Re: Reasons to choose CLISP over other free implementations

On 5202 September 1993, Javier wrote:
Well, in my experience, and with this graphs:
http://sbcl.boinkor.net/bench/

Very interesting, I haven't seen those before.

Here you can see that CLISP is still about 2x faster than SBCL
compiling things, and not only compiling, but a lot of other useful
stuff, like defining CLOSS classes and instantiating them, and
printing strings (useful for web serving).

This is only one half of the truth. The other half is where the
benchmarks shows Clisp to be a lot slower.

Of course, SBCL is much faster for fixnum and float operations. But if
your application do not rely very much on arithmetic operations, CLISP
is a very capable Lisp implementation.

I don't doubt the correctness of your conclusion, but in combination
with your precondition you give a misleading picture of the situation,
at least when you cite the above mentioned benchmark as the source of
your conclusion. In those benchmarks, Clisp seems to be slower than
CMUCL or SBCL on a variety of tasks that have nothing to do with
arithmetic operations. For instance, hashing seems to be considerably
slower, regardless whether you're hashing integers or strings. As
another example, CLOS methodcalls also seems to be significantly
slower.

Consider CLISP for web, string manipulation, and CLOS. It is a very
capable Lisp distribution for these areas.

Looking at the benchmark, I have considerable doubts to follow your
recommendation. Instead I would follow the usual advice and use
whatever I'm used to or feel most comfortabl with until I run into
some problem performance wise and defer premature optimization until
it's necessary.

Holger

--
--- http://hillview.bugwriter.net/ ---
"Ich hasse es, mich nachts an Warmes zu schmiegen, um morgens auf
platter Katze zu erwachen."
-- Cristalle Sabon in de.alt.sysadmin.recovery
.

Relevant Pages

  • Re: Comparison about Win32 / DotNet / CSharp on Delphi, and a wish
    ... .NET is slower, all it shows is that C# MONO is slower. ... E.g. if you do a lot of string concatenations they are commonly way faster in Delphi Win32 as in Delphi .NET. ... we should not do the benchmarks that are worst for .NET?? ...
    (borland.public.delphi.non-technical)
  • Re: scripting languages vs statically compiled ones
    ... > benchmarks described by the author in chapters 25-26. ... > means the language is slower. ... VBA in MS Office is interpreted and your C dll will ... which showed VB as native code running only 50% slower than C. ...
    (comp.lang.python)
  • Re: scripting languages vs statically compiled ones
    ... > benchmarks described by the author in chapters 25-26. ... > means the language is slower. ... VBA in MS Office is interpreted and your C dll will ... which showed VB as native code running only 50% slower than C. ...
    (comp.programming)
  • Re: AMD-64 systems and Java performance
    ... > Does anyone have any comparisons or benchmarks for compiling and running ... > I'm considering buying an AMD-64 notebook which will run Windows XP Pro. ... The main advantage of 64-bit CPUs on any sane platform is getting access ... never saw any actual benchmarks.. ...
    (comp.lang.java.programmer)
  • Re: Why is C# 450% slower than C++ on nested loops ??
    ... > | I know why but I don't care, these benchmarks are clueless and they are ... > exclude the JIT process if you are comparing execution times as, ... > Joanna Carter ... (compiling the static di-unction takes less than a millisec. ...
    (microsoft.public.dotnet.languages.csharp)