"I think the real question is: why are these people not working in research institutes? Image recognition is a hard problem. It's baffling that someone with that kind of talent would be working for spammers instead of in a tenured university position."
So, I have a Ph.D. and know how to write this kind of software (well, I know how to go about writing this kind of software and have done it for other domains). Here's why I'm not working at a research institute or pursing a tenured university position:
First off, research institutes don't really exist anymore. There are a few corporate labs left, but they all focus on medium term product development (5 years out). The national labs still exist, but they're managed like businesses now and it's more difficult to do pure research at them. University "institutes" are just glorified research labs. If you're not the PI, you're either a post-doc, grad student, or tech, none of which is a viable long-term career option.
To get tenure, you have to spend 4-8 years working non-stop writing grants to fund students to do research so you can build up a publication record that impresses the tenure committee. Note that grants and pubs are both necessary: grants show you can bring money into the university, publications get the approval of the committee members outside your domain who only know how to assess research abilities by impact factors.
During this time, all your research is done by graduate students, who are often at the beginning of the careers and have limited technical abilities. They may be brilliant, but they are not the most efficient workers. So, not only do you have to publish, but your labor pool consists of people with 1-3 years experience.
Before tenure, you'll also only pull in about $60-90k/yr (and I know two very smart people who worked for free their first year as "visiting professors" just to get their foot in the door). At the end of this, if you don't get tenure, you're unemployable until you build up some marketable skills.
Contrast this with industry positions. While you don't get to work on whatever you want, there are some very interesting problems out there if you take your time to find a good position. At work, you're hired to do a job, not chase down funding, so you can spend more time working on the fun stuff. The hours are reasonable, so you have time in the evenings for other projects/hobbies (you don't have free time in academia). If you're selective in your employer, you'll also work with people with a broad range of experience and skills. You'll also make more money. And, if you're good and publish from time to time, you can get a tenured position later in life without having to go through the tenure process.
Of course, if you're evil, you can also find work breaking CAPTCHAs and building bot nets.
Note that though this sounds bitter, I'm not... I had a blast going back to school and highly recommend it to people mid-career (hint: go to the mid-west where it's cheap to live and your quality-of-life will remain about the same). But, modern academic environments just don't present an enticing career path.
"...debugger integration is not really their strongest point."
You should really take a sideways look at the iPhone SDK. The debugger integration is solid and almost up there with Visual Studio for memory and thread debugging.
While xcode is technically just a wrapper on top of GCC, Apple has done an enormous amount of work to integrate all elements of the toolchain into the environment in a way that enhances developer productivity.
I used xcode when it first came out and was underwhelmed - it was really just a simple gcc wrapper back then. But, it's evolved significantly and makes the GNU tools it's built on actually efficient to use (think using the CLI version of gdb for debugging compiled, multi-threaded code on remote devices... sure, you can do it, but it's a time sink).
"How can you possibly evaluate supercomputers in any other way except how many mathematical operations can be performed in some reference time? "
It's much more subtle than that. Most programs, including weather simulations, use a large amount of data stored on disk and in RAM. The problem with LINPACK as a benchmark is that, for all practical purposes, it ignores this cost by using a few very specific linear algebra operations that have very low communication/computation ratios. The LINPACK number is only relevant if your program is primarily based on operations that have this characteristic.
Unfortunately, most scientific codes (weather simulations included, of course), have evolved past simple implementations based on dense matrix-matrix multiplication (the particular kernel that gives the peak performance number) and include a number of steps that perform closer to the speed of the memory bus than the speed of the processor (sparse matrix operations, which make simulations tractable with millions of variables work this way). There's also the simple fact that very few programmers are even aware of the techniques required to achieve even 50% of peak performance on a kernel as simple as matrix-matrix multiplication. And, the cost of getting past 50% in programmer time is rather high. So, even if scientific codes could be optimally implemented, there's almost no chance they are.
Most people in HPC (myself included) have reached the point where the Top 500 list is a fun curiosity, but has little relevance to actual practice of supercomputing. Optimizing memory bandwidth and interconnects is much more important than raw FLOPS.
Still, I applaud the Roadrunner project. They took some serious risks to pull it off and created a very impressive computer. It's too bad that it will most likely be a one-off implementation (yeah, you can buy QS-22s from IBM, but I doubt they'll be around for too long).
"I'm surprised that opengl was never really 'open'."
Remember, that 'open' in the FOSS sense is one of many meanings of the term in software. When OpenGL was created, the "Open" part referred to the standard and standards process being open. This meant that anyone could volunteer to join the working groups, pay their dues, and contribute to the design. It also meant that anyone could read the standard. Here, Open has nothing to do with the actual code. It had everything to do with it being an industry standard that was not proprietary to one company (of course, we can debate forever if SGI had undue influence over the standard, but in theory they shouldn't have).
This is a completely valid use of the word open and is pretty common among industry standards groups (even C++ works more or less this way, minus the dues paying part).
My biggest concern as a consumer of GIS data has always been access to the high-quality, tax-payer funded data (which is usually aerial, not satellite). Exclusivity deals are fine as long as any data gathered from the instruments for tax-payer funded programs remains accessible without restrictions.
I'm not sure how deals such as Google's will affect this, but as the parent pointed out, there are already many sources of high-quality data from government sources.
If there is something to fear from Google Maps/Earth, it's the spatial imagery mono-culture developing around consumer and media GIS applications. Google's approach is by no means the best approach for all geospatial data, it just happens to work well for navigating large data sets. But, as we've learned from Microsoft, if enough people are using a solution, the level of technology present in dominant solution becomes the "state-of-the-art" even if it isn't.
Most good developers are can pick up a new language fairly quickly, so it shouldn't be a hurdle.
Objective C is a little strange at first, since it mixes "real" (i.e., Smalltalk-esqe, message-based) object oriented programming concepts into a statically typed language. But, once you separate the C and the Objective C mentally, it's kinda like programing Python and writing the occasional C extension.
Cocoa Touch has a large set of support libraries that cover most tasks. Once you understand the basic patterns they use, it's pretty easy to pick up a new library and run with it. That said, the basic data structures take a little getting used to since it takes a little longer to develop a good mental model of how they work under the hood (whereas C++ Standard Library components are easier for C programmers to grok).
Anyway, don't let the language stop you. If nothing else, you might expand your knowledge of what is possible with programming languages.
The switch to Beamer in science is unfortunate. The best PowerPoint presentations I've seen are from scientists with good graphics, good data, and a good story to tell. PowerPoint is by far the most productive tool for generating slide shows that integrate non-textual information, especially where good layout helps communicate complex topics (think biology and physics). With good research, a scientist can productively build a quality presentation using PowerPoint (by productively I mean: not have to recompile every time a change is made to see what it will look like).
All academic Beamer presentations I've seen over the last five years have been people's notes with a few GNU Plot generated charts. The presenters read their bullet points, point out the charts as if their data is self evident, and fail miserably to effectively communicate their work to a broader audience. But, they used a cool LaTeX-based presentation system, so like every scientist who used "Java" or "Linux", they must be on the cutting edge. (and don't even get me started on the annoying nav bar everyone feels compelled to include in their Beamer presentations - it's fine for distributed copies, but in a conference room, the content matters, not the widgets).
The PPU is a striped down PowerPC chip. It is a simple, in-order implementation of the PowerPC ISA, a little less capable than the G4s used in older Macs, but in the same ballpark. It is fine for running the OS and performing basic tasks, but is woefully underpowered compared to most recent general purpose processors.
The SPUs actually are general-purpose processors. They support scalar operations and have branch instructions and can run arbitrary applications. Of course, the real power in the SPUs comes from the fact that they are also vector processors that can work on up to four 32-bit operands in one instruction.
There are some other features that make the SPUs even more interesting (specifically the user-programmable local store and fast interconnect between cores). But for the sake of general purpose computing, every core on the chip is a general purpose processor.
It depends on your program's definition of Computational Biology. Traditionally, Computational Biology and Bioinformatics have referred to the same general type of work, but more recently each has taken on more precise meanings, especially in industry. Computational Biology primarily refers to ab-initio, in-silico modeling of biological systems (conformation, docking, systems biology simulations). Bioinformatics refers to the analysis of biological data, primarily from the various "omics" (genomics, proteomics, transcriptomics, etc).
In industry, computational biology is viewed with a healthy dose of skepticism. The promise of in-silico modeling has been just around the corner since the 70s and no system has yet delivered on the promise. If this is what you're doing, stay in academia and keep working on the dream (but, apply to D.E. Shaw Research on the off chance you can get a job there, they're building a supercomputer for this). Or, if you would like a bigger paycheck and more predictable work, switch over to computational chemistry, which is much more accepted and an important component of most drug discovery pipelines.
Bioinformatics, on the other hand, is basis of many product and research groups. The most important skills are the ability to communicate with biologists and experience with genomic databases, genomic search tools, and statistical modeling along with the ability to tie it all together programatically. This often includes developing data mining pipelines and creating nice Web interfaces for the scientists to access them with. Good CS and programming skills can give you a leg up over the people with bioinformatics degrees. If this is what you want to do and you're young/single, biotech startups or bioinformatics startups are a lot of fun. You'll work hard and the company will probably go under, but it's a great way to get deep experience in the field and make connections for your next job.
Right now, the most exciting industrial work is probably around next-gen sequencing platforms. Look at 454/Roche, Applied Biosystems (SOLiD), or Illumina. Lots of really interesting high-performance computing, algorithmic, and scientific challenges.
Actually, with a tab separated text file, you don't even need to do regexps to separate fields. If you have cygwin (or are on Linux/Unix/etc), you can use a combination of (e)grep and cut to do most basic queries.
Say you have a tab delimited file with records of the form: key\tv1\tv2\tv3\t...\tvn. To find the record foo and get the key plus the first two columns:
egrep '^foo' myfile.txt | cut -d '\t' -f2,3
It doesn't meet the original criteria of Windows (at least w/o cygwin), but this pattern is very useful when you don't want to write a Perl/Python/Tcl/etc script.
As one of those "graduate students" who got some good publications and a thesis out of the Cell, I can tell you that the processor is definitely good for that.:)
I want to take issue with one of your negatives, however. The programmable local store is one of the best features of the Cell and actually simplifies programming for performance significantly. Once you get used to moving data and code from main memory to local store (admittedly a foreign concept for most programmers weened on Java/Ruby/PHP/etc), you wonder why you ever relied on the cache in the first place.
With the local store, there is no guessing when your data is in L1 or playing prefetching tricks (which is where most development time is spent on performance critical code). You load the data you need and it's there. No surprises. The model is actually very similar to working on a client/server system (think servlet loading data from database), but in this case the client is a core on your processor and the server is your main memory.
On your point about the PPE: it wasn't designed to be used heavily by application code. The PPE was designed to run essential OS services. All application code was supposed to be developed for the SPEs. Unfortunately, IBM never really publicized this point and a lot of developers got burned trying to use the PPE to dispatch work to SPEs.
"Right, so we'll just use the emulator forever, shall we?"
Kinda like those of us still waiting for access to the beta program? We have no choice but to continue to use the emulator forever, even if we did go out and buy devices specifically for development.
Seriously, it's annoying enough that Apple limited the beta to a lottery, but it's even worse to hear the lucky few complain that they're unable to continue development because the beta software broke their phone.
The whole point of a beta program is to test software and procedures on users willing to put up potentially buggy software to get valuable feedback prior to a real release. Things like this should be expected. If you don't want to be a good participant in a beta program, don't sign up for one (and release space for those of us who are fine beta testing things and made proper preparations for the inevitable bugs that will turn up).
I remember back in the late 90s when a fun email version of Scrabble along with all the Boggle sites were shut down by Hasbro. Just a new generation of programmers learning about trademark laws from Hasbro.
Bummer, since the free games are great advertising for the few people who don't already own hardcopies of the games. But I guess that's the problem - everyone already owns the game, so the the only way to increase revenues is charge for the online versions.
Dimensions (width x height x depth) 259x208x22 mm Laptop weight (incl. battery) 1,22 kg
For the Air:
Height: 0.16-0.76 inch (0.4-1.94 cm) Width: 12.8 inches (32.5 cm) Depth: 8.94 inches (22.7 cm) Weight: 3.0 pounds (1.36 kg)
According to those numbers, the Vaio is smaller (note that depth for the Vaio is what Apple is calling height). Unfortunately, Sony doesn't have the specs online anymore, so it's just my google search vs. yours.:)
"Then get a MacBook. Sorry but you are not going to fit it into that form factor."
"Do you want a laptop that is 0.16" to 0.76" thick? Go grab a ruler and put that in perspective. There is no way in hell you're going to do that with a standard external battery."
So, my two laptops prior to my current MacBook were a Sony Vaio PCG-505 TR (smaller than the MacBook Air) and a Toshiba Portege (about the same size as the Air). Both of these had replaceable batteries and very clever engineering designs to make them work with the form factor. It's definitely possible to do it.
I have currently have a black MacBook and two batteries. I don't leave home w/o them both fully charged and usually go through them both over the course of a day of travel. I fly all the time and it's still rare to find working outlets (at least on US flights).
Are they serious? No way to swap out a battery halfway through a 10 hour flight? No way to take it out at security check points (or if it catches fire)?
Please tell me I'm misinterpreting that phrase. Want to buy one now, but that's a deal breaker. Argh!!!
Submarines have had very small, very safe reactors for decades. Unfortunately, the technology is highly classified and will most likely never be made available for commercial uses
My brother-in-law was on an attack sub and I got to tour it (my father in-law actually got to drive it!). We weren't allowed aft passed a certain point, but give where we were in the sub, you could get an idea of how small the reactor was. Always thought it would be fun to put those reactors to use for domestic power generation, even if just for special purposes like powering server farms.
10 years ago, yes. But, seriously, it takes only about $30k to build a tera-scale system with commodity parts. And, if single precision is OK, $2400 will get you 900 "gig-flops" worth of PS3s. Last time I went through Bahrain, you could buy those in the airport for your kids, so they shouldn't be too hard for the Iranian government to buy.
I've recently discovered Nokia's Series 60 development tools. Even though portions are closed, they provide support for almost every common language (I prefer their handy Python implementation - yes, Python on a cell phone that can make calls, talk to bluetooth devices, draw images/complex UI, take pictures, record audio, etc), making it trivial to develop apps for Nokia phones.
I think the US carriers are the real problem, not the lack of good dev tools. T-mobile lets me put my apps on my phone, but I'm not sure every carrier is this permissive (and even t-mobile has some limits if I want to easily distribute my apps.
I did purposely ignore the PPUs. Unfortunately the PPU doesn't help that much for performance. The OS and network stack keep it pretty busy. Also, touching memory from the PPU that an SPU may use later significantly decreases performance (touching memory on the PPU brings it into cache, triggering a very expensive dirty update operation if the SPU reads the same memory later). The general rule for Cell programming is to pretend the PPU doesn't exist and code everything for the SPUs.
As for the GPU, no one has figured out how to use it yet (on the PS3) and even so, GP-GPU programs aren't nearly as useful as we've been led to believe. They work great for streaming graphics operations, but as another poster pointed out (in a strangely hostile response to my original post), special processors significantly limit the types of applications the chip is good for. While the Cell is proving to be useful for a number of applications (though not all - any pointer chasing algorithm (e.g. graph theoretic) is not going to perform well on the Cell), it's not perfect for everything. Adding the GPU to the mix further narrows the scope.
If you're curious about the overall usefulness of GP-GPU algorithms, go back through the literature and pay close attention to how they constrained each application to work on the GPU. In most cases, it won't scale out to real world uses. Incidentally, the same is true for most Cell applications at this point (even the ones I've written:) ).
8 PS3s gives you 1.2 teraflops of single-precision performance or a similar number if you stick to integer operations (6 SPUs/PS3 gives ~150 GFlops). 1.2 teraflops is a supercomputer in my book.
Using Jack Dongerra's single-precision algorithms that do half the work in single and the other half in double precision, you can maintain a high level of performance and precision. And, the unique architecture of the Cell opens up some interesting algorithmic research issues, allowing scientists to publish twice for the same work: once for the science results, once for the computer science results.:)
On the flip side, the Gigabit ethernet on the PS3s isn't really 1GB - the PPU can barely keep up. So, extra care must be taken around communication points. And, a similar Intel/AMD-based rack would run about $20k and is much easier to develop for, so if your labor is expensive (i.e., you're not in academia), PS3 clusters may not make much sense.
Slashdot Mirror
"I think the real question is: why are these people not working in research institutes? Image recognition is a hard problem. It's baffling that someone with that kind of talent would be working for spammers instead of in a tenured university position."
So, I have a Ph.D. and know how to write this kind of software (well, I know how to go about writing this kind of software and have done it for other domains). Here's why I'm not working at a research institute or pursing a tenured university position:
First off, research institutes don't really exist anymore. There are a few corporate labs left, but they all focus on medium term product development (5 years out). The national labs still exist, but they're managed like businesses now and it's more difficult to do pure research at them. University "institutes" are just glorified research labs. If you're not the PI, you're either a post-doc, grad student, or tech, none of which is a viable long-term career option.
To get tenure, you have to spend 4-8 years working non-stop writing grants to fund students to do research so you can build up a publication record that impresses the tenure committee. Note that grants and pubs are both necessary: grants show you can bring money into the university, publications get the approval of the committee members outside your domain who only know how to assess research abilities by impact factors.
During this time, all your research is done by graduate students, who are often at the beginning of the careers and have limited technical abilities. They may be brilliant, but they are not the most efficient workers. So, not only do you have to publish, but your labor pool consists of people with 1-3 years experience.
Before tenure, you'll also only pull in about $60-90k/yr (and I know two very smart people who worked for free their first year as "visiting professors" just to get their foot in the door). At the end of this, if you don't get tenure, you're unemployable until you build up some marketable skills.
Contrast this with industry positions. While you don't get to work on whatever you want, there are some very interesting problems out there if you take your time to find a good position. At work, you're hired to do a job, not chase down funding, so you can spend more time working on the fun stuff. The hours are reasonable, so you have time in the evenings for other projects/hobbies (you don't have free time in academia). If you're selective in your employer, you'll also work with people with a broad range of experience and skills. You'll also make more money. And, if you're good and publish from time to time, you can get a tenured position later in life without having to go through the tenure process.
Of course, if you're evil, you can also find work breaking CAPTCHAs and building bot nets.
Note that though this sounds bitter, I'm not... I had a blast going back to school and highly recommend it to people mid-career (hint: go to the mid-west where it's cheap to live and your quality-of-life will remain about the same). But, modern academic environments just don't present an enticing career path.
-Chris
"...debugger integration is not really their strongest point."
You should really take a sideways look at the iPhone SDK. The debugger integration is solid and almost up there with Visual Studio for memory and thread debugging.
While xcode is technically just a wrapper on top of GCC, Apple has done an enormous amount of work to integrate all elements of the toolchain into the environment in a way that enhances developer productivity.
I used xcode when it first came out and was underwhelmed - it was really just a simple gcc wrapper back then. But, it's evolved significantly and makes the GNU tools it's built on actually efficient to use (think using the CLI version of gdb for debugging compiled, multi-threaded code on remote devices... sure, you can do it, but it's a time sink).
-Chris
"How can you possibly evaluate supercomputers in any other way except how many mathematical operations can be performed in some reference time? "
It's much more subtle than that. Most programs, including weather simulations, use a large amount of data stored on disk and in RAM. The problem with LINPACK as a benchmark is that, for all practical purposes, it ignores this cost by using a few very specific linear algebra operations that have very low communication/computation ratios. The LINPACK number is only relevant if your program is primarily based on operations that have this characteristic.
Unfortunately, most scientific codes (weather simulations included, of course), have evolved past simple implementations based on dense matrix-matrix multiplication (the particular kernel that gives the peak performance number) and include a number of steps that perform closer to the speed of the memory bus than the speed of the processor (sparse matrix operations, which make simulations tractable with millions of variables work this way). There's also the simple fact that very few programmers are even aware of the techniques required to achieve even 50% of peak performance on a kernel as simple as matrix-matrix multiplication. And, the cost of getting past 50% in programmer time is rather high. So, even if scientific codes could be optimally implemented, there's almost no chance they are.
Most people in HPC (myself included) have reached the point where the Top 500 list is a fun curiosity, but has little relevance to actual practice of supercomputing. Optimizing memory bandwidth and interconnects is much more important than raw FLOPS.
Still, I applaud the Roadrunner project. They took some serious risks to pull it off and created a very impressive computer. It's too bad that it will most likely be a one-off implementation (yeah, you can buy QS-22s from IBM, but I doubt they'll be around for too long).
-Chris
"I'm surprised that opengl was never really 'open'."
Remember, that 'open' in the FOSS sense is one of many meanings of the term in software. When OpenGL was created, the "Open" part referred to the standard and standards process being open. This meant that anyone could volunteer to join the working groups, pay their dues, and contribute to the design. It also meant that anyone could read the standard. Here, Open has nothing to do with the actual code. It had everything to do with it being an industry standard that was not proprietary to one company (of course, we can debate forever if SGI had undue influence over the standard, but in theory they shouldn't have).
This is a completely valid use of the word open and is pretty common among industry standards groups (even C++ works more or less this way, minus the dues paying part).
-Chris
C++'s template system is actually a Turing-complete functional programming language.
-Chris
My biggest concern as a consumer of GIS data has always been access to the high-quality, tax-payer funded data (which is usually aerial, not satellite). Exclusivity deals are fine as long as any data gathered from the instruments for tax-payer funded programs remains accessible without restrictions.
I'm not sure how deals such as Google's will affect this, but as the parent pointed out, there are already many sources of high-quality data from government sources.
If there is something to fear from Google Maps/Earth, it's the spatial imagery mono-culture developing around consumer and media GIS applications. Google's approach is by no means the best approach for all geospatial data, it just happens to work well for navigating large data sets. But, as we've learned from Microsoft, if enough people are using a solution, the level of technology present in dominant solution becomes the "state-of-the-art" even if it isn't.
-Chris
Most good developers are can pick up a new language fairly quickly, so it shouldn't be a hurdle.
Objective C is a little strange at first, since it mixes "real" (i.e., Smalltalk-esqe, message-based) object oriented programming concepts into a statically typed language. But, once you separate the C and the Objective C mentally, it's kinda like programing Python and writing the occasional C extension.
Cocoa Touch has a large set of support libraries that cover most tasks. Once you understand the basic patterns they use, it's pretty easy to pick up a new library and run with it. That said, the basic data structures take a little getting used to since it takes a little longer to develop a good mental model of how they work under the hood (whereas C++ Standard Library components are easier for C programmers to grok).
Anyway, don't let the language stop you. If nothing else, you might expand your knowledge of what is possible with programming languages.
-Chris
The switch to Beamer in science is unfortunate. The best PowerPoint presentations I've seen are from scientists with good graphics, good data, and a good story to tell. PowerPoint is by far the most productive tool for generating slide shows that integrate non-textual information, especially where good layout helps communicate complex topics (think biology and physics). With good research, a scientist can productively build a quality presentation using PowerPoint (by productively I mean: not have to recompile every time a change is made to see what it will look like).
All academic Beamer presentations I've seen over the last five years have been people's notes with a few GNU Plot generated charts. The presenters read their bullet points, point out the charts as if their data is self evident, and fail miserably to effectively communicate their work to a broader audience. But, they used a cool LaTeX-based presentation system, so like every scientist who used "Java" or "Linux", they must be on the cutting edge. (and don't even get me started on the annoying nav bar everyone feels compelled to include in their Beamer presentations - it's fine for distributed copies, but in a conference room, the content matters, not the widgets).
-Chris
The PPU is a striped down PowerPC chip. It is a simple, in-order implementation of the PowerPC ISA, a little less capable than the G4s used in older Macs, but in the same ballpark. It is fine for running the OS and performing basic tasks, but is woefully underpowered compared to most recent general purpose processors.
The SPUs actually are general-purpose processors. They support scalar operations and have branch instructions and can run arbitrary applications. Of course, the real power in the SPUs comes from the fact that they are also vector processors that can work on up to four 32-bit operands in one instruction.
There are some other features that make the SPUs even more interesting (specifically the user-programmable local store and fast interconnect between cores). But for the sake of general purpose computing, every core on the chip is a general purpose processor.
-Chris
It depends on your program's definition of Computational Biology. Traditionally, Computational Biology and Bioinformatics have referred to the same general type of work, but more recently each has taken on more precise meanings, especially in industry. Computational Biology primarily refers to ab-initio, in-silico modeling of biological systems (conformation, docking, systems biology simulations). Bioinformatics refers to the analysis of biological data, primarily from the various "omics" (genomics, proteomics, transcriptomics, etc).
In industry, computational biology is viewed with a healthy dose of skepticism. The promise of in-silico modeling has been just around the corner since the 70s and no system has yet delivered on the promise. If this is what you're doing, stay in academia and keep working on the dream (but, apply to D.E. Shaw Research on the off chance you can get a job there, they're building a supercomputer for this). Or, if you would like a bigger paycheck and more predictable work, switch over to computational chemistry, which is much more accepted and an important component of most drug discovery pipelines.
Bioinformatics, on the other hand, is basis of many product and research groups. The most important skills are the ability to communicate with biologists and experience with genomic databases, genomic search tools, and statistical modeling along with the ability to tie it all together programatically. This often includes developing data mining pipelines and creating nice Web interfaces for the scientists to access them with. Good CS and programming skills can give you a leg up over the people with bioinformatics degrees. If this is what you want to do and you're young/single, biotech startups or bioinformatics startups are a lot of fun. You'll work hard and the company will probably go under, but it's a great way to get deep experience in the field and make connections for your next job.
Right now, the most exciting industrial work is probably around next-gen sequencing platforms. Look at 454/Roche, Applied Biosystems (SOLiD), or Illumina. Lots of really interesting high-performance computing, algorithmic, and scientific challenges.
Good luck!
-Chris
Actually, with a tab separated text file, you don't even need to do regexps to separate fields. If you have cygwin (or are on Linux/Unix/etc), you can use a combination of (e)grep and cut to do most basic queries.
Say you have a tab delimited file with records of the form: key\tv1\tv2\tv3\t...\tvn. To find the record foo and get the key plus the first two columns:
egrep '^foo' myfile.txt | cut -d '\t' -f2,3
It doesn't meet the original criteria of Windows (at least w/o cygwin), but this pattern is very useful when you don't want to write a Perl/Python/Tcl/etc script.
-Chris
As one of those "graduate students" who got some good publications and a thesis out of the Cell, I can tell you that the processor is definitely good for that. :)
I want to take issue with one of your negatives, however. The programmable local store is one of the best features of the Cell and actually simplifies programming for performance significantly. Once you get used to moving data and code from main memory to local store (admittedly a foreign concept for most programmers weened on Java/Ruby/PHP/etc), you wonder why you ever relied on the cache in the first place.
With the local store, there is no guessing when your data is in L1 or playing prefetching tricks (which is where most development time is spent on performance critical code). You load the data you need and it's there. No surprises. The model is actually very similar to working on a client/server system (think servlet loading data from database), but in this case the client is a core on your processor and the server is your main memory.
On your point about the PPE: it wasn't designed to be used heavily by application code. The PPE was designed to run essential OS services. All application code was supposed to be developed for the SPEs. Unfortunately, IBM never really publicized this point and a lot of developers got burned trying to use the PPE to dispatch work to SPEs.
-Chris
"Right, so we'll just use the emulator forever, shall we?"
Kinda like those of us still waiting for access to the beta program? We have no choice but to continue to use the emulator forever, even if we did go out and buy devices specifically for development.
Seriously, it's annoying enough that Apple limited the beta to a lottery, but it's even worse to hear the lucky few complain that they're unable to continue development because the beta software broke their phone.
The whole point of a beta program is to test software and procedures on users willing to put up potentially buggy software to get valuable feedback prior to a real release. Things like this should be expected. If you don't want to be a good participant in a beta program, don't sign up for one (and release space for those of us who are fine beta testing things and made proper preparations for the inevitable bugs that will turn up).
-Chris
I remember back in the late 90s when a fun email version of Scrabble along with all the Boggle sites were shut down by Hasbro. Just a new generation of programmers learning about trademark laws from Hasbro.
Bummer, since the free games are great advertising for the few people who don't already own hardcopies of the games. But I guess that's the problem - everyone already owns the game, so the the only way to increase revenues is charge for the online versions.
-Chris
Specs I have for the Vaio PCG-505TR are (from http://www.notebook-pcs.com/sony/1074.html):
:)
Dimensions (width x height x depth) 259x208x22 mm
Laptop weight (incl. battery) 1,22 kg
For the Air:
Height:
0.16-0.76 inch (0.4-1.94 cm)
Width:
12.8 inches (32.5 cm)
Depth:
8.94 inches (22.7 cm)
Weight:
3.0 pounds (1.36 kg)
According to those numbers, the Vaio is smaller (note that depth for the Vaio is what Apple is calling height). Unfortunately, Sony doesn't have the specs online anymore, so it's just my google search vs. yours.
-Chris
Actually, I'd suggest a Sony Vaio or Toshiba Portege. Both lines have notebooks with similar form factors that didn't make as many tradeoffs.
-Chris
"Then get a MacBook. Sorry but you are not going to fit it into that form factor."
"Do you want a laptop that is 0.16" to 0.76" thick? Go grab a ruler and put that in perspective. There is no way in hell you're going to do that with a standard external battery."
So, my two laptops prior to my current MacBook were a Sony Vaio PCG-505 TR (smaller than the MacBook Air) and a Toshiba Portege (about the same size as the Air). Both of these had replaceable batteries and very clever engineering designs to make them work with the form factor. It's definitely possible to do it.
-Chris
I have currently have a black MacBook and two batteries. I don't leave home w/o them both fully charged and usually go through them both over the course of a day of travel. I fly all the time and it's still rare to find working outlets (at least on US flights).
-Chris
(from tech specs page on apple.com)
"Integrated 37-watt-hour lithium-polymer battery"
Are they serious? No way to swap out a battery halfway through a 10 hour flight? No way to take it out at security check points (or if it catches fire)?
Please tell me I'm misinterpreting that phrase. Want to buy one now, but that's a deal breaker. Argh!!!
-Chris
Submarines have had very small, very safe reactors for decades. Unfortunately, the technology is highly classified and will most likely never be made available for commercial uses
My brother-in-law was on an attack sub and I got to tour it (my father in-law actually got to drive it!). We weren't allowed aft passed a certain point, but give where we were in the sub, you could get an idea of how small the reactor was. Always thought it would be fun to put those reactors to use for domestic power generation, even if just for special purposes like powering server farms.
-Chris
10 years ago, yes. But, seriously, it takes only about $30k to build a tera-scale system with commodity parts. And, if single precision is OK, $2400 will get you 900 "gig-flops" worth of PS3s. Last time I went through Bahrain, you could buy those in the airport for your kids, so they shouldn't be too hard for the Iranian government to buy.
Not sure what the story is here...
-Chris
Huh? I've been using free (as in beer and sometimes as in speech) languages and libraries for Nokia's S60 platform (built on Symbian) for a while now:
http://www.forum.nokia.com/
http://en.wikipedia.org/wiki/S60_platform
-Chris
Mod the parent up!
I've recently discovered Nokia's Series 60 development tools. Even though portions are closed, they provide support for almost every common language (I prefer their handy Python implementation - yes, Python on a cell phone that can make calls, talk to bluetooth devices, draw images/complex UI, take pictures, record audio, etc), making it trivial to develop apps for Nokia phones.
I think the US carriers are the real problem, not the lack of good dev tools. T-mobile lets me put my apps on my phone, but I'm not sure every carrier is this permissive (and even t-mobile has some limits if I want to easily distribute my apps.
-Chris
I did purposely ignore the PPUs. Unfortunately the PPU doesn't help that much for performance. The OS and network stack keep it pretty busy. Also, touching memory from the PPU that an SPU may use later significantly decreases performance (touching memory on the PPU brings it into cache, triggering a very expensive dirty update operation if the SPU reads the same memory later). The general rule for Cell programming is to pretend the PPU doesn't exist and code everything for the SPUs.
As for the GPU, no one has figured out how to use it yet (on the PS3) and even so, GP-GPU programs aren't nearly as useful as we've been led to believe. They work great for streaming graphics operations, but as another poster pointed out (in a strangely hostile response to my original post), special processors significantly limit the types of applications the chip is good for. While the Cell is proving to be useful for a number of applications (though not all - any pointer chasing algorithm (e.g. graph theoretic) is not going to perform well on the Cell), it's not perfect for everything. Adding the GPU to the mix further narrows the scope.
If you're curious about the overall usefulness of GP-GPU algorithms, go back through the literature and pay close attention to how they constrained each application to work on the GPU. In most cases, it won't scale out to real world uses. Incidentally, the same is true for most Cell applications at this point (even the ones I've written :) ).
-Chris
8 PS3s gives you 1.2 teraflops of single-precision performance or a similar number if you stick to integer operations (6 SPUs/PS3 gives ~150 GFlops). 1.2 teraflops is a supercomputer in my book.
:)
Using Jack Dongerra's single-precision algorithms that do half the work in single and the other half in double precision, you can maintain a high level of performance and precision. And, the unique architecture of the Cell opens up some interesting algorithmic research issues, allowing scientists to publish twice for the same work: once for the science results, once for the computer science results.
On the flip side, the Gigabit ethernet on the PS3s isn't really 1GB - the PPU can barely keep up. So, extra care must be taken around communication points. And, a similar Intel/AMD-based rack would run about $20k and is much easier to develop for, so if your labor is expensive (i.e., you're not in academia), PS3 clusters may not make much sense.
-Chris