nLeyten: Recent Comments

Comment on Use Ocaml, save the world

2008-08-09T21:21:48Z

Hi, Well, when performance matters, simply use native code! As you noted, with Ocaml 3.10, it is not possible to dynamically load libraries when using native code. However, version 3.11 fixes this, so there's nothing holding you back...

Comment on Simple benchmarks on the Ocsigen server

2008-08-09T21:11:46Z

Hi Ben,

You can use either bytecode or native-code, but whichever your choice, all components must be using the same mode, because you cannot mix bytecode and native code.

Comment on Simple benchmarks on the Ocsigen server

2008-08-07T14:36:52Z

This is 100% completely usefull information. As newbie I still have some questions. When you talk of bytecode vs native compiled do you mean
- the ocsigen environment?
- the eliom framework?
- your application?

Comment on Use Ocaml, save the world

2008-08-07T14:30:00Z

interesting article. One question though: I think when you speak on Ocamls performance you always mean native compiled programs right? Isn't there a problem regarding bigger applications because there you often have to load libraries? For example mod_caml or the ocsigen webserver isn't this bytecode compiled? In the shout out example you refering to, Ocaml comes after ruby if compiled to bytecode! http://shootout.alioth.debian.org/debian/benchmark.php?test=all&lang=all&calc=Calculate&xfullcpu=1&xmem=0&xloc=5&binarytrees=1&chameneosredux=1&fannkuch=1&fasta=1&knucleotide=1&mandelbrot=1&nbody=1&nsieve=1&nsievebits=1&partialsums=1&pidigits=1&recursive=1®exdna=1&revcomp=1&spectralnorm=1&hello=0&sumcol=1&threadring=1

Comment on Use Ocaml, save the world

2008-05-24T09:29:54Z

You should also take into consideration "how much the language is able to be understood and processed by a computer". In other word, how much lines of the written program can be reprocessed by a computer to produced a more energy efficient version for example.

This is a general "legacy code" problem. For example, COBOL is not a regular language and there is at least 10 to 20% of written COBOL code that need to be reprocessed by hand to migrate the code from a mainframe to a UNIX, for example. The cost of this migration is directly proportional to the time spent to write the initial program. For 10% of lines not understood in the initial program, you will spent ~15% of the initial time to rewrite your program.

This parameters should be taken into account using "computer tersness" of the language, ability to find efficient programmers and number of bugs left in the code.

"computer tersness" defines how much a computer reading your code will understand your aim...

Sylvain Le Gall
(not so tongue in cheek)

Comment on Peak Oil revisited

2008-05-23T19:39:52Z

Thanks Alan. Don't underestimate the power of denial...

Comment on Peak Oil revisited

2008-05-23T16:50:53Z

Thank you for a very insightful article. I just find it amazing it's taking this long to reach the public awareness.

Comment on Simple benchmarks on the Ocsigen server

2008-04-22T12:26:47Z

Hi Berke,

I'll give it a spin, comparing Ocsigen vs Lighttpd serving static pages. Soon!

Comment on Simple benchmarks on the Ocsigen server

2008-04-21T21:57:08Z

Hello Dario,

Would it be possible for you to run similar benchmarks using, say, Apache and/or Lighttpd with static pages on the same machines?

Also, I assume siege doesn't accept gzip-compressed data (unless you add a header for it).

Comment on Simple benchmarks on the Ocsigen server

2008-04-21T19:44:23Z

Might be useless to you, but certainly not to me and others. Sure, I knew beforehand that native code would be faster, but I had no idea of how much faster. Would it be faster enough to be worth the extra trouble of using Ocaml CVS HEAD? Faster enough to advocate that -dlcode be enabled by default on 3.11 for AMD64?

Now I know.

As for comparisons with Rails or Django, those would indeed be most welcome. However, I am certainly not going to be wasting my time learning those frameworks (I chose Ocsigen for reasons other than speed). I published the Eliom code I used. If someone would do the same for Rails or Django, and I would be most glad to help in setting up a comparison page.