Ok I am quite familiar with SISD(Single Instruction Single Data), this is the basic abstraction that pretty much every computer out there uses.
SIMD(Single Instruction Multiple Data) is the hot 'new' kid on the block and the basic abstraction/concept behind altivec, MMX, SSE, VIS, etc.
MIMD(Multiple Instruction Multiple Data) would (IMNSHO) be just a misnomer for VLIW(Very large instruction word) which is almost the same thing as EPIC(Explicitly parallel instruction set computing) aka Itanic... er... ia64... er... itanium(2)... er... whatever.
Now what exactly would a MISD(multiple instructions single data) system be?! And, can anyone point to an example of such a system?
My guess:
If you look out into space and a ratio(lets say L:B) of little asteroids to big asteroids and if you make the assumption that for an individual asteroid its size does not effect its chances of hitting a planet then: X/Y=L/B => XB=YL => Y=XB/L Now if you want to get fancy and remove the above assumption then it wouldn't be all that difficult to run a monte carlo simulation where you initialize a solar system with different populations of asteroids and apply the laws of gravity to simulate a few million years and use the output statistics to replace the above ratios. Hmmm maybe I should write up a grant proposal and see if anyone would pay for that cluster I've always wanted;)
Im not so sure that this is really practical in the sense you meant, but just being near a large enough blackhole would provide a simple one way time machine (cf twins paradox).
So if they use TCP/IP for the transport/network layer do they need to do a 3-way handshake?...<groan>. Ok, ok, they are probably using UDP/IP in which case there is no actual handshake... hmmm.. very matrixy;)
What is the power coefficient? There is a graph with this heading on page 7 but absolutely NO explanation... not even the usual two line benchmark comments;) The values range from 6% to 117% and the power supplies that seem to be better have higher scores... but the graph doesn't even have a comment along the lines of 'higher is better'
Actually just one component of the northbridge, the memory controller, was moved into the CPU. Other things such as: AGP interface, interface to southbridge, etc (hmmm is there anything else?) still need to be on a separate chip.
Since the trait is so recessive, the extinction of redheads is predicted to be sometime in the late 21st or early 22nd century, due to population implosion of the native populations of northern Europe and the traits being lost due to interbreeding with those with dominant dark-haired traits.
Oh boy... damn biologists can't do math! Unless there is some environmental presure working against the phenotype the percentage of genotype is (more or less) preserved across generations. If mating were truly random across the globe then in a few generations we would have just as many redheads (as a percentage of the GLOBAL population)... they would just be spread out across the globe better.
Type O blood is a recessive, BUT the percentage of type O phenotypes is not exactly going up or down b/c of type A, type B, or type AB phenotypes.
I think the term you are looking for is CAM: Content Adressable Memory.
It is used commonly to implement things like caches and switches/routers. CAM tends to have scalability issues, as you pointed out. This scalability is usually salved via a hierarchy (cf L1, L2, L3... caches). The HBSS currently solves this via an external hierarchy: HBSS -> PDA -> Computer -> Internet -> HyperInteligentSecretAlienOverseers (Ooops that was supposed to be secret.. oh well;).
Now the idea of adding another level of heirarchy(in the form of some extra GreyMatter) closer to the existing HBSS would definately improve throughput and decrease latencies (cf cpus w/ on die cache tend to perform better).
Actually it measures angular rates... you can integrate to get angles. It has a range of either +- 150 deg/sec or +- 300deg/sec depending on model. the 'noise density' is.05deg/sec/sqrt(Hz) meaning (if i am reading this correctly) that the faster you measure the more accurate each measurement is, but you have to integrate more... btw i dont usually work w/ hardware so I might be misreading the specs a little. So with it saturating(or going nuts) beyond 25rpm or 50rpm this will be just fine to measure the orientation of a robotic arm but will not keep track of how fast your spinbot is spinning. Also remember that this is a single axis gyro so that you need at least 3(4 if you want to avoid the dreaded gimbal lock) if you are interested in a complete orienation.
Re:UT2k3 - linux impressions
on
UT2003 LiveCD
·
· Score: 2
Actually I didnt expect it to run until...to my surprise... I was able to run it on w2k... I am quite disapointed that linux faired so much worse on my system, I really do hope that either I screwed something up or it is a peculiarity of my system or the beta level software.
I do get similar SPECViewperf numbers b/w the two partitions. So maybe this is just a bad intereaction w/ my system and the beta level demo.
Re:UT2k3 - linux impressions
on
UT2003 LiveCD
·
· Score: 2
Ok sorry I didn't give enough info. I am using the brand spanking new 1.0-3123 NVIDIA drivers
If it helps my complete specs are: Celeron333a - ie the one WITH a cache and the 'blazing' 66mhz fsb 440bx based mobo which 192mb of pc66 sdram - hey its a celeron and it goes nuts if I OC the fsb AT ALL! 7200rpm 40gig drive for linux 7200rpm 30gig drive for w2k - fat32 partition for easier interoperability SoundBlaster pci128 - aka es1370 based sound card. DEC de4x5 based nic and to top things of an ISA scsi card (please dont laugh out loud) to connect my scanner.
UT2k3 - linux impressions
on
UT2003 LiveCD
·
· Score: 3, Informative
I downloaded and installed UT2k3 demo on both my linux partition (up to date gentoo) and my W2k partition. I am running an admitidly unbalanced celeron333a w/ 192mb ram and a gf4ti4200-64mb. Under linux I get very choppy sound and an almost-kinda-sorta-playable slideshow after setting all the options to their min (ok I left the res at 800x600... 320x240 was in the list!).
While w2k gives me a reasonably playable game... heck its not completly unplayable at 1600x1200.
Actually there were two very valid reasons for doing this back when the original Mac UI was being developed.
The cheesy reason: It saves screen space, on a 3 or 4 inch low res monitor screen space is very valuable.
The good reason: 'targetability' With the menu always at the top of the screen it has an effectivly infinite height making it easier for the user to get to the menu (ie a quick flick upwards of the wrist always gets the mouse over the menu).
Clearly the first reason is no longer valid on todays systems, but the second still has some merit. But on the other hand if I wish to 'target' a menu item in a different document window things get much more cumbersome... I guess they just optimized for the common case at the expense of the uncommon one... not neccisarily bad but still very confusing to those of us (myself inculded) coming from a[n] [X]windows background.
I LOVE my hp48gx sitting next to me right now. It takes some getting used to the "Reverse Polish Notation" but now I find 'normal' calcs very dificult to use. It is my understanding that the line got updated to the hp49 and the HP gave up on further development of its calculator line, which is a shame since they are so much better (IMNSHO) than the TI's.
Disclaimor: My work tends toward the 'scientific computing' end of the spectrum so my calculator needs might be a bit different than the 'average' programmers. In fact I use my calc several times a day, often while I have matlab up and running too... sometimes a calculator is just much easier to use.
Sony said it would continue to offer repairs and manufacture tapes for the format, adding the move would not affect its Betacam products for the broadcasting industry.
I doubt this could be done (at least with any kind of reliability) at the driver level. It definately could be done with a little creative rewiring, but I would not be the right person to ask how to actually accomplish this.
Just watch out when doing 'multi-pipe', things can get nauseating real fast. If the monitors do not update in unison then the 'jitter' can be disconcerting enough to cause nausea. This is well documented with profesional flight sims. This is why many profesional($$$$) cards come with genlock connectors that force all the monitors to refresh simultaneously.
Slashdot Mirror
Ok I am quite familiar with SISD(Single Instruction Single Data), this is the basic abstraction that pretty much every computer out there uses.
... er ... ia64 ... er ... itanium(2) ... er ... whatever.
SIMD(Single Instruction Multiple Data) is the hot 'new' kid on the block and the basic abstraction/concept behind altivec, MMX, SSE, VIS, etc.
MIMD(Multiple Instruction Multiple Data) would (IMNSHO) be just a misnomer for VLIW(Very large instruction word) which is almost the same thing as EPIC(Explicitly parallel instruction set computing) aka Itanic
Now what exactly would a MISD(multiple instructions single data) system be?! And, can anyone point to an example of such a system?
My guess: ;)
If you look out into space and a ratio(lets say L:B) of little asteroids to big asteroids and if you make the assumption that for an individual asteroid its size does not effect its chances of hitting a planet then:
X/Y=L/B => XB=YL => Y=XB/L
Now if you want to get fancy and remove the above assumption then it wouldn't be all that difficult to run a monte carlo simulation where you initialize a solar system with different populations of asteroids and apply the laws of gravity to simulate a few million years and use the output statistics to replace the above ratios. Hmmm maybe I should write up a grant proposal and see if anyone would pay for that cluster I've always wanted
Im not so sure that this is really practical in the sense you meant, but just being near a large enough blackhole would provide a simple one way time machine (cf twins paradox).
this might be waht you want ... The documentation wasn't immediately clear to me so I might be off the mark.
honey beer ... wouldn't that be mead?!? Well, I guess that would technically be honey wine, but hey its close enough.
So if they use TCP/IP for the transport/network layer do they need to do a 3-way handshake? ...<groan>. Ok, ok, they are probably using UDP/IP in which case there is no actual handshake ... hmmm .. very matrixy ;)
What is the power coefficient? There is a graph with this heading on page 7 but absolutely NO explanation ... not even the usual two line benchmark comments ;) The values range from 6% to 117% and the power supplies that seem to be better have higher scores ... but the graph doesn't even have a comment along the lines of 'higher is better'
Actually just one component of the northbridge, the memory controller, was moved into the CPU. Other things such as: AGP interface, interface to southbridge, etc (hmmm is there anything else?) still need to be on a separate chip.
Ah ... dont forget about the periodic table table
Shouldn't this be called an iMod ... <groan>
Oh boy
Type O blood is a recessive, BUT the percentage of type O phenotypes is not exactly going up or down b/c of type A, type B, or type AB phenotypes.
I think the term you are looking for is CAM: Content Adressable Memory.
... caches). The HBSS currently solves this via an external hierarchy: HBSS -> PDA -> Computer -> Internet -> HyperInteligentSecretAlienOverseers (Ooops that was supposed to be secret .. oh well ;).
It is used commonly to implement things like caches and switches/routers. CAM tends to have scalability issues, as you pointed out. This scalability is usually salved via a hierarchy (cf L1, L2, L3
Now the idea of adding another level of heirarchy(in the form of some extra GreyMatter) closer to the existing HBSS would definately improve throughput and decrease latencies (cf cpus w/ on die cache tend to perform better).
Actually it measures angular rates ... you can integrate to get angles. It has a range of either +- 150 deg/sec or +- 300deg/sec depending on model. the 'noise density' is .05deg/sec/sqrt(Hz) meaning (if i am reading this correctly) that the faster you measure the more accurate each measurement is, but you have to integrate more ... btw i dont usually work w/ hardware so I might be misreading the specs a little. So with it saturating(or going nuts) beyond 25rpm or 50rpm this will be just fine to measure the orientation of a robotic arm but will not keep track of how fast your spinbot is spinning.
Also remember that this is a single axis gyro so that you need at least 3(4 if you want to avoid the dreaded gimbal lock) if you are interested in a complete orienation.
C'mon you make this too easy...
nwn is almost here
as is WCIII
Actually I didnt expect it to run until ...to my surprise... I was able to run it on w2k ... I am quite disapointed that linux faired so much worse on my system, I really do hope that either I screwed something up or it is a peculiarity of my system or the beta level software.
I do get similar SPECViewperf numbers b/w the two partitions. So maybe this is just a bad intereaction w/ my system and the beta level demo.
Ok sorry I didn't give enough info.
I am using the brand spanking new 1.0-3123 NVIDIA drivers
If it helps my complete specs are:
Celeron333a - ie the one WITH a cache and the 'blazing' 66mhz fsb
440bx based mobo which 192mb of pc66 sdram - hey its a celeron and it goes nuts if I OC the fsb AT ALL!
7200rpm 40gig drive for linux
7200rpm 30gig drive for w2k - fat32 partition for easier interoperability
SoundBlaster pci128 - aka es1370 based sound card.
DEC de4x5 based nic
and to top things of an ISA scsi card (please dont laugh out loud) to connect my scanner.
Normal Person: "Er.. OK. Show me The Sims next!"
Sure!! Right here.
I downloaded and installed UT2k3 demo on both my linux partition (up to date gentoo) and my W2k partition. I am running an admitidly unbalanced celeron333a w/ 192mb ram and a gf4ti4200-64mb. ... 320x240 was in the list!). ... heck its not completly unplayable at 1600x1200.
Under linux I get very choppy sound and an almost-kinda-sorta-playable slideshow after setting all the options to their min (ok I left the res at 800x600
While w2k gives me a reasonably playable game
Actually there were two very valid reasons for doing this back when the original Mac UI was being developed.
... not neccisarily bad but still very confusing to those of us (myself inculded) coming from a[n] [X]windows background.
The cheesy reason: It saves screen space, on a 3 or 4 inch low res monitor screen space is very valuable.
The good reason: 'targetability' With the menu always at the top of the screen it has an effectivly infinite height making it easier for the user to get to the menu (ie a quick flick upwards of the wrist always gets the mouse over the menu).
Clearly the first reason is no longer valid on todays systems, but the second still has some merit. But on the other hand if I wish to 'target' a menu item in a different document window things get much more cumbersome... I guess they just optimized for the common case at the expense of the uncommon one
Uh... NO
VPC is a virtual x86 box while MOL requires a PPC. It sounds like it is very similar to VMware/plex86 except for PPC.
I LOVE my hp48gx sitting next to me right now. It takes some getting used to the "Reverse Polish Notation" but now I find 'normal' calcs very dificult to use. It is my understanding that the line got updated to the hp49 and the HP gave up on further development of its calculator line, which is a shame since they are so much better (IMNSHO) than the TI's.
... sometimes a calculator is just much easier to use.
Disclaimor: My work tends toward the 'scientific computing' end of the spectrum so my calculator needs might be a bit different than the 'average' programmers. In fact I use my calc several times a day, often while I have matlab up and running too
How does an optical mouse work
I doubt this could be done (at least with any kind of reliability) at the driver level. It definately could be done with a little creative rewiring, but I would not be the right person to ask how to actually accomplish this.
How often does a programmer actually stand up? I do most of my thinking either sitting in a meeting or sitting in front of a terminal. ;)
Just watch out when doing 'multi-pipe', things can get nauseating real fast. If the monitors do not update in unison then the 'jitter' can be disconcerting enough to cause nausea. This is well documented with profesional flight sims. This is why many profesional($$$$) cards come with genlock connectors that force all the monitors to refresh simultaneously.