Domain: netlib.org
Stories and comments across the archive that link to netlib.org.
Stories · 6
-
Full Details Uncovered on Chinese Tianhe-2 Supercomputer
An anonymous reader writes "With help from a draft report (PDF) from Oak Ridge National Laboratory's Jack Dongarra, who also spearheads the process of verifying the top of the pack supercomputer, we get a detailed look at China's Tianhe-2 system. As noted previously, the system will be housed at the National Supercomputer Center in Guangzhou and has been aimed at providing an open platform for research and education and to provide a high performance computing service for southern China. From Jack's details: '... was sent results showing a run of HPL benchmark using 14,336 nodes, that run was made using 50 GB of the memory of each node and achieved 30.65 petaflops out of a theoretical peak of 49.19 petaflops, or an efficiency of 62.3% of theoretical peak performance taking a little over 5 hours to complete.The fastest result shown was using 90% of the machine. They are expecting to make improvements and increase the number of nodes used in the test.'" -
Slashback: Diebold, Cluster, Radiation
Slashback tonight brings you word on the less-spectacular-than-advertised solar storm earlier in the week, Mandrake's response (a good one) to the problems their new release had with LG brand CD drives, more Diebold madness, and more, including a lengthy rebuttal to Slashdot's review of Eclipse in Action. Read on for the details, and check your costume in the mirror before leaving the house.Copies files in under 17 minutes, I bet. Eug writes "The latest supercomputer list (Oct. 26) has Apple/VT's G5 Power Mac cluster at 9555 Gflops/s, which puts it into third place overall. This list is hosted here. This new score is interesting for a number of reasons, besides placing them in third place:
- It is now ahead of the 1.5 GHz Itanium 2 cluster, which is composed of 1936 CPUs and which achieves 8633 Gflops/s.
- On a per CPU basis, the G5 2.0 is also ahead of the Itanium 2. The G5 2.0 scores 4.52 Gflops/s per CPU, while the Itanium 2 1.5 scores 4.46 Gflops/s per CPU.
- If one extrapolates from the score of NetworX's Xeon 2.4 cluster (2304 CPUs at 7623 Tflops/s), a G5 2.0 would be as fast as a Xeon 3.28 GHz.
- Efficiency of the G5 clusters is now at 57%, which is considerably higher than the IBM POWER4 clusters in the top twenty. (The G5 is a derivative of the POWER4.)
- Virginia Tech's cluster is now in shouting distance of 10 Teraflops/s, and there are still a few weeks left to optimize the system. (They've gained over 2 Teraflops/s in the last 2 weeks.
- They have utilized only 2112 CPUs (1056 dual Power Macs), despite having supposedly purchased 2200."
eGovOS 3 cancelled due to EC funding withdrawal jaruz writes "Due to the unexpected withdrawal of EC funding for the eGovOS conference from the University of Maastricht's MERIT's FLOSSPOLS EC contract, the conference is now cancelled."
I prefer conspiracy theories, myself. MyNameIsFred writes "Slashdot recently discussed White House Website Limits Iraq-Related Crawling. It turns out The Dead Parrot Society got an explanation for their behavior. They used the unprecedented approach of asking someone at the White House. White House spokesman Jimmy Orr stated the blocking of search engines is not an attempt to ensure future revisions will remain undetected. Rather, he explained, they "have an Iraq section [of the website] with a different template than the main site." Thus, for example, a press release on a meeting between President Bush and Special Envoy Bremer is available in the Iraq template (blocked from being indexed by search engines) or the normal White House template (available for indexing by search engines). The attempt, Mr. Orr said, was that when people search, they should not get multiple copies of the same information. It was also reported that the White House recently asked the The Internet Archive to do a thorough scan of everything on its website."
My dad can beat up your burst of solar radiation. Earth survives solar storm. kurth writes "A major solar flare unleashed Tuesday punished Earth's protective magnetic field early Wednesday, but the planet and its high-tech communication systems appear to have weathered the worst of the storm."
eggfellow writes "here's an article in the WashPost about the geomagnetic storm that pounded Earth (with little disruption) [Tuesday]. What I want to know is why the predicted pounding-time was 12 hours later than actual. Can't these scientist do their math?"
Sounds like a nice feature. News.OSDir.com is reporting that Mandrake is re-releasing it's 9.2 ISOs and CDs after the unfortunate LG CD drive incident earlier this week. "The problem was that the kernel would send a FLUSH_CACHE command to the LG CD-ROM drive which would make the drive inoperable by overwriting its firmware....A new kernel (2.4.22-21mdk) has been released that fixes this problem in the kernel, although the CD-ROM devices are still not up to specification. New CDs and ISOs will be available shortly to correct these problems; they will come with the new kernel."
Maybe they should stick with safes and such. The work of the Swarthmore rebels is paying dividends, (they now have 17 mirrors of the Diebold memos set up). Meanwhile Scoop is reporting how one of the memos deals with an incident in which a single memory card from a precinct of just 600 voters managed to subtract 16022 votes from Al Gore in Florida, nearly lead to his concession of presidency. You can read more about this in Bev Harris's "Black Bov Voting" Chapter11 (PDF) also available here & here."
More on the Diebold front: cananian writes "Two students at MIT (I'm one of them) received cease-and-desist letters from Diebold today for mirroring Diebold's incriminating internal memos, which reveal (among other things) -16,000 votes being credited to Gore in Florida in the 2000 presidential election, how the vote could have been rigged by changing the audit logs or creating a manager card, etc. Students at Amherst also received cease-and-desist letters today. Diebold claims we are infringing its copyrights, but there is good precedent for the legality of the publication. The EFF has in is support: "Wendy Seltzer, an attorney for the Electronic Frontier Foundation [...] encouraged them to defy the Diebold cease-and-desist letters.""
... because making text cross-platform is Unamerican. David H. Rothman writes "Convert Lit, the program that lets you crack Microsoft Reader to make backups as part of Fair Use, has moved to a Polish host to escape the tyrannies of the new EU-style DMCAism in the United Kingdom and elsewhere. Meanwhile, in the wake of a new Copyright Office ruling on the DMCA, lawyer Robin Gross at IP Justice warns not to think that the DMCA peril has passed."
But how do you really feel? In reaction to our ealier review of Eclipse in Action, wobbet writes "I've started using Eclipse at work and consistently feel that there is more sophistication and power hiding underneath the obvious and wanted a book that would help me find and fine tune the goodies under the covers. I read a previous review of this book on Slashdot that prompted my purchase. If that review had not been as positive I probably would not have been so disappointed and moved to post my own review.
When I read a technical book I ask myself how well it stays on topic, how thoroughly it addresses the topic and whether it meets my expectations. In this instance I find that the book stays on topic about half of the time and that it is thorough about half of the time. Unfortunately that half of the time I really didn't care about and thereofre my expectations were unmet. To be honest - after reading the book and then re-reading the back cover I should have not even purchased the book because the objectives set forth on the back cover would have warned me that this book was not what I was looking for.
I found the first half of the book to be simply horrible. A supposed introduction to actually using Eclipse this section concentrates more on the "Agile" toolset that all competent, well-informed Java developers that care about the quality of their code, products and development process should already be using. Well, that's what all the books say anyway.
If I wanted a book on Agile tools for Java developers I would purchase Java Tools For Extreme Programming . Is it a great book? No, but it is honest about what it is - a survey of tools. Despite what Mr. Chappell says about Eclipse In Action, I did not find the authors' "...TDD evangelism, skillfully disguised as Eclipse usage instruction. ." Instead I found the first half of the book to be TDD Evangelism thinly disguised as poor Eclipse usage instructions. I did not learn a single thing about USING Eclipse that I hadn't already figured out from randomly selecting menu items over the past two months.
The second half of the book seemed to be a decent introduction to the development of Eclipse plug-ins. If I cared I probably would have found it interesting in its discussion of the API, the perspectives, views and even editors. Those of you that do care may find the second half of the book to be worth skipping the first half of the book." -
Factual 'Big Mac' Results
danigiri writes "Finally Varadarajan has put some hard facts on the speed of the VT 'Big Mac' G5 cluster. Undoubtedly after some weeks of tuning and optimization, the home-brewn supercluster is happily rolling around at 9.555 TFlops in LINPACK. The revelations were made by the parallel computing voodoo master himself at the O'Reilly Mac OS X conference. It seems they are expecting and additional 10% speed boost after some more tweaking. Srinidhi received standing ovations from the audience. Wired news is also running a cool news piece on it. Lots of juicy technical and cost details not revealed before. Myth dispelling redux: yes, VT paid full price, yes, it's running Mac OS X Jaguar (soon Panther), yes, errors in RAM are accounted for, Varadarajan was not an Apple fanboy in the least... read the articles for more booze." -
Big Mac Benchmark Drops to 7.4 TFlops
coolmacdude writes "Well it seems that the early estimates were a bit overzealous. According to preliminary test results (in postscript format) on the full range of CPUs at Virginia Tech, the Rmax score on Linpack comes in at around 7.4 TFlops. This puts it at number four on the Top 500 List. It also represents an efficiency of about 44 percent, down from the previous result of 80 achieved on a subset of the computers. Perhaps in light of this, apparantly VT is now planning to devote an additional two months to improve the stability and efficiency of the system before any research can begin. While these numbers will no doubt come as a disappointment for Mac zealots who wanted to blow away all the Intel machines, it should still be noted that this is the best price/performance ratio ever achieved on a supercomputer. In addition, the project was successful at meeting VT's goal of developing an inexpensive top 5 machine. The results have also been posted at Ars Technica's openforum." -
Use of Math Languages and Packages in Research?
CEHT asks: "As a research programmer at the university, I have encountered numerous times when I need to choose which language(s) or package(s) to use for different projects. Tradeoffs and performance issues have to be considered: results from one package may be more compatible with the data from other researchers, another package may find the solution faster and use less resources, and so forth. Maple, Matlab, Magma, and Mathematica are among the most well-known packages. Libraries such as IMSL is also popular. Of course, there are smaller (and mostly free) packages that tend to target specific types of problem, such as LiDIA, Singular, and LAPACK. The question is, how useful are these [and other] math packages? Do researchers use only one or two packages for most of their projects? Or do people like to mix things a little by pulling the strength of different packages together to solve a math problem? If not, do researchers write C/C++ programs and use GMP or Matpack to solve math problems?" -
Is FORTRAN Still Kicking?
Algorithm wrangler queries: "I'm beginning to wonder if I should invest the time in learning FORTRAN. Although it is, arcane it seems to be the best tool when it comes to demanding optimization tasks and heavy computations. C/C++ does not cut it for me - it is simply too easy to make mistakes and I find myself using half of my time hunting bugs unrelated to the problem at hand. Additionally, although tools like Matlab exist they don't provide the power that justify the huge price tag they carry. I find any script based language (Matlab, Numeric Python, Scilab) to be inadequate as soon as it is necessary to use loops to describe a problem and using such tools for recursive systems can be a real pain. As another data-point, the Netlib repository seems to be very FORTRAN oriented, and it is a true gold mine when it comes to free routines for solving almost any computing task. What bothers me though is that FORTRAN code is really ugly and the language lacks almost any modern day language feature (I know about Fortran 90 but it is not much nicer than F77, and no one seems to use it). Can it really be true that the best tool we have for heavy duty computing is a 25 year old language, or have you found anything better - free or non-free?"