every since I read the write up, I've been humming his tune: noooothing compaaaares 2u. I like the dichotomy between the long pronounciation and short spelling.
And, to remain marginally on topic, I think "tho" is first in line for legitimisation. "Light" -> "lite" isn't enough of a savings to be worthwhile. For no particular reason, I'm not nearly as close to accepting "thru". No idea why...
HP's Dynamo is a _great_ technology: basically, it is trace-based binary translator. It acheives impressive speedups by 1) only doing the expensive optimisations when deemed likely to be useful, and 2) discovering dynamic code paths than can be turned into straight-line code, and then re-optimised as a unit.
Dynamo achieved 20% speedups when run with source==target architectures (ie, it beat native code in a apples-to-apples test).
I'm kinda despairing tho; the paper is several years old, and any 27 obvious applications have failed to use it.
Re:Not just Stephenson...
on
0wnz0red
·
· Score: 1
erm... maybe I do.
Blood Music was back when Bear was good. However, between Vitals and Darwin's Radio, I'm none to happy with him---c'mon two books in a row with so-so plot devices, and absolutely no endings? I honestly think he owes me money back for Vitals, it was so bad.
I'm currently re-reading Eon to see whether I was just young and impressionable when I decided I liked him.
Egan still kicks ass tho, even if he didn't write the book I alledged.
Re:Not just Stephenson...
on
0wnz0red
·
· Score: 2
I thought it somewhat similar to Egan's "Blood Music" (which goes in a different direction), and very similar to Barnes's [sic] "Mother of Storms" (*). In the latter book, the protagonist's thought processes are enhanced, and he eventually discovers how to control exactly these autonomous processes.
It's also a damn good book: Barnes either sucks or rocks. This book is in the latter category.
for $1500 a month/head you're not sharing an appt, but a palace. You're talking marble stairs and sub-zero fridges.
even in boston.
You live well, alone, for $1100; sharing should be cheaper - high end shares right downtown will not go much higher than $1000, all included. Accept a 10 minute commute and you're easily at the $800 mark.
even under M$, I would have thought that the browser's memory would have been swapped to disk, thereby taking up no physical memory, when you need all the RAM you can get.
Are there any parallels to FFT? A FFT goes between the frequency and the temporal domains for a signal. These cameras allow us to trade off resolution in the spatial domain for resolution in the temporal domain.
I'm wondering if we couldn't basically reduce the camera to one _very_ fast pixel, and then FFT to retrieve a sequence high resolution image. Of course, just one pixel wouldn't work, but I'm throwing that out there as an extreme to illustrate what I lack the vocabulary to express.
In switzerland, they just plaster up the frames on the wall at equal spacing, and then flash a stroboscope at the proper frequency. Instead of moving the film over the projector, the projector moves over the film...
What's your take on the two level approach: things like NEPL or Python + NumPy?
The idea being that you write the complicated bits in a high-level language that can be debugged and understood by mortals, and the really hairy parallel bits in a low-level language. The assumption being that the complicated bits tend to take relatively little CPU time, but alot of programmer time.
The punchline is that the low level bits tend to always be matrix ops, so you just wrap BLAS or some other hyper-optimized code (heh! Fastest FF in the West perhaps) in a suitable foreign function interface and call that directly from the high-level language. So you don't need to write low level code ever...
This always seemed to me to be the most appealing way to write numerical programs, but since I _never_ need numerical code, what do I know?
about 10**-2 * c**2 Joules worth of damage. (assuming total conversion, which is WAY optimistic) c ~ 3*10**8 , so 9 * 10**14 Joules. According to http://pointa.autodesk.com/local/eng/portal/resour ces/cad/formulas.jsp?po=eng a tonne of TNT has 4.2 * 10**9, so about 2 * 10**5 tonnes TNT, or in human speak, about 200 kilotonnes.
I hope that you don't regret putting your phone number out publically on a radical forum such as this.
While I applaud your openness, I can't help but fear that you'll end up leaving it off the hook to avoid the loonies(*) amoung us, thereby also avoiding the sanies as well.
(*) I was going to say something vague and drastic about family and emergencies and whatnot, but who doesn't have a semi-private cellphone these days? Perhaps you're right. These sorts of purposes may be exactly what you have a work line for.
you don't want the probes to survive longer than planned. You want them to be like F1 race cars: ideally, the engine should explode _just_ over the finish line. Only then have you maximized tolerances. However, due to uncertainty, you engineer in a margin of safety.
A 30 year margin doesn't indicate good design, it indicates a MASSIVE misjudgemnt of the tolerances involved. Fine. these were the first probes built, so noone knew the margins needed.
It's misguided to continue insisting on such ludicrous margins. If you want a long-living probe, then that becomes a design consideration, but this _moves the finish line_, rather than increasing the margins necessary.
The long life of the probes is indicative of good engineers making conservative choices in the face of uncertainty rather than good design.
aside:
the only reason why fast-cheap-cheerful isn't a handsdown winner is that each probe's cost is augmented by the cost of launch, which makes even a free probe an expensive mission. Thus, there is economic gain from a bit of overengineering, as the cost of the hardware isn't really a large part of the total cost, so any bonus functionality you get is worth the price, to a limit.
The real loss if the ISS is shut down will be that they could have built a rail-gun to fire largely unpowered probes on long-term missions for basically free.
Great stuff. includes the guy from earthing, which too is great stuff. If you like that, then Fila Brazilia is not far behind. However, I recommmend Telepopmusik because it is broad enough to appeal to all but the terminally closed minded.
Not forgetting the always cool BeOS solution, where the vinyl was a special time-code record, which was sampled by the computer, and which caused corresponding scratching to be performed on an mp3 file.
The cool thing with that setup was that you could needle drop anywhere into the track, in addition to using whichever player you had at hand.
Of course, I imagine they are dead in the water, what with the proliferation of BeOS platforms to run on and what not.
Slashdot Mirror
Silk won a few years ago, and that is based on C (or is it C++?). May be spelled Cilk, now that I think about it
my money's on finland.
every since I read the write up, I've been humming his tune: noooothing compaaaares 2u. I like the dichotomy between the long pronounciation and short spelling.
And, to remain marginally on topic, I think "tho" is first in line for legitimisation. "Light" -> "lite" isn't enough of a savings to be worthwhile. For no particular reason, I'm not nearly as close to accepting "thru". No idea why...
oh, for mod points... Bravo!
HP's Dynamo is a _great_ technology: basically, it is trace-based binary translator. It acheives impressive speedups by 1) only doing the expensive optimisations when deemed likely to be useful, and 2) discovering dynamic code paths than can be turned into straight-line code, and then re-optimised as a unit.
Dynamo achieved 20% speedups when run with source==target architectures (ie, it beat native code in a apples-to-apples test).
I'm kinda despairing tho; the paper is several years old, and any 27 obvious applications have failed to use it.
erm... maybe I do.
Blood Music was back when Bear was good. However, between Vitals and Darwin's Radio, I'm none to happy with him---c'mon two books in a row with so-so plot devices, and absolutely no endings? I honestly think he owes me money back for Vitals, it was so bad.
I'm currently re-reading Eon to see whether I was just young and impressionable when I decided I liked him.
Egan still kicks ass tho, even if he didn't write the book I alledged.
I thought it somewhat similar to Egan's "Blood Music" (which goes in a different direction), and very similar to Barnes's [sic] "Mother of Storms" (*). In the latter book, the protagonist's thought processes are enhanced, and he eventually discovers how to control exactly these autonomous processes.
It's also a damn good book: Barnes either sucks or rocks. This book is in the latter category.
All of Egan's work is highly recommended.
quoth lotussuper7:
> I'm NOT a "Car guy"
hrm. With a name and sig like that, we'd be forgiven for making that assumption
for $1500 a month/head you're not sharing an appt, but a palace. You're talking marble stairs and sub-zero fridges.
even in boston.
You live well, alone, for $1100; sharing should be cheaper - high end shares right downtown will not go much higher than $1000, all included. Accept a 10 minute commute and you're easily at the $800 mark.
still get's stuck in the email wizard. Completely impossible to back out of! Great engineering...
even under M$, I would have thought that the browser's memory would have been swapped to disk, thereby taking up no physical memory, when you need all the RAM you can get.
Are there any parallels to FFT? A FFT goes between the frequency and the temporal domains for a signal. These cameras allow us to trade off resolution in the spatial domain for resolution in the temporal domain.
I'm wondering if we couldn't basically reduce the camera to one _very_ fast pixel, and then FFT to retrieve a sequence high resolution image. Of course, just one pixel wouldn't work, but I'm throwing that out there as an extreme to illustrate what I lack the vocabulary to express.
In switzerland, they just plaster up the frames on the wall at equal spacing, and then flash a stroboscope at the proper frequency. Instead of moving the film over the projector, the projector moves over the film...
I _think_ that dvi-d is a dual channel dvi connector, as per the earlier comment that 1600x1200 was pretty much all that one DVI channel is good for.
I'm suprised you're getting any picture at all, using dvi-i
Interesting.
What's your take on the two level approach: things like NEPL or Python + NumPy?
The idea being that you write the complicated bits in a high-level language that can be debugged and understood by mortals, and the really hairy parallel bits in a low-level language. The assumption being that the complicated bits tend to take relatively little CPU time, but alot of programmer time.
The punchline is that the low level bits tend to always be matrix ops, so you just wrap BLAS or some other hyper-optimized code (heh! Fastest FF in the West perhaps) in a suitable foreign function interface and call that directly from the high-level language. So you don't need to write low level code ever...
This always seemed to me to be the most appealing way to write numerical programs, but since I _never_ need numerical code, what do I know?
It's just so hokey, tho, the way you have to plug the TV sound output into the microphone jack to have it sampled.
It just doesn't give a very solid impression
*pedantic*
r ces/cad/formulas.jsp?po=eng
about 10**-2 * c**2 Joules worth of damage. (assuming total conversion, which is WAY optimistic) c ~ 3*10**8 , so 9 * 10**14 Joules.
According to http://pointa.autodesk.com/local/eng/portal/resou
a tonne of TNT has 4.2 * 10**9, so about 2 * 10**5 tonnes TNT, or in human speak, about 200 kilotonnes.
Nice yield for half a gram worth of payload.
I hope that you don't regret putting your phone number out publically on a radical forum such as this.
While I applaud your openness, I can't help but fear that you'll end up leaving it off the hook to avoid the loonies(*) amoung us, thereby also avoiding the sanies as well.
(*) I was going to say something vague and drastic about family and emergencies and whatnot, but who doesn't have a semi-private cellphone these days? Perhaps you're right. These sorts of purposes may be exactly what you have a work line for.
bravo!
That was a truly terrible joke. I can't believe you beat me to it.
Good show!
No!
you don't want the probes to survive longer than planned. You want them to be like F1 race cars: ideally, the engine should explode _just_ over the finish line. Only then have you maximized tolerances. However, due to uncertainty, you engineer in a margin of safety.
A 30 year margin doesn't indicate good design, it indicates a MASSIVE misjudgemnt of the tolerances involved. Fine. these were the first probes built, so noone knew the margins needed.
It's misguided to continue insisting on such ludicrous margins. If you want a long-living probe, then that becomes a design consideration, but this _moves the finish line_, rather than increasing the margins necessary.
The long life of the probes is indicative of good engineers making conservative choices in the face of uncertainty rather than good design.
aside:
the only reason why fast-cheap-cheerful isn't a handsdown winner is that each probe's cost is augmented by the cost of launch, which makes even a free probe an expensive mission. Thus, there is economic gain from a bit of overengineering, as the cost of the hardware isn't really a large part of the total cost, so any bonus functionality you get is worth the price, to a limit.
The real loss if the ISS is shut down will be that they could have built a rail-gun to fire largely unpowered probes on long-term missions for basically free.
How are things in mp3 land?
Are you guys developing any new mp3 players? Other cool gadgets?
I'm thinking a slimp3 with _wireless_ ethernet would be pretty cool. Also with spdif output.
Even more fun would be something I hadn't expected.
I love that band!
Bare bone nest is one of 5 records that keeps the LP player hooked up to the stereo.
telepopmusik.
Great stuff. includes the guy from earthing, which too is great stuff. If you like that, then Fila Brazilia is not far behind. However, I recommmend Telepopmusik because it is broad enough to appeal to all but the terminally closed minded.
isn't that what's normally called an EP?
Not forgetting the always cool BeOS solution, where the vinyl was a special time-code record, which was sampled by the computer, and which caused corresponding scratching to be performed on an mp3 file.
The cool thing with that setup was that you could needle drop anywhere into the track, in addition to using whichever player you had at hand.
Of course, I imagine they are dead in the water, what with the proliferation of BeOS platforms to run on and what not.