I looked at that code (which hasn't been updated in a while) assuming I might be able to fix some of the glaring HTML errors (I mean, who is allowed to check in HTML code with validator errors?), but gave up after realizing it wasn't complete.
There is java wrapped into the back-end server infrastructure too. That was evidenced by the 3 page long java exceptions wrapped in JSON it was returning yesterday.
I got 36 HTML errors on the front page running it through the wc3 validator. Its chuck full of misspelled attributes, duplicate attributes (aka two hrefs in an A tag, etc).
There is no doctype on the login page, and it throws another 61 warnings (mostly because without the doctype the validator turns errors into warnings).
Plus there were a fair number of jqeury errors.
Finally I stopped at the login because the security question AJAX URL literally returned a 500 line java/jboss exception pile of crap wrapped in JSON data.
You have got to wonder who is creating a site where they can't even get the markup correct.
I lived in NH, now I live in Austin city limits, where much of what you describe is illegal (shooting guns for example).
Anyway, my point is a lot of what you describe is great if everyone has 10+ acres. In my neighborhood, with houses 6" apart those kinds of things can get annoying. So everyone has to understand that and try to be good neighbors. That means bondfires blowing smoke into my house, or whatever are a bad idea. Like for example my neighbor with the rotting cars in the front yard. Everyone else on the street drops little hints like "does that thing still run" or "you can donate that for a tax writeoff" but he doesn't get it and he has a car that hasn't moved for 7+ years. The tires went flat 5+ years ago.
Your article points out much of that gap has to do with house prices.
Which I agree with, and would like to point out that the housing market is similarly fragmented. There are a lot of places in the US where the house values continue to decline, and others where the house values are exceeding the pre bubble values. But the one trend everywhere is that the houses in the upper 1/4rd of the desirability scale are the ones going up. Its the white flight syndrome but it seems to have more to do with "class" and the ranking of individual schools and other factors than race. Basically those that were rich are even richer without any effort while everyone else is busy trying to scramble over each other to get there.
They have been coming to my house (in Austin) since the uverse upgrade a couple years ago. I asked them "can i have it without TV" and the answer was no. Now its possible, but TW is giving me 30/5 for less than their 20/1 service.
A rather cute little girl knocked on my door the other day, but as soon as I saw the AT&T shirt, I told her I wasn't interested, when she was persistent I basically told her where to put it and closed the door in her face.
Anyway, it took TW 5 years to roll out DOCIS 3 after everyone else did it, its still only a single channel upstream though. I suspect TW's plan in Austin is to ignore google until they actually start to see a significant impact in their business. Which may never happen if google just wires up a couple neighborhoods.
I think its TW's model, do the absolute minimum, charge the maximum and only upgrade/lower prices if another competitor has a better offering in the same area. Which is basically AT&T at the moment since its impossible to choose Grande if you have TW because of the monopoly rules in Austin.
Anyway, AT&T's model seems to be charge whatever they want for garbage service, advertise like the dickens, and pull in suckers.
OTOH, grande is suddenly advertising a 100/5 service in Austin (up from 15/1 or some crap) I can only assume to get some good will before google hits.
Which is really just a load of political double speak BS. The loss of pensions is just one of many moves to suppress the cost of labor in the US.
First, pension programs are designed so that a company puts money in an account which grows until retirement, and then gets drawn down. Same as a 401k. When run properly, the company doesn't incur "future" obligations _AT ALL_.
So, the finer points here are worth a book, but. Some other points.
Expected rates of return, and how long someone lives tells you how much money should be invested each year.
A sliding scale of risk tells you the likely-hood that the numbers you pick for rates of return and life expectancy will provide a surplus or deficit in the long run.
In the past, these things were regulated by the government.
The government regulated them because they were also running the pension guarantee fund and are on the hook for pension plans that failed.
This regulation setup standards for the riskiness of investments and what could be considered valid rates of return based on conservative estimates.
The federal government started changing these laws in ways that allowed more risky behavior, mostly at the behest of large corporations and wall street.
Corporations also saw that it was easy to fleece their employees by moving the risk to the employee using 401ks.
Part of moving this risk means that they can play more games with the actual amount of money in play. Hiring a worker with a pension at X salary costs Y, vs hiring a worker at A+B costs C, where Y>C because the employee is told that the extra "B" is going to provide them a similar or larger retirement.
Lots of this was happening during the 80's and 90's where everyone was told that 401Ks were returning double digits every year. So the numbers appeared to work as long as the long term risks weren't calculated.
Short term rewards at many companies put them in positions where they were on the hook for pension deficits. Corporate bankruptcy law does little to punish decision makers at large companies/boards that can walk away with massive amounts of money, while leaving the government or other parties on the hook even on fairly small timescales (see hostess).
Oh the list could go on...
The bottom line is that pensions are a workable method for building a retirement, and many companies actually included the amount they were saving for each employee in offer letters/etc as part of the "total compensation" package details (making offers with lower salary potentially more attractive).
Cloaking it in BS like you regurgitated was a way to confuse the wage suppression.
And with the introduction of ACPI and soft power on the PC was also the introduction of the PC that needed to by physically unplugged or have the batteries removed to power it off.
It also introduced the PC that sucks down 3-7 watts of power even when turned off.
BTW: I had to removed power again just a few days ago, bmc went crazy on a HP and holding the power button down/etc failed to power it off. Had to reach behind the rack and disconnect mains power. I've seen laptops do it to, linux or something makes some EFI call hits a bug and boom the system won't power off without the battery being removed.
Mostly because some asshole wants to be able to power on his PC via the keyboard? Or sometimes even via the LAN. I myself do this, but outside of wake from standby, i'm pretty convinced using an external device is a better option. I still have an external smart power switch to power off my monitors/etc when the PC goes off. Saves me ~25 watts because of all the USB devices with wall warts and monitors attached to my desktop.
You haven't known pain until you try to get Ultima 7 to run on a system with a Proaudio Spectrum 16
IIRC, with that version of ultima it wasn't your PAS that was the problem, it was the game. That darn game was a buggy POS even a couple years after release.
My game machines from that era always had the latest Sound Blaster (even though I also owned a PAS and a Gravis (actually still have the gravis)) because then tended to "just work". That is until PCI came out, in which case nothing really worked for a couple years.
Or you can realize that we're talking about building a *new* computer, with *new* hardware.
Which doesn't do him a lick of good if he wants the new computer to run his old $5000 data acquisition hardware that only has XP drivers. Or dozens of other pieces of hardware which may have newer versions supported by newer OS's but the price of replacement is significantly more than the computer.
Yawn, I've heard this before, and its not that big a deal with to change the scaling and it mostly works. I have a laptop with a 180PPI display running windows XP from ~2006 and it works fine with display scaling. Te problem is the removal of exactly x2 mode in win7 (added back with 8.1).
Heck an OEM could probably rig a deal with the GPU manufacture to produce a driver that 2x scales the display to windows same as Apple does and white-list applications that understand higher resolutions displays with a GDI hook even without Microsoft support. With Microsoft support adding a HIGH DPI flag in the applications manifest or similar to allow native scaling can't possibly be that hard. In fact I bet there is a way for an application to bypass the windows scaling today (if nothing else you detect it enabled and resize components as necessary).
True, it's horrible that review sites benchmark CPUs using the kind of programs people actually run on them.
Yes, and no. If your a gamer, obviously having a CPU that the games are optimized for is a big win. But don't extrapolate general performance from a single benchmark. Especially when one of the CPU vendors is providing "free" performance help for the game/application.
At this point, its pretty clear that choosing the Intel is the correct choice for big name games.
We will see if this changes over the next few years with the consoles being AMD based. The game companies are going to optimizing for those platforms, we will see if any of that carries over into game benchmarks for desktop machines. In many cases its possible to get a 2x delta by optimizing for a particular CPU/platform at the expense of the general case.
As for SSE, AMD has always been a little behind with SSE (well duh, they follow whatever intel is doing and it takes them a while to catchup), so if your application is built for the latest version of SSE, and its gaining something from it, then running a similar version without hurts. Recently seen with SSE 4.1/4a where there were a couple useful instructions for some code paths that didn't exist on the AMD and hurt it on benchmarks using SSE4.
The spec benchmarks tell a different story, and tend to be more representative because each vendor does their best rather than intel/nvidia providing "free" performance enhancement advice for game companies.
So, from my own experience the Amd/Intel story is a lot closer than some of these benchmarks might lead you to believe. Especially for server applications.
Its pretty easy with modern CPU's to make fairly small tweaks that give advantage to one CPU or another. We have a bunch of microbenchmarks for our application, and things like memcpy performance can be swung 2x-4x. Or even the depth of loop unrolling for some things. In one loop the intel it may like 2x and the AMD like 4x unroll. With each one tuned to run best on the platform the bottom line performance is often quite similar, but run the AMD optimized one on the intel, or the reverse and suddenly one or the other CPU appears to be trouncing the other.
That document fails to point out the largest problem with the linux driver model. It basically forces the distribution maintainer to support a gobsnot pile of devices by backporting driver fixes, or it requires the users to upgrade their kernel every-time something fails to work. Which is basically continuously.
Neither solution is particularly helpful, as upgrading the kernel means its just as likely some other subsystem fails.
The model of only fixing what is broken, rather than throwing everything away and trying the latest version is NOT as robust a solution.
Frankly it also ignores the fact that over the last couple years the majority of the code changes in each kernel release are fixes for drivers. Heck didn't Linus just say that the current kernel RC was 70 something percent driver changes? So with that many bug fixes/changes going into drivers its amazing that anyone can claim the kernel drivers are stable pieces of software.
So basically, your hoping all the driver writers get their act together one day and release a kernel where all the drivers actually work for a given system configuration. Otherwise. your forever trying to find a combination that actually works for your PC. Or you learn to program in C, and start maintaining the drivers for your particular piece of hardware (what I end up doing as I also do this professionally). I have yet to find a linux distribution that works as well without extensive hacking as a fairly basic windows install. There always seems to be something or another that doesn't work.
My most recent laptop, with the most recent version of my distribution of choice failed to load the correct firmware for my bluetooth adapter (cause it didn't understand the PCI device id, and fell back to stupid mode), failed to understand the LCD brightness controls and has a serious bug in the HD4000 drivers GL SL 1.4 causing it to fail to shade polygons correctly in a number of games that require a recent version of GL SL. Plus, there was a fair amount of fun getting the EFI configuration correct, along with a number of other issues.
None of these would have been serious issues if the OEM customized a version of linux for the machine (aka someone could have fixed it), but as it was that job fell to me. I'm not really sure how your average slashdot user actually gets linux working correctly. I suspect that most people installing ubuntu or whatever don't actually get it working properly with most of their hardware. Instead setting for laptops that dont resume properly, or suck up unnecessary battery life running the LCD at 100%, or running the graphics card at a small percentage of its performance because the driver failed to setup the PCIe port properly, etc.
Basically, the driver model is busted and the arguments for keeping it the way it is boil down to religious wars rather than an attempt to fix a glaring problem.
Tape drives like T10kD and LTO6 have sustained transfer rates that are individually larger than single (non SSD) hard drives. Combined into tape libraries, they are capable of outrunning the fastest RAID arrays on the market.
That of course assumes you are using them as a backup/archive medium and streaming data from a RAID to the tape drive rather than trying to use them in random access mode.
A few years ago, I was involved in the conversion of the Stanford AI lab tape archive to modern media. This involved reading thousands of reels of 1/2" magnetic tape. It was a slow process. Volunteers were loading a tape onto a tape drive every 15 minutes for weeks.
Tape has changed a lot since the early 80's. Modern tape drives can sustain over 500MB/sec compressed each. And they tend to live in automated tape libraries that load and unload the tapes without human intervention. A library like that can do TByte/sec of IO.
That is true for something like SD, but modern tape drives can do well over 500MB/sec in compressed streaming mode, and have native uncompressed capacities of 8.5TB.
That means that even with a midrange tape library with 56 drives, your talking a read/write bandwidth of 27GB/sec (aka 1/3 Tbit/sec). Tape bandwidth scales linearly with the number of drives in the library, and things like the the SL8500 from STK can support up to 640 drives.
It still faster if you have a PB of data you need shipped from NY to LA to write it to tape, put it on a plane and read it back in LA. Plus, it all fits nicely into a big suitcase. Furthermore, even for smaller amounts of data (say 10-20TB a day) the cost of a tape drive and an next day delivery is going to be significantly less than the Gbit/sec or so of bandwidth required to ship a similar amount of data in most places in the US.
fortran code usually still seems to give faster executables than comparable C++ code on numerical applications
I don't think this is true anymore. C++ is pretty much the only language that has BLAS libraries that can actually beat the fortran ones. The latest C++ template libraries are using SSE/etc vector intrinsics and are capable of meeting if not exceeding the fortran performance for many applications.
But, if you have a bunch of code in fortran, its probably not worth the trouble to convert it.
Someone who is skilled in C can pick up Perl and C++ quite quickly
Spoken like someone who hasn't actually programmed in C++. I was very good in object pascal, C, and quite a number of other languages when I first started programming in C++.
Lets just say that "quickly" isn't a description I would have for myself or anyone I've ever met when it comes to how long it takes to "pick up C++". Sure, you might be able to write some code in C++ after a couple days, but that code probably isn't worth the price of the storage your using for it. I would guess it takes upwards of 3 or 4 years before a C++ programer is good enough not to be making absolutely stupid mistakes on a daily basis. This is especially true for Java programmers who waltz in and think they know how to program in C++ because the syntax is so similar. The results are usually buggy garbage that compiles but won't run for long.
Heck, after nearly 20 years, I still look at code I write on a daily basis and go WTF was I thinking yesterday when I decided to do that! Or my co workers do it for me during code reviews.
Then explain why would you use C++ instead of C? If you are just going to cut it down to things C does? If you are going to do things C can't, then you will have the performance penalities. Your statments are not really valid.
Because there are cases where you are manually creating a construct in C, that is handled automatically by the compiler in C++.
Take virtual methods for example. The linux kernel is chuck full of structures filled with function pointers. That is a virtual method in C++, except in C++ you don't have to worry at runtime if the function your calling is NULL because the compiler assured that during compile time. This allows micro-optimization, and more natural error handling. Plus, the syntax is standard, so that every 3rd driver writer isn't creating their own version of the same thing.
Then there is generic data structure management. The kernel is full of macros for RB trees, linked lists/etc. Using templates for this allows better micro optimization without the programmer having to get involved.
There are a lot of reasons, and most of the negatives can be answered with, a the simple statement, don't use that feature.
The people at itworld are less IBM literate than the WSJ, because they keep repeating "POWER mainframe".
Repeat after me, POWER is _NOT_ IBM's mainframe line. The mainframe line is the zSeries and runs on proprietary processors clocked at 5.5Ghz. POWER processors are in the pSeries and iSeries machines.
Now, that said, in many ways the high end pSeries stuff is better than the mainframe hardware, but in no way is it considered "mainframe" grade to the IBM sales guys.
All that said, RHEL and SLES both run on pSeries and zSeries machines.
With PC gaming its really important that the graphics drivers are easy to upgrade.
I recently purchased the humble bundle and tried running a couple of the games on a linux install on a recently purchased laptop. While the distro I was running was supported by the games in question. The drivers needed are apparently new enough they didn't make it into the most recent version of that distribution. So, instead what I got was a GL SL v4 system where the shaders didn't work well enough to actually play the games in question.
Uh, what a PITA. All the games just worked in windows 8 on the same machine, but the install process to upgrade the intel HD 4000 graphics drivers is basically dependent on the distribution maintainer. The process for doing it by hand quickly unmotivated me enough to give up. I don't need to waste 4-5 hours recompiling 1/2 of the linux graphics stack to just upgrade my drivers. This from a guy who maintains large parts of an embedded linux system at work.
I can't really imagine how anyone really runs linux who isn't a kernel hacker. I had to manually hack the blutooth driver on that machine, write a custom script to disable the touchpad, and only after a few days discovered how to control the LCD brightness because the built in KDE/etc tools weren't working. All this on a laptop that actually has worked better in linux than any I've previously owned/used.
Bottom line, linux needs to create a proper driver ABI and dump the kernel symbol versions for something more like C++ name mangling. Then the drivers will only fail to load if the kernel API's actually changed rather than because someone happened to recompile the given kernel version on a different machine.
That is not 100% accurate because privoxy can strip the google analytics calls out of the javascript in the source pages. It can also degrade https to http for questionable domains.
And it can also blacklist whole domains SSL or not, so i don't think there are any cases where you get more functionality with a DNS blacklist.
Privoxy can remove a lot more than just ads served from a given domain/server. It can remove ads served by the same domain/server as the source website, as well as a number of other features that make it pretty nice for speeding up browsing on devices that don't have ad/javascript blockers.
Slashdot Mirror
I looked at that code (which hasn't been updated in a while) assuming I might be able to fix some of the glaring HTML errors (I mean, who is allowed to check in HTML code with validator errors?), but gave up after realizing it wasn't complete.
There is java wrapped into the back-end server infrastructure too. That was evidenced by the 3 page long java exceptions wrapped in JSON it was returning yesterday.
Its really amazing it works at all.
I got 36 HTML errors on the front page running it through the wc3 validator. Its chuck full of misspelled attributes, duplicate attributes (aka two hrefs in an A tag, etc).
There is no doctype on the login page, and it throws another 61 warnings (mostly because without the doctype the validator turns errors into warnings).
Plus there were a fair number of jqeury errors.
Finally I stopped at the login because the security question AJAX URL literally returned a 500 line java/jboss exception pile of crap wrapped in JSON data.
You have got to wonder who is creating a site where they can't even get the markup correct.
I lived in NH, now I live in Austin city limits, where much of what you describe is illegal (shooting guns for example).
Anyway, my point is a lot of what you describe is great if everyone has 10+ acres. In my neighborhood, with houses 6" apart those kinds of things can get annoying. So everyone has to understand that and try to be good neighbors. That means bondfires blowing smoke into my house, or whatever are a bad idea. Like for example my neighbor with the rotting cars in the front yard. Everyone else on the street drops little hints like "does that thing still run" or "you can donate that for a tax writeoff" but he doesn't get it and he has a car that hasn't moved for 7+ years. The tires went flat 5+ years ago.
So, I understand both sides of the coin..
See subject
Your article points out much of that gap has to do with house prices.
Which I agree with, and would like to point out that the housing market is similarly fragmented. There are a lot of places in the US where the house values continue to decline, and others where the house values are exceeding the pre bubble values. But the one trend everywhere is that the houses in the upper 1/4rd of the desirability scale are the ones going up. Its the white flight syndrome but it seems to have more to do with "class" and the ranking of individual schools and other factors than race. Basically those that were rich are even richer without any effort while everyone else is busy trying to scramble over each other to get there.
And the 250GB month cap...
They have been coming to my house (in Austin) since the uverse upgrade a couple years ago. I asked them "can i have it without TV" and the answer was no. Now its possible, but TW is giving me 30/5 for less than their 20/1 service.
A rather cute little girl knocked on my door the other day, but as soon as I saw the AT&T shirt, I told her I wasn't interested, when she was persistent I basically told her where to put it and closed the door in her face.
Anyway, it took TW 5 years to roll out DOCIS 3 after everyone else did it, its still only a single channel upstream though. I suspect TW's plan in Austin is to ignore google until they actually start to see a significant impact in their business. Which may never happen if google just wires up a couple neighborhoods.
I think its TW's model, do the absolute minimum, charge the maximum and only upgrade/lower prices if another competitor has a better offering in the same area. Which is basically AT&T at the moment since its impossible to choose Grande if you have TW because of the monopoly rules in Austin.
Anyway, AT&T's model seems to be charge whatever they want for garbage service, advertise like the dickens, and pull in suckers.
OTOH, grande is suddenly advertising a 100/5 service in Austin (up from 15/1 or some crap) I can only assume to get some good will before google hits.
incurring future unsustainable pension obligations
Which is really just a load of political double speak BS. The loss of pensions is just one of many moves to suppress the cost of labor in the US.
First, pension programs are designed so that a company puts money in an account which grows until retirement, and then gets drawn down. Same as a 401k. When run properly, the company doesn't incur "future" obligations _AT ALL_.
So, the finer points here are worth a book, but. Some other points.
Expected rates of return, and how long someone lives tells you how much money should be invested each year.
A sliding scale of risk tells you the likely-hood that the numbers you pick for rates of return and life expectancy will provide a surplus or deficit in the long run.
In the past, these things were regulated by the government.
The government regulated them because they were also running the pension guarantee fund and are on the hook for pension plans that failed.
This regulation setup standards for the riskiness of investments and what could be considered valid rates of return based on conservative estimates.
The federal government started changing these laws in ways that allowed more risky behavior, mostly at the behest of large corporations and wall street.
Corporations also saw that it was easy to fleece their employees by moving the risk to the employee using 401ks.
Part of moving this risk means that they can play more games with the actual amount of money in play. Hiring a worker with a pension at X salary costs Y, vs hiring a worker at A+B costs C, where Y>C because the employee is told that the extra "B" is going to provide them a similar or larger retirement.
Lots of this was happening during the 80's and 90's where everyone was told that 401Ks were returning double digits every year. So the numbers appeared to work as long as the long term risks weren't calculated.
Short term rewards at many companies put them in positions where they were on the hook for pension deficits. Corporate bankruptcy law does little to punish decision makers at large companies/boards that can walk away with massive amounts of money, while leaving the government or other parties on the hook even on fairly small timescales (see hostess).
Oh the list could go on...
The bottom line is that pensions are a workable method for building a retirement, and many companies actually included the amount they were saving for each employee in offer letters/etc as part of the "total compensation" package details (making offers with lower salary potentially more attractive).
Cloaking it in BS like you regurgitated was a way to confuse the wage suppression.
And with the introduction of ACPI and soft power on the PC was also the introduction of the PC that needed to by physically unplugged or have the batteries removed to power it off.
It also introduced the PC that sucks down 3-7 watts of power even when turned off.
BTW: I had to removed power again just a few days ago, bmc went crazy on a HP and holding the power button down/etc failed to power it off. Had to reach behind the rack and disconnect mains power. I've seen laptops do it to, linux or something makes some EFI call hits a bug and boom the system won't power off without the battery being removed.
Mostly because some asshole wants to be able to power on his PC via the keyboard? Or sometimes even via the LAN. I myself do this, but outside of wake from standby, i'm pretty convinced using an external device is a better option. I still have an external smart power switch to power off my monitors/etc when the PC goes off. Saves me ~25 watts because of all the USB devices with wall warts and monitors attached to my desktop.
You haven't known pain until you try to get Ultima 7 to run on a system with a Proaudio Spectrum 16
IIRC, with that version of ultima it wasn't your PAS that was the problem, it was the game. That darn game was a buggy POS even a couple years after release.
My game machines from that era always had the latest Sound Blaster (even though I also owned a PAS and a Gravis (actually still have the gravis)) because then tended to "just work". That is until PCI came out, in which case nothing really worked for a couple years.
Or you can realize that we're talking about building a *new* computer, with *new* hardware.
Which doesn't do him a lick of good if he wants the new computer to run his old $5000 data acquisition hardware that only has XP drivers. Or dozens of other pieces of hardware which may have newer versions supported by newer OS's but the price of replacement is significantly more than the computer.
They can't because people are still on Windows
Yawn, I've heard this before, and its not that big a deal with to change the scaling and it mostly works. I have a laptop with a 180PPI display running windows XP from ~2006 and it works fine with display scaling. Te problem is the removal of exactly x2 mode in win7 (added back with 8.1).
Heck an OEM could probably rig a deal with the GPU manufacture to produce a driver that 2x scales the display to windows same as Apple does and white-list applications that understand higher resolutions displays with a GDI hook even without Microsoft support. With Microsoft support adding a HIGH DPI flag in the applications manifest or similar to allow native scaling can't possibly be that hard. In fact I bet there is a way for an application to bypass the windows scaling today (if nothing else you detect it enabled and resize components as necessary).
2560x1600 display.
Dell, Hp, Asus, etc????? Hello!!!!!!
Idiots.
True, it's horrible that review sites benchmark CPUs using the kind of programs people actually run on them.
Yes, and no. If your a gamer, obviously having a CPU that the games are optimized for is a big win. But don't extrapolate general performance from a single benchmark. Especially when one of the CPU vendors is providing "free" performance help for the game/application.
At this point, its pretty clear that choosing the Intel is the correct choice for big name games.
We will see if this changes over the next few years with the consoles being AMD based. The game companies are going to optimizing for those platforms, we will see if any of that carries over into game benchmarks for desktop machines. In many cases its possible to get a 2x delta by optimizing for a particular CPU/platform at the expense of the general case.
As for SSE, AMD has always been a little behind with SSE (well duh, they follow whatever intel is doing and it takes them a while to catchup), so if your application is built for the latest version of SSE, and its gaining something from it, then running a similar version without hurts. Recently seen with SSE 4.1/4a where there were a couple useful instructions for some code paths that didn't exist on the AMD and hurt it on benchmarks using SSE4.
The spec benchmarks tell a different story, and tend to be more representative because each vendor does their best rather than intel/nvidia providing "free" performance enhancement advice for game companies.
So, from my own experience the Amd/Intel story is a lot closer than some of these benchmarks might lead you to believe. Especially for server applications.
Its pretty easy with modern CPU's to make fairly small tweaks that give advantage to one CPU or another. We have a bunch of microbenchmarks for our application, and things like memcpy performance can be swung 2x-4x. Or even the depth of loop unrolling for some things. In one loop the intel it may like 2x and the AMD like 4x unroll. With each one tuned to run best on the platform the bottom line performance is often quite similar, but run the AMD optimized one on the intel, or the reverse and suddenly one or the other CPU appears to be trouncing the other.
That document fails to point out the largest problem with the linux driver model. It basically forces the distribution maintainer to support a gobsnot pile of devices by backporting driver fixes, or it requires the users to upgrade their kernel every-time something fails to work. Which is basically continuously.
Neither solution is particularly helpful, as upgrading the kernel means its just as likely some other subsystem fails.
The model of only fixing what is broken, rather than throwing everything away and trying the latest version is NOT as robust a solution.
Frankly it also ignores the fact that over the last couple years the majority of the code changes in each kernel release are fixes for drivers. Heck didn't Linus just say that the current kernel RC was 70 something percent driver changes? So with that many bug fixes/changes going into drivers its amazing that anyone can claim the kernel drivers are stable pieces of software.
So basically, your hoping all the driver writers get their act together one day and release a kernel where all the drivers actually work for a given system configuration. Otherwise. your forever trying to find a combination that actually works for your PC. Or you learn to program in C, and start maintaining the drivers for your particular piece of hardware (what I end up doing as I also do this professionally). I have yet to find a linux distribution that works as well without extensive hacking as a fairly basic windows install. There always seems to be something or another that doesn't work.
My most recent laptop, with the most recent version of my distribution of choice failed to load the correct firmware for my bluetooth adapter (cause it didn't understand the PCI device id, and fell back to stupid mode), failed to understand the LCD brightness controls and has a serious bug in the HD4000 drivers GL SL 1.4 causing it to fail to shade polygons correctly in a number of games that require a recent version of GL SL. Plus, there was a fair amount of fun getting the EFI configuration correct, along with a number of other issues.
None of these would have been serious issues if the OEM customized a version of linux for the machine (aka someone could have fixed it), but as it was that job fell to me. I'm not really sure how your average slashdot user actually gets linux working correctly. I suspect that most people installing ubuntu or whatever don't actually get it working properly with most of their hardware. Instead setting for laptops that dont resume properly, or suck up unnecessary battery life running the LCD at 100%, or running the graphics card at a small percentage of its performance because the driver failed to setup the PCIe port properly, etc.
Basically, the driver model is busted and the arguments for keeping it the way it is boil down to religious wars rather than an attempt to fix a glaring problem.
Tape drives like T10kD and LTO6 have sustained transfer rates that are individually larger than single (non SSD) hard drives. Combined into tape libraries, they are capable of outrunning the fastest RAID arrays on the market.
That of course assumes you are using them as a backup/archive medium and streaming data from a RAID to the tape drive rather than trying to use them in random access mode.
A few years ago, I was involved in the conversion of the Stanford AI lab tape archive to modern media. This involved reading thousands of reels of 1/2" magnetic tape. It was a slow process. Volunteers were loading a tape onto a tape drive every 15 minutes for weeks.
Tape has changed a lot since the early 80's. Modern tape drives can sustain over 500MB/sec compressed each. And they tend to live in automated tape libraries that load and unload the tapes without human intervention. A library like that can do TByte/sec of IO.
That is true for something like SD, but modern tape drives can do well over 500MB/sec in compressed streaming mode, and have native uncompressed capacities of 8.5TB.
That means that even with a midrange tape library with 56 drives, your talking a read/write bandwidth of 27GB/sec (aka 1/3 Tbit/sec). Tape bandwidth scales linearly with the number of drives in the library, and things like the the SL8500 from STK can support up to 640 drives.
It still faster if you have a PB of data you need shipped from NY to LA to write it to tape, put it on a plane and read it back in LA. Plus, it all fits nicely into a big suitcase. Furthermore, even for smaller amounts of data (say 10-20TB a day) the cost of a tape drive and an next day delivery is going to be significantly less than the Gbit/sec or so of bandwidth required to ship a similar amount of data in most places in the US.
fortran code usually still seems to give faster executables than comparable C++ code on numerical applications
I don't think this is true anymore. C++ is pretty much the only language that has BLAS libraries that can actually beat the fortran ones. The latest C++ template libraries are using SSE/etc vector intrinsics and are capable of meeting if not exceeding the fortran performance for many applications.
But, if you have a bunch of code in fortran, its probably not worth the trouble to convert it.
Someone who is skilled in C can pick up Perl and C++ quite quickly
Spoken like someone who hasn't actually programmed in C++. I was very good in object pascal, C, and quite a number of other languages when I first started programming in C++.
Lets just say that "quickly" isn't a description I would have for myself or anyone I've ever met when it comes to how long it takes to "pick up C++". Sure, you might be able to write some code in C++ after a couple days, but that code probably isn't worth the price of the storage your using for it. I would guess it takes upwards of 3 or 4 years before a C++ programer is good enough not to be making absolutely stupid mistakes on a daily basis. This is especially true for Java programmers who waltz in and think they know how to program in C++ because the syntax is so similar. The results are usually buggy garbage that compiles but won't run for long.
Heck, after nearly 20 years, I still look at code I write on a daily basis and go WTF was I thinking yesterday when I decided to do that! Or my co workers do it for me during code reviews.
Then explain why would you use C++ instead of C? If you are just going to cut it down to things C does? If you are going to do things C can't, then you will have the performance penalities. Your statments are not really valid.
Because there are cases where you are manually creating a construct in C, that is handled automatically by the compiler in C++.
Take virtual methods for example. The linux kernel is chuck full of structures filled with function pointers. That is a virtual method in C++, except in C++ you don't have to worry at runtime if the function your calling is NULL because the compiler assured that during compile time. This allows micro-optimization, and more natural error handling. Plus, the syntax is standard, so that every 3rd driver writer isn't creating their own version of the same thing.
Then there is generic data structure management. The kernel is full of macros for RB trees, linked lists/etc. Using templates for this allows better micro optimization without the programmer having to get involved.
There are a lot of reasons, and most of the negatives can be answered with, a the simple statement, don't use that feature.
The people at itworld are less IBM literate than the WSJ, because they keep repeating "POWER mainframe".
Repeat after me, POWER is _NOT_ IBM's mainframe line. The mainframe line is the zSeries and runs on proprietary processors clocked at 5.5Ghz. POWER processors are in the pSeries and iSeries machines.
Now, that said, in many ways the high end pSeries stuff is better than the mainframe hardware, but in no way is it considered "mainframe" grade to the IBM sales guys.
All that said, RHEL and SLES both run on pSeries and zSeries machines.
With PC gaming its really important that the graphics drivers are easy to upgrade.
I recently purchased the humble bundle and tried running a couple of the games on a linux install on a recently purchased laptop. While the distro I was running was supported by the games in question. The drivers needed are apparently new enough they didn't make it into the most recent version of that distribution. So, instead what I got was a GL SL v4 system where the shaders didn't work well enough to actually play the games in question.
Uh, what a PITA. All the games just worked in windows 8 on the same machine, but the install process to upgrade the intel HD 4000 graphics drivers is basically dependent on the distribution maintainer. The process for doing it by hand quickly unmotivated me enough to give up. I don't need to waste 4-5 hours recompiling 1/2 of the linux graphics stack to just upgrade my drivers. This from a guy who maintains large parts of an embedded linux system at work.
I can't really imagine how anyone really runs linux who isn't a kernel hacker. I had to manually hack the blutooth driver on that machine, write a custom script to disable the touchpad, and only after a few days discovered how to control the LCD brightness because the built in KDE/etc tools weren't working. All this on a laptop that actually has worked better in linux than any I've previously owned/used.
Bottom line, linux needs to create a proper driver ABI and dump the kernel symbol versions for something more like C++ name mangling. Then the drivers will only fail to load if the kernel API's actually changed rather than because someone happened to recompile the given kernel version on a different machine.
That is not 100% accurate because privoxy can strip the google analytics calls out of the javascript in the source pages. It can also degrade https to http for questionable domains.
And it can also blacklist whole domains SSL or not, so i don't think there are any cases where you get more functionality with a DNS blacklist.
Privoxy can remove a lot more than just ads served from a given domain/server. It can remove ads served by the same domain/server as the source website, as well as a number of other features that make it pretty nice for speeding up browsing on devices that don't have ad/javascript blockers.