Read my comments to the other person who replied about the tasks that I do.
And, the only instances where I've found GCC to yield code of equal speed, or just narrowly faster is in programs that use little FP math, and few arrays. But that's for the kind of code I work with. Could be different when working with less math-heavy code(And I _did_ state that my findings were for FP-heavy code)
As for it being the most bug-tested and solid, I can't agree. For example, 3.3 has choked on me when using libraries such as BLAS and ATLAS.
As for it being "Free", I don't really care, since it doesn't work well enough for me to be happy with it.
Hmmmm, what error message do you get with ICC 7.1? Version 3.40 compiles without problem on my RH8.0 box and ICC 7.1. It also compiles without problems when using Metrowerk's compiler.
Yes, I've tested the 3.3 version of GCC lately, compared to Metrowerks compiler in CodeWarrior Pro 8, ICC 7.1 and VS.Net's compiler, and for x86 my target platform is a dual Xeon 1.9GHz with 512kb cache each, and 2GB RAM. For Mac, I've only used GCC 3.3 and Metrowerk's compiler, with a G4 800MHz and 512MB RAM as the target platform. For MIPS, I've used GCC 3.3, Metrowerks latest, as well as MIPSPro 7.4, on an Origin 300 with 2 R14k@600MHz, 2MB L2 cache each, and 4GB RAM.
The stuff I do is airflow simulation in buildings, some CFD calculations, particle simulation and animation rendering for architects(often with radiosity) with some custom tools etc as a consultant.
Most often, the difference between the fastest and the slowest is measured in about an hour over a week-long run, but over a 2-week long run, it can be measured in days, depending on how complex it is, and how well one can use vector instructions and similar. The hardest part is actually finding good libraries that give you all the stuff you need in an optimized way.
That would be a first. Everyone I know who write FP-heavy code other than for playing around or proof-of-concept type of apps tries to avoid GCC due to it's low performance, just like I do. On x86, Microsoft's, Intel's and Metrowerk's compilers all outperform it(Don't know about Inprise's compiler, haven't used it in a long while). On Mac's, Metrowerks(The only one I've used on Mac's) compiler outperforms GCC. On MIPS, it's MIPSPro or Metrowerks that counts, GCC is right out, the performance so abyssmal that one sometimes thinks it's interpreted code, rather than compiled.
And GCC's focus is not on getting maximum performance, but to be an Open Source compiler. Lots of target platforms, too many cooks involved in the soup, so performance will never be optimal, but you will find it on many platforms instead.
With 2D, for basic 24bit, no alpha, it's basic requirement is image file size X 3, and then add overhead for effects such as filters etc. For 3D graphics.....Depends on the complexity of the scene. I've worked with scenes that took less than 4MB RAM to render, and I've worked with scenes were 2GB is not enough RAM, so it has to start swapping, and that's not fun. Of course, I had 42 light sources, reflections, particles(with reflection), diffraction and refraction, around 4M polygons after the NURBS were tesselated, as well as some 15-layer shaders...
Actually, 3D graphics, video editing, image editing, compositing, encoding, and some sound apps all do use lots of FP maths. And those areas are very much inside Apple's domain, and not very isolated, like you make them out to be.
This whole suggestion is ridiculous, and distasteful.
Let's say that we have OpenGL. It's an open standard. If this suggestion was implemented, anyone who develops their own API and libraries, which perform better or give better image quality for their specific application would be punished, just because that individual or organization is then trying to push a solution that does not conform to the open standard.
If Open Source claims that it competes through all the programmers contributing, and their bug-checking, why should we then push for a law that banishes competition and innovation?
Innovation is creating something new, that hasn't been done before. Not imitation, and all conforming to the same thing. Innovators have always been those who have broken away from the rest of the people. But instead, the more I see of the Open Source community, the more I see a community, stagnant in it's way and beliefs, and intent on gettings it's way, and preventing other choices. In effect forming a sort of monopoly, but in this case an ideological one, which goes against freedom of choice.
Please, compete through promoting the open standards, instead of working for a ban against innovation, because, for many solutions, not conforming to a standard can give much better results. The only thing an open standard helps with is ease of programming, and ease of immitating... And I'd also like it if the open source community would start to try and innovate, instead of only immitating what others have already done before. Browsing through SourceForge and Freshmeat, I've never seen anything inventive, only a different kind of implementation, or a direct immitation.
Personally, I use what works best for me, and what I need to do. I use Apache, Dreamweaver, PHP, Maya, CodeWarrior for programming C++ and Java(No, pundits, GCC is nowhere near CodeWarrior in speed, on any platform it supports. Especially not MIPS).
Well, 625 DV is all fine and dandy for amateur video and editing, but if you edit compressed your material for anything above amateur-level, except for news broadcasts.If you do post-production or anything on compressed material, then you're the muppet.
Compression is the LAST thing you do with the material, for digital media and archiving.
"as the macrokernel system and crusty old ASCII-pipe-based GNU tools would have to be remade."
Don't you mean the crusty old Crack-pipe-based GNU tools?;)
And yes, I agree. Perpetuating UNIX doesn't really help in developing anything new.
Unfortunately, the OSS movement in general, which proclaims to want to see development and evolution happen, is actually helping to maintain the status quo by wanting to make UNIX-like systems the be-all and end-all of computing.
Yes, SGI's are nice when you work with huge scenes or datasets, especially when you work with cad and similar, and absolutely need that 8x8 supersampled antialiasing. Unfortunately, they cost a little too much.
Compared to the GeForce, the FireGL X11 will do quite well. And, well, games are also rendering, it's just more optimized for fillrates, and also DirectGFX, while the real pro-cards are aimed at OpenGL(For example, the Wildcat 4 series supports DX7, and the Wildcat VP's support DX8.1 and pixel shaders 1.1 and 1.2)
Against the Wildcats? Depends on what models... Against the VP's... Well, they would still be slower, especially since ATI focuses on DX, and their OpenGL-drivers are, well, crap. One example: I still haven't got the entire SpecViewPerf to run all tests without crashing on a computer with an ATI card. The so highly vaunted price/performance(On/. at least) is actually in favour of 3DLabs cards, if you do pro-work. If you're an amateur, who also do a lot of gaming, and want DX9 support, go with the ATI. If you as an amateur want good OpenGL-support, and want to play games, buy the nVidia card.
And, neither ATI nor Nvidia has anything that compares to the Wildcat 4's, other than in marketing drivel and clueless peoples imaginations.
Actually, movies don't require full screen raytracing. Most special effects are STILL rendered with raycasting, and if raytraced effects are thought to be needed, you use a special rayserver, as it's called, to render only the shaders that have the raytracing flag set.
If you are really interested in pro work, 3DLabs offerings are miles ahead of both ATI's and nVidia's offerings, and the Wildcat VP870 is even cheaper than the GF Quadro4 900XGL, and at the high-end, the Wildcat IV series is quite good, the top-end model having 128MB Frame Buffer and 256MB texture buffer. Also, the image quality is great.
I've tested the cards under Maya and SoftImage XSI, as well as SolidEngineer, and all I can say is that the Wildcat VP870 is definitely worth it's price if you're interested in pro-level graphics, and don't care about games, and the Wildcat IV 7210 is a wet dream. Wish I could afford the latter one though =(.
Sounds like a clear example of not bothering to learn how to work with efficiently, and then judging it against a system that you have learned to work with efficiently. I've got working knowledge of both Unix permissions and Windows ACL's(As well as some of the ACL stuff in Trusted IRIX), and I have to say that I prefer the Windows and IRIX ACL stuff over the traditional Unix permissions system.
Uhm, it's the compiler developers task to make sure that their compiler supports the architecture, not the other way around. And, GCC is absolute crap on MIPS, even the latest version.
"We made a benchmark comparing an SGI Origin and a linux Ahtlon cluster, the athlon needed only two nodes to beat the origin and with all 16 nodes where about 10 times faster... SGIs are just overpriced, for 99.999% (that's 5 nines) PCs can do the job and even do it better and especially do it much cheaper."
Then you're running small tasks that require little memory, little I/O and don't use much cache, and a substandard compiler. I've got a particle simulation going right now, the Origin 300 with 2 R14000A@600Mhz and 2MB L2 cache and 4GB RAM, using MipsPro compiler, that I have access to outperforms the dual Xeon 1.9GHz with 512kB L2 cache using both VS and Intel's own compiler. The difference in time is measured in days. It's the same thing with a cluster of athlons(And if you run a task where the task isn't easily parallellized, and need to keep in synch with the others, a node crash might ruin a lot of work and force you to start over)
Uh, yes. Bredbandsbolaget here in Sweden, and Bostream also. The problem is that they only cover relatively small areas, and Bredbandsbolaget can only be had in large apartment complexes etc.
No they didn't. In fact, the US Army set up a base at Dien Bien Phu, the US commander reasoning that the Vietnamese couldn't get artillery onto the mountains surrounding it, just like the french did before him.... Different army, even more stupid, since it didn't bother to learn from previous events.
Slashdot Mirror
Read my comments to the other person who replied about the tasks that I do.
And, the only instances where I've found GCC to yield code of equal speed, or just narrowly faster is in programs that use little FP math, and few arrays. But that's for the kind of code I work with. Could be different when working with less math-heavy code(And I _did_ state that my findings were for FP-heavy code)
As for it being the most bug-tested and solid, I can't agree. For example, 3.3 has choked on me when using libraries such as BLAS and ATLAS.
As for it being "Free", I don't really care, since it doesn't work well enough for me to be happy with it.
Hmmmm, what error message do you get with ICC 7.1?
Version 3.40 compiles without problem on my RH8.0 box and ICC 7.1. It also compiles without problems when using Metrowerk's compiler.
Yes, I've tested the 3.3 version of GCC lately, compared to Metrowerks compiler in CodeWarrior Pro 8, ICC 7.1 and VS .Net's compiler, and for x86 my target platform is a dual Xeon 1.9GHz with 512kb cache each, and 2GB RAM. For Mac, I've only used GCC 3.3 and Metrowerk's compiler, with a G4 800MHz and 512MB RAM as the target platform. For MIPS, I've used GCC 3.3, Metrowerks latest, as well as MIPSPro 7.4, on an Origin 300 with 2 R14k@600MHz, 2MB L2 cache each, and 4GB RAM.
The stuff I do is airflow simulation in buildings, some CFD calculations, particle simulation and animation rendering for architects(often with radiosity) with some custom tools etc as a consultant.
Most often, the difference between the fastest and the slowest is measured in about an hour over a week-long run, but over a 2-week long run, it can be measured in days, depending on how complex it is, and how well one can use vector instructions and similar. The hardest part is actually finding good libraries that give you all the stuff you need in an optimized way.
That would be a first. Everyone I know who write FP-heavy code other than for playing around or proof-of-concept type of apps tries to avoid GCC due to it's low performance, just like I do. On x86, Microsoft's, Intel's and Metrowerk's compilers all outperform it(Don't know about Inprise's compiler, haven't used it in a long while). On Mac's, Metrowerks(The only one I've used on Mac's) compiler outperforms GCC. On MIPS, it's MIPSPro or Metrowerks that counts, GCC is right out, the performance so abyssmal that one sometimes thinks it's interpreted code, rather than compiled.
And GCC's focus is not on getting maximum performance, but to be an Open Source compiler. Lots of target platforms, too many cooks involved in the soup, so performance will never be optimal, but you will find it on many platforms instead.
With 2D, for basic 24bit, no alpha, it's basic requirement is image file size X 3, and then add overhead for effects such as filters etc. For 3D graphics.....Depends on the complexity of the scene. I've worked with scenes that took less than 4MB RAM to render, and I've worked with scenes were 2GB is not enough RAM, so it has to start swapping, and that's not fun. Of course, I had 42 light sources, reflections, particles(with reflection), diffraction and refraction, around 4M polygons after the NURBS were tesselated, as well as some 15-layer shaders...
Actually, 3D graphics, video editing, image editing, compositing, encoding, and some sound apps all do use lots of FP maths. And those areas are very much inside Apple's domain, and not very isolated, like you make them out to be.
This whole suggestion is ridiculous, and distasteful.
Let's say that we have OpenGL. It's an open standard. If this suggestion was implemented, anyone who develops their own API and libraries, which perform better or give better image quality for their specific application would be punished, just because that individual or organization is then trying to push a solution that does not conform to the open standard.
If Open Source claims that it competes through all the programmers contributing, and their bug-checking, why should we then push for a law that banishes competition and innovation?
Innovation is creating something new, that hasn't been done before. Not imitation, and all conforming to the same thing. Innovators have always been those who have broken away from the rest of the people. But instead, the more I see of the Open Source community, the more I see a community, stagnant in it's way and beliefs, and intent on gettings it's way, and preventing other choices. In effect forming a sort of monopoly, but in this case an ideological one, which goes against freedom of choice.
Please, compete through promoting the open standards, instead of working for a ban against innovation, because, for many solutions, not conforming to a standard can give much better results. The only thing an open standard helps with is ease of programming, and ease of immitating... And I'd also like it if the open source community would start to try and innovate, instead of only immitating what others have already done before. Browsing through SourceForge and Freshmeat, I've never seen anything inventive, only a different kind of implementation, or a direct immitation.
Personally, I use what works best for me, and what I need to do. I use Apache, Dreamweaver, PHP, Maya, CodeWarrior for programming C++ and Java(No, pundits, GCC is nowhere near CodeWarrior in speed, on any platform it supports. Especially not MIPS).
In bits and bytes, I guess......
Well, 625 DV is all fine and dandy for amateur video and editing, but if you edit compressed your material for anything above amateur-level, except for news broadcasts.If you do post-production or anything on compressed material, then you're the muppet.
Compression is the LAST thing you do with the material, for digital media and archiving.
"as the macrokernel system and crusty old ASCII-pipe-based GNU tools would have to be remade."
;)
Don't you mean the crusty old Crack-pipe-based GNU tools?
And yes, I agree. Perpetuating UNIX doesn't really help in developing anything new.
Unfortunately, the OSS movement in general, which proclaims to want to see development and evolution happen, is actually helping to maintain the status quo by wanting to make UNIX-like systems the be-all and end-all of computing.
Ooops, add a comma after 22 in the last sum...
Your figures are very off....
That stream would require about 22.118MB/s, which is too much for even a 100Mb/s network.
Number of pixels:
640*480=307200
At 24bit, it's 3 Bytes/pixel
307200*3=921600 Bytes for one image
Multiply that by 24 images to get the throughput in MB/s
921600*24=22118400MB/s
Well, there are other places, that don't just share crap such as the D20 system....
Definitely. Many books can be found in PDF format if you know where to look...
Yes, SGI's are nice when you work with huge scenes or datasets, especially when you work with cad and similar, and absolutely need that 8x8 supersampled antialiasing. Unfortunately, they cost a little too much.
/. at least) is actually in favour of 3DLabs cards, if you do pro-work. If you're an amateur, who also do a lot of gaming, and want DX9 support, go with the ATI. If you as an amateur want good OpenGL-support, and want to play games, buy the nVidia card.
Compared to the GeForce, the FireGL X11 will do quite well. And, well, games are also rendering, it's just more optimized for fillrates, and also DirectGFX, while the real pro-cards are aimed at OpenGL(For example, the Wildcat 4 series supports DX7, and the Wildcat VP's support DX8.1 and pixel shaders 1.1 and 1.2)
Against the Wildcats? Depends on what models... Against the VP's... Well, they would still be slower, especially since ATI focuses on DX, and their OpenGL-drivers are, well, crap. One example: I still haven't got the entire SpecViewPerf to run all tests without crashing on a computer with an ATI card. The so highly vaunted price/performance(On
And, neither ATI nor Nvidia has anything that compares to the Wildcat 4's, other than in marketing drivel and clueless peoples imaginations.
Whoops... Haven't noticed that 3DLabs has released the VP970 also...
Gotta get my hands on one of those to test it.
Actually, movies don't require full screen raytracing. Most special effects are STILL rendered with raycasting, and if raytraced effects are thought to be needed, you use a special rayserver, as it's called, to render only the shaders that have the raytracing flag set.
If you are really interested in pro work, 3DLabs offerings are miles ahead of both ATI's and nVidia's offerings, and the Wildcat VP870 is even cheaper than the GF Quadro4 900XGL, and at the high-end, the Wildcat IV series is quite good, the top-end model having 128MB Frame Buffer and 256MB texture buffer. Also, the image quality is great.
I've tested the cards under Maya and SoftImage XSI, as well as SolidEngineer, and all I can say is that the Wildcat VP870 is definitely worth it's price if you're interested in pro-level graphics, and don't care about games, and the Wildcat IV 7210 is a wet dream. Wish I could afford the latter one though =(.
Oooops, I guess I've always misunderstood the point about Sun and their Big Iron then....
Sounds like a clear example of not bothering to learn how to work with efficiently, and then judging it against a system that you have learned to work with efficiently. I've got working knowledge of both Unix permissions and Windows ACL's(As well as some of the ACL stuff in Trusted IRIX), and I have to say that I prefer the Windows and IRIX ACL stuff over the traditional Unix permissions system.
"Decent gcc-3.x support"
Uhm, it's the compiler developers task to make sure that their compiler supports the architecture, not the other way around. And, GCC is absolute crap on MIPS, even the latest version.
"We made a benchmark comparing an SGI Origin and a linux Ahtlon cluster, the athlon needed only two nodes to beat the origin and with all 16 nodes where about 10 times faster... SGIs are just overpriced, for 99.999% (that's 5 nines) PCs can do the job and even do it better and especially do it much cheaper."
Then you're running small tasks that require little memory, little I/O and don't use much cache, and a substandard compiler. I've got a particle simulation going right now, the Origin 300 with 2 R14000A@600Mhz and 2MB L2 cache and 4GB RAM, using MipsPro compiler, that I have access to outperforms the dual Xeon 1.9GHz with 512kB L2 cache using both VS and Intel's own compiler. The difference in time is measured in days. It's the same thing with a cluster of athlons(And if you run a task where the task isn't easily parallellized, and need to keep in synch with the others, a node crash might ruin a lot of work and force you to start over)
Uh, yes. Bredbandsbolaget here in Sweden, and Bostream also. The problem is that they only cover relatively small areas, and Bredbandsbolaget can only be had in large apartment complexes etc.
Not really. Tritium is applied to the iron sights, so that they are visible in the dark. Night vision scopes are electronic in nature
No they didn't. In fact, the US Army set up a base at Dien Bien Phu, the US commander reasoning that the Vietnamese couldn't get artillery onto the mountains surrounding it, just like the french did before him.... Different army, even more stupid, since it didn't bother to learn from previous events.
The Octane was the one using a Crossbar Switch, and it was announced in late 96. The O2 used UMA which is somewhat different.