The SPEC benchmarks are real-world. That's the point of them, and they've been used over the last 10 years to judge the real performance of a processor.
Code size matters because *cache* isn't cheap. Worse, you can't make L1 cache arbitrarily fast without slowing down your chip big time.
This is irrelevant with the trace cache in the Pentium4. Instructions are decoded into micro-ops, by "traces" which are sequences of executed ops, and stored in the cache. Compact x86 CISC instructions are not stored in the trace cache.
Check the latest SPEC CPU benchmarks. The Itanium2 has the fastest floating-point score and is no slouch in the integer tests either. It will improve. Linus will eat his words in a few years.
Google does not evict anything out of their cache. They just keep adding capacity. Hence Google can already see changes to websites. Granted I'm sure that this data isn't durable though.
They have a cache of nearly every internet document. Even 4gb of memory in each of those 15,000 boxes is 60 terabytes. Maybe that's enough, but I doubt it because they _never_ throw away stuff from the cache. The Google engineer actually addressed this issue because they have a lot of "power". For example, they can look at how websites change over time if they wanted to.
Also, they actually use the filesystem for communication between nodes. Messages are written to a log on some type of distributed file system.
Ok. Good point. I guess when I think "embedded", I think either "small" or "real-time". Why didn't your motion control applications just use an industrial PC? If they did, did you consider that an embedded system?
That could be useful in embedded applications. Running the entire app in the cache, at higher speed than in main memory, could be a Very Good Thing.
Sorry dude. Embedded apps usually aren't starved for CPU time. They are often real-time and have to be very predictable. Caches actually make things less predictable because the execution time of a high-priority task can substantially differ depending on the cache hit rate at a given execution.
And any embedded application that needs an Athlon, and can afford to dissipate 64 watts of power, will surely be larger than 640kb in footprint.
In addition, Itanium performance for CPU-bound applications is bad.
Last time I checked the Spec CPU benchmarks, Itanium2 was the leader for floating-point performance. Check them out...they may not be the leader right now but Itanium2 is no slouch.
I recently attended a talk by Google's chief engineer. They have approximately 15,000 x86 machines running Linux at seven data centers in the United States.
Weird failures occur so often, such as disks returning garbage without the controller informing the OS, that Google does a checksum on _every_ data structure in their user-level software. He also talked about how Linux is good enough for them, but it doesn't perform well with respects to I/O under heavy load. He says they like Linux because they have the source-code and that they minimize excessive I/O loads on their machines. Nobody asked why they don't use FreeBSD but I suspect its because Linux has better hardware support and Google builds their own machines with numerous different components based on the latest technology.
Another version of FORTRAN! Yeah! Now if I can just find a card punch machine and reader on eBay, I'll have hutled into the 1970s!
The state of the art supercomputers in this country, the ASCI machines, primarily run Fortran codes. I would say that half of all new scientific applications are still written in Fortran. Your post isn't funny, its ignorant.
Moron? Do you really think that Compaq stocks and sends replacement capacitors to customers? For someone as pompous as yourself, why are you using a Compaq?
Compaq tech support is designed to help people like my mother who struggles to figure out how the power button works.
Read my post again. We have people license our code who are not academic. We can't give them code, that uses QT, because it would then violate an academic license.
Why? GTK libraries are LGPL. Before I get flamed, listen to this story:
I am a researcher at a large university. We develop software for numerous entities including national labs, industry, and acadamia. About a year ago, we wanted to start a new tool with a better user interface (we've been using Tcl/Tk for other projects). We started developing with QT but quickly had to abandon after our supervisor pointed out the GPL issue. You see, we can't GPL our code even though we nearly always open-source it under our own license. QT libraries our either GPL or you buy a license. Our funding sources prevent the use of GPL libraries because we often sign agreements with them such that they can use our code in commercial products. We have the money to purchase QT licenses, but dealing with the budgets (bureaucracy) and licences with other organizations is just to complex and unwanted by most involved (except for some developers).
This is the exact same reason why Gnome has more commercial support (at least in the U.S.) than KDE. Sad but true.
Believe it. Maybe because its a Dell machine. My home-built Athlon machine also runs Win2k but it does indeed crash occasionally (once a month or so). My Dell is a P3 866 MHz. It is used heavily for games, photoshop work, and standard office-type stuff. In my 2 years of having it, it seriously has never ever crashed. Applications yes have crashed. The OS has never froze or blue-screened and gets rebooted only a few times a year (vacations)
I've run Linux, FreeBSD, NetBSD, Solaris and HP-UX. I've seen a Linux kernel panic and a HP-UX panic. I've never seen Solaris or a BSD crash. I'm sure it depends on how hard you stress a machine though.
Windows XX - For the clueless masses, and often a neccassary evil (esp. games)
Or for those who use a computer and don't tinker with them.
These type of comments make you guys look clueless to ordinary people.
I'm not clueless and can handle Linux better than most. I've developed Linux device drivers, built my Linux system from scratch before (ugh), and am a seasoned Unix systems developer in industry. I use Windows 2000 and I like it. Fast, never-ever crashed before, and works with all of my hardware.
Then if the system is compromised somehow, any debugger/program/back-orifice can attach to Opera and read its memory thus obtaining access to your "RAM-only" cached contents.
What really irks me is that the French people can't stand Americans even though we bailed them out of both WWI and WWII. Yeah, 60 years ago..right. Tell that to my relatives who were killed in France.
This guy is a freaking idiot. Who the hell modded him up to 4 for this line of BS. Verizon is not cellular other than in the sense that every OTHER carrier is.
Sorry dude. Your the idiot if you don't understand that the term "cellular" has multiple meanings in the context of mobile phones. Cellular can refer to the concept of using frequency reuse or in other words, using "cells" of channels. Or it can refer to the 800MHz spectrum used by AMPS, CDMA, and maybe others I'm not aware of. PCS is commonly referred to as the 1900MHz spectrum used by Sprint (CDMA) and Cingular (GSM).
I'll admit that my comment regarding tower placement is questionable. However tower placement does depend on the frequencies being used. The original cell phone service in the U.S. was all AMPS running in the 800MHz Cellular frequencies. When they decided to use this frequency band for CDMA, they likely just upgraded the equipment in the same base station.
Its called "subsidy lock". Your Sprint PCS phone is fully capable of using a Verizon network (assuming it is tri-band). This is programmed into the software by the phone manufacturer and is why "unlocked" phones are highly sought after.
Get ahold of an unlocked phone and switch between Verizon and Sprint freely. They can be had on eBay. If you own your own phone, you don't have to sign a yearly contract. This is what I do with Verizon. They don't advertise this, but ask them.
Also, Verizon doesn't use subsidy locks and will activate nearly any CDMA phone on their network.
But they might be able to cause the antenna in a GSM phone to resonate causing measurable disturbance.
The SPEC benchmarks are real-world. That's the point of them, and they've been used over the last 10 years to judge the real performance of a processor.
Code size matters because *cache* isn't cheap. Worse, you can't make L1 cache arbitrarily fast without slowing down your chip big time.
This is irrelevant with the trace cache in the Pentium4. Instructions are decoded into micro-ops, by "traces" which are sequences of executed ops, and stored in the cache. Compact x86 CISC instructions are not stored in the trace cache.
Check the latest SPEC CPU benchmarks. The Itanium2 has the fastest floating-point score and is no slouch in the integer tests either. It will improve. Linus will eat his words in a few years.
Google does not evict anything out of their cache. They just keep adding capacity. Hence Google can already see changes to websites. Granted I'm sure that this data isn't durable though.
They have a cache of nearly every internet document. Even 4gb of memory in each of those 15,000 boxes is 60 terabytes. Maybe that's enough, but I doubt it because they _never_ throw away stuff from the cache. The Google engineer actually addressed this issue because they have a lot of "power". For example, they can look at how websites change over time if they wanted to.
Also, they actually use the filesystem for communication between nodes. Messages are written to a log on some type of distributed file system.
It's too bad Google doesn't have one of those things where you can watch everyone's search scrolling down the screen live.
No, but you can have Google scroll the results of your search!! Try this, and other "Google Experiments" at http://labs.google.com
Ok. Good point. I guess when I think "embedded", I think either "small" or "real-time". Why didn't your motion control applications just use an industrial PC? If they did, did you consider that an embedded system?
That could be useful in embedded applications. Running the entire app in the cache, at higher speed than in main memory, could be a Very Good Thing.
Sorry dude. Embedded apps usually aren't starved for CPU time. They are often real-time and have to be very predictable. Caches actually make things less predictable because the execution time of a high-priority task can substantially differ depending on the cache hit rate at a given execution.
And any embedded application that needs an Athlon, and can afford to dissipate 64 watts of power, will surely be larger than 640kb in footprint.
In addition, Itanium performance for CPU-bound applications is bad.
Last time I checked the Spec CPU benchmarks, Itanium2 was the leader for floating-point performance. Check them out...they may not be the leader right now but Itanium2 is no slouch.
I recently attended a talk by Google's chief engineer. They have approximately 15,000 x86 machines running Linux at seven data centers in the United States.
Weird failures occur so often, such as disks returning garbage without the controller informing the OS, that Google does a checksum on _every_ data structure in their user-level software. He also talked about how Linux is good enough for them, but it doesn't perform well with respects to I/O under heavy load. He says they like Linux because they have the source-code and that they minimize excessive I/O loads on their machines. Nobody asked why they don't use FreeBSD but I suspect its because Linux has better hardware support and Google builds their own machines with numerous different components based on the latest technology.
Another version of FORTRAN! Yeah! Now if I can just find a card punch machine and reader on eBay, I'll have hutled into the 1970s!
The state of the art supercomputers in this country, the ASCI machines, primarily run Fortran codes. I would say that half of all new scientific applications are still written in Fortran. Your post isn't funny, its ignorant.
Lets not forget that Linux can't be considered Unix. It is a Unix clone. It is not Unix.
Moron? Do you really think that Compaq stocks and sends replacement capacitors to customers? For someone as pompous as yourself, why are you using a Compaq?
Compaq tech support is designed to help people like my mother who struggles to figure out how the power button works.
Read my post again. We have people license our code who are not academic. We can't give them code, that uses QT, because it would then violate an academic license.
Why? GTK libraries are LGPL. Before I get flamed, listen to this story:
I am a researcher at a large university. We develop software for numerous entities including national labs, industry, and acadamia. About a year ago, we wanted to start a new tool with a better user interface (we've been using Tcl/Tk for other projects). We started developing with QT but quickly had to abandon after our supervisor pointed out the GPL issue. You see, we can't GPL our code even though we nearly always open-source it under our own license. QT libraries our either GPL or you buy a license. Our funding sources prevent the use of GPL libraries because we often sign agreements with them such that they can use our code in commercial products. We have the money to purchase QT licenses, but dealing with the budgets (bureaucracy) and licences with other organizations is just to complex and unwanted by most involved (except for some developers).
This is the exact same reason why Gnome has more commercial support (at least in the U.S.) than KDE. Sad but true.
Believe it. Maybe because its a Dell machine. My home-built Athlon machine also runs Win2k but it does indeed crash occasionally (once a month or so). My Dell is a P3 866 MHz. It is used heavily for games, photoshop work, and standard office-type stuff. In my 2 years of having it, it seriously has never ever crashed. Applications yes have crashed. The OS has never froze or blue-screened and gets rebooted only a few times a year (vacations)
I've run Linux, FreeBSD, NetBSD, Solaris and HP-UX. I've seen a Linux kernel panic and a HP-UX panic. I've never seen Solaris or a BSD crash. I'm sure it depends on how hard you stress a machine though.
Windows XX - For the clueless masses, and often a neccassary evil (esp. games)
Or for those who use a computer and don't tinker with them.
These type of comments make you guys look clueless to ordinary people.
I'm not clueless and can handle Linux better than most. I've developed Linux device drivers, built my Linux system from scratch before (ugh), and am a seasoned Unix systems developer in industry. I use Windows 2000 and I like it. Fast, never-ever crashed before, and works with all of my hardware.
Then if the system is compromised somehow, any debugger/program/back-orifice can attach to Opera and read its memory thus obtaining access to your "RAM-only" cached contents.
Good luck with recreating the beautiful text rendering and responsiveness of Windows XP.
If so, that explains why they ask for handouts.
What really irks me is that the French people can't stand Americans even though we bailed them out of both WWI and WWII. Yeah, 60 years ago..right. Tell that to my relatives who were killed in France.
Gobe Productive is meant to be a lightweight office suite, correct? Then why the %&^*$ does the Linux beta require Gnome libraries?!
Maybe you should ground your kids for saying "Linux" instead of "GNU/Linux".
This guy is a freaking idiot. Who the hell modded him up to 4 for this line of BS. Verizon is not cellular other than in the sense that every OTHER carrier is.
Sorry dude. Your the idiot if you don't understand that the term "cellular" has multiple meanings in the context of mobile phones. Cellular can refer to the concept of using frequency reuse or in other words, using "cells" of channels. Or it can refer to the 800MHz spectrum used by AMPS, CDMA, and maybe others I'm not aware of. PCS is commonly referred to as the 1900MHz spectrum used by Sprint (CDMA) and Cingular (GSM).
I'll admit that my comment regarding tower placement is questionable. However tower placement does depend on the frequencies being used. The original cell phone service in the U.S. was all AMPS running in the 800MHz Cellular frequencies. When they decided to use this frequency band for CDMA, they likely just upgraded the equipment in the same base station.
Its called "subsidy lock". Your Sprint PCS phone is fully capable of using a Verizon network (assuming it is tri-band). This is programmed into the software by the phone manufacturer and is why "unlocked" phones are highly sought after.
Get ahold of an unlocked phone and switch between Verizon and Sprint freely. They can be had on eBay. If you own your own phone, you don't have to sign a yearly contract. This is what I do with Verizon. They don't advertise this, but ask them.
Also, Verizon doesn't use subsidy locks and will activate nearly any CDMA phone on their network.