Essentially they created a flash bytecode backend for LLVM. Now create a flash frontend (i.e. converts flash source code -> LLVM intermediate representation) and you'd have a LLVM-optimized flash compiler. Or for that matter, a flash -> native compiler (since the native backend already exists).
Um, no. ZFS is a local file system, AFS (though you probably mean AFP, but curiously it doesn't matter in this case) is a protocol for sharing files over a network.
Not necessarily. NFS4 supports kerberos encryption (just like CIFS in AD mode). If root su:s to another user he doesn't have the other users kerberos keys, hence no go.
Of course, if you have root, and the other user is logged in, you can compromise his key store and impersonate him. But that's no different from CIFS.
They're not that different. The network is compatible, the physical design is compatible, e.g. you can slot in XT4 nodes into a XT5 enclosure (which is what Cray recommends if you have jobs with high bandwidth requirements).
From an end-user perspective the only difference is that XT5 nodes have twice the number of cores (2 sockets per node instead of 1).
Hasn't Red Storm switched to Linux yet? Seems most of the Cray XT sites have done so already. One reason for developing CNL (Compute Node Linux, the Linux kernel they run on the nodes) IIRC was that multicore support in Catamount was sort of a kludge. And OS buffering, and..
Then again, the catamount developers are at Sandia, so maybe they have some emotional attachment to it.:)
CNL still doesn't allow dynamic linking, though it's much less alien than catamount.
I think the point was that machines like BG or the Cray XT series are MPP:s, not clusters. But yeah, a MPP is essentially a cluster, so the distinction is not that useful.
Fortran 95 when programmed in a "modern Fortran" style is actually a pretty modern, easy to use, and safe language. "Modern Fortran" style usually means having all procedures in modules (giving automatic argument checking and a lot of other nice things), derived types (same as structs in C), avoiding old error-prone constructs like common and equivalence, and so forth. The Fortran 90+ array features alone are worth dying for (similar to the array syntax in matlab).
My girlfriend is finishing her M.Sc thesis in economics (finance). As one might guess, they use Excel a lot, but at least in her school more sophisticated statistics is done almost exclusively with R. If you haven't checked out R yet, it's a really fantastic tool.
Personally, I'm in computational physics. I mainly use Fortran, C, matlab and python.
Huh? I don't find Fortran multidimensional arrays to be any more difficult than C. Fortran is column major, or generalizing to multiple dimensions, inner-most indexes are adjacent in memory, whereas C is the opposite, i.e. outer-most indexes adjacent. I.e. in C you have
foo[c][b][a]
in Fortran the equivalent element is
foo(a,b,c)
(assuming you start the Fortran array dimensions at 0, if you want to be pedantic).
Slashdot Mirror
Essentially they created a flash bytecode backend for LLVM. Now create a flash frontend (i.e. converts flash source code -> LLVM intermediate representation) and you'd have a LLVM-optimized flash compiler. Or for that matter, a flash -> native compiler (since the native backend already exists).
Um, no. ZFS is a local file system, AFS (though you probably mean AFP, but curiously it doesn't matter in this case) is a protocol for sharing files over a network.
Not necessarily. NFS4 supports kerberos encryption (just like CIFS in AD mode). If root su:s to another user he doesn't have the other users kerberos keys, hence no go.
Of course, if you have root, and the other user is logged in, you can compromise his key store and impersonate him. But that's no different from CIFS.
They're not that different. The network is compatible, the physical design is compatible, e.g. you can slot in XT4 nodes into a XT5 enclosure (which is what Cray recommends if you have jobs with high bandwidth requirements).
From an end-user perspective the only difference is that XT5 nodes have twice the number of cores (2 sockets per node instead of 1).
He criticized a piece of software, your carrying on like he insulted your god(s), Get a grip...
At least the software exists..
USS Ticonderoga (CVA-14) in the Pacific Ocean - an A4E armed with a hydrogen bomb rolls off the deck of the carrier in 16,000ft of water
Yeah, those hotshots should have listened to their driving instructor; never forget the parking brake!
Only another order of magnitude to go and they are were Linux is today (4096).
Uh, you can do that today with LVM and ext3.
Let me get this straight -- in rocket science, you can spend small and get great performance?
At least you can do it the other way around -- spend big and get crappy performance. Just look at the space shuttle.
Hasn't Red Storm switched to Linux yet? Seems most of the Cray XT sites have done so already. One reason for developing CNL (Compute Node Linux, the Linux kernel they run on the nodes) IIRC was that multicore support in Catamount was sort of a kludge. And OS buffering, and..
Then again, the catamount developers are at Sandia, so maybe they have some emotional attachment to it. :)
CNL still doesn't allow dynamic linking, though it's much less alien than catamount.
And after a few years when Microsoft follows VMWare, we'll have Microsoft DataCenter OS, abbreviated MS-DOS.
I think the point was that machines like BG or the Cray XT series are MPP:s, not clusters. But yeah, a MPP is essentially a cluster, so the distinction is not that useful.
These days Linux market share is about 85% of the top500, iirc (you can find stats on the top500.org site).
I love git, and it really is the most advanced SCM today, but if git gui wouldn't be a ugly TK app, maybe more ppl would jump on that wagon.
qgit?
No, it's just you who's getting older.
Actually, PEP-8, the official python style guide, strongly recommends spaces over tabs.
In other news, Firefox 3.1 and some future version of Opera, will have built-in support of Ogg/Theora:
http://weblogs.mozillazine.org/roc/archives/2008/08/why_ogg_matters.html
Which supercomputers might that be? This is the first time I've heard of it.
(Cray uses Flourinert for some machines, which is not a liquid metal).
So is ZFS, genius...
Fortran 95 when programmed in a "modern Fortran" style is actually a pretty modern, easy to use, and safe language. "Modern Fortran" style usually means having all procedures in modules (giving automatic argument checking and a lot of other nice things), derived types (same as structs in C), avoiding old error-prone constructs like common and equivalence, and so forth. The Fortran 90+ array features alone are worth dying for (similar to the array syntax in matlab).
My girlfriend is finishing her M.Sc thesis in economics (finance). As one might guess, they use Excel a lot, but at least in her school more sophisticated statistics is done almost exclusively with R. If you haven't checked out R yet, it's a really fantastic tool.
Personally, I'm in computational physics. I mainly use Fortran, C, matlab and python.
Huh? I don't find Fortran multidimensional arrays to be any more difficult than C. Fortran is column major, or generalizing to multiple dimensions, inner-most indexes are adjacent in memory, whereas C is the opposite, i.e. outer-most indexes adjacent. I.e. in C you have
foo[c][b][a]
in Fortran the equivalent element is
foo(a,b,c)
(assuming you start the Fortran array dimensions at 0, if you want to be pedantic).
It's "Terra Preta", btw.
One of the biologists they interviewed mentioned it in his blog:
http://scienceblogs.com/pharyngula/2007/08/im_gonna_be_a_movie_star.php
One man mentions "neo-Darwinism" which I've never heard about before.
See http://en.wikipedia.org/wiki/Neo-Darwinism
Especially Richard Dawkins lecture about neo-darwinism (linked above) is informative: http://video.google.com/videoplay?docid=4633079169415752395