Slashdot Mirror

I work for a company that develops applications for high performance computing clusters (aka beowulf). In addition to application development, we provide turn-key cluster solutions. Meaning, we act as a scientific cluster technology integrator. Also, we only use linux!

This said, we have been benchmarking the Intel C compilers (under Linux) for sometime on the Pentium4. We are getting some great results with the Intel compiler on the Pentium4, although there are some usability issues that intel is working through; but speedup is excellent.

Now, given that VectorC can make use of parallelism inherent in some C codes on multiple x86 architectures (SSE, SSE2, 3dNow), I could see the availability of VectorC for Linux as being a big win for us. Especially if it can give the Intel compiler a run for it's money on the Pentium4!

Ok, as a side note we resell Portland Group Compilers, so our customers will pay for damn good compilers!

I really hope codeplay decides to throw some developers at porting VectorC to Linux. I can be reached via email at: msnitzer@plogic.com

I look forward to hearing more on Linux offerings from Codeplay in the future.

If you have a serious interest in Beowulf then you may want to read this link

Many people do not know this but Dell sells very nice turn-key clusters. They are working with a company called Paralogic These guys know what the are doing.

We build turn-key Beowulf systems for a living. You may want to consult:

Beowulf.org

Beowulf FAQ

Beowulf Underground

Beowulf Quick Start

In addition, you may want to contact us directly about helping you convince your management that turn-key and supportable Linux Beowulf Clusters are available and do useful things. Take a look at our success stories.

Doug Eadline,
deadline@plogic.com
Paralogic, Inc.

Having written and received a parallel computing patent, let me first define what a patent is: The right to take someone to court if they use your idea. Note that it does not say that you will win because you have a patent.

The test of a patent is how well it will hold up in court. Which often means which ever side has the most money. Obvious patents (in which patent examiners have allowed patents due to careful wording) are often an offensive business strategy. Patent everything, scare away the small guys, and make anyone who wants to use it pays in legal fees.

Finally, parallel computing methods have been around for many years. I doubt there is really anything in this patent that has not been tried or implemented. However, by being the first to claim a method, you restrict anyone else from doing the same and do not have to go to court and pay to break the patent, if you need to use the method. This is more a defensive strategy.

What is all comes down to is If you have the money, patent everything you do, and let the courts settle it later. Also, did I mention that many companies give bonuses for those employees that obtain patents.

Douglas Eadline
Paralogic, Inc.

No offense intended here; this comment is meant to be about 25 deg C, and is just a clarification for what seems to be a common misconception.

No application can make use of a parallel computer without a programmer first identifying the inherent concurrency. There are pre-processors that attempt to do this for you, such as BERT 77, but IMO that cannot replace the intuition and basic understanding of the problem that a human programmer provides (with all due respect to Doug & everyone else at Paralogic). Once that concurrency is identified, then implementing it using the MPI or PVM API, or using HPF directives, or using pthreads (depending on your platform) should be fairly straight-forward.

Computers are not magic! You cannot feed a program written for a sequential computer into a parallel computer and expect speedup, period. At run-time, the computer just doesn't have sufficient view of the application to even attempt to identify the concurrency for you, and it certainly doesn't have the cycles to spare to decide what aspects of the concurrency should or shouldn't be parallelized. The best you can do if you won't or can't parallelize your code is to submit multiple instances of the application, with different inputs, to the batch scheduler and obtain a greater throughput.

I looked at the web site. Too many buzz words. What they propose seems to have been done (Linda) and may not be the general solution to parallel computing. But the lack of a clear concise statement of what they propose is troubling.

There is no free lunch in parallel computing. The closest thing I have seen to a true "parallel compiler" is BERT

I just read the article. As a manufacturer of "turn-key" Beowulf systems, here was my reply to the author:

Stephen,

I just read your story about Beowulf systems. While the story was well written and informative, there are some points that you have missed.

1) The definition of a Beowulf requires and "open source" OS (See "How to Build a Beowulf" by Sterling, Becker, Salmon, Savarese) Therefore, systems built from True 64 are NOT Beowulf systems.

2) You missed my company, Paralogic Inc. We sell turnkey Beowulf systems. In fact rather than "several" as reported by IBM, we have several dozens of installed production systems at companies like Lucent, Amerada Hess, Conoco, Procter and Gamble, government sites like NASA, NRL, and the Air Force, and many Universities. (see www.xtreme-machines.com)

3) There is a rather huge barrier to entry because of the technical nature of these machines. As far as I know, we are the only company who will offer support for Beowulf clusters. Without support the market can never enter the mainstream.

4) There have been quite a few other people who contributed quite a lot of effort to the Beowulf technology other than IBM and VA Linux. Although all contributions are welcome, these guys are a little late to the party and we hope they stay.

Sincerely

Douglas Eadline, Ph.D.
President

Paralogic, Inc.
PEAK PARALLEL PERFORMANCE

Isn't it interesting just how small the world seems to be some days?