a potential lisp convert, and interpreting the shootout
in [Lisp]
|
Prev: Pocket Lisp Machine
Next: Next Generation of Language
From: Jon Harrop on 30 Sep 2006 21:53 Henry Bigelow wrote: > it looks like you have a lot of experience. so, benchmark politics > aside, do you find lisp to be fast enough to do heavy-duty computing, > (in my case, a bayesian network with hundreds of nodes and lots of > floating point addition, but also written in a very extensible, > abstract style) You may be interested in my ray tracer benchmarks: http://www.ffconsultancy.com/free/ray_tracer/languages.html On my computer, those benchmarks show Lisp to be roughly twice as slow and twice as verbose as OCaml. However, the benchmark is quite specific. It only tests data structures (trees) and floating point performance (ray sphere). Also, timings vary considerably between architectures. Now that Intel has a decent CPU, I'll be interested to see a Core Duo version of this benchmark. -- Dr Jon D Harrop, Flying Frog Consultancy Objective CAML for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists
From: Jon Harrop on 2 Oct 2006 10:58 Isaac Gouy wrote: > Jon Harrop wrote: >> Only Isaac Gouy has tried to deceive people, AFAIK. > > That is no more than malicious personal abuse - GV From my point of view, you implied that I had plagiarised the ray tracer (following Alex Goldman) and then revoked my access. > you contributed similar baseless personal attacks to the shootout > mailing-list. I don't believe so. > Jack Andrews created "The shootin" partly in response to criticisms of > the shootout. I think it's sad that in the past year you haven't taken > the opportunity to be constructive and contribute to "The shootin". Fortunately, my business is sufficiently busy that I don't have spare time to invest in such projects. Indeed, I have little time to improve my own benchmarking site. I think we'll all agree that benchmarking "across the board" is very difficult. You've tried. I disagree with the fundamental approach taken by your shootout (particularly the unnecessary subjectivity). I tried. Juho kindly contributed a lot of Lisp code to my ray tracer benchmark but, I believe, he disagrees with the way I presented the results. So it goes on... -- Dr Jon D Harrop, Flying Frog Consultancy Objective CAML for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists
From: Jon Harrop on 2 Oct 2006 16:13 Isaac Gouy wrote: >> saying this, i realize it might be a lot more work for the maintainers, >> and possibly create problems. > > You seem to be suggesting that we should show every program that has > ever been contributed, for ever. Do you understand that we continually > update the language implementations and re-measure the programs? Why don't you reinstate the ray tracer benchmark? I still think it is the best designed benchmark on the shootout (spectral-norm is my second favourite). PS: If your answer is going to be "because the different language's implementations use different algorithms" then I'll repeat "so do several of your other benchmarks". -- Dr Jon D Harrop, Flying Frog Consultancy Objective CAML for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists
From: Isaac Gouy on 2 Oct 2006 16:24 Jon Harrop wrote: > Isaac Gouy wrote: > >> saying this, i realize it might be a lot more work for the maintainers, > >> and possibly create problems. > > > > You seem to be suggesting that we should show every program that has > > ever been contributed, for ever. Do you understand that we continually > > update the language implementations and re-measure the programs? > > Why don't you reinstate the ray tracer benchmark? I still think it is the > best designed benchmark on the shootout (spectral-norm is my second > favourite). > > PS: If your answer is going to be "because the different language's > implementations use different algorithms" then I'll repeat "so do several > of your other benchmarks". > > -- > Dr Jon D Harrop, Flying Frog Consultancy > Objective CAML for Scientists > http://www.ffconsultancy.com/products/ocaml_for_scientists Let no one accuse you of modesty.
From: Jon Harrop on 2 Oct 2006 16:24
Juho Snellman wrote: > My point was just that people really shouldn't be using the Shootout > for making decisions about what languages to use, since it's not > unlikely that a program for Blub has been submitted by someone who is > still a newbie at programming Blub. Given that poor (slow) implementations get replaced, the lack of a good implementation can be taken to imply either an unwillingness or inability of that community. I'm not saying that is a good idea, but it is all the reader has to go on. On the other hand, maybe Lispers think that the shootout is biased and, consequently, will not bother working on it whereas the D programmer might love the shootout because it is a low-cost advertisement. -- Dr Jon D Harrop, Flying Frog Consultancy Objective CAML for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists |