NASA Scientists Simulate Black Hole Collision

Krishna Dagli writes to tell us Yahoo! News is reporting that NASA scientists have managed to simulate the merger of two massive orbiting black holes. Using technology from Silicon Graphics, Inc. built from 20 SGI Altix systems the team was able to show how the resulting gravitational waves would interact with surrounding space.

Re:didn't know processors have memory

[eliot1785]

They actually have several layers of memory (registers, L1, L2...), it's just not called that.

I don't think that's what they meant though.

Re:didn't know processors have memory

[Anonymous Coward]

What they mean is that they were able to join the address spaces of the different machines to form a single, unified one. Without any extra software, a processor can thus execute load and stores to the memory of a different machine simply by using an address that is mapped to the memory of another SGI Altix - the machine does all the rest.

In other words, you're able to use shared-memory forms of multi-processor programming, such as threads, instead of message-passing, as is used e.g. by clusters (think pthreads instead of MPI).

Re:How?

[amRadioHed]

The thing we don't understand about black holes is the singularity itself. The observable behavior of the black hole is mostly understood and I think that is what is being simulated. The stuff that is truly mysterious is hidden away within the event horizon.

Re:Dup?

[DoctorBit]

Yep, http://science.slashdot.org/article.pl?sid=06/04/1 9/1443210 [slashdot.org]

Re:didn't know processors have memory

[prefect42]

I think you need to relax your terminology a little. On an Altix box you have what used to be called C-Bricks. Basically a unit that contains processors and RAM. Those all link together over NUMAFlex (with appropriate routers) to form your large shared memory machine. But the RAM is still localised (as it's a NUMA architecture). So 'main memory' should be considered as 'owned' by a processor (or processors). If you'd made an OpenMOSIX cluster to match you'd refer to it as a machine's memory, but since all these C-Bricks form a single machine whole, you can't do that.

Some more info.

[Stoutlimb]

I actually went to a seminar years back by one of the individuals working on this. The equation alone filled pages, and was something he had to derive by hand. He showed us a cgi video of the results. The 2 black holes approached, snapped together, and the resulting larger black hole temporarily oscillated. The strange part was partway through the oscillation, the black hole just popped out of existence, and then reappeard several seconds later.

In the question and answer period, a student asked why this gap in the calculations. The professor explained there was no gap in the calculations, but rather, the result of the calculations was non-euclidean in nature, so it was physically impossible to display it in a 3d model. At about that time, half of the undergrad audience whispered a Keanu Reeves style "whoah..."

Don't ask me any of the details, this was years ago in a course on stellar astrophysics that I have mostly forgot. This is just something anecdotal. Astrophysicists have been working on this black hole merger thing for a very long time. The computer labs at the time had P133's running. I'd love to see what they're doing now, but that site wasn't very big on actual information.

Re:How?

[Anonymous Coward]

I'm a physics student and have to complain a bit.

The "purely hypothetical mathematical model" is known as General Relativity - a theory which has sustained every test physicist could conceive so far.
Blacks holes itself not only were found to be the fate of any superdense object in this theory but are widely accepted by almost any physicist as real.

Wikipedia gets it right:
http://en.wikipedia.org/wiki/Black_hole [wikipedia.org]:

The existence of black holes in the universe is well supported by astronomical observation, particularly from studying X-ray emission from X-ray binaries and active galactic nuclei.