Of course, it sounds pretty wonky, although there are some precedents. For instance, that new Gnome mailing agent (can't think of the name offhand) doesn't have actual folders for your emails. Rather, all emails are stored in a single monolithic database, and virtual folders are constructed using queries. Even if you just wanted "Unread mail" and "Already read mail", you could construct that with queries.
Done properly, this behaves EXACTLY as he was saying -- the "folders" automatically "pull emails" into themselves. At least, that's how it might look to a user. Particularly if the user changes their folder parameters and all of the emails automatically appear where they need to be. The physical layout doesn't require the emails to actually move. Rather, the database merely processes the query when the virtual folder is opened.
It's a very powerful model, and a data-store structured in this manner would be very useful. I stopped short of saying filesystem, because I still feel a filesystem, at least in the sense we have today, should not be that heavily abstracted. Rather, if the filesystem is not an appropriate level of abstraction for a user's data, then a coherent software layer should construct one above that. Sorta like how your accelerator pedal and shifter on an automatic don't directly move the throttle or shift gears -- they state the intentions of the driver and the car does what's required.
Uhm, [Alt-F] pops open the File menu. Not what you wanted. Consulting WordPad under Wine, it appears [Ctrl-H] is what you wanted to press.
And, once you press [Enter], that'll get you though the first "find". If you want to "Replace All", you'll have to do a little more work, such as pressing [Alt-A].
Also, you forgot the 2-stroke penalties for the "[Control] reach" and "[Alt] reach". Both of them are "reaches" on PC keyboards because the Control key is off in the corner, rather than where it should be. And, both are more expensive than hitting [Esc].:-)
The NT kernel itself may be wonderful, but nobody's seen it since the Microsoft Backwards Compatiblity Dumptruck unloaded Win32 all over it.
This isn't meant to be a troll. Windows NT would be a far better platform if they'd just drop the "Windows is part of the core OS" part of it. Put it in user space completely.
(In other words, I agree with you, I'm just expounding.)
I read through this article, and I have to agree with these educators: Bringing the cutesy video-game world of Windows and the MTV-esqe Internet (not the meaty content that experienced surfers go for, but the eye candy kids will gravitate to naturally) would be little better than having kids watch cartoons all day in class.
I got a computer of my own for my eighth birthday. Prior to that, I had used other people's computers to program, both at school and at friends' houses. I learned quite a lot on that machine, because it was a machine that did little on its own. It was raw clay, and I got to learn how to sculpt. How could you deny that that's valuable to a child?
Sure, there were game cartridges, and yes, I played them. (Moon Patrol anyone?) But kids have N64 or Dreamcast or PS2 or whatever nowadays, and so don't need the computer for that. Most of the value I derived from my computer was learning how to make it do things. It was like a box of Legos, only the building blocks were program statements and the structures I built were on a TV screen.
Today's computers aren't like that. Rather, they're like TV. Force feed eye candy. They exist for "wow" and "fluff." I personally had started falling into that trap in the PC world. I got pulled out of that trap when I went to college and learned Unix. Now, whenever I go to use a PC running Windows, I feel like I'm watching MTV or something. It's all so uselessly flashy and relatively devoid of content compared to its volume.
It's really sad.
I intend to keep my Apple ][e's, Commodore 64's, TI-99/4A's, and so on, to give my kids machines to learn on. When they're old enough, I'll give them logins on my Linux network and start teaching them C or some other structured language, before BASIC's brain-rot sets in too heavily -- you're ok if you catch them by puberty.
Sitting a kid in front of a web browser does not teach computing. Showing a kid how to make the computer do things it's not already trained to do (ie. program) opens the door for true creative exploration.
No comments about posture though... (as I slouch heavily into my chair).
FUD: Term used to describe misinformation / disinformation that's intended to strike fear, uncertainty and doubt within the target with respect to some competitor or adversary. For instance, Microsoft spreads FUD about Linux, and rabid myopic Linux zealots do the same in reverse.
F.U.D. is related to ad hominem attacks, only they're directed at objects, not people, and the focus is on distorting the facts rather than just diverting from revelant arguments.
I paid top dollar for a high-end Gateway box (G6-300 in 1997), and I got stuck with a Gateway OEM rev of an otherwise standard Intel mobo. The result is that I can't flash the BIOS to recognize disks larger than 8GB. Sure, Linux copes with it fine, but I can't boot even an MS-DOS floppy, let alone Windows, if I have IDE enabled and my 17GB drive plugged in.
That's NOT user error, except possibly the error of buying a machine whose mobo has a dead-end upgrade path.
Heck, if my 6 & 9 year old kids can figure it out, it can't be that hard.
Reminds me of a quote I once heard: "A child of age 5 could understand this... Quick, fetch me a child of age 5!"
Seriously, if your kids can understand it, that's NOT a good indicator that other adults will understand it. I know -- I used to be that "6 year old whiz" myself about 18 years ago, and none of the adults even tried understand any of the stuff I was learning.
You don't have to be. Anime in Japanese with subtitles is often far better than an overdub, since you don't lose the expressiveness that was in the original voice acting. Besides, the overdubs I've seen just suck -- not that I'm a huge anime freak or anything.
I guess this goes to show yet another way that region-coding can screw customers in multiple markets. In this case, there is obviously demand in both Region 1 and Region 2 markets, but since the folks that serve Region 2 aren't ready to ship their copies, the folks in Region 1 have to suffer too.
This is a load of BS. Without region coding, everyone would be able to get the full edition when it was first pressed. The only reason not to serve the demand is to artificially raise prices by artificially limiting supply -- eg. gouging the consumer.
Region coding is just a thinly veiled collusive agreement between manufacturers to divide up the market ahead of time to reduce global competition and therefore artificially inflate prices. They claim its to better serve each market without damaging other markets, since each region can afford different prices, etc. etc. etc.. Sorry, but really that's saying "With the status quo, we think we'll make more dollars than if we allow the markets to level themselves and let natural market forces actually determine prices."
Who the f**k moderated this 100% valid and relevant question as a troll?
There are some good sites out there for keeping your Junkbuster block lists up to date. Although I can't vouch personally for the following, here's what my blocklist has to say: (I actually got this file from the second link below. The comments below are from the block-list's author.)
ack... I killed a <BR> somewhere, and I forgot a comma before INT_MAX. Ugh.
Slashdot needs to allow the <PRE> tag, and just limit areas w/in the <PRE> tag to 80 columns or so. That'd make posting code much easier. Either that, or implement a pseudo-tag which does all the hard work for you.
Filling a post with 's, <BR>'s and so on gets pretty old after awhile. Also, typing an informative, thought-out post in a teeny little 50x10 editor pane is just annoying! Can Slashdot please make the TEXTAREA size configurable on the user-preferences page?
Sorry, you lose. Those will give you a free life if you have EXACTLY 10000 points, but in many games, your score will not necessarily be EXACTLY 10000 points.
How about:
int extra_life(int score) { static int prev_score = 0; int extra = prev_score < 10000 && score >= 10000;
prev_score = score; return extra; }
Of course, in a real game, they have multiple 1ups at different score thresholds. So, you might instead do something like this:
int next_1up_tbl[] = { 10000, 30000, 50000, 70000 INT_MAX }; int next_1up = 0;
int extra_life(int score) { static int prev_score = 0; int extra = prev_score < next_1up_tbl[next_1up] && score >= next_1up_tbl[next_1up];
But then, we're being silly. After all, all of these games were programmed in Z80 or 6502 assembler. I'll leave it as an exercise for the reader how to do this in said assembly languages.
--Joe --
Re:Real link to Packard Bell article
on
1.21 Quickiewatts
·
· Score: 1
At 0.18um going to 0.15um? No f***ing way! The main reason (aside from cooling) that 386s and 486s are used on space flights and not anything newer is that the finer geometries are too sensitive to radiation. 386s and 486s have large enough transistors and wide enough wires to have a hope of being resistant to the radiation.
The problem is, as you shrink the transistors, the amount of charge required to flip a gate goes way down. They say when you get down to these finer geometries, you get to know your electrons personally.;-) (IIRC, at 0.07um, they figure you have around 100 electrons to flip a transistor.) The odds of a random alpha particle causing a calculation error in one of the newer chips is therefore far greater than it is in one of the older chips.
2.8.1 may be old, but it's the most recent "stable" release of the compiler. In fact, many still consider 2.7.2 to be the most recent "stable" version for various reasons, although many of these people consider that due to the fact that 2.8.1 breaks their (broken) code, not due to 2.8.1 being unstable.
Is this a Intel PC only concern? The only heatsink in here is on the Proc.
It may be. The PC mobo chipset has to provide all of the legacy support crap in a PC. Apple's been a little more successful, I think, in throwing off cruft in each spin of their machine, due to their tighter control of the platform.
Also, in the PC space, you have more rabid head-to-head competition between vendors, and most are competing directly on performance and cost, so they don't care if they burn a few extra Watts. The Mac can get away from that a bit since the competition within the Mac space is less cutthroat, and the competition between the Mac and PC spaces is an apples-to-oranges comparison (eg. you can't just post a 0.5 FPS difference on Quake 3 timedemo scores and knowingly point at the mobo chipset as the reason).
In the laptop space, this is of course less of an issue, although all of the legacy cruft still gets dragged along. Transmeta's software solution allows them to also virtualize much of the lesser-used legacy crap in software, thereby giving them even better bang/buck and bang/Watt ratios.
How much power could this draw, though, vis-a-vis the LCD or the DVD drive[?]
Not really sure. My wild-ass-guess is that the PCI stuff could burn as much as 50% as much power the CPU does in a laptop (much lower ratio in a desktop), but it's really just a WAG. How that factors into the overall system's equation is even more of a mystery.
You're right about the display being the really "sucky" element. I can't wait until low-power LCD goggles are available. (I know of someone who's working on such a toy...)
I must be an 'e-tard'. 2.8.1 does seem to have the newer Sparc backend, in direct contrast to my statement above. The short test I ran to check I forgot to compile with optimization, so obviously I saw bogus info. Moderate away.:-P
Oh, to reply to my own message, here are some interesting and worthwhile links related to GCC performance:
NullStone compiler optimization benchmarks. On this page, they give some comparisons between GCC and some other compilers.
The Stepanov Benchmark page at KAI. The Stepanov Benchmark measures C++ abstraction penalty. GCC sucks wind on this one (as do most C++ compilers), whereas Apples MrC compiler slices through the abstraction and gets a "perfect score" (eg. no penalty).
The GCC site contains many interesting tidbits, some of which I mentioned above. For instance, news on the Sparc backend details some of the issues I described above. Note, although that rewrite occurred in late 1998, I still see the older backend's behavior in gcc 2.8.1, which is what we have at work.
Slashdot's story on the Compaq Alpha compiler. Some good data down in the comments.
I personally find that GCC tends to be the same or better than the SparcWorks compiler on "integer" control codes, but it sucks wind on floating point codes. (GCC is also simply a much more forgiving compiler.) In most of the other anecdotal evidence I've heard about other platforms (Alpha, mainly), it's a similar story -- integer is ok, floating point sucks. Here on Slashdot, we've seen multiple stories about Compaq's Alpha compiler and math libraries and how they outperform GCC. GCC 2.95 and newer might be better given several recent developments, but I'm not in a position to test it presently.
Until recently, GCC lacked several optimizations, such as software pipelining and even accurate pipelining modeling for straight-line scheduling, so it did a poor job of keeping highly parallel pipes full. (Note that on x86, this isn't a problem since those CPUs have relative narrow pipes these days.) For instance, GCC's architecture description primitives weren't expressive enough to describe how to order Sparc instructions so as to generate a schedule that would issue four instructions per cycle. Many fixes have occurred as part of the EGCS / GCC 3.00 project that's now in progress, but that's not the mainstream GCC currently.
These days, vendor compilers are pretty good, and GCC is reasonable. When GCC 2.0 came out, GCC pulled ahead of many vendors, as I understand it, and in the meantime the vendors have caught up and/or have gotten ahead. I'm hoping that with GCC 3.0, GCC pulls ahead again.
PS. To give you an idea of how bad GCC is on some platforms, at least for awhile, the native build of the Alpha RC5 client was SLOWER than than the Windows version of the client running in FX32!. IIRC, the native version was a GCC build.
You're forgetting that the TM5400 also integrates some of the other components, such as one of the halves of the PCI bridge (North Bridge, IIRC). That too is a big power-eater.
Take a lid off a PC sometime and look for that other heatsink over the mobo chipset.
You seem to have a bone to pick with Al Viro personally. What, did he flame you publicly and now you need to get back at him? When I read the threads he participates in, I find he's occasionally abrasive, but somehow he and whoever he's talking with always seem to get down to business. Anyway.... Back to your argument:
because the VFS in its current state is really a "Virtual Ext2 Filesystem".
You keep saying this. It doesn't make a huge amount of sense in light of the facts though. Linux has a huge number of supported filesystems (I daresay as many or more than I've seen in any other popular OS). So already, there's a few leaks in your argument.
The VFS implements a generalized interface to a filesystem which supports Unix semantics, with some extensions to those semantics. Well, guess what, EXT2 happens to support that full set, and other filesystems support some subset. Does that mean VFS is inherently EXT2 biased? Sure. Is it biased that way in a bad way? I don't see how.
Is it the other filesystems' fault if they don't implement all of the EXT2 extensions? No.
Is it VFS's fault if the filesystems don't map perfectly onto Unix semantics? No.
Is it the VFS's fault if it doesn't support all possible extensions to Unix semantics? No.
I will grant you that there's still some heavy EXT2-specific (and other filesystem-specific) crude hanging around in VFS-internal structures, but if you check out the traffic on linux-fsdevel, work is continually underway to clean things up. But the argument that "filesystems need to look like EXT2 to be supported by VFS" is a non-starter to me, because if a filesystem looks vaguely Unix-like, it probably looks a lot like EXT2. (You know, concepts like "inodes" and "hard links" aren't the sole province of EXT2.)
The common journalling API is a good thing. This API is not meant to provide a "grand unified way to implement a journal", it's to provide a "grand unified way for the kernel to respond to a journalling filesystem." The latter is quite different. The VM needs to behave properly in the presence of a journalling FS, and Tweedie and the ReiserFS guys are actually working actively on solving these issues, Hans' outbursts notwithstanding.
I think the real problem is that we're all in violation of that old saying that those who enjoy sausage should not watch while it's being made. To extend that metaphor to Linux, we're watching while it's being made, and are complaining when it upsets our stomach.
What you fail to address, though, is the part of the napster debate I'm mystified by: where is it written that you have to have freakin' recorded music AT ALL? If you don't like the game the record companies are playing, don't play it at all. Why isn't "I gave up on music because the music industry sucks" considered to be a legitimate choice in this debate? It seems to be "the music industry sucks, therefore I'll steal from them" is the/only/ choice.
I certainly could play that game of self-denial, but honestly, I would rather not. Sure, I don't need to own a music collection or listen to music at all, but why be a martyr against a Big Business that doesn't give a sh*t?
It's sorta like saying "I don't like who's running for office, so I won't vote," only listening to music is far more enjoyable and far more difficult to put aside than, say, voting.
Yes, people have a sense of entitlement to music and other creative works. In fact, until recent years they were entitled to those works after a limited period of copyright expired. Nowadays, with the deep pocketed lobbyists pushing the copyright term to +infinity, we are being made into criminals for acts which were originally constitutionally protected.
You're damn right I have a sense of entitlement. I want my rights back, thank you.
Slashdot Mirror
Of course, it sounds pretty wonky, although there are some precedents. For instance, that new Gnome mailing agent (can't think of the name offhand) doesn't have actual folders for your emails. Rather, all emails are stored in a single monolithic database, and virtual folders are constructed using queries. Even if you just wanted "Unread mail" and "Already read mail", you could construct that with queries.
Done properly, this behaves EXACTLY as he was saying -- the "folders" automatically "pull emails" into themselves. At least, that's how it might look to a user. Particularly if the user changes their folder parameters and all of the emails automatically appear where they need to be. The physical layout doesn't require the emails to actually move. Rather, the database merely processes the query when the virtual folder is opened.
It's a very powerful model, and a data-store structured in this manner would be very useful. I stopped short of saying filesystem, because I still feel a filesystem, at least in the sense we have today, should not be that heavily abstracted. Rather, if the filesystem is not an appropriate level of abstraction for a user's data, then a coherent software layer should construct one above that. Sorta like how your accelerator pedal and shifter on an automatic don't directly move the throttle or shift gears -- they state the intentions of the driver and the car does what's required.
--Joe--
Uhm, [Alt-F] pops open the File menu. Not what you wanted. Consulting WordPad under Wine, it appears [Ctrl-H] is what you wanted to press.
And, once you press [Enter], that'll get you though the first "find". If you want to "Replace All", you'll have to do a little more work, such as pressing [Alt-A].Also, you forgot the 2-stroke penalties for the "[Control] reach" and "[Alt] reach". Both of them are "reaches" on PC keyboards because the Control key is off in the corner, rather than where it should be. And, both are more expensive than hitting [Esc]. :-)
--Joe--
You forgot the trailing 'g', as in:
[ESC]:%s/foo/bar/g[Enter]
Without the 'g', it'll only change the first instance on each line.
--Joe--
Or better, comma-separated value (CSV) format. But, from the sounds of things, they have fancy macros and such in this "timecard". Ick.
My employer's solution is to make it irrelevant -- we're all salaried, which means the company owns us.
--Joe--
The NT kernel itself may be wonderful, but nobody's seen it since the Microsoft Backwards Compatiblity Dumptruck unloaded Win32 all over it.
This isn't meant to be a troll. Windows NT would be a far better platform if they'd just drop the "Windows is part of the core OS" part of it. Put it in user space completely.
(In other words, I agree with you, I'm just expounding.)
--Joe--
I read through this article, and I have to agree with these educators: Bringing the cutesy video-game world of Windows and the MTV-esqe Internet (not the meaty content that experienced surfers go for, but the eye candy kids will gravitate to naturally) would be little better than having kids watch cartoons all day in class.
I got a computer of my own for my eighth birthday. Prior to that, I had used other people's computers to program, both at school and at friends' houses. I learned quite a lot on that machine, because it was a machine that did little on its own. It was raw clay, and I got to learn how to sculpt. How could you deny that that's valuable to a child?
Sure, there were game cartridges, and yes, I played them. (Moon Patrol anyone?) But kids have N64 or Dreamcast or PS2 or whatever nowadays, and so don't need the computer for that. Most of the value I derived from my computer was learning how to make it do things. It was like a box of Legos, only the building blocks were program statements and the structures I built were on a TV screen.
Today's computers aren't like that. Rather, they're like TV. Force feed eye candy. They exist for "wow" and "fluff." I personally had started falling into that trap in the PC world. I got pulled out of that trap when I went to college and learned Unix. Now, whenever I go to use a PC running Windows, I feel like I'm watching MTV or something. It's all so uselessly flashy and relatively devoid of content compared to its volume.
It's really sad.
I intend to keep my Apple ][e's, Commodore 64's, TI-99/4A's, and so on, to give my kids machines to learn on. When they're old enough, I'll give them logins on my Linux network and start teaching them C or some other structured language, before BASIC's brain-rot sets in too heavily -- you're ok if you catch them by puberty.
Sitting a kid in front of a web browser does not teach computing. Showing a kid how to make the computer do things it's not already trained to do (ie. program) opens the door for true creative exploration.
No comments about posture though... (as I slouch heavily into my chair).
--Joe--
FUD: Term used to describe misinformation / disinformation that's intended to strike fear, uncertainty and doubt within the target with respect to some competitor or adversary. For instance, Microsoft spreads FUD about Linux, and rabid myopic Linux zealots do the same in reverse.
F.U.D. is related to ad hominem attacks, only they're directed at objects, not people, and the focus is on distorting the facts rather than just diverting from revelant arguments.
--Joe--
I paid top dollar for a high-end Gateway box (G6-300 in 1997), and I got stuck with a Gateway OEM rev of an otherwise standard Intel mobo. The result is that I can't flash the BIOS to recognize disks larger than 8GB. Sure, Linux copes with it fine, but I can't boot even an MS-DOS floppy, let alone Windows, if I have IDE enabled and my 17GB drive plugged in.
That's NOT user error, except possibly the error of buying a machine whose mobo has a dead-end upgrade path.
--Joe--
But they still send question marks for quotes and ticks... :-)
--Joe--
Reminds me of a quote I once heard: "A child of age 5 could understand this ... Quick, fetch me a child of age 5!"
Seriously, if your kids can understand it, that's NOT a good indicator that other adults will understand it. I know -- I used to be that "6 year old whiz" myself about 18 years ago, and none of the adults even tried understand any of the stuff I was learning.
--Joe--
You don't have to be. Anime in Japanese with subtitles is often far better than an overdub, since you don't lose the expressiveness that was in the original voice acting. Besides, the overdubs I've seen just suck -- not that I'm a huge anime freak or anything.
--Joe--
I guess this goes to show yet another way that region-coding can screw customers in multiple markets. In this case, there is obviously demand in both Region 1 and Region 2 markets, but since the folks that serve Region 2 aren't ready to ship their copies, the folks in Region 1 have to suffer too.
This is a load of BS. Without region coding, everyone would be able to get the full edition when it was first pressed. The only reason not to serve the demand is to artificially raise prices by artificially limiting supply -- eg. gouging the consumer.
Region coding is just a thinly veiled collusive agreement between manufacturers to divide up the market ahead of time to reduce global competition and therefore artificially inflate prices. They claim its to better serve each market without damaging other markets, since each region can afford different prices, etc. etc. etc.. Sorry, but really that's saying "With the status quo, we think we'll make more dollars than if we allow the markets to level themselves and let natural market forces actually determine prices."
No sir, I don't like it.
--Joe--
Who the f**k moderated this 100% valid and relevant question as a troll?
There are some good sites out there for keeping your Junkbuster block lists up to date. Although I can't vouch personally for the following, here's what my blocklist has to say: (I actually got this file from the second link below. The comments below are from the block-list's author.)
# I got this from http://mind.learning.cs.cmu.edu/blockfile
# and changed it a little bit. Note that my junkbuster is compiled
# to understand full Posix regular expressions.
# Send suggestions to boldt (at) math.ucsb.edu.
# Home page: http://math-www.uni-paderborn.de/~axel/
# Other blockfiles are available elsewhere, try searching
# documents that mention "junkbuster" and are called "blocklist"
# altavista.digital.com/cgi-bin/query?pg=q&what=web
Hope that helps.
--Joe--
ack... I killed a <BR> somewhere, and I forgot a comma before INT_MAX. Ugh.
Slashdot needs to allow the <PRE> tag, and just limit areas w/in the <PRE> tag to 80 columns or so. That'd make posting code much easier. Either that, or implement a pseudo-tag which does all the hard work for you.
Filling a post with 's, <BR>'s and so on gets pretty old after awhile. Also, typing an informative, thought-out post in a teeny little 50x10 editor pane is just annoying! Can Slashdot please make the TEXTAREA size configurable on the user-preferences page?
--Joe--
Sorry, you lose. Those will give you a free life if you have EXACTLY 10000 points, but in many games, your score will not necessarily be EXACTLY 10000 points.
How about:
int extra_life(int score)
{ static int prev_score = 0;
int extra = prev_score < 10000 && score >= 10000;
prev_score = score;
return extra;
}
Of course, in a real game, they have multiple 1ups at different score thresholds. So, you might instead do something like this:
int next_1up_tbl[] = { 10000, 30000, 50000, 70000 INT_MAX };int next_1up = 0;
int extra_life(int score)
{ static int prev_score = 0;
int extra = prev_score < next_1up_tbl[next_1up] && score >= next_1up_tbl[next_1up];
next_1up+= extra; prev_score = score;
return extra;
}
But then, we're being silly. After all, all of these games were programmed in Z80 or 6502 assembler. I'll leave it as an exercise for the reader how to do this in said assembly languages.
--Joe--
Yeah, I was wondering how they picked him...
--Joe--
At 0.18um going to 0.15um? No f***ing way! The main reason (aside from cooling) that 386s and 486s are used on space flights and not anything newer is that the finer geometries are too sensitive to radiation. 386s and 486s have large enough transistors and wide enough wires to have a hope of being resistant to the radiation.
The problem is, as you shrink the transistors, the amount of charge required to flip a gate goes way down. They say when you get down to these finer geometries, you get to know your electrons personally. ;-) (IIRC, at 0.07um, they figure you have around 100 electrons to flip a transistor.) The odds of a random alpha particle causing a calculation error in one of the newer chips is therefore far greater than it is in one of the older chips.
--Joe--
2.8.1 may be old, but it's the most recent "stable" release of the compiler. In fact, many still consider 2.7.2 to be the most recent "stable" version for various reasons, although many of these people consider that due to the fact that 2.8.1 breaks their (broken) code, not due to 2.8.1 being unstable.
--Joe--
It may be. The PC mobo chipset has to provide all of the legacy support crap in a PC. Apple's been a little more successful, I think, in throwing off cruft in each spin of their machine, due to their tighter control of the platform.
Also, in the PC space, you have more rabid head-to-head competition between vendors, and most are competing directly on performance and cost, so they don't care if they burn a few extra Watts. The Mac can get away from that a bit since the competition within the Mac space is less cutthroat, and the competition between the Mac and PC spaces is an apples-to-oranges comparison (eg. you can't just post a 0.5 FPS difference on Quake 3 timedemo scores and knowingly point at the mobo chipset as the reason).
In the laptop space, this is of course less of an issue, although all of the legacy cruft still gets dragged along. Transmeta's software solution allows them to also virtualize much of the lesser-used legacy crap in software, thereby giving them even better bang/buck and bang/Watt ratios.
Not really sure. My wild-ass-guess is that the PCI stuff could burn as much as 50% as much power the CPU does in a laptop (much lower ratio in a desktop), but it's really just a WAG. How that factors into the overall system's equation is even more of a mystery.
You're right about the display being the really "sucky" element. I can't wait until low-power LCD goggles are available. (I know of someone who's working on such a toy...)
--Joe--
I must be an 'e-tard'. 2.8.1 does seem to have the newer Sparc backend, in direct contrast to my statement above. The short test I ran to check I forgot to compile with optimization, so obviously I saw bogus info. Moderate away. :-P
--Joe--
Oh, to reply to my own message, here are some interesting and worthwhile links related to GCC performance:
- NullStone compiler optimization benchmarks. On this page, they give some comparisons between GCC and some other compilers.
- The Stepanov Benchmark page at KAI. The Stepanov Benchmark measures C++ abstraction penalty. GCC sucks wind on this one (as do most C++ compilers), whereas Apples MrC compiler slices through the abstraction and gets a "perfect score" (eg. no penalty).
- The GCC site contains many interesting tidbits, some of which I mentioned above. For instance, news on the Sparc backend details some of the issues I described above. Note, although that rewrite occurred in late 1998, I still see the older backend's behavior in gcc 2.8.1, which is what we have at work.
- Slashdot's story on the Compaq Alpha compiler. Some good data down in the comments.
--Joe--
I personally find that GCC tends to be the same or better than the SparcWorks compiler on "integer" control codes, but it sucks wind on floating point codes. (GCC is also simply a much more forgiving compiler.) In most of the other anecdotal evidence I've heard about other platforms (Alpha, mainly), it's a similar story -- integer is ok, floating point sucks. Here on Slashdot, we've seen multiple stories about Compaq's Alpha compiler and math libraries and how they outperform GCC. GCC 2.95 and newer might be better given several recent developments, but I'm not in a position to test it presently.
Until recently, GCC lacked several optimizations, such as software pipelining and even accurate pipelining modeling for straight-line scheduling, so it did a poor job of keeping highly parallel pipes full. (Note that on x86, this isn't a problem since those CPUs have relative narrow pipes these days.) For instance, GCC's architecture description primitives weren't expressive enough to describe how to order Sparc instructions so as to generate a schedule that would issue four instructions per cycle. Many fixes have occurred as part of the EGCS / GCC 3.00 project that's now in progress, but that's not the mainstream GCC currently.
These days, vendor compilers are pretty good, and GCC is reasonable. When GCC 2.0 came out, GCC pulled ahead of many vendors, as I understand it, and in the meantime the vendors have caught up and/or have gotten ahead. I'm hoping that with GCC 3.0, GCC pulls ahead again.
PS. To give you an idea of how bad GCC is on some platforms, at least for awhile, the native build of the Alpha RC5 client was SLOWER than than the Windows version of the client running in FX32!. IIRC, the native version was a GCC build.
--Joe--
You're forgetting that the TM5400 also integrates some of the other components, such as one of the halves of the PCI bridge (North Bridge, IIRC). That too is a big power-eater.
Take a lid off a PC sometime and look for that other heatsink over the mobo chipset.
--Joe--
You seem to have a bone to pick with Al Viro personally. What, did he flame you publicly and now you need to get back at him? When I read the threads he participates in, I find he's occasionally abrasive, but somehow he and whoever he's talking with always seem to get down to business. Anyway.... Back to your argument:
You keep saying this. It doesn't make a huge amount of sense in light of the facts though. Linux has a huge number of supported filesystems (I daresay as many or more than I've seen in any other popular OS). So already, there's a few leaks in your argument.
The VFS implements a generalized interface to a filesystem which supports Unix semantics, with some extensions to those semantics. Well, guess what, EXT2 happens to support that full set, and other filesystems support some subset. Does that mean VFS is inherently EXT2 biased? Sure. Is it biased that way in a bad way? I don't see how.
I will grant you that there's still some heavy EXT2-specific (and other filesystem-specific) crude hanging around in VFS-internal structures, but if you check out the traffic on linux-fsdevel, work is continually underway to clean things up. But the argument that "filesystems need to look like EXT2 to be supported by VFS" is a non-starter to me, because if a filesystem looks vaguely Unix-like, it probably looks a lot like EXT2. (You know, concepts like "inodes" and "hard links" aren't the sole province of EXT2.)
The common journalling API is a good thing. This API is not meant to provide a "grand unified way to implement a journal", it's to provide a "grand unified way for the kernel to respond to a journalling filesystem." The latter is quite different. The VM needs to behave properly in the presence of a journalling FS, and Tweedie and the ReiserFS guys are actually working actively on solving these issues, Hans' outbursts notwithstanding.
I think the real problem is that we're all in violation of that old saying that those who enjoy sausage should not watch while it's being made. To extend that metaphor to Linux, we're watching while it's being made, and are complaining when it upsets our stomach.
--Joe--
I certainly could play that game of self-denial, but honestly, I would rather not. Sure, I don't need to own a music collection or listen to music at all, but why be a martyr against a Big Business that doesn't give a sh*t?
It's sorta like saying "I don't like who's running for office, so I won't vote," only listening to music is far more enjoyable and far more difficult to put aside than, say, voting.
Yes, people have a sense of entitlement to music and other creative works. In fact, until recent years they were entitled to those works after a limited period of copyright expired. Nowadays, with the deep pocketed lobbyists pushing the copyright term to +infinity, we are being made into criminals for acts which were originally constitutionally protected.
You're damn right I have a sense of entitlement. I want my rights back, thank you.
--Joe--