4 ms·

Mods- put a (2006) in there please. The number-crunching ecosystem has changed drastically in 10 years. He also didn't mention what compiler he's using. Intel's

Mods- put a (2006) in there please. The number-crunching ecosystem has changed drastically in 10 years. He also didn't mention what compiler he's using. Intel's ICC had a huge advantage over gcc 2.95.3 (or whatever was used in '06) especially in numerical methods. (Though, I'm not sure when Intel's MKL became semi-freeware; at which point there was a significant performance jump if one opted to use the lib appropriately in tandem with gcc.) This guy also could have at least linked to a repo of his code... though none of that fancy github stuff was around, Sourceforge and CVSweb was around back then. Without the source we can't really accurately judge, well, much of anything. Compiler flags, JVM tunes (was it warm?), source code, data sets, or hell, even mention of compilers. Edit: Though his point held hm, not quite 'true' categorically, but you could argue the point. Around '06, Fortran was still largely used, at least bioinformatics (sequencing, attempts at heuristic folding .. stuff), weather prediction, and was a perfect fit for linear algebra.[1] It was accessible and fast enough to be used by post-docs who didn't write software for a living, so presumably the ecosystem was founded then continued to develop within Fortran. Even if it was possible to outperform your simulations in C++ (I'm sure you could get at least fairly close), all of the work you were building off of (e.g. with your colleagues, advisors, etc) were Fortran-based. [1] This is all hearsay from my father's / his doctoral students work from the late '90s/early '00s when I was in high-school, so I can't really offer sources on hand, but if someone really wants me to, I can fire off an email to him to gather specifics.

satysin 10y ago
Indeed. This post is nothing more than "I did this and look at the results" but offers nothing for us to repeat his tests. Plus yeah it is a decade old, things have changed a hell of a lot in the past decade.
- YeGoblynQueenne 10y ago
  >> This post is nothing more than "I did this and look at the results" but offers nothing for us to repeat his tests. That's a valid gripe to have, but it should be said that most scientific papers that report simulation or algorithm results do exactly the same. Reproducing results is totally left to whomever feels like (and has the time to) dispute the claims. Which is why you don't often see much in the way of disputation of performance claims, btw (yeah, ok, also because people take part in competitions where you get the chance to put your code where your mouth is).
  - semi-extrinsic 10y ago
    But scientific papers aren't comparing Language X v.2.7.5 vs Language Y v. 16.9.1. They're comparing Algorithm X vs Algorithm Y, and usually also give theoretical results. Unless there is a big dramatic change in computer hardware, results comparing some algorithms using good implementations in a single well-optimized language will likely stand the test of time well.
    - jblow 10y ago
      No, because the comparisons are usually "baseline version of algorithm I want to beat" vs "highly optimized and hand-tweaked version of the algorithm I have a vested interest in."
      - lorenzhs 10y ago
        At least in my (and my colleagues') areas of CS, you usually ask the original authors of a paper you want to compare to. A large percentage of them will be glad to provide you with their code, for it goes a long way towards ensuring that their algorithms aren't grossly misrepresented. If they don't (because the code is a mess, they can't find it or get it to work any more, or for whatever reason), you'll have to re-implement it. We take great care to match their performance and have, on several occasions, beat it (compared to the published results, running on similar hardware). It's a lot more work, but in our group we really try to make fair comparisons. Maybe not everyone is so inclined, but I'm sick of reading these stereotypes over and over again.
        YeGoblynQueenne 10y ago
        >> Maybe not everyone is so inclined, but I'm sick of reading these stereotypes over and over again. I'm sorry if my comment upset you. I'm really not trying to advance any stereotypes and rather just report what my experience is so far, as a Masters student (rather than a researcher). I've also discussed this with one of my professors, who does research mainly in NLP and he told me the same thing. Personally I find it hard to understand why it's not compulsory for academics to point to a public repository where anyone can find the code that goes along with their papers. I've heard arguments for and against sharing code and particularly data, so I'm not saying that I am necessarily right, but it's really just weird to read "our system beats the best results reported so far" with nothing but a few figures in a table to back that up. And with so much infrastructure and tools around to support sharing code (just think of github) I really don't see what's stopping people.
gus_massa 10y ago
Fortran is still used in Physics. They have a lot of libraries, and they usually have a some program of a previous paper that almost do what they want now, and they know the technology stack, so many physicist are still not planning to change. (I work with a Physics group. Sometimes I found programs with an old part that still use numbered lines for the "do". We are still trying to get rid of numbered lines ...)
- rhodysurf 10y ago
  I work for a company working with hydrodynamics (Potential Flow) and all of our main hydro code is written in Fortran and we don;t plan on moving away from it. Its too easy for non-software oriented engineers to write performant code in with a low barrier to entry.
- beezle 10y ago
  Not to mention that Fortran has come a long ways from your father's (or my) Fortran IV.
  - rbobby 10y ago
    RatFor for the win... :)... :(
- jjdredd 10y ago
  Particle physicists began to move on to C++: Pythia, GEANT4, root. EDIT: Also, lattice QCD people are using C++ nowdays (frankly, I haven't seen them using anything else).
coliveira 10y ago
Here is the problem, it is possible to match or outperform Fortan in C++. But it is not simple. You need to use specific libraries and templates. Fortran can do this with no effort. If you're writing numeric code it is still the place to go for high performance.
ksk 10y ago
> The number-crunching ecosystem has changed drastically in 10 years. How has it gotten better? What has changed? And why are those changes relevant?

satysin 10y ago

Indeed. This post is nothing more than "I did this and look at the results" but offers nothing for us to repeat his tests. Plus yeah it is a decade old, things have changed a hell of a lot in the past decade.

YeGoblynQueenne 10y ago
>> This post is nothing more than "I did this and look at the results" but offers nothing for us to repeat his tests. That's a valid gripe to have, but it should be said that most scientific papers that report simulation or algorithm results do exactly the same. Reproducing results is totally left to whomever feels like (and has the time to) dispute the claims. Which is why you don't often see much in the way of disputation of performance claims, btw (yeah, ok, also because people take part in competitions where you get the chance to put your code where your mouth is).
- semi-extrinsic 10y ago
  But scientific papers aren't comparing Language X v.2.7.5 vs Language Y v. 16.9.1. They're comparing Algorithm X vs Algorithm Y, and usually also give theoretical results. Unless there is a big dramatic change in computer hardware, results comparing some algorithms using good implementations in a single well-optimized language will likely stand the test of time well.
  - jblow 10y ago
    No, because the comparisons are usually "baseline version of algorithm I want to beat" vs "highly optimized and hand-tweaked version of the algorithm I have a vested interest in."
    - lorenzhs 10y ago
      At least in my (and my colleagues') areas of CS, you usually ask the original authors of a paper you want to compare to. A large percentage of them will be glad to provide you with their code, for it goes a long way towards ensuring that their algorithms aren't grossly misrepresented. If they don't (because the code is a mess, they can't find it or get it to work any more, or for whatever reason), you'll have to re-implement it. We take great care to match their performance and have, on several occasions, beat it (compared to the published results, running on similar hardware). It's a lot more work, but in our group we really try to make fair comparisons. Maybe not everyone is so inclined, but I'm sick of reading these stereotypes over and over again.
      - YeGoblynQueenne 10y ago
        >> Maybe not everyone is so inclined, but I'm sick of reading these stereotypes over and over again. I'm sorry if my comment upset you. I'm really not trying to advance any stereotypes and rather just report what my experience is so far, as a Masters student (rather than a researcher). I've also discussed this with one of my professors, who does research mainly in NLP and he told me the same thing. Personally I find it hard to understand why it's not compulsory for academics to point to a public repository where anyone can find the code that goes along with their papers. I've heard arguments for and against sharing code and particularly data, so I'm not saying that I am necessarily right, but it's really just weird to read "our system beats the best results reported so far" with nothing but a few figures in a table to back that up. And with so much infrastructure and tools around to support sharing code (just think of github) I really don't see what's stopping people.

gus_massa 10y ago

Fortran is still used in Physics. They have a lot of libraries, and they usually have a some program of a previous paper that almost do what they want now, and they know the technology stack, so many physicist are still not planning to change. (I work with a Physics group. Sometimes I found programs with an old part that still use numbered lines for the "do". We are still trying to get rid of numbered lines ...)

rhodysurf 10y ago
I work for a company working with hydrodynamics (Potential Flow) and all of our main hydro code is written in Fortran and we don;t plan on moving away from it. Its too easy for non-software oriented engineers to write performant code in with a low barrier to entry.
beezle 10y ago
Not to mention that Fortran has come a long ways from your father's (or my) Fortran IV.
- rbobby 10y ago
  RatFor for the win... :)... :(
jjdredd 10y ago
Particle physicists began to move on to C++: Pythia, GEANT4, root. EDIT: Also, lattice QCD people are using C++ nowdays (frankly, I haven't seen them using anything else).

coliveira 10y ago

Here is the problem, it is possible to match or outperform Fortan in C++. But it is not simple. You need to use specific libraries and templates. Fortran can do this with no effort. If you're writing numeric code it is still the place to go for high performance.

ksk 10y ago

> The number-crunching ecosystem has changed drastically in 10 years. How has it gotten better? What has changed? And why are those changes relevant?