Hmm, I originbally thought along the same lines (taking SCO's CEO talk about "the complete program" at face value), but then I thought: "Well, most likely he is not a tech. person, just an administrator, giving interview to BusinessWeek, why on the Earth would he care to distinguish between a 'program' and a 'source file'". Quote: "... it's not just a line or two, it's an entire section -- and in some cases, an entire program." --- would YOU say 'section' in this context, or would you say 'function' or 'block' instead?
You don't want the employees that you have to monitor continuously.
Hmm, for any company that needs Kevin's help in securing itself (and can pay for it!) it would be rather prudent to "monitor" pretty closely not only him, but all other employees as well...
I believe the original poster was referring to this presentation by one of Intel guys:
http://www.intel.com/research/mrl/library/micro3 2k eynote.pdf (check out slide #8) comparing power density on CPU surface with other hot objects. Hot plate power density was surpassed by Pentium II, now we are moving pretty fast to nuclear reactor, and Sun surface is there on the plot too!
It is Motion JPEG, essentially the same as MPEG inside the frame, but without motion estimation (and resulting frame-to-frame dependance). I've heard that this is what professional video editors use (for the reasons you stated above).
Also, as a shameless plug, a friend of mine builds hardware MJPEG encoder/decoder boards for Linux. Check out http://www.linuxmedialabs.com/
Well, in satellite business the hardware that actually performs useful function is "payload" (to distinuguish from the rest of the satellite which just supports it). Maybe in weather balloon business it is the same.
Another example, in SONET packet actual user data is "payload", the rest (control messages, sync, etc., is "overhead")
Paul B.
Nope, air is FASTER! :)
on
1KM 802.11b @ 2MB
·
· Score: 3, Informative
The plastic in your coax (or insulation on your twisted pair) has epsilon of more than 1, slowing signal down. Light is fastest in vacuum, in air it's just a notch slower, but when you start adding solid insulation, it can get you down quite a bit.
Paul B.
It's NOT disk space, it's memory footprint, silly!
on
Ark Linux
·
· Score: 2
You compile all apps -static and you get 50 copies of malloc() in your memory (RAM+swap, but eventually you would want them to be in RAM) simultaneously! (actually, ps -ax|wc showed my that I have 89 processes running now, and I suppose most of them would use malloc(), right?)
Next, factor in cache trashing -- that shared malloc() code could've occupied couple lines in a corner of your instruction cache, now you need to flush/reload them every time application context is switched.
Assuming that you would want to have your video equipment plugged in all the time and want to hijack its bandwidth only when it's transmitting blank stream, you can hook up (entirely in software) a weird setup which would dump data packets over digital video link (as if they were streams of pixels) and decode them on the other end, assuming that your video streams pass thru some kind of video caption cards inside more or less general-purpose computer...:)
Yeah, it might be slow, but as a next step you might consider some steganography: why not use couple LSBs in each pixel for data and the rest for video?
(If anyone cares to moderate this, please mod it as "funny", not "informative";-) )
Do what UNIX passwd does: generate a couple-bytes random number, concat it with the IP address before MD5ing AND write it in clear-text in front of the hash. Given an IP we can always make the same hash, but the required dictionary size grows exponentially with thge number of bytes in seed.
The funniest quote from the original news story at: http://www.newsfactor.com/perl/story/19980.ht ml
Intel cited a Harris Interactive (Nasdaq: HPOL) Latest News about Harris Interactive study it commissioned that showed the vast majority of computer users do more than one task at a time on their machines, with 50 percent saying they play video games while also burning CDs, for instance. Many of those same users say older machines purchased three or more years ago can have difficulty performing several tasks at once.
So, how EXACTLY hyperthreading will help an I/O bound problem (burning a CD) AND a GPU-bound problem (rendering OpenGL) simultaneously?;) If anything, it will make worst-case response for these real-time tasks worse!
Just some stupid marketoid speaking... I like Apple's straightforward approach ("Supercomputer on your desk/lap") way better... At least Apple is not trying to convince people that a fine (for many applications which know how to use it!) feature in their CPU can really help mundane tasks like burning a CD.
Two points: They built a "half-smp" inside a single processor.
It's more of smp squared (or at least doubled...;) ) -- threads in multithreaded app share instructon AND data caches (no cache coherency problems which are going to plague SMP implementations more and more), as well as register files (fast data exchange which does not require access to the main memory nor even to the inter-processor bus).
one should code multi-threaded for apps that can benefit from parallelism
One? Or the compiler? You use s/w available in source code, right?;-)
The area where it's useful to use a cluster is in problems where you can split the calculation in small pieces.
hmm, not to sound too much like a purist, but it is useful to use a cluster when you can split the calculation in LARGE pieces. When pieces are SMALL you need communication often and get bit by latency...
Actually I remember that IP-over-SCSI driver was implemented back in pre-1.0 kernel days! I doubt it would be easier to port it to 2.5.xx than to write it from scratch though...
Actually I had the pleasure of meeting fathers of the original Beowulf (Don Becker and Tom Sterling) and their story goes that it was originally built exactly like this: a bunch of 4-port cards connected in hypercube configuration. By the way, for this case you can scale it to 2^4=16 nodes with 3 hops worst case latency.
The reason for this was that Don at the time was writing a linux driver for that particular card and needed some justification for that activity...:)
Problem with this approach is that you do message routing in software running on your computational nodes, not too efficient compared to dedicated hardware on a switch. Thus, switches were used ever since...
nice program, but it looks like it is more for accumulation of THE SAME image to clean up noise than for alignment of a mosaic of images. Also, it works only with B/W BMP files, not that I mind converting to BMP, but B/W is a serious limitation.
Must be a good choice for the astronomy though, mod parent up!
On a related topic, I'm wondering what do you people use for image stitching?
Image stitching is taking a bunch of low-res images from, say small cam and aligning them automatically to produce large hi-res image. Good program would have to compensate for lens distortion and tries to balance color between different sub-images.
There are Panotools which is GPLed but I've never made it to work for me (I just need to stitch one image every two months:) ), another one is PanaVue, commercial Windows-only SW available as shareware (but adds nasty logos all over if free as beer version is used).
Any other suggestions/projects? (this is that time in 2-month cycle when I need it...)
1) Most of your pictures are standard size, right? Come up with a grid to lay them on your scanner. Stick to that grid.
2) get ppm tools (bunch of image manipulation programs callable from command line or scripts). I assume you have access to a UNIX/Linux command line, huh?:)
3) Find the tool which makes cuts from larger image, write a four line script to automatically cut out, say, all four corners of your large scan.
4) Run another script to apply "aggressive crop" to cut out any remaining white margins (you did not place original photos too close to each other on step one, right?)
Nope, only about couple hundred GHz. You are limited by the gap voltage of superconductors, at too high frequency electron pairs which carry supercurrent break down. Nb goes to about 700 GHz for simplest digital divide-by-two operations.
Even tougher limit is speed of light in your wires.
Believe me, I do this fdor fun and paycheck too!:)
IANAL, but as far as I know, in the US the law
system is based on precedents. If one court
finds (for whatever reason) that GPL is
unenforceable it establishes a precedent, which
can (and will) be referred to by "bad guys" in
all subsequent cases.
Of course if you are dealing with small claims court matters you can as well do everything yourself, but for something like defending GPL it is soo much better to ask for help from EFF or FSF lawers: the stakes are just too high...
I guess I read this warning on one of "How to beat your traffic ticket" sites...
Paul B.
Slashdot Mirror
I've started to enjoy PDFs recently for a couple of reasons..
...
;-)
a) its better for handing off documentation to customers in an unmodifiable form, especially project plans
If you have customers who can edit a PostScript print-out of your project plan BY HAND ---
LUCKY YOU!!!
Paul B.
Hmm, I originbally thought along the same lines (taking SCO's CEO talk about "the complete program" at face value), but then I thought: "Well, most likely he is not a tech. person, just an administrator, giving interview to BusinessWeek, why on the Earth would he care to distinguish between a 'program' and a 'source file'". Quote: "... it's not just a line or two, it's an entire section -- and in some cases, an entire program." --- would YOU say 'section' in this context, or would you say 'function' or 'block' instead?
Paul B.
You don't want the employees that you have to monitor continuously.
Hmm, for any company that needs Kevin's help in securing itself (and can pay for it!) it would be rather prudent to "monitor" pretty closely not only him, but all other employees as well...
PaulBu
I believe the original poster was referring to
3 2k eynote.pdf
this presentation by one of Intel guys:
http://www.intel.com/research/mrl/library/micro
(check out slide #8)
comparing power density on CPU surface with other
hot objects. Hot plate power density was
surpassed by Pentium II, now we are moving
pretty fast to nuclear reactor, and Sun surface
is there on the plot too!
Anyway, check it out.
Paul Bu.
Was I the only one to read the title ;-)
as desALIENation plant? (As in desinfection,
but cleaning up aliens and their remains
from crashed UFOs)
Anyway...
Paul B.
It is Motion JPEG, essentially the same as MPEG inside the frame, but without motion estimation (and resulting frame-to-frame dependance). I've heard that this is what professional video editors use (for the reasons you stated above).
Also, as a shameless plug, a friend of mine builds hardware MJPEG encoder/decoder boards for Linux. Check out http://www.linuxmedialabs.com/
Maybe this is what the original poster needed?
Paul B.
Well, in satellite business the hardware that actually performs useful function is "payload" (to distinuguish from the rest of the satellite which just supports it). Maybe in weather balloon business it is the same.
Another example, in SONET packet actual user data is "payload", the rest (control messages, sync, etc., is "overhead")
Paul B.
The plastic in your coax (or insulation on your twisted pair) has epsilon of more than 1, slowing signal down. Light is fastest in vacuum, in air it's just a notch slower, but when you start adding solid insulation, it can get you down quite a bit.
Paul B.
You compile all apps -static and you get 50 copies of malloc() in your memory (RAM+swap, but eventually you would want them to be in RAM) simultaneously! (actually, ps -ax|wc showed my that I have 89 processes running now, and I suppose most of them would use malloc(), right?)
Next, factor in cache trashing -- that shared malloc() code could've occupied couple lines in a corner of your instruction cache, now you need to flush/reload them every time application context is switched.
Paul B.
Assuming that you would want to have your video equipment plugged in all the time and want to hijack its bandwidth only when it's transmitting blank stream, you can hook up (entirely in software) a weird setup which would dump data packets over digital video link (as if they were streams of pixels) and decode them on the other end, assuming that your video streams pass thru some kind of video caption cards inside more or less general-purpose computer... :)
;-) )
Yeah, it might be slow, but as a next step you might consider some steganography: why not use couple LSBs in each pixel for data and the rest for video?
(If anyone cares to moderate this, please mod it as "funny", not "informative"
Paul B.
Do what UNIX passwd does: generate a couple-bytes random number, concat it with the IP address before MD5ing AND write it in clear-text in front of the hash. Given an IP we can always make the same hash, but the required dictionary size grows exponentially with thge number of bytes in seed.
Paul B
Well, actually their Linux commercials/ads
. gu erilla.idg/
a /1 1ADCO.html
:) )
were quite funny!
See:
http://www.cnn.com/2001/TECH/industry/04/19/ibm
or
http://www.nytimes.com/2002/01/11/business/medi
("He just loves the game..."
Paul B.
The funniest quote from the original news story at:t ml
;) If anything, it will make worst-case response for these real-time tasks worse!
http://www.newsfactor.com/perl/story/19980.h
Intel cited a Harris Interactive (Nasdaq: HPOL) Latest News about Harris Interactive study it commissioned that showed the vast majority of computer users do more than one task at a time on their machines, with 50 percent saying they play video games while also burning CDs, for instance. Many of those same users say older machines purchased three or more years ago can have difficulty performing several tasks at once.
So, how EXACTLY hyperthreading will help an I/O bound problem (burning a CD) AND a GPU-bound problem (rendering OpenGL) simultaneously?
Just some stupid marketoid speaking... I like Apple's straightforward approach ("Supercomputer on your desk/lap") way better... At least Apple is not trying to convince people that a fine (for many applications which know how to use it!) feature in their CPU can really help mundane tasks like burning a CD.
Paul B.
Two points:
;) ) -- threads in multithreaded app share instructon AND data caches (no cache coherency problems which are going to plague SMP implementations more and more), as well as register files (fast data exchange which does not require access to the main memory nor even to the inter-processor bus).
;-)
They built a "half-smp" inside a single processor.
It's more of smp squared (or at least doubled...
one should code multi-threaded for apps that can benefit from parallelism
One? Or the compiler? You use s/w available in source code, right?
The area where it's useful to use a cluster is in problems where you can split the calculation in small pieces.
hmm, not to sound too much like a purist, but it is useful to use a cluster when you can split the calculation in LARGE pieces. When pieces are SMALL you need communication often and get bit by latency...
Paul B.
Actually I remember that IP-over-SCSI driver
was implemented back in pre-1.0 kernel days!
I doubt it would be easier to port it to 2.5.xx
than to write it from scratch though...
Paul B.
Actually I had the pleasure of meeting fathers
:)
of the original Beowulf (Don Becker and Tom Sterling)
and their story goes that it was originally built
exactly like this: a bunch of 4-port cards
connected in hypercube configuration. By the
way, for this case you can scale it to 2^4=16
nodes with 3 hops worst case latency.
The reason for this was that Don at the time was
writing a linux driver for that particular card
and needed some justification for that
activity...
Problem with this approach is that you do message routing in software running on your computational
nodes, not too efficient compared to dedicated
hardware on a switch. Thus, switches were used
ever since...
Paul B.
Yeah, checked their datasheets, very neat indeed!
... and clickable transparent keys on top to
:) Maybe we should ;-)
provide tactile feedback.
Even better if screen is backlit and each key
has a built-in lens to focus whatever is
on screen on slightly matted keycap surface.
Sounds like a patentable idea!
give it a try?
Paul B.
nice program, but it looks like it is more
for accumulation of THE SAME image to clean up
noise than for alignment of a mosaic of images.
Also, it works only with B/W BMP files, not that
I mind converting to BMP, but B/W is a serious
limitation.
Must be a good choice for the astronomy though,
mod parent up!
On a related topic, I'm wondering what do you
:) ), another
people use for image stitching?
Image stitching is taking a bunch of low-res
images from, say small cam and aligning them
automatically to produce large hi-res image.
Good program would have to compensate for lens
distortion and tries to balance color between
different sub-images.
There are Panotools which is GPLed but I've
never made it to work for me (I just need to
stitch one image every two months
one is PanaVue, commercial Windows-only SW
available as shareware (but adds nasty logos
all over if free as beer version is used).
Any other suggestions/projects? (this is that
time in 2-month cycle when I need it...)
Paul B.
1) Most of your pictures are standard size, right? Come up with a grid to lay them on your scanner. Stick to that grid.
:)
2) get ppm tools (bunch of image manipulation programs callable from command line or scripts).
I assume you have access to a UNIX/Linux command line, huh?
3) Find the tool which makes cuts from larger image, write a four line script to automatically cut out, say, all four corners of your large scan.
4) Run another script to apply "aggressive crop" to cut out any remaining white margins (you did not place original photos too close to each other on step one, right?)
5) Burn that CD!
Paul B.
Nope, only about couple hundred GHz. You are limited by the gap voltage of superconductors, at too high frequency electron pairs which carry supercurrent break down. Nb goes to about 700 GHz for simplest digital divide-by-two operations.
:)
Even tougher limit is speed of light in your wires.
Believe me, I do this fdor fun and paycheck too!
Paul B.
Would not help you about patents: even if you make the same "invention" entirely independently an the clean room, the patent holder can still sue you.
Paul B.
IANAL, but as far as I know, in the US the law system is based on precedents. If one court finds (for whatever reason) that GPL is unenforceable it establishes a precedent, which can (and will) be referred to by "bad guys" in all subsequent cases.
Of course if you are dealing with small claims court matters you can as well do everything yourself, but for something like defending GPL it is soo much better to ask for help from EFF or FSF lawers: the stakes are just too high...
I guess I read this warning on one of "How to beat your traffic ticket" sites...
Paul B.