0.) For any architecture, an arbitrarily conceived benchmark can prove it optimal.
1.) Only L2 cache is shared.
2.) A modern scheduler wont give any type of guarantee that when one process is on a CPU, a certain process is on another.
3.) Even if it did, as soon as a user process blocks to the kernel for network IO, the kernel will reschedule a differant process to run on the first CPU, with a high probability of wiping much of the cache.
4.) Having two differant processes sharing the same cache can lead to many scenarios where one process is battling the other in terms of overwriting cache lines. L2 on Intel is something on the order of 8-way associative, if I remember correctly, so conflict misses will happen quite often.
Don't read this the wrong way, this is the way of the future, but in architecture, few problems are simple.
I really struggled with this for a long time. My final solution was to buy a regular bookshelf, and then went to Lowes and had even more shelves made for it. I buy jewel cases for every CD worth keeping. If it doesnt rank spending 10 cents on a jewel case, I throw it out. At the same time I did all this, I bought a labeling kit, and printed labels for all of the burned CDs. The labels cause the burned ones to last much longer, as they protect the top side of the disk, which is the much more easily damamged side.
When I was all done, I am able to store between 600 and 700 CDs on the book case. Most of them are music, so I put those in alphabetic order. Software gets sorted by type, and DVDs and games just kindof go wherever. People laugh when they see the bookcase completely full, but I think most people dont realize how much media we have now.
I have rowed at the University level (Purdue in the U.S.) and can honestly say that out of all cardio exercises, I find rowing to be the easiest on the body. When done intensely and incorrectly, it can be hard on the back and knees. This is rare, and more likely a result of bad form.
A rowing machine (most often called an erg), is a lot more than just a piece of exercise equipment. Many work very hard on achieving certain goals, such as 1 million meters, or rowing a marathon. At the university level, we always concentrated on 2k, 5k, 6k and long distances. During the winter, it wouldn't be uncommon to sit down and row 40k in a practice just to build endurance. Others days we would do 12 one minute on, one minute off pieces, and go home and just crash. Because rowing is so low impact, most rowers peak quite a bit older than in other sports. The Men's 2k world record is held by a person in the 30-39 age group. $800 is cheaper than a gym membership over time, and I recommend either a stereo or TV to go with the thing. The YMCA near me has several Concept2 erg's, and I would recommend looking at a couple of sites for some technique information, and try them out.
The great thing about rowing, is that if you enjoy it, pretty much any city with a river will have a club. Here you can meet other people interested, find coaches, and use their boats. In some select cities (Philly, Boston) their is a country club atmosphere to these places. Anywhere else, they are just normal people, like a bike club or gym.
Even though I no longer row competitively, I still use an erg for a good warmup and cool down, no matter what my exercise routine for the day is.
Just finishing up my end of semester projects, had to analyze several CPUs I designed this semester. One very simple one used 1048 logic cells, but could do integer arithmetic, jumps, branches, and memory operations. Not quite sure how many transistors that translates too, but normally an FPGA cell is a binary operation. Removing several of the odd arithmetic operations would lower that size quite a bit more. I also designed a 5 stage pipelined CPU with 32 word instruction cache, and 32 word data cache in about 2300 logic cells.
When the first vacuum tube based computers were invented, I'll built the designers felt like they were implementing a CPU of this size in Legos. It seems funny now, but this analogy probably holds a lot of water.
You most likely can use a PIII. I have an asus 440BX dual board that was designed for PII processors, but I put two 700s in it no problem. Check the FSB and multiplier maximums. Sometimes its a factor of what rev of the board you have, and what version bios is installed. On mine, Asus P2B-D, the owners manual doesnt say it's possilbe, but it's been my desktop for 6 years or so.
A neat trick on 440BX is that when you overclock the FSB to 133 MHz, the pci cards go back to their correct 33 MHz, but the AGP slot doesnt get corrected. This let me get extra CPU speed, high data tranfer to my video card, but didnt lock up due to crappy PCI cards.
The one thing to remember about these systems though is, a dual PIII with PC100 memory is incredibly memory bound. The memory just cant feed the CPU fast enough, and you lose a lot of cycles. On the other hand, the PIII is an incredibly efficient architecture, with only 10 pipeline stages.
At any rate, before buying PIIs, make sure you cant put some newer PIIIs in there.
Also known as HKN is a mix between a co-ed social fraternity and a student organization for Electrical and Computer Engineers. One must pledge to join, must be at least a Junior, and must be above a certain class rank. The organization has been around for over 100 years. My chapter (Beta Chapter at Purdue), was an underground society for a period of time. We run the lounge and snackbar in the EE building at Purdue, perform community service, have get togethers most fridays, have professionals come in and speak to us. It is quite an active organization, found at most top engineering schools. Remember, HKN, "The Nice Guys".
Well, modern cpus are clocked so much faster for two reasons. 1, the transistors are smaller, and therefore faster. 2, modern cpus do less per instruction per clock, due to instructions moving through the cpu's pipeline. While a sinclair might execute an entire instruction in a single clock, a modern Pentium will break that instruction into close to 30 clock cycles, doing a very small portion of the instruction in each pipeline stage. The downside to this, is that when a jump or branch occurs, which is very common, the pipeline must be cleared. Modern chips use branch prediction to try and use which way a branch went the last time it was evaluated, to predict which way it will go in the future. This does add tranistors, which increases power consumption, due to smaller transistor, I thikn the change is negligible. In addition, modern cpus provice floating point capabilities, larger word width, and things like dma controllers, page frame pointers and cache built into hardware. The sinclair probably only had a stack pointer.
It is true that a modern sinclair with smaller transistors would use less power, but so much has been done in this field in the last 25 years, that a modern design could do so much better. Look at something like a motorola hc12 for a modern version of a cpu similar to the sinclais. Even this has a 2 stage pipeline if I recall correctly.
Furthermore, the ISA of old chips is quite limited.
Industry doesnt change designs unless valid. x86 is a great example of this. It is a pretty much messed up design, some instructions are 17 words long, but it can be made to work. Industry moved away from the sinclair because a better solution was found.
No, because the throttle is connected to a spring, that returns it to neutral on one side, and the gas pedal linkage pulls it the opposite direction. If the throttle is stuck, releasing pressure on the pedal linkage isnt going to move it, if it would, then it wouldnt be stuck. The hope is that by pressing the throttle further, and then releasing, the spring has more force trying to make it contract, thus overcoming whatever caused the throttle to stick. In any case, putting the car in neutral (or use the clutch in a manual) will overcome a stuck throttle.
When Syllable was forked from AthOS, a lot of time and effort went into choosing a name. IIRC, a long submission period was conducted, with a voting on all submissions. This narrowed down the pool to a few names, which were voted on again. Syllable was the winner. It is a lot harder than one can imagine to choose a good name that doesnt violate something else's name.
Alright, Ill bite. My experience is in MR, PET and CT imaging. Ultrasound has many of the same concerns listed, but on a smaller level, as it is less computationaly intense (I'm told anyway).
These systems aren't embedded. They have 4 and 8 gigs of main memory, huge disk arrays. 2 or 4 processors in the host alone. Dual monitors,and several attached storage devices. This is just for the console the operator sits at. Some systems will have more than one workstation as part of the console. Connected to this over high speed interfaces can be hundreds of FPGA's coupled to embedded boards running VxWorks, QNX, Nucleos, or just about anything else. Vendors such as Motorola, Mercury, and custom rolled boards are the norm. One board produced by Mercury contains 4 G3 processors and a gig of ram. Systems will have 2 to 8 of these, each running an embedded OS. There are a lot of hardware interconnects, IPC mechanisms, and little pieces of software to break.
Industrial operating systems struggle with these loads. For years, the standard was Solaris or IRIX, but it is quickly becoming Linux, with vendors rolling their own distributions. Windows is used by some, due to programmer availability, ability to run Java, and it is stable when used with tested and validated device drivers. Another advantage of Windows software is that laptops are all over hospitals, so if you write an image viewer application for Windows, suddenly images can be pulled up all over the hospital. Most of the problems that hospitals are having with virus's and what not are from improperly implemented firewalls (and some stupid users)
On the time required to validate: The FDA requires a strict, audited validation. Plans to accomplish this for an entire systems can be thousands of pages long, and take many weeks to complete. On top of this, every configuration must be tested, and there are plenty of differant of options available. If validation is not completed, the manufacturer cannot guarantee the equipment, and is in violation of FDA rules and regulations. It is not like the manufacturers are trying to screw the customers. It is a fiercely competitive business, and every little edge is worth fighting for.
The software that powers these machines has evolved over 15 and 20 years in some cases. It is incredibly complex, and not easy to just all of a sudden say "Oh, I guess we should switch platforms."
Image quality and features are paramount, hospitals are willing to pay for this, because this is what helps to save lives.
I'd also recommend that you don't feed your computer. Computers are _inatimate_objects_, not to be confused with pets that need food and water. I know you might think you'll get an extra MHz or 2, but that food is _really_ unneccessary...
I agree, changing all the clocks is just a pain. The current system has worked well so far, why change? I have yet to find software that can't cope with it, and it's nice to not have to change the VCR, your watch, the clock in your car, the clock in your other car, the wall clock, the microwave, the oven, the alarm clock, the answering machine, and whatever else has a clock these days.
I hear that. I will never buy another ATI card again. I had a Radeon 9000, I ended up trading it for a far less superior NVidia card, purely becuase the drivers didnt suck as bad. I bought the 9000 for its dual screen capability, and yet I have not found one person that has gotten dual screen to work in Linux. I found plenty of instability, and the driver reporting that it supported features that it didnt, so that when programs tried to use them, they crashed. I think this binary module idea is crap. If you think the source code to the driver is gonna give that much insight into how these things work, than why are only two companies actively competing? Furthermore, many open source drivers do exist, and no rampant intellectual property loss has occured.
Yeah, I'd like to believe you, but I've seen people get away with murder in source code before. Open source coders worry a lot more about things like indentation, and filenames that make sense. In closed source shops, a lot of times what is quickly coded as a prototype becomes the shipping product, and things like indent cant be used because it breaks diffs. As much as I'd like to look with my own eyes, this sounds like one of the things it would be best if I just ignored it.
Im glad Purdue is fast to somewhere, cause I live off campus, and it's faster to pull stuff from North Carolina than from csociety, 3 blocks away. Csociety incidently is actually just a dell pc sitting on top of a filing cabinet in the IEEE lounge, in the basement of the EE building. If you plug into the right subnet in EE, its fast.
I know on slashdot, there is always someone who will prove you wrong. Today, I am that guy. I'm 21 years old, live on a college campus with a fat pipe. I pretty much don't remember when we didnt have MP3s. I own between 500-600 cd's, and I feel that it is money well spent:
l: It's not illegal. 2: A hard drive crash doesnt erase my collection. Burned cd's, backups, what have you get scratched, and aren't reliable. My factory made cd's will last much longer. 3: I can legally rip them at a high-bitrate in whatever the common format is. 4: A lot of my collection is indie / small label punk, these bands probably make less than I do, stealing their cds instead of buying really does affect them. 5: The main reason I buy cds is that when I rip them, there are no pops, none of my tracks are cut short, there are no duplicates, and the tags are 100% correct. I can put them in a database, and magically all the songs by the same artist end up together. When you buy cds, you get much better quality.
If I do use an MP3 service, it is just to see if a cd I'm thinking about buying is any good. I generally use Limewire, and store what I download in a seperate folder away from my collection, so I can easily delete it.
The RIAA does some stupid things, but I still think it is worth the money to actually buy the CD, and I view boycotts as one of the most in-effective tools to combat the RIAA. I think a well-written letter will do so much more than 1% of the population boycotting cds.
Except you block the damn bathrooms from the first floor, and I always, without fail, go up the wrong staircase. And there is this great sign telling you, no through traffic. One day, I got mad, and I walked through anyway, and I got away with it.
I'll bet like 5 people who read this article will have any idea about which bvilding I'm talking about. Those who do, dont you fell my pain?
Slashdot Mirror
A couple points:
0.) For any architecture, an arbitrarily conceived benchmark can prove it optimal.
1.) Only L2 cache is shared.
2.) A modern scheduler wont give any type of guarantee that when one process is on a CPU, a certain process is on another.
3.) Even if it did, as soon as a user process blocks to the kernel for network IO, the kernel will reschedule a differant process to run on the first CPU, with a high probability of wiping much of the cache.
4.) Having two differant processes sharing the same cache can lead to many scenarios where one process is battling the other in terms of overwriting cache lines. L2 on Intel is something on the order of 8-way associative, if I remember correctly, so conflict misses will happen quite often.
Don't read this the wrong way, this is the way of the future, but in architecture, few problems are simple.
Somewhere in here is a joke about free as in beer just waiting to be told.
I really struggled with this for a long time. My final solution was to buy a regular bookshelf, and then went to Lowes and had even more shelves made for it. I buy jewel cases for every CD worth keeping. If it doesnt rank spending 10 cents on a jewel case, I throw it out. At the same time I did all this, I bought a labeling kit, and printed labels for all of the burned CDs. The labels cause the burned ones to last much longer, as they protect the top side of the disk, which is the much more easily damamged side.
When I was all done, I am able to store between 600 and 700 CDs on the book case. Most of them are music, so I put those in alphabetic order. Software gets sorted by type, and DVDs and games just kindof go wherever. People laugh when they see the bookcase completely full, but I think most people dont realize how much media we have now.
I have rowed at the University level (Purdue in the U.S.) and can honestly say that out of all cardio exercises, I find rowing to be the easiest on the body. When done intensely and incorrectly, it can be hard on the back and knees. This is rare, and more likely a result of bad form.
A rowing machine (most often called an erg), is a lot more than just a piece of exercise equipment. Many work very hard on achieving certain goals, such as 1 million meters, or rowing a marathon. At the university level, we always concentrated on 2k, 5k, 6k and long distances. During the winter, it wouldn't be uncommon to sit down and row 40k in a practice just to build endurance. Others days we would do 12 one minute on, one minute off pieces, and go home and just crash. Because rowing is so low impact, most rowers peak quite a bit older than in other sports. The Men's 2k world record is held by a person in the 30-39 age group. $800 is cheaper than a gym membership over time, and I recommend either a stereo or TV to go with the thing. The YMCA near me has several Concept2 erg's, and I would recommend looking at a couple of sites for some technique information, and try them out.
The great thing about rowing, is that if you enjoy it, pretty much any city with a river will have a club. Here you can meet other people interested, find coaches, and use their boats. In some select cities (Philly, Boston) their is a country club atmosphere to these places. Anywhere else, they are just normal people, like a bike club or gym.
Even though I no longer row competitively, I still use an erg for a good warmup and cool down, no matter what my exercise routine for the day is.
Just finishing up my end of semester projects, had to analyze several CPUs I designed this semester. One very simple one used 1048 logic cells, but could do integer arithmetic, jumps, branches, and memory operations. Not quite sure how many transistors that translates too, but normally an FPGA cell is a binary operation. Removing several of the odd arithmetic operations would lower that size quite a bit more. I also designed a 5 stage pipelined CPU with 32 word instruction cache, and 32 word data cache in about 2300 logic cells.
When the first vacuum tube based computers were invented, I'll built the designers felt like they were implementing a CPU of this size in Legos. It seems funny now, but this analogy probably holds a lot of water.
You most likely can use a PIII. I have an asus 440BX dual board that was designed for PII processors, but I put two 700s in it no problem. Check the FSB and multiplier maximums. Sometimes its a factor of what rev of the board you have, and what version bios is installed. On mine, Asus P2B-D, the owners manual doesnt say it's possilbe, but it's been my desktop for 6 years or so.
A neat trick on 440BX is that when you overclock the FSB to 133 MHz, the pci cards go back to their correct 33 MHz, but the AGP slot doesnt get corrected. This let me get extra CPU speed, high data tranfer to my video card, but didnt lock up due to crappy PCI cards.
The one thing to remember about these systems though is, a dual PIII with PC100 memory is incredibly memory bound. The memory just cant feed the CPU fast enough, and you lose a lot of cycles. On the other hand, the PIII is an incredibly efficient architecture, with only 10 pipeline stages.
At any rate, before buying PIIs, make sure you cant put some newer PIIIs in there.
Also known as HKN is a mix between a co-ed social fraternity and a student organization for Electrical and Computer Engineers. One must pledge to join, must be at least a Junior, and must be above a certain class rank. The organization has been around for over 100 years. My chapter (Beta Chapter at Purdue), was an underground society for a period of time. We run the lounge and snackbar in the EE building at Purdue, perform community service, have get togethers most fridays, have professionals come in and speak to us. It is quite an active organization, found at most top engineering schools. Remember, HKN, "The Nice Guys".
Well, modern cpus are clocked so much faster for two reasons. 1, the transistors are smaller, and therefore faster. 2, modern cpus do less per instruction per clock, due to instructions moving through the cpu's pipeline. While a sinclair might execute an entire instruction in a single clock, a modern Pentium will break that instruction into close to 30 clock cycles, doing a very small portion of the instruction in each pipeline stage. The downside to this, is that when a jump or branch occurs, which is very common, the pipeline must be cleared. Modern chips use branch prediction to try and use which way a branch went the last time it was evaluated, to predict which way it will go in the future. This does add tranistors, which increases power consumption, due to smaller transistor, I thikn the change is negligible. In addition, modern cpus provice floating point capabilities, larger word width, and things like dma controllers, page frame pointers and cache built into hardware. The sinclair probably only had a stack pointer.
It is true that a modern sinclair with smaller transistors would use less power, but so much has been done in this field in the last 25 years, that a modern design could do so much better. Look at something like a motorola hc12 for a modern version of a cpu similar to the sinclais. Even this has a 2 stage pipeline if I recall correctly.
Furthermore, the ISA of old chips is quite limited.
Industry doesnt change designs unless valid. x86 is a great example of this. It is a pretty much messed up design, some instructions are 17 words long, but it can be made to work. Industry moved away from the sinclair because a better solution was found.
No, because the throttle is connected to a spring, that returns it to neutral on one side, and the gas pedal linkage pulls it the opposite direction. If the throttle is stuck, releasing pressure on the pedal linkage isnt going to move it, if it would, then it wouldnt be stuck. The hope is that by pressing the throttle further, and then releasing, the spring has more force trying to make it contract, thus overcoming whatever caused the throttle to stick. In any case, putting the car in neutral (or use the clutch in a manual) will overcome a stuck throttle.
When Syllable was forked from AthOS, a lot of time and effort went into choosing a name. IIRC, a long submission period was conducted, with a voting on all submissions. This narrowed down the pool to a few names, which were voted on again. Syllable was the winner. It is a lot harder than one can imagine to choose a good name that doesnt violate something else's name.
Alright, Ill bite. My experience is in MR, PET and CT imaging. Ultrasound has many of the same concerns listed, but on a smaller level, as it is less computationaly intense (I'm told anyway).
These systems aren't embedded. They have 4 and 8 gigs of main memory, huge disk arrays. 2 or 4 processors in the host alone. Dual monitors,and several attached storage devices. This is just for the console the operator sits at. Some systems will have more than one workstation as part of the console. Connected to this over high speed interfaces can be hundreds of FPGA's coupled to embedded boards running VxWorks, QNX, Nucleos, or just about anything else. Vendors such as Motorola, Mercury, and custom rolled boards are the norm. One board produced by Mercury contains 4 G3 processors and a gig of ram. Systems will have 2 to 8 of these, each running an embedded OS. There are a lot of hardware interconnects, IPC mechanisms, and little pieces of software to break.
Industrial operating systems struggle with these loads. For years, the standard was Solaris or IRIX, but it is quickly becoming Linux, with vendors rolling their own distributions. Windows is used by some, due to programmer availability, ability to run Java, and it is stable when used with tested and validated device drivers. Another advantage of Windows software is that laptops are all over hospitals, so if you write an image viewer application for Windows, suddenly images can be pulled up all over the hospital. Most of the problems that hospitals are having with virus's and what not are from improperly implemented firewalls (and some stupid users)
On the time required to validate: The FDA requires a strict, audited validation. Plans to accomplish this for an entire systems can be thousands of pages long, and take many weeks to complete. On top of this, every configuration must be tested, and there are plenty of differant of options available. If validation is not completed, the manufacturer cannot guarantee the equipment, and is in violation of FDA rules and regulations. It is not like the manufacturers are trying to screw the customers. It is a fiercely competitive business, and every little edge is worth fighting for.
The software that powers these machines has evolved over 15 and 20 years in some cases. It is incredibly complex, and not easy to just all of a sudden say "Oh, I guess we should switch platforms."
Image quality and features are paramount, hospitals are willing to pay for this, because this is what helps to save lives.
It is sad, but in big corporations, this is often a notable differance. An engineer's signature doesnt count for much anymore.
Isn't Cray just a division of SGI now? It seems that they are moving more and more in the direction of the old Cray.
Those of us who are American still don't know who Al Franken is. From context, I presume a Democrat as well?
I'd also recommend that you don't feed your computer. Computers are _inatimate_objects_, not to be confused with pets that need food and water. I know you might think you'll get an extra MHz or 2, but that food is _really_ unneccessary...
I think my sig says all that is needed...
I agree, changing all the clocks is just a pain. The current system has worked well so far, why change? I have yet to find software that can't cope with it, and it's nice to not have to change the VCR, your watch, the clock in your car, the clock in your other car, the wall clock, the microwave, the oven, the alarm clock, the answering machine, and whatever else has a clock these days.
I hear that. I will never buy another ATI card again. I had a Radeon 9000, I ended up trading it for a far less superior NVidia card, purely becuase the drivers didnt suck as bad. I bought the 9000 for its dual screen capability, and yet I have not found one person that has gotten dual screen to work in Linux. I found plenty of instability, and the driver reporting that it supported features that it didnt, so that when programs tried to use them, they crashed. I think this binary module idea is crap. If you think the source code to the driver is gonna give that much insight into how these things work, than why are only two companies actively competing? Furthermore, many open source drivers do exist, and no rampant intellectual property loss has occured.
Yeah, I'd like to believe you, but I've seen people get away with murder in source code before. Open source coders worry a lot more about things like indentation, and filenames that make sense. In closed source shops, a lot of times what is quickly coded as a prototype becomes the shipping product, and things like indent cant be used because it breaks diffs. As much as I'd like to look with my own eyes, this sounds like one of the things it would be best if I just ignored it.
no shit, I kept waiting for the funny part, and it never came. Lighten up buddy.
Im glad Purdue is fast to somewhere, cause I live off campus, and it's faster to pull stuff from North Carolina than from csociety, 3 blocks away. Csociety incidently is actually just a dell pc sitting on top of a filing cabinet in the IEEE lounge, in the basement of the EE building. If you plug into the right subnet in EE, its fast.
I know on slashdot, there is always someone who will prove you wrong. Today, I am that guy. I'm 21 years old, live on a college campus with a fat pipe. I pretty much don't remember when we didnt have MP3s. I own between 500-600 cd's, and I feel that it is money well spent:
l: It's not illegal.
2: A hard drive crash doesnt erase my collection. Burned cd's, backups, what have you get scratched, and aren't reliable. My factory made cd's will last much longer.
3: I can legally rip them at a high-bitrate in whatever the common format is.
4: A lot of my collection is indie / small label punk, these bands probably make less than I do, stealing their cds instead of buying really does affect them.
5: The main reason I buy cds is that when I rip them, there are no pops, none of my tracks are cut short, there are no duplicates, and the tags are 100% correct. I can put them in a database, and magically all the songs by the same artist end up together. When you buy cds, you get much better quality.
If I do use an MP3 service, it is just to see if a cd I'm thinking about buying is any good. I generally use Limewire, and store what I download in a seperate folder away from my collection, so I can easily delete it.
The RIAA does some stupid things, but I still think it is worth the money to actually buy the CD, and I view boycotts as one of the most in-effective tools to combat the RIAA. I think a well-written letter will do so much more than 1% of the population boycotting cds.
"I'm really tired of the Universities on the West/East coasts pissing on the Midwest Universities."
Reminds me of a joke...
"Which way do I leave from?"
"Here at Harvard, we don't use prepositions to end our sentences."
"Alright. Which way do I leave from, asshole?"
(Purdue, thinking were better than the people who think they are better than us since 1869.)
You're free to explore the potential of the language and come up with new instructions and invent new ways to use existing instructions.
Strategery?
Except you block the damn bathrooms from the first floor, and I always, without fail, go up the wrong staircase. And there is this great sign telling you, no through traffic. One day, I got mad, and I walked through anyway, and I got away with it.
I'll bet like 5 people who read this article will have any idea about which bvilding I'm talking about. Those who do, dont you fell my pain?
This is one of the most insightful comments I have read on here in a while. Definently +1 insightful.