Re:C++ not useful for numerical computing
on
Why not Ruby?
·
· Score: 1
C++ isn't horrible for numerical computing, it's just harder to do it right in C++. If you're writing a specialized numerical integrator, so solve a highly tuned set of fluid equations for example, you're probably better off with Fortran. However, if you need a toolkit to perform research and analysis C++ can be wonderful. I speak from experience. I've spent the last 3 months extending the blitz++ library. I've built a very powerful toolkit for analyzing enormous data sets. I slice, dice, chop, rearrange, transform, and abuse my data. The memory management is rock solid. The code is fast (admittedly not as fast as some C prototyping that we've done, but close), and the whole OO paradigm has worked very well. Yes, C++ is not useful for quick one off numeric solutions, but it is great for building extensible research toolkits if you take the time to think your solution out and do it right.
To get back on topic, I use Ruby too. I've built parsers to analyze code and automate some very hairy analysis tasks. We've written patents and sold projects with Ruby code. I think that Python would next best language that I could have done that work with, but I felt that Ruby gave me much more freedom.
Hi Shawn,
I am very impressed with how hard you are working towards commercial Linux applications while still supporting the ideas of free software. I've noticed you on several mailing lists, and you seem to offer advice as much as you seek it. How do you find the time to run a company, develop software, and participate in the comminity?
The Template Metaprogramming mentioned in other posts illustrate the power of templates. While there may not be a facility for looping in the "linear iterative process" (see "Structure and Interpretation of Computer Programs for a definition) sense, there is in the "linear recursive process" sense. What happens is the proprocessor expands the templates. If a template has a template as a parameter, the embedded template is also expanded. This continues until an atomic (terminating) parameter is passed or until the preprocessor runs out of steam. Of course, for this to work on compilers like gcc, you need to increase the -ftemplate-depth-xx to obscene values. Since the preprocessor expands the code into new, legal, C++ code, you can write very fast loops. This is because the loop is unrolled, and you have explicit code (that the preprocessor wrote from your instructions) that usually smokes its competitors. Now for the obligatory link: Go to the Blitz++ web page to see a nice example (and some interesting papers on the subject).
The thing I find amusing about this is that languages like Lisp provide this exact functionality without the truly cryptic syntax (although you could argue that parentheses are replaced by angle brackets).
The original Castle Wolfenstein and Beyond Castle Wolfenstein. I spent many hours playing these games. Although I liked the pace of Castle Wolfenstein better, I liked having the knife in Beyond Castle Wolfenstein. It really added variability to the difficulty of the game. I prided myself of being able to silently assassinate all of the guards in the bunker with my knife. No bullets!
It also reminds me of the first time that a computer game creeped me out. In the final room where Hitler is marching back and forth in front of his officers, when everyone yelled 'Heil!' for the first time I nearly jumped out of my skin. Amazingly enough, this was done with the crappy Apple ][ speaker.
I whip up my own Linux distribution. I'll call it Lap-Linux. Everything that comes on the CD is licsnsed under the GPL, except for one item: the installer. I provide all the sources, for the GPL'ed items on the CD, and they are accessible without installing the system. Now my installer has the alleged behavior of Windows XP. It scans the system, calls home, and allows one install on one machine. My licensing reflects this. If I built my installer from scratch, using only my code, I have the right to do this.
I haven't violated the GPL, but I impose a one license, one seat restriction with respect to the installer. I haven't taken away your GPL guaranteed freedoms; the GPLed source is there for the taking. You can do whatever you want with it. You just can't touch my installer.
That is the simplified view of what Caldera is doing. Whether it helps or hurts their business, they have a right to do that.
It seems like there is a lot of misunderstanding about what Caldera is doing, what they want to acomplish, and if what they are doing is even allowed by the GPL. I'll touch on every one of those issues.
1) They want to charge liscensing fees for their software. They are putting a whole bunch of code they they developed onto a CD and selling it under some sort of single desktop license. Good for them. Why?
2) They want to make money. One way to make money off of software is by selling licenses. Nearly every player in the software world does this. There is nothing wrong with it. If you don't want to buy it then don't buy it. There is no need to get your panties in a twirl. But you say. . .
3) What about the GPL? Well, what about it? Nothing Caldera is doing violates the GPL (at least, I can't prove it). Is their proprietary stuff derivatie of GPL work? Probably not. Are they restricting your freedoms as guaranteed by the GPL? On GPL'd software, probably not. Remember, they only need to make the source available for you, as well as any changes they have made to it. If they do that, they have lived up to their end of the bargain. You can take the source off their CD, repackage it, roll around in printouts of it naked, so on and so forth.
This certainly isn't new. Take a look at the SuSE license. You can make copies of SuSE's proprietary code and give it to all of your friends. The moment you try to sell copied ISO's, SuSE legal will come down on you like the hammer of god (ok, maybe not that hard). They are out there to make money. They produce value added software and sell it (along with service). It's common, and not anything to get excited about.
The great thing about the I86 computing world is that you have a choice. You can choose Red Hat, Debian, Caldera, FreeBSD, Windows 2000. ..anything that you want! If you like Caldera (I tried it, didn't like it), good for you. Use it. If you don't like their politics, good for you. Don't use it.
You're right. My sentiments about not having a modern Fortran compiler rose two years ago, when I graduated and had to give up my SGI O2. That is when I started to use Linux, and I was very dissapointed that I couldn't compile my thesis work on my PC at home. I've moved on since then, and haven't worked in Fortran since my graduation. However, I still attribute part of my success as a scientific programmer (right now I'm writing innovative tranformation software in C++, ugh) to the skills I picked up in graduate school.
I never would have used Fortran if my academic advisor (and numerical analysis teacher) hadn't forced me to. C and C++ were banned from his course. He claimed that students spent too much time worrying about semi-colons, overloaded operators, and managing memory than writing good software that explored the ideas he taught in his class. This just goes to show, you should use the right tools for the job.
Fortran does not suck. The Fortran you are thinking of is F77 and it's younger siblings. The strange formatting (a throwback to the punch card days), the awful syntax, and the CAPS. For my thesis I used Fortran 90/95. It was a fantastic language. I started my research in Matlab, but found that it was too slow. I wanted to port my work to a compiled language as fast as I could, and F90 had the perfect syntax for it. What would have taken me weeks in any other language took me three days in F90. It was fast, easy to extend (add new functions, new integrators, etc), and let me use a rich set of prewritten tools that I didn't need to worry about debugging.
If you're writing the newest and best desktop, use C++, or Java, or something else well suited for the job. If you want to integrate complex fluid systems or model intricate mechanical systems, grab yourself a Fortran compiler and go to town.
One of my biggest complaints about Linux is that there isn't a modern open source Fortran compiler for it.
I just tried to download KAI++ and install it on my SuSE box. Amazingly enough, Intel fucked themselved (and their customers) by releasing versions of their compiler that were dependent upon the RedHat 6.2 version of egcs, and the RedHat 7.1 version of GCC (2.9unstable). GCC 3.0 will come out, all of the distros will get up to speed, and I'll be able to use an optimizing compiler to help my fast code run faster. Slightly off topic, but not really.
If you want your students to be code monkeys, teach them C++, Java, C, or anything else like that right off the bat. They will gain early experience on how to build their syntax handling skills, and can put C++, Java, C, etc. . . on their resumes.
However, if you want to teach them ideas, and techniques, and how to think critically about computing problems, languages like Lisp and Scheme are the way to go. Impractical? Most students feel that way. After all, how many companies out there want people who can write in Lisp? It pays off in the long run, though. Lisp and Scheme have minimal syntax, but allow you to express complex ideas. You can spend more time on teaching computer science fundamentals and less time on syntax. This pays off when then students learn C++, or Java, or Perl. They are capable of learning the syntax in the context of "doing it right," rather than just in the context of "getting the damn thing to compile."
Now, to further complicate the issue. C should be taught before C++/Java. Learn syntax, structure building, and clean programming practices. Then teach C++/Java in the context of OO Design and Design Patterns. After all, those areas are where C++/Java shine, so the languages should be taught in that context to prevent the formation of bad habits (more so in C++ than in Java).
The main problem with my philosopy is that it takes planning and discipline to implement it. Most people are too short sighted to handle it. I was lucky. For the most part I took the path described above. It lead me to a lead developer position in a research and development firm. I feel like my work is (mostly) effortless, yet I constantly amaze people with what I produce.
However, you shoot yourself in the foot here. You should not have to read a 300-page tome about the operating system before installing it for the first time.
Never had to do that with Linux. Nope. Never.
To stay somewhat on topic, I'm running OS 9, OS X, and SuSE 7.1 on my Blue and White G3. SuSE with Windowmaker and Mac on Linux blows OS X away (except in the "ohhhh, pretty" contest").
I think that KDE is great. It's slick, well thought out, and complete. Having said that, I don't use it. Simply put Windowmaker stikes a very good balance between usability, eye candy, and speed on my Blue and White G3. When I actually need to get work done (at places like work), I prefer to use PWM. It loads very fast, handles windows well (I like have 9 or 10 xterms in one frame), and for the most part stays out of my way. Speed isn't an issue here; I develop on a dual PIII. What I would like to see from the KDE developers is some sort of KDE Lite. Then again, I guess that I could just join the development team and do it myself.
What we really need is a secure television remote control. My old room mate once bought a watch that doubled as a TV remote. He programmed it to work with the house TV then proceeded to mess with everyone in the house. Imagine what marketers, or the FBI, or some haX0R could do to the televisions of the world. When will we wake up and see the real threats that face us?
Try "The Chronicles of Thomas Covenant, The Unbeliever." This is a great fantasy series. The protaginist has Leprosy, and is somehow summoned to a world where magic works and his disease can be cured. However, he doesn't believe that the world is real, and so through the entire series he acts like a bastard. After the first 50 pages of this series you find it really hard to like Covenant, but you keep reading to see what happens. It ain't Tolkein, but it is good.
He came to my university as part of the Assembly Series, and gave a very good talk on protecting the environment, wrapped in a funny story about a quirky bird from New Zealand that mates only once every seven years (and is thus, very endangered).
Ironically enough, that same year the former Prime Minister of New Zealand (no, I don't know his name) was the key note speaker for commencement. He gave a very good talk on protecting the environment, wrapped in a funny story about a quirky bird from New Zealand that mates only once every seven years.
The funniest part about this is that when our degrees were conferred by the Chancellor, the university released thousands of balloons up into the air.
As the proud owner of a shiny HP 932C (what one might consider a mid-range ink jet printer), and a user of Linux (when I bought the printer I was running Mac OS 9), I can say that HPs supposed support of Linux is dissapointing. The community developed drivers that are sketchy at best give better performance than the official HP drivers. If I want to print a resume, I need a resolution of at least 600x600 dpi. HPs drivers on Source Forge give me 300x300 on plain paper. In my eyes this is nothing more than a marketing ploy to trick people into buying their technologically superior (I like HP printers much more than I do Epson or Lexmark), but Linux inferior products. This has an impact on my buying decisions. I won't buy HP products any more, and will refuse to until they stand behind their marketing and support my platform of choice. I just spent $150 on an Epson product specifically because of this.
What hardware are you running? I've been running LinuxPPC, Yellow Dog Linux, and SuSE Linux PPC on various boxes over the last year and a half. Linux PPC ran great on my PowerPC 8600 (or some piece of shit like that), better than OS 9. Yellow Dog 1.2 screamed on my iBook (which I later sold for a nice profit). My Blue and White G3 with 400mb of memory (memory is cheap!) runs SuSE 7.1 much faster than in runs OS X. Linux of PPC hardware is getting pretty good. OS X, well, blows.
Shit Phil! I remember reading you site a couple of years ago and seeing you claim how awesome and bulletproof your site (photo.net) was. Oh, but then again, narcissists aren't always right about themselves (or their work).
I like the current C++ standard. It is very expressive and powerful. Just look at the amazing things that have been developed using templates, like the Blitz++ library . Having said that, it's very frustrating to have so many compilers (read, none that I know of) that support the entire standard. By doing this, Stroutrup et. al. are simple going to create a new divide for vendors (including the FSF ) to play catchup with.
Their goals of avoiding stagnation and fossilisation are admirable, but as an active developer (and student) of C++, I find this to be counterproductive. Thinking about new constructs and definitions is always a good goal; redefining the standard that can hardly be considered standard if divisive and stupid.
Slashdot Mirror
as well as KINK - KINK Is Not KIllustrator.
To get back on topic, I use Ruby too. I've built parsers to analyze code and automate some very hairy analysis tasks. We've written patents and sold projects with Ruby code. I think that Python would next best language that I could have done that work with, but I felt that Ruby gave me much more freedom.
Like the illuminati, or the masons, or the stonecutters. . .
Hi Shawn, I am very impressed with how hard you are working towards commercial Linux applications while still supporting the ideas of free software. I've noticed you on several mailing lists, and you seem to offer advice as much as you seek it. How do you find the time to run a company, develop software, and participate in the comminity?
The thing I find amusing about this is that languages like Lisp provide this exact functionality without the truly cryptic syntax (although you could argue that parentheses are replaced by angle brackets).
It also reminds me of the first time that a computer game creeped me out. In the final room where Hitler is marching back and forth in front of his officers, when everyone yelled 'Heil!' for the first time I nearly jumped out of my skin. Amazingly enough, this was done with the crappy Apple ][ speaker.
I could never get in to the whole 3D FPS thing.
I whip up my own Linux distribution. I'll call it Lap-Linux. Everything that comes on the CD is licsnsed under the GPL, except for one item: the installer. I provide all the sources, for the GPL'ed items on the CD, and they are accessible without installing the system. Now my installer has the alleged behavior of Windows XP. It scans the system, calls home, and allows one install on one machine. My licensing reflects this. If I built my installer from scratch, using only my code, I have the right to do this.
I haven't violated the GPL, but I impose a one license, one seat restriction with respect to the installer. I haven't taken away your GPL guaranteed freedoms; the GPLed source is there for the taking. You can do whatever you want with it. You just can't touch my installer.
That is the simplified view of what Caldera is doing. Whether it helps or hurts their business, they have a right to do that.
1) They want to charge liscensing fees for their software. They are putting a whole bunch of code they they developed onto a CD and selling it under some sort of single desktop license. Good for them. Why?
2) They want to make money. One way to make money off of software is by selling licenses. Nearly every player in the software world does this. There is nothing wrong with it. If you don't want to buy it then don't buy it. There is no need to get your panties in a twirl. But you say. . .
3) What about the GPL? Well, what about it? Nothing Caldera is doing violates the GPL (at least, I can't prove it). Is their proprietary stuff derivatie of GPL work? Probably not. Are they restricting your freedoms as guaranteed by the GPL? On GPL'd software, probably not. Remember, they only need to make the source available for you, as well as any changes they have made to it. If they do that, they have lived up to their end of the bargain. You can take the source off their CD, repackage it, roll around in printouts of it naked, so on and so forth.
This certainly isn't new. Take a look at the SuSE license. You can make copies of SuSE's proprietary code and give it to all of your friends. The moment you try to sell copied ISO's, SuSE legal will come down on you like the hammer of god (ok, maybe not that hard). They are out there to make money. They produce value added software and sell it (along with service). It's common, and not anything to get excited about.
The great thing about the I86 computing world is that you have a choice. You can choose Red Hat, Debian, Caldera, FreeBSD, Windows 2000. . .anything that you want! If you like Caldera (I tried it, didn't like it), good for you. Use it. If you don't like their politics, good for you. Don't use it.
Having said that, I'm going to relive my days as a DNA mapping software author and download this program.
I never would have used Fortran if my academic advisor (and numerical analysis teacher) hadn't forced me to. C and C++ were banned from his course. He claimed that students spent too much time worrying about semi-colons, overloaded operators, and managing memory than writing good software that explored the ideas he taught in his class. This just goes to show, you should use the right tools for the job.
If you're writing the newest and best desktop, use C++, or Java, or something else well suited for the job. If you want to integrate complex fluid systems or model intricate mechanical systems, grab yourself a Fortran compiler and go to town.
One of my biggest complaints about Linux is that there isn't a modern open source Fortran compiler for it.
I just tried to download KAI++ and install it on my SuSE box. Amazingly enough, Intel fucked themselved (and their customers) by releasing versions of their compiler that were dependent upon the RedHat 6.2 version of egcs, and the RedHat 7.1 version of GCC (2.9unstable). GCC 3.0 will come out, all of the distros will get up to speed, and I'll be able to use an optimizing compiler to help my fast code run faster. Slightly off topic, but not really.
However, if you want to teach them ideas, and techniques, and how to think critically about computing problems, languages like Lisp and Scheme are the way to go. Impractical? Most students feel that way. After all, how many companies out there want people who can write in Lisp? It pays off in the long run, though. Lisp and Scheme have minimal syntax, but allow you to express complex ideas. You can spend more time on teaching computer science fundamentals and less time on syntax. This pays off when then students learn C++, or Java, or Perl. They are capable of learning the syntax in the context of "doing it right," rather than just in the context of "getting the damn thing to compile."
Now, to further complicate the issue. C should be taught before C++/Java. Learn syntax, structure building, and clean programming practices. Then teach C++/Java in the context of OO Design and Design Patterns. After all, those areas are where C++/Java shine, so the languages should be taught in that context to prevent the formation of bad habits (more so in C++ than in Java).
The main problem with my philosopy is that it takes planning and discipline to implement it. Most people are too short sighted to handle it. I was lucky. For the most part I took the path described above. It lead me to a lead developer position in a research and development firm. I feel like my work is (mostly) effortless, yet I constantly amaze people with what I produce.
Two cents; take it or leave it.
Never had to do that with Linux. Nope. Never.
To stay somewhat on topic, I'm running OS 9, OS X, and SuSE 7.1 on my Blue and White G3. SuSE with Windowmaker and Mac on Linux blows OS X away (except in the "ohhhh, pretty" contest").
What we really need is a secure television remote control. My old room mate once bought a watch that doubled as a TV remote. He programmed it to work with the house TV then proceeded to mess with everyone in the house. Imagine what marketers, or the FBI, or some haX0R could do to the televisions of the world. When will we wake up and see the real threats that face us?
Welcome to DARPA, where Research is our middle name.
Try "The Chronicles of Thomas Covenant, The Unbeliever." This is a great fantasy series. The protaginist has Leprosy, and is somehow summoned to a world where magic works and his disease can be cured. However, he doesn't believe that the world is real, and so through the entire series he acts like a bastard. After the first 50 pages of this series you find it really hard to like Covenant, but you keep reading to see what happens. It ain't Tolkein, but it is good.
Ironically enough, that same year the former Prime Minister of New Zealand (no, I don't know his name) was the key note speaker for commencement. He gave a very good talk on protecting the environment, wrapped in a funny story about a quirky bird from New Zealand that mates only once every seven years.
The funniest part about this is that when our degrees were conferred by the Chancellor, the university released thousands of balloons up into the air.
The title of this interview reminds me of The Onion series, "Ask A," in which a gut shot cop and Navy Seal answer reader questions. Should we expect the same sort of thing from "Ask a Lawyer" on Slashdot?
What hardware are you running? I've been running LinuxPPC, Yellow Dog Linux, and SuSE Linux PPC on various boxes over the last year and a half. Linux PPC ran great on my PowerPC 8600 (or some piece of shit like that), better than OS 9. Yellow Dog 1.2 screamed on my iBook (which I later sold for a nice profit). My Blue and White G3 with 400mb of memory (memory is cheap!) runs SuSE 7.1 much faster than in runs OS X. Linux of PPC hardware is getting pretty good. OS X, well, blows.
Kinda like the "You can't put the crap back in the dog" law.
Actually, my dog eats her own crap all the time.
Good luck with your troubles.
Their goals of avoiding stagnation and fossilisation are admirable, but as an active developer (and student) of C++, I find this to be counterproductive. Thinking about new constructs and definitions is always a good goal; redefining the standard that can hardly be considered standard if divisive and stupid.
Just my $0.02.