Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?

Comment: Re:How about Jumping the Cuda? (Score 1) 226

by inflamed (#48722713) Attached to: Red Hat Engineer Improves Math Performance of Glibc

How about compiler being able to leverage Cuda hardware if found during execution? Might make for fatter binaries, but..... if it found GPU, it might help in some cases?

Well, the system of execution is often different from the system of compilation. So, you'd want to have the compiler-user specify that a CUDA-specific binary should be compiled alongside (afaik cuda/x86 fat binaries aren't a thing right now). And that falls into the realm of build system, not compiler. From what I've seen the build system (make in this case) relies on nvidia's compiler to builld the cuda libraries subsequent linking into the cuda executable along with the c* and fortran code.

Comment: The forms are man-made, but the life within is not (Score 1) 221

In countless ways, spatiotemporally proximal organisms interact bi-directionally. For an extreme example: A student tends a dish of cells and the student's subsequent experiences (including chances of reproducing) are affected.

A succesful student gains momentum for a career boost, rocketing towards a tenure-track job-chase. Unsuccessful students are more likely to reproduce.

Comment: Re:The obvious solution (Score 1) 348

by inflamed (#47874043) Attached to: When Scientists Give Up

There's a wealth of useful equipment on eBay and other places, big expensive equipment is not out of the reach of the dedicated researcher. Ben Krasnow has three (I think) electron microscopes. I personally own a UV/VIS spectrophotometer. a microgram scale, and a Weston cell.

The idea that "research can only be done at the behest of government" or "is only associated with university" is a modern fiction. Government would *like* you to believe that everything depends on their whim and largesse, but it's not the only, nor even the best way.

Build a lab and start tinkering, or join a hackerspace. Lots of people do it. Lots of good science is done this way.

Electron microscopes are pricey. UV/VIS specs, mmg balances, and weston cells, not so much. High field NMR spectrometers and x-ray crystallography setups? You're dreaming. Thanks for playing!

Comment: Re:Estrogen? (Score 1) 109

by inflamed (#47286907) Attached to: Scientists Successfully Grow Full Head of Hair On Bald Man

The chemical layout of Tofacitinib looks fairly similar to estrogen. We've known for ages that giving MPB-afflicted men estrogen will result in hair regrowth. Unfortunately, it also makes them grow breasts, but that's besides the point.

Throwing my moderations in this disccusion to reply to this - there ought to be a "Wrong" mod option ;) Nope, it's nothing like estrogen. One might as well conclude the structures are related to LSD: I am an organic chemist currently working in drug design and would conclude these compounds are, from a molecular standpoint, extremely disparate.

Comment: Re:Trust networks can fix this (Score 1) 120

by inflamed (#47185023) Attached to: Whom Must You Trust?

God this sounds familiar..... and that's because I wrote a PhD thesis about building a system to do something a lot like this. It involved a fairly mediocre web interface wrapping a database of trust relationships specified by end users. A trusts B for 0.7 and B trusts C for 0.6 then you can put together a trust level between A and C by multiplying those together with some user-tweakable distance dropoff. Those trust levels were then measured against the levels required for access to shared data. Maybe you would allow anyone with a 0.7 or higher to read a given document and a 0.9 or higher to contribute to it. It was an interesting idea, but man did I get tired of it by the end. If for some bizarre reason anyone wants to read bits of it google books has some indexed and I probably have a pdf laying around somewhere....

I figured it could be quite useful, but I was so fed up with the work in mid-2007 that I never looked back at it.

Thanks for laboring through a thesis on the topic, it's an occasional daydream of mine and I would love a copy. :-)

Comment: Re:Trust networks can fix this (Score 1) 120

by inflamed (#47185017) Attached to: Whom Must You Trust?

Which would work well if you could trust people to consistently submit "trust statements" truthfully and accurately. Sometimes people lie when they tell you who they trust and who they don't.

People lie but no so much when their lies are detremental to them. Such a web of trust could only be conned by 'fake' nodes which would have a very hard time developing any links to 'real' nodes.

Comment: Re:How is Burying Africa Under PCs Going to Help? (Score 1) 201

by inflamed (#46970797) Attached to: $7 USB Stick Aims To Bring Thousands of Poor People Online

Right. Most people in Africa have no electricity. Gotcha. And those without electricity are the ones paying for used PCs, $15k per container, to dump them to save Americans recycling dollars. Gotcha. The urban electrification rate in Africa is 59%. Nigeria had 6.9 million households with televisions in 2006. You are more likely to be hit by a Mercedes than to die from a machete or burning computer. This e-waste hoax never stops giving.

I don't know what your intent is in that statement but you forgot to mention that the distribution of Africa's population is 2:3 urban:rural, so the overall electrification rate is 24%. If your intent was to highlight that Africa has far bigger needs (and needs coverage of more pressing issues than) live USB thumb drives, you should have mentioned that.

Comment: Re: We're Not (Score 1) 634

by inflamed (#46966907) Attached to: Why Scientists Are Still Using FORTRAN in 2014

In your own domain -- Molecular Dynamics -- you might wish to send your initial configuration (position and velocities) to a colleague/reviewer who is using a different compiler. He could, in principle, reproduce your trajectory exactly. (Otherwise, there is a compiler error.)

You might not wish to routinely run with IEEE arithmetic, because it is slower. But for those folks who need it, it is right there at their fingertips and totally and completely (ANSI/ISO) standard.

The position and velocities or seed value will allow for reproduction of trajectories up to a certain point. Accumulation-derived errors take quite some time to develop and are acceptable. This is why relatively lossy GPU hybrid single-precision / double precision accumulation codes for CUDA are acceptable (and a game changer). Over an extended time the simulation will sample the same phase space. If the same phase space isn't sampled the simulation isn't run long enough or the model is bunk :-)

If it has syntax, it isn't user friendly.