Go 1.1beta1 now on the benchmarks game

http://benchmarksgame.alioth.debian.org/u64/benchmark.php?test=all&lang=go&lang2=v8&data=u64

by jondot • 13 years ago

30 59 13 years ago

benchmarksgame.alioth.debian.org

melling• 13 years ago

It would be nice to compare it to the old version. At any rate, it seems like it's a lot more competitive with Java in performance and it uses a lot less memory:

http://benchmarksgame.alioth.debian.org/u64/benchmark.php?te...

It's certainly much faster than the fastest "scripting" language:

http://benchmarksgame.alioth.debian.org/u64/benchmark.php?te...

igouy22• 13 years ago

>>It would be nice to compare it to the old version.<<

12 days ago

https://news.ycombinator.com/item?id=5494113

vardump• 13 years ago

I'd love to see how Go 1.1 compares against modern JITted scripting languages, because I think those will be a major part of Go's current and future competition.

Hopefully the benchmarks game will one day include those JIT variants. V8 is there, but sadly for example PyPy, HHVM (Facebook's PHP TJIT) and LuaJIT are not.

igouy22• 13 years ago

"If you're interested in something not shown then please take the program source code and the measurement scripts and publish your own measurements."

http://benchmarksgame.alioth.debian.org/play.php#languagex

dilap2• 13 years ago

Yeah, but the request isn't to include some random language, it's to use the fast implementations of any given language. Why wouldn't you, in a site that benchmarks speed?

I think not to do so is unhelpful, actually -- if you don't realize they're are JITs for Python and Lua that aren't being included in the benchmarks, you will come away with a completely wrong impression about the possible performance of those languages.

edit: what spion said in their comment

spion7• 13 years ago

I think we're in dire need of a new benchmarks game. The author:

removed PyPy

removed the clear overview table that showed all languages, keeping only the barely visible gray bar charts with vertical text (why not horizontal bar charts, they can be scrolled and read without head tilting)

refuses to add LuaJIT and other languages

responds to most change inquiries with "publish your own version" which probably means that he doesn't really care about the project anymore

As a result, the spirit of the benchmarks game was lost a long time ago as its made less fun with each iteration.

igouy22• 13 years ago

Not only that, but "the author" has for many years provided the source code for the scripts to make the measurements -- so anyone who actually wants to can easily start a new benchmarks game.

Go do!

http://benchmarksgame.alioth.debian.org/play.php#languagex

rdtsc2• 13 years ago

That was there. PyPy was there at least. LuaJIT too I think. All go yanked out. The benchmarks are bogus anyway unless what you plan on working on n-body or fasta search or a short other list of things they run.

fabriceleal• 13 years ago

I believe Go is compiled [1][2]. Isn't a comparison with something like Haskell [3] more fair?

[1] http://golang.org/doc/faq#Implementation

[2] http://golang.org/cmd/go/#hdr-Compile_and_run_Go_program

[3] http://benchmarksgame.alioth.debian.org/u64/benchmark.php?te...

elliotlai• 13 years ago

The Glorious Glasgow Haskell Compilation System is truly glorious

sudhirj• 13 years ago

Why the spectacular loss to V8 in the Regex tests? Is Go's regex engine a naive implementation as of now?

chaosphere2112• 13 years ago

Russ Cox, in response to a regex benchmark vs. Python/Ruby.

"You assume the benchmark is worth something.

First of all, Ruby and Python are using C implementations of the regexp search, so Go is being beat by C, not by Ruby.

Second, Go is using a different algorithm for regexp matching than the C implementations in those other languages. The algorithm Go uses guarantees to complete in time that is linear in the length of the input. The algorithm that Ruby/Python/etc are using can take time exponential in the length of the input, although on trivial cases it typically runs quite fast."[1]

[1] https://groups.google.com/d/msg/golang-nuts/6d1maef4jQ8/yg4N...

baq• 13 years ago

As a member of the masses of programmers that typically deal with truckloads of trivial problems, his reply is worth as much as the benchmark he criticizes. I've never used anything but a trivial regexp (for some values of trivial) in production code (one-shot problems don't count) and I don't give a damn what is being benchmarked here - I only care if my Python script runs fast.

igouy22• 13 years ago

>>a different algorithm for regexp matching<<

Here's a C++ program that uses RE2

http://benchmarksgame.alioth.debian.org/u64/program.php?test...

CatMtKing• 13 years ago

Looks like there's a reimplementation of RE2 in Go that might we worth benchmarking against. https://code.google.com/p/sre2/

igouy22• 13 years ago

So what approach is implemented in the Go distribution?

simonster• 13 years ago

V8 JIT-compiles regular expressions.

EDIT: Since apparently someone doesn't believe me: http://blog.chromium.org/2009/02/irregexp-google-chromes-new... ("After optimization we generate native machine code which uses backtracking to try different alternatives.") See also the yarr JIT, which jsc and SpiderMonkey use for regular expressions: http://trac.webkit.org/browser/trunk/Source/JavaScriptCore/y....

cnlwsu• 13 years ago

Isn't comparing against a scripting language apples and oranges? Compiled vs interpreted one would assume to always look like this. Although the modern sophisticated JITs have definitely narrowed the playing field.

igouy22• 13 years ago

http://benchmarksgame.alioth.debian.org/dont-jump-to-conclus...

zobzu• 13 years ago

I dunno about you but I don't use Go where I could use javascript. And also, I use multicore machines only.

So I prefer this URL: http://benchmarksgame.alioth.debian.org/u64q/benchmark.php?t...

tptacek• 13 years ago

Go vs. v8 is a reasonable comparison. Node.js is probably the closest substitute for the way Go is most often used today (as API servers or backend app components for web applications).

I have no comment as to whether this "benchmarks game" stuff tells us anything meaningful.

burntsushi• 13 years ago

... That's using Go 1.0.3.

mseepgood2• 13 years ago

Why are some benchmarks using Go 1.0.3 and some 1.1beta1? And why is the PCRE program broken? http://benchmarksgame.alioth.debian.org/u64/program.php?test...

As far as I understand it just needs a "go get github.com/glenn-brown/golang-pkg-pcre/src/pkg/pcre"

xwei• 13 years ago

you prefer to do multi-thread programming in C? serious?

alecbenzer• 13 years ago

I'd be interested in seeing compile times. In a talk posted recently Rob Pike seemed to be kind of emphasizing #include inefficiencies and bytes read by compiler to bytes of source input ratios for C/C++, and how was better at this.

Go 1.1beta1 now on the benchmarks game

Related Stories