.
[-] Took: 2703043 ns.
[-] Took: 2702262 ns.
[-] Took: 2703352 ns.
[-] Took: 2703532 ns.
[-] Took: 2703112 ns.
[-] Took: 2702533 ns.
When taking the geometric mean of all the Python benchmarks I carried out for this article on the AMD Ryzen 9 5950X, Python 3.11 Beta was about 41% faster overall than the current Python 3.10.4 stable release or 45% over the aging Python 3.8 series.