Not to mention that, while the number of new jobs created was pretty small, at least it was positive. Or that unemployment fell from 5.6% to 5.5. That's pretty low to be called 'bad.'
The small number of jobs created, 32,000, is insufficient to keep up with inflation of the working-age population. The zero number you should use is 150,000.
The unemployment rate only measures the number of people collecting unemployment insurance, and not the number of people without work. When your benefits expire, you are considered nonworking, but not unemployed.
Re:Swap nessisary in low mem (128M) systems
on
Is Swap Necessary?
·
· Score: 1
Gigabit ethernet might give you enough bandwith for swapping, but the latency will kill you.
Huh? Hard drive seek times are 5-15 ms. 100 Mbit Ethernet round trip times are about 0.2 ms. Myrinet can bring that down to about 0.01 ms.
One thing I noticed with one of my GCCs is that -fomit-frame-pointer was not activated on -O3, even though the manual says it is...
-fomit-frame-pointer is only enabled with -O on architectures that support debugging without a frame pointer. For example, it is enabled on x86-64 but not on x86-32.
sum is merely a utility function, also found in imperative languages such as Python: sum(xrange(1,9)) . Using fold demonstrates the power of functional languages. Python has a weak version of fold called reduce, which may be removed in future versions because it is not idiomatic and slower than the imperative equivalent due to tradeoffs in Python's implementation. One cannot treat operators as functions and is forced to use operator.add or the lambda equivalent: reduce(lambda x, y: x + y, xrange(1,9)) .
You could make a reasonable argument that this year's winner wasn't decided purely on programming skill, and an even more reasonable argument that doing well in one such contest doesn't say nearly as much as doing well in multiple consecutive ones with different problems to solve.
Andrew Hudson has gone to the ACM International Collegiate Programming Contest twice (the maximum allowed) and placed 21st and 27th, done well in local university contests, and is ranked highly on TopCoder. While it is unfortunate that the contest organizers picked a problem that favored those with expensive hardware, his contest performance has been consistent.
the amount of energy that goes into making a [solar] tile is far more than you'll ever get out of it.
This has been false for quite some time. Modern panels pay for themselves from an energy perspective in roughly 5 years, given a sunny location and decent gear. The article "PV Payback" in Home Power #80, pg. 42 claims 2.9-6.5 years, but I have not had the chance to critically analyze their data. There are reasons to prefer centralized generation of power such as nuclear plants, but this is not one of them.
the regular old light bulb is still most efficient overall when compared to fluorescents or neobulbs. The amount of energy, resources, and pollution that goes into something has to be taken into account if you're really looking at reducing your impact on the environment.
Given your previous statement, I would like to see evidence to back up this assertion. While CFL are more expensive to produce on a unit basis, they last longer and have cheaper secondary costs such as shipping and stocking because fewer bulbs are required.
I believe the tablet PC provides adaquate proof that material from Microsoft Research makes it into production software products. They developed the "digital ink" concept and the ability to use "ink" in e-mail/other document types.
Apple's Newton opearting system had these innovations in 1994.
Re:Small range increases mean BIGGER area increase
on
New Look at ADSL2
·
· Score: 3, Insightful
You wrote:
Remember your high-school geometry , area of a circle is pi(r^2)
So the 6% ( ? ) increase in range translates to a more than 12% increase in coverage area. It's not as small as first it appears.
From article:
For example, on longer phone lines, ADSL2 will provide a data rate increase of 50 kbps--a significant increase. This data rate increase also produces an increase in reach of about 600 feet, which translates to an increase in coverage area of about six percent, or 2.5 square miles.
An increase of 600 ft, or 3%, in distance yields a 6% increase in area. You can work it out yourself if you add to the maximum ADSL distance of 17500 ft.
Just introduce branch delay slots and you lose a whole lot of bubbles and complexity. Just imagine how simple a CPU without branch prediction would be...
How many branch delay slots were you planning to add? The trend to lengthen the pipeline increases the slots you need to have; you might be able to fill 1 or 2 branch delay slots, but can you fill 5 or 10? Also, do you want to change your ISA every time you add pipeline stages? Branch delay slots are considered a design wart on modern processors (MIPS, SPARC, any others?).
I think you overestimate the cost of branch prediction. All modern processors have added branch prediction, including the ones with branch delay slots.
I question the value of a high TopCoder ranking for purposes of finding employment, but it is far from a sham. I get money to solve sometimes interesting problems and compete with my friends. The money isn't great compared to a real job (I average $20 per competition hour), but it's worth it if you enjoy solving these kinds of problems. TopCoder changed their compensation scheme about a month ago, which lowered most people's average winnings.
As for the buzzword compliance, remember that TopCoder is a business and recruiters are its customers. TopCoder is only providing what they are asking for.
Another good deal for tuition is the University of Texas at Austin. It ranks at the bottom of the top ten for computer science(*) and is US$5500 for out of state and US$2500 for in state students per semester. A lot of the other departments are mediocre, so be sure you want to be a computer scientist (there's a LOT of attrition in our department) if you come here.
(*) This is important to some people and employers. Since I've never been to a "lesser" university I can't give an honest assessment of whether it is valuable. I will say that the professors are competent and there is lots of interesting research to get involved in. There is a big difference between the best and average students, but I suspect this is true at all schools.
Fortran is a non-existant language, as far as college students are concerned. At the University of Texas at Austin, the required(*) languages are C++, Java, some assembly, and usually one functional language (Haskell or Scheme). The better programmers usually learn one or more scripting languages on their own (Perl, Python, and sh being most popular). There is a remarkable lack of multi-lingual students; I've often wondered if this hurts them (thinking about problems more abstractly with higher-order functions, recursion, and coroutines).
(*) Our department has a policy of teaching concepts, not languages or technologies. Required is a minomer, but most students learn these in the introductory classes.
The performance hit for using the smaller pages is mostly unrelated to paging. When a CPU loads an virtual address (all addressing in "protected mode" is virtual), there is a translation to a physical address before data can be accessed. This table is stored in memory and the CPU breaks into kernel mode to do the translation. To avoid this cost, there is a cache of translations (managed by the kernel) in the Translation Look-aside Buffer (TLB). Most of the entries in this cache are for 4kb pages, but there are a few 4mb pages which are generally used for kernel memory (I am unsure if any OSes use the big pages for user programs).
That said, there should be a modest performance hit. Bigger pages can store more data, which results in fewer TLB misses. Hopefully someone will post benchmarks.
How likely it is that one will rewrite the same sector one million times in a row? 1/(num_sectors ** (10 ** 6)). Flash is meant for permanent storage; it is unlikely that one's data is so unimportant that it would be rewritten almost immediately after use.
Python 2.1 (#1, Jun 20 2001, 17:56:17)
[GCC 2.95.2 19991024 (release)] on linux2
Type "copyright", "credits" or "license" for more information.
>>> 'Python also has this feature. ' * 3
'Python also has this feature. Python also has this feature. Python also has this feature. '
I guess Seymour Cray was a moron because he didn't use cache in his computer designs
Seymour Cray never designed any consumer-grade products. Cost was not nearly the issue for him as for systems designers today. His market dictated that a 10% increase in performance was worth a doubling the cost.
Are there STILL people out there that think their processor is the bottleneck of their system?
Your ignorance is blinding; memory speed is _not_ the problem with today's computers. Large/smart caches are a far superior solution than fast memory in any calculation that involves cost. Caches are a great solution in the sense that virtual memory is; optimization for the common case.
I fail to see why the bus systems and processors can't make the speed transition at the same time. The extra time in the CPU cycle can be spent improving data pipes between components and increasing the number of processors.
Then you fail to see why fast memory busses are expensive. It's much cheaper to take advantage of spatial and temporal locality of reference by using caches than to build a faster memory bus.
Imagine if the standard system today was four, 128-bit processors running at 133MHz on a bus of 133MHz?
You'd pay about $10,000 and there would be no speedup. A single Pentium III/Athlon running at 600 Mhz would probably be faster. Why? Most (95+%) of all memory accesses are found in L1 or L2 cache. The limiting factor in your system would be the excruciating slow processors. If there were no caches within the processor (as you seem to imply), you would see a Pentium/133 blowing the doors off your mythical system. Why? Bus contention between the 4 processors.
I don't apoligize for my condescending attitude; you are spouting off about something you obviously have no knowledge of. "Bigger/faster = better" often isn't. Tradeoffs are necessary to hit price/performance goals. The hardware industry is extremely demanding; everyone is working their asses off to hit Moore's law in an industry of collapsing prices.
Followups to Hennessy and Patterson's excellent Computer Architecture: A Quantitative Approach, chapters 1 and 5.
Newton hardware, as great as it was for its time, is poorly documented in Apple tradition. The hardware is also relatively expensive (~$500 used) . Besides, porting Linux to Windows CE allows one to 'liberate' another platform from Microsoft.;-)
3,333,360 points, as reported by the article, divided by 255 = 13072 points/level. Can anyone verify this as the maximum number of points you can score on a level?
Does anyone else think the limiting factor is the control of the actual gameboy games? There's no great way to emulate the GameBoy directional pad with the current hardware buttons. Sure, it's a neat trick to get the games working on PalmOS, but getting to play them is an entirely different story.
Hopefully Handspring will add a directional pad to their units, as they are supposedly going after the consumer market.
Slashdot Mirror
The intel 945G chipset for Atom is fully documented and has quite good open source 3d drivers.
It sucks up 22W+ by itself though, and is very old. It's nothing compared to the VX800 or CN896.
The desktop 945 uses 22 W but the mobile 945GSE uses 4W: http://www.heise-online.co.uk/news/Intel-introduces-the-Atom-230-and-Atom-N270--/110855
The unemployment rate only measures the number of people collecting unemployment insurance, and not the number of people without work. When your benefits expire, you are considered nonworking, but not unemployed.
A neat paper on fold is A tutorial on the universality and expressiveness of fold.
Haskell: foldl (+) 0 [1..8]
Intuitive and obvious if you are comfortable with higher order functions.
Apple's Newton opearting system had these innovations in 1994.
How many branch delay slots were you planning to add? The trend to lengthen the pipeline increases the slots you need to have; you might be able to fill 1 or 2 branch delay slots, but can you fill 5 or 10? Also, do you want to change your ISA every time you add pipeline stages? Branch delay slots are considered a design wart on modern processors (MIPS, SPARC, any others?).
I think you overestimate the cost of branch prediction. All modern processors have added branch prediction, including the ones with branch delay slots.
As for the buzzword compliance, remember that TopCoder is a business and recruiters are its customers. TopCoder is only providing what they are asking for.
Another good deal for tuition is the University of Texas at Austin. It ranks at the bottom of the top ten for computer science(*) and is US$5500 for out of state and US$2500 for in state students per semester. A lot of the other departments are mediocre, so be sure you want to be a computer scientist (there's a LOT of attrition in our department) if you come here.
(*) This is important to some people and employers. Since I've never been to a "lesser" university I can't give an honest assessment of whether it is valuable. I will say that the professors are competent and there is lots of interesting research to get involved in. There is a big difference between the best and average students, but I suspect this is true at all schools.
Fortran is a non-existant language, as far as college students are concerned. At the University of Texas at Austin, the required(*) languages are C++, Java, some assembly, and usually one functional language (Haskell or Scheme). The better programmers usually learn one or more scripting languages on their own (Perl, Python, and sh being most popular). There is a remarkable lack of multi-lingual students; I've often wondered if this hurts them (thinking about problems more abstractly with higher-order functions, recursion, and coroutines).
(*) Our department has a policy of teaching concepts, not languages or technologies. Required is a minomer, but most students learn these in the introductory classes.
According to your numbers, it's a ~0.3% performance difference, which is still insignificant.
The performance hit for using the smaller pages is mostly unrelated to paging. When a CPU loads an virtual address (all addressing in "protected mode" is virtual), there is a translation to a physical address before data can be accessed. This table is stored in memory and the CPU breaks into kernel mode to do the translation. To avoid this cost, there is a cache of translations (managed by the kernel) in the Translation Look-aside Buffer (TLB). Most of the entries in this cache are for 4kb pages, but there are a few 4mb pages which are generally used for kernel memory (I am unsure if any OSes use the big pages for user programs).
That said, there should be a modest performance hit. Bigger pages can store more data, which results in fewer TLB misses. Hopefully someone will post benchmarks.
How likely it is that one will rewrite the same sector one million times in a row? 1/(num_sectors ** (10 ** 6)). Flash is meant for permanent storage; it is unlikely that one's data is so unimportant that it would be rewritten almost immediately after use.
Python 2.1 (#1, Jun 20 2001, 17:56:17)
[GCC 2.95.2 19991024 (release)] on linux2
Type "copyright", "credits" or "license" for more information.
>>> 'Python also has this feature. ' * 3
'Python also has this feature. Python also has this feature. Python also has this feature. '
Just to state the obvious, fluorescent lights use less power AND give off less waste heat, which makes my air conditioner run less often.
This summer I'm going to put aluminum foil over one of my apartment windows (there's no shade at all), has anyone tried this?
> (And NewsWatcher and its derivatives are still the best news readers
> on any platform. Nothing compares on Unix or Win32.)
Not only is this ignorant, its simply not wise to make such a sweeping statement. Software `goodness' is largely a matter of preference.
-- a SLRN user
http://space.mit.edu/~davis/slrn.html
Seymour Cray never designed any consumer-grade products. Cost was not nearly the issue for him as for systems designers today. His market dictated that a 10% increase in performance was worth a doubling the cost.
Your ignorance is blinding; memory speed is _not_ the problem with today's computers. Large/smart caches are a far superior solution than fast memory in any calculation that involves cost. Caches are a great solution in the sense that virtual memory is; optimization for the common case.
Then you fail to see why fast memory busses are expensive. It's much cheaper to take advantage of spatial and temporal locality of reference by using caches than to build a faster memory bus.
You'd pay about $10,000 and there would be no speedup. A single Pentium III/Athlon running at 600 Mhz would probably be faster. Why? Most (95+%) of all memory accesses are found in L1 or L2 cache. The limiting factor in your system would be the excruciating slow processors. If there were no caches within the processor (as you seem to imply), you would see a Pentium/133 blowing the doors off your mythical system. Why? Bus contention between the 4 processors.
I don't apoligize for my condescending attitude; you are spouting off about something you obviously have no knowledge of. "Bigger/faster = better" often isn't. Tradeoffs are necessary to hit price/performance goals. The hardware industry is extremely demanding; everyone is working their asses off to hit Moore's law in an industry of collapsing prices.
Followups to Hennessy and Patterson's excellent Computer Architecture: A Quantitative Approach, chapters 1 and 5.
Newton hardware, as great as it was for its time, is poorly documented in Apple tradition. The hardware is also relatively expensive (~$500 used) . Besides, porting Linux to Windows CE allows one to 'liberate' another platform from Microsoft. ;-)
3,333,360 points, as reported by the article, divided by 255 = 13072 points/level. Can anyone verify this as the maximum number of points you can score on a level?
Does anyone else think the limiting factor is the control of the actual gameboy games? There's no great way to emulate the GameBoy directional pad with the current hardware buttons. Sure, it's a neat trick to get the games working on PalmOS, but getting to play them is an entirely different story.
Hopefully Handspring will add a directional pad to their units, as they are supposedly going after the consumer market.