| IDL User Group Meeting Huge Success: Correction [message #63006] |
Fri, 17 October 2008 13:55 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Folks,
I wrote in my User Group Meeting review yesterday:
> Peter Messmer continues to up the performance
> of GPULib, using graphical processors to increase computationally
> expensive operations from 20-100%.
Unfortunately, I got this wrong. The numbers are not 20-100
*percent*, but 20-100 *times* as fast!! Someone who knows
more about this than I do, had this to say:
The GPULib performance improvements described by Peter Messmer
from Tech-X Corp are on the order of 20 to 100 *times* (versus
percent). If you take this a step further and write custom
kernels using the NVIDIA CUDA library directly, you can achieve
results many times even this remarkable number. GPU processing
has the potential to open up an entire universe of problem sets
that researchers would not have even considered up until now
because the computations would have otherwise required
supercomputing time.
I understand Mort Canty tried this on a PCA analysis on just
a regular run-of-the-mill graphics card and found a 17 *times*
improvement in performance. When you are working with large
satellite images, as I am, this gets your heart beating a
little faster. The 100% numbers Peter was talking about came,
as I understand it, on high-end graphics processors that cost
in the neighborhood of $1500. Not very much to pay for near-
supercomputer processing power.
Sorry for the confusion.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: IDL User Group Meeting Huge Success [message #63009 is a reply to message #63006] |
Fri, 17 October 2008 09:36  |
ed.schmahl
Messages: 11
Registered: October 2008
|
Junior Member |
|
|
On Oct 16, 11:42 pm, David Fanning <n...@dfanning.com> wrote:
> Folks,
>
> I thought you might like an update on today's IDL User Group
> meeting in Boulder. What a HUGE success! Kudo's to Harold Cline,
> Bill Okubo, and all the fine folks at ITTVIS for a wonderful
> day of exciting and stimulating presentations.
>
> The festivities actually started yesterday, when a dozen or
> so "early testers" of the new IDL 7.1 Workbench were invited
> to do a series of exercises in front of the IDL engineering
> staff. I was one of the "testers", and boy did they get
> an earful about my ideas for iTools! But not once was I
> told, "Now, hang on, Fanning, you are WAY out of line!"
> Instead, everyone wrote down my suggestions and seemed
> genuinely interested in those things that irritate the hell
> out of me about iTools.
>
> I can't tell you how unprecedented and refreshing this approach
> seemed to me. Imagine listening to what your customers like
> and don't like about your product. It was great. And after
> I finished ranting, I felt like I was heard, and--even more
> important to me--that the folks at ITTVIS really cared. I can't
> say I've converted into a big fan of the direction IDL is
> moving in, but I will say I am impressed with the people
> currently running the show at ITTVIS. I think they are
> doing some very, very important things right. So much so,
> that I've even decided to give iTools another chance. (Not
> that I have much choice, but I've signed an agreement not
> to give away any secrets. :-)
>
> Today was jam-packed with presentations, all of them
> excellent. Goodness, people do some interesting things
> with IDL! But even better than that was seeing in person
> some of my friends from this newsgroup. Ken Bowman was
> there, and Mark Buie. Kelly Dean, who is here occasionally,
> and with whom I have carried on numerous e-mail conversations
> over the years. Amara Grap of IDL Wavelets fame was there.
> Ronn Kling and Rob Dimeo, two of the best students I ever
> had in IDL classes were there, with IDL programs that put
> mine to shame. Vince Hradil, the young man who appears to be
> taking over my place on this newsgroup flew in from Chicago for
> the day,and I am happy to report you are going to be in good hands
> for a long time to come. David Stern even showed up, looking
> for a job, I guess, what with the current economic crisis
> playing havoc with all the money he made selling RSI.
>
> Lots and lots of fun.
>
> I'm going to be short-changing a lot of people here, but
> here were some of the outstanding highlights for me.
>
> Rob Dimeo has some amazing quantum mechanical visualizations
> that he runs in real-time on a honking Dell computer, that he
> claims is faster than the fastest Supercomputer in 1990. It
> was an impressive (if back-breaking to carry around in an
> airport) computer. Peter Messmer continues to up the performance
> of GPULib, using graphical processors to increase computationally
> expensive operations from 20-100%.
>
> Richard Azuah's presentation of how it took him only 3+ years
> to write an iTools application (6 months before he could write
> a line of code, I heard him say, as he basically had to re-write
> the entire set of documentation) not only confirmed everything
> I believe about iTools, but pushed it into the nightmare category.
> Richard is the only person I know, outside of ITTVIS, who has
> ever successfully written an iTool application, and my hat is
> off to him. I think we will all be writing iTool applications
> now that he has forged the way. ;-)
>
> (It was probably not a coincidence that when I asked an ITTVIS
> sales rep how IDL sales were fairing he said the real gain had
> been in software consulting. Yeah, I guess...)
>
> Ronn Kling demonstrated some amazingly good software for fitting
> ellipses in 2, 3, or N-dimensions(!), and *extremely* fast
> software for telling if a point (or, actually, as he demonstrated,
> MANY points) are inside a 2D polygon or 3D polyhedra or not. Very,
> very slick, but not surprising if you get one of Ronn's Christmas
> Cards every year. (I'm not clear where you can find these, whether
> on Ronn's web page or in the IDL Contrib Library on the ITTVIS
> web page. Maybe he can clear that up. I know I want to use these
> programs.)
>
> Finally, the absolute highlight for me was Eduardo Iturrate
> demonstrating how to use his Revolution IDL program. This
> program was mentioned here a year or so ago, but didn't make
> much of a splash at the time. It allows the user to quickly
> build a 3D graphics scene from stock "templates". Then, and
> this is the best point, you can save the IDL object graphics
> code to a file.
>
> I thought it was a neat idea at the time, and I downloaded
> it, but I was busy at the time and didn't follow up too much.
> I am *definitely* giving this program another look! OK, Eduardo
> is a master at the little keyboard short-cuts you need to know
> to move around in this program effectively, but this was the
> most amazing, spectacular demo of IDL functionality I have
> ever seen.
>
> The best thing about it is that is solves the number one
> problem with object graphics: it takes a DAMN long time to
> get even one object coded up, and there are at least 1000
> ways to do it wrong. With at least 999 of those ways, you
> see absolutely nothing on your display. So it is massively
> frustrating to learn how to code object graphics. (Ok, Ronn
> Kling's book Power Graphics in IDL helps. A lot.) But
> with this program, you can add an object in about 10 seconds,
> and 2 seconds later, you are looking at the IDL code that
> does it right! This will squash the object graphics learning
> curve practically flat.
>
> I came home *really* jazzed about writing IDL object
> graphics programs again for the first time in a LONG time.
> This program is named perfectly. It could cause a revolution
> in your IDL programming. It is available on the ITTVIS
> User Contrib web page.
>
> Alright, I've left a lot of good things out, but you
> get the idea. ITTVIS is hoping they can get this going
> on an annual basis. I can tell you this, I am not going
> to ever miss one. This was terrific. Hats off to the
> great crew at ITTVIS for putting on a most productive day.
>
> Cheers,
>
> David
>
> --
> David Fanning, Ph.D.
> Fanning Software Consulting, Inc.
> Coyote's Guide to IDL Programming:http://www.dfanning.com/
> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
Great summary of the IDL user's group, David!
In his usual self-deprecating way, David neglected to mention his own
talk about direct object graphics. In about 15 minutes, he managed to
demonstrate all the basic parts of an IDL object, and how to create an
object on the command line. To one like myself who uses objects all
the time, but only has a hazy understanding of how they are created,
this was quite a revelation. Now I can go back to reading Dave's "IDL
Programing Techniques", and start making my own objects.
|
|
|
|
| Re: IDL User Group Meeting Huge Success [message #63010 is a reply to message #63006] |
Fri, 17 October 2008 08:02 |
jeffnettles4870
Messages: 111
Registered: October 2006
|
Senior Member |
|
|
On Oct 17, 1:42 am, David Fanning <n...@dfanning.com> wrote:
> Folks,
>
> I thought you might like an update on today's IDL User Group
> meeting in Boulder. What a HUGE success! Kudo's to Harold Cline,
> Bill Okubo, and all the fine folks at ITTVIS for a wonderful
> day of exciting and stimulating presentations.
>
> The festivities actually started yesterday, when a dozen or
> so "early testers" of the new IDL 7.1 Workbench were invited
> to do a series of exercises in front of the IDL engineering
> staff. I was one of the "testers", and boy did they get
> an earful about my ideas for iTools! But not once was I
> told, "Now, hang on, Fanning, you are WAY out of line!"
> Instead, everyone wrote down my suggestions and seemed
> genuinely interested in those things that irritate the hell
> out of me about iTools.
>
> I can't tell you how unprecedented and refreshing this approach
> seemed to me. Imagine listening to what your customers like
> and don't like about your product. It was great. And after
> I finished ranting, I felt like I was heard, and--even more
> important to me--that the folks at ITTVIS really cared. I can't
> say I've converted into a big fan of the direction IDL is
> moving in, but I will say I am impressed with the people
> currently running the show at ITTVIS. I think they are
> doing some very, very important things right. So much so,
> that I've even decided to give iTools another chance. (Not
> that I have much choice, but I've signed an agreement not
> to give away any secrets. :-)
>
> Today was jam-packed with presentations, all of them
> excellent. Goodness, people do some interesting things
> with IDL! But even better than that was seeing in person
> some of my friends from this newsgroup. Ken Bowman was
> there, and Mark Buie. Kelly Dean, who is here occasionally,
> and with whom I have carried on numerous e-mail conversations
> over the years. Amara Grap of IDL Wavelets fame was there.
> Ronn Kling and Rob Dimeo, two of the best students I ever
> had in IDL classes were there, with IDL programs that put
> mine to shame. Vince Hradil, the young man who appears to be
> taking over my place on this newsgroup flew in from Chicago for
> the day,and I am happy to report you are going to be in good hands
> for a long time to come. David Stern even showed up, looking
> for a job, I guess, what with the current economic crisis
> playing havoc with all the money he made selling RSI.
>
> Lots and lots of fun.
>
> I'm going to be short-changing a lot of people here, but
> here were some of the outstanding highlights for me.
>
> Rob Dimeo has some amazing quantum mechanical visualizations
> that he runs in real-time on a honking Dell computer, that he
> claims is faster than the fastest Supercomputer in 1990. It
> was an impressive (if back-breaking to carry around in an
> airport) computer. Peter Messmer continues to up the performance
> of GPULib, using graphical processors to increase computationally
> expensive operations from 20-100%.
>
> Richard Azuah's presentation of how it took him only 3+ years
> to write an iTools application (6 months before he could write
> a line of code, I heard him say, as he basically had to re-write
> the entire set of documentation) not only confirmed everything
> I believe about iTools, but pushed it into the nightmare category.
> Richard is the only person I know, outside of ITTVIS, who has
> ever successfully written an iTool application, and my hat is
> off to him. I think we will all be writing iTool applications
> now that he has forged the way. ;-)
>
> (It was probably not a coincidence that when I asked an ITTVIS
> sales rep how IDL sales were fairing he said the real gain had
> been in software consulting. Yeah, I guess...)
>
> Ronn Kling demonstrated some amazingly good software for fitting
> ellipses in 2, 3, or N-dimensions(!), and *extremely* fast
> software for telling if a point (or, actually, as he demonstrated,
> MANY points) are inside a 2D polygon or 3D polyhedra or not. Very,
> very slick, but not surprising if you get one of Ronn's Christmas
> Cards every year. (I'm not clear where you can find these, whether
> on Ronn's web page or in the IDL Contrib Library on the ITTVIS
> web page. Maybe he can clear that up. I know I want to use these
> programs.)
>
> Finally, the absolute highlight for me was Eduardo Iturrate
> demonstrating how to use his Revolution IDL program. This
> program was mentioned here a year or so ago, but didn't make
> much of a splash at the time. It allows the user to quickly
> build a 3D graphics scene from stock "templates". Then, and
> this is the best point, you can save the IDL object graphics
> code to a file.
>
> I thought it was a neat idea at the time, and I downloaded
> it, but I was busy at the time and didn't follow up too much.
> I am *definitely* giving this program another look! OK, Eduardo
> is a master at the little keyboard short-cuts you need to know
> to move around in this program effectively, but this was the
> most amazing, spectacular demo of IDL functionality I have
> ever seen.
>
> The best thing about it is that is solves the number one
> problem with object graphics: it takes a DAMN long time to
> get even one object coded up, and there are at least 1000
> ways to do it wrong. With at least 999 of those ways, you
> see absolutely nothing on your display. So it is massively
> frustrating to learn how to code object graphics. (Ok, Ronn
> Kling's book Power Graphics in IDL helps. A lot.) But
> with this program, you can add an object in about 10 seconds,
> and 2 seconds later, you are looking at the IDL code that
> does it right! This will squash the object graphics learning
> curve practically flat.
>
> I came home *really* jazzed about writing IDL object
> graphics programs again for the first time in a LONG time.
> This program is named perfectly. It could cause a revolution
> in your IDL programming. It is available on the ITTVIS
> User Contrib web page.
>
> Alright, I've left a lot of good things out, but you
> get the idea. ITTVIS is hoping they can get this going
> on an annual basis. I can tell you this, I am not going
> to ever miss one. This was terrific. Hats off to the
> great crew at ITTVIS for putting on a most productive day.
>
> Cheers,
>
> David
>
> --
> David Fanning, Ph.D.
> Fanning Software Consulting, Inc.
> Coyote's Guide to IDL Programming:http://www.dfanning.com/
> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
Glad it was such a sucess! Wish I could've been there. Also very glad
to hear some promising news about iTools. Sounds like some neat stuff
is being done with IDL and I'm glad to hear Eduardo's Revolution IDL
getting some spotlight time. Way overdue in my opinion - a LOT of
potential avenues to explore there I think.
|
|
|
|
|
|
|
|
| Re: IDL User Group Meeting Huge Success [message #63014 is a reply to message #63006] |
Fri, 17 October 2008 01:17 |
Mort Canty
Messages: 134
Registered: March 2003
|
Senior Member |
|
|
It's morning here in Germany, David, so I get to be the first to thank
you for a great summary.
Best
Mort
David Fanning schrieb:
> Folks,
>
> I thought you might like an update on today's IDL User Group
> meeting in Boulder. What a HUGE success! Kudo's to Harold Cline,
> Bill Okubo, and all the fine folks at ITTVIS for a wonderful
> day of exciting and stimulating presentations.
>
> The festivities actually started yesterday, when a dozen or
> so "early testers" of the new IDL 7.1 Workbench were invited
> to do a series of exercises in front of the IDL engineering
> staff. I was one of the "testers", and boy did they get
> an earful about my ideas for iTools! But not once was I
> told, "Now, hang on, Fanning, you are WAY out of line!"
> Instead, everyone wrote down my suggestions and seemed
> genuinely interested in those things that irritate the hell
> out of me about iTools.
>
> I can't tell you how unprecedented and refreshing this approach
> seemed to me. Imagine listening to what your customers like
> and don't like about your product. It was great. And after
> I finished ranting, I felt like I was heard, and--even more
> important to me--that the folks at ITTVIS really cared. I can't
> say I've converted into a big fan of the direction IDL is
> moving in, but I will say I am impressed with the people
> currently running the show at ITTVIS. I think they are
> doing some very, very important things right. So much so,
> that I've even decided to give iTools another chance. (Not
> that I have much choice, but I've signed an agreement not
> to give away any secrets. :-)
>
> Today was jam-packed with presentations, all of them
> excellent. Goodness, people do some interesting things
> with IDL! But even better than that was seeing in person
> some of my friends from this newsgroup. Ken Bowman was
> there, and Mark Buie. Kelly Dean, who is here occasionally,
> and with whom I have carried on numerous e-mail conversations
> over the years. Amara Grap of IDL Wavelets fame was there.
> Ronn Kling and Rob Dimeo, two of the best students I ever
> had in IDL classes were there, with IDL programs that put
> mine to shame. Vince Hradil, the young man who appears to be
> taking over my place on this newsgroup flew in from Chicago for
> the day,and I am happy to report you are going to be in good hands
> for a long time to come. David Stern even showed up, looking
> for a job, I guess, what with the current economic crisis
> playing havoc with all the money he made selling RSI.
>
> Lots and lots of fun.
>
> I'm going to be short-changing a lot of people here, but
> here were some of the outstanding highlights for me.
>
> Rob Dimeo has some amazing quantum mechanical visualizations
> that he runs in real-time on a honking Dell computer, that he
> claims is faster than the fastest Supercomputer in 1990. It
> was an impressive (if back-breaking to carry around in an
> airport) computer. Peter Messmer continues to up the performance
> of GPULib, using graphical processors to increase computationally
> expensive operations from 20-100%.
>
> Richard Azuah's presentation of how it took him only 3+ years
> to write an iTools application (6 months before he could write
> a line of code, I heard him say, as he basically had to re-write
> the entire set of documentation) not only confirmed everything
> I believe about iTools, but pushed it into the nightmare category.
> Richard is the only person I know, outside of ITTVIS, who has
> ever successfully written an iTool application, and my hat is
> off to him. I think we will all be writing iTool applications
> now that he has forged the way. ;-)
>
> (It was probably not a coincidence that when I asked an ITTVIS
> sales rep how IDL sales were fairing he said the real gain had
> been in software consulting. Yeah, I guess...)
>
> Ronn Kling demonstrated some amazingly good software for fitting
> ellipses in 2, 3, or N-dimensions(!), and *extremely* fast
> software for telling if a point (or, actually, as he demonstrated,
> MANY points) are inside a 2D polygon or 3D polyhedra or not. Very,
> very slick, but not surprising if you get one of Ronn's Christmas
> Cards every year. (I'm not clear where you can find these, whether
> on Ronn's web page or in the IDL Contrib Library on the ITTVIS
> web page. Maybe he can clear that up. I know I want to use these
> programs.)
>
> Finally, the absolute highlight for me was Eduardo Iturrate
> demonstrating how to use his Revolution IDL program. This
> program was mentioned here a year or so ago, but didn't make
> much of a splash at the time. It allows the user to quickly
> build a 3D graphics scene from stock "templates". Then, and
> this is the best point, you can save the IDL object graphics
> code to a file.
>
> I thought it was a neat idea at the time, and I downloaded
> it, but I was busy at the time and didn't follow up too much.
> I am *definitely* giving this program another look! OK, Eduardo
> is a master at the little keyboard short-cuts you need to know
> to move around in this program effectively, but this was the
> most amazing, spectacular demo of IDL functionality I have
> ever seen.
>
> The best thing about it is that is solves the number one
> problem with object graphics: it takes a DAMN long time to
> get even one object coded up, and there are at least 1000
> ways to do it wrong. With at least 999 of those ways, you
> see absolutely nothing on your display. So it is massively
> frustrating to learn how to code object graphics. (Ok, Ronn
> Kling's book Power Graphics in IDL helps. A lot.) But
> with this program, you can add an object in about 10 seconds,
> and 2 seconds later, you are looking at the IDL code that
> does it right! This will squash the object graphics learning
> curve practically flat.
>
> I came home *really* jazzed about writing IDL object
> graphics programs again for the first time in a LONG time.
> This program is named perfectly. It could cause a revolution
> in your IDL programming. It is available on the ITTVIS
> User Contrib web page.
>
> Alright, I've left a lot of good things out, but you
> get the idea. ITTVIS is hoping they can get this going
> on an annual basis. I can tell you this, I am not going
> to ever miss one. This was terrific. Hats off to the
> great crew at ITTVIS for putting on a most productive day.
>
> Cheers,
>
> David
>
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
