>I don't know if the kernel supports more than two IDE/EIDE controllers.
I think recent kernels allow 4 controllers "out of the box", and can be made to support 8 with very little work...
>If you have that many devices, frankly: >GET YOURSELF A SCSI SETUP! STOP THE EIDE INSANITY!
Agreed. I used to have more than 2 EIDE controllers, but it was 'cause I had an ATAPI streamer that didn't cooperate very well with any other device, and that had to install with its own dumb controller as a last resort...:(
>IDE I/O is very CPU intensive, you'll find it to be faster to go SCSI.
That used to be true, but DMA modes aren't CPU intensive anymore. SCSI still has advantages over IDE (being able to hang lots of devices out of a single controller, as you mentionned, is one of them), but CPU usage advantage is a thing of the past.
Actually TrueType fonts should be "enhanced" with "hints" that allow them to be rendered on low res devices and still be readable. I guess the problem is that very few fonts designers take the time to add those "hints", and that all but very standard fonts are ugly at some resolutions.
Mmmh, if IE for Linux is just as bloated and slow as IE for the other *nix OSes (Solaris and HPUX right now, iirc), then using IE on Linux will be a good way to be disgusted of Linux from the very start. And I really don't see how it could be otherwise; Starting a native *nix version (no more megs of "win32 compatilibity libs") from scratch would cost too much of development time, for a low priority platform... IE is fast because it is integrated (you pay the loading time when the GUI start, where it is less noticable), so if you want IE, you'll have to use Windows or be very disappointed by performance. IE was not designed for portability, mind you.
Personally, I'm don't feel a huge need to replace Netscape bugs with Microsoft ones on my desktop. IE parsers must be horrible cludges (ever tried an URL with "§=" as a substring ? just an example of what shouldn't happen). I hate how IE promotes the use of tables with fixed sizes everywhere. Can anyone write pages "best viewed in 680x800" so it looks good on my browser ?). And so on.
>After it found it, the install went >relatively smoothly, except for the fact that all >the packages were just listed by a single letter, >with no explanations, and being a Linux novice, >"D" did not mean anything to me.
It's not the problem of being a novice, it's a problem with reading the docs. Last time I installed a Slackware distro ('95), it came with a small "book" that explained, among things, what each package set was about (D is development tools, iirc). I must add that Slackware is not the easiest distro to start with, its motto is KISS (Keep It Simple Stupid), so you have to have some hacking abilities to enjoy it.
The patch does not violate any locking requirements in the kernel, because the kerne lock could have been dropped at any time anyways when the copy_from_user slept to swap a page in. (I'm not sure if running a published benchmark with such a patch is fair though. On the other hand Microsoft did some many hidden changes in their service packs that probably everything is allowed;)
Another problem is that Linux 2.2 per default uses only 1GB of memory. This can be tuned by changing the PAGE_OFFSET constant in include/asm/page.h and arch/i386/vmlinux.lds from 0xc0000000 to 0x80000000 or so and recompiling (the tradeoff is that that it limits the per process virtual memory to ~1.8GB, but increases the overall physical memory that can be mapped). [...] Message-ID: Date: Tue, 18 May 1999 19:44:25 +0200 From: Juergen Schmidt To: Andi Kleen CC: linux-kernel@vger.rutgers.edu Subject: Re: Bad apache perfomance wtih linux SMP References:
Andi Kleen wrote: > One culprit is most likely that the data copy for TCP sending runs completely > serialized. This can be fixed by doing replacing the > > skb->csum = csum_and_copy_from_user(from, > skb_put(skb, copy), copy, 0, &err); > > in tcp.c:tcp_do_sendmsg with > > unlock_kernel(); > skb->csum = csum_and_copy_from_user(from, > skb_put(skb, copy), copy, 0, &err); > lock_kernel();
Bingo !!!
This raised performance from 270 rps to 802 rps when 64 clients were pulling a 4k HTML-page. Single CPU perfomance lies by 890 rps -- but the new numbers are just from a very short run. (BTW: NT/IIS on 4 CPUs deliver 840 rps:-)
Or with apaches ab:
ab -c 8 -t 120 127.0.0.1/4k.html
produces:
2.2.8 4 CPUs: 350.95 2.2.8 4 CPUs with patch: 1334.19 2.2.8 no SMP: 1540.22 [...]
If a two lines patch fixed 95% of the problem, I suppose a full de-serialization of the tcp stack (and vfs/filesystems ? please) would allow Linux to beat NT on that test... This is a much more complex task, but that kind of news help me keeping the faith in Linux kernel developers... Kernel hackers rocks !!
Journaling is not about not loosing data (that would requires no-delay write to disk, etc), but about being able to return to a consistent state fast. You may loose several seconds of updates (or more), but you are then guarantee to have all your files as they were supposed to be (from system and applications point-of-vue) at that time, not a "hopefully rightly fixes fs" like what most *fsck can do.
Netscape scrolling repaint is plain trash too. And ImageMagic repaint is all but optimized... Try moving a window in "OpaqueMove" mode over it...:( Or is it an X problem ? Off topic anyway.
Re:Relax and enjoy school and Life. Work can wait.
on
The Dark Side of IT
·
· Score: 1
No wonder why you find so much unmaintainable code all around. People think that because they can write code they're done.
Tried with a PII 350 + 192Mb + Voodoo2 CPU usage was between 45% and 60% most of the time, with peaks at 100% when entering rooms/enabling new visual effects (like "quake power" effect). q3 also used 85Mb during play, so it may be playable with a voodoo2 and enough ram on your box.
You missed this interview with Linus some time ago where he said (sort of late answer to Tanenbaum's "Linux is obsolete"), mainly, that microkernels were the panacea for researchers only, but that in real world you could have most of microkernel's benefits without the overhead with a monolithic kernel and good code design...
That it's not surprizing that those who get caught are mostly dumb, since smart ones usually manage to stay unseen...
So "All those hackers [examples of lately caught hand-in-sack people there] are dumb" does not prove anything, nor is helpful in evaluating the damage done by crackers.
I read it as a "nice" way to downplay the involvment of Word macros in industrial piracy;(
Most of the time, it's the netscape's event queing, that uses a mere Unix pipe, that is overflowing (and blocking)... Stanislav Meduna posted a workaround on the linux-kernel mailing list the 23rd of January, look for the "nspipepatch".
It solved 95% of my problems (Communicator still "freezes" at times 'till I move the mouse, but it occurs much less often now; It still randomly crashes in java/javascript stuff, though).
> Hmmm... I thought that at some point Intel made > it possible to do 36 bit memory addressing on the > i686 processors. This would allow access to 8 GB > of memory. Of course I could totaly be off my > rocker.
Yes, but that involves gross hacks, thanks to yet-another-f*cked-up-design-from-youknowwho.
NT mm engineers screamed "yuck" loudly, but will add support for that because they will be paid for the job.
Linus (and "official kernel team") won't mangle Linux mm to support this. 3rd parties may, but anyone needing such amounts of memory should use clean 64 bit architectures anyway.
>...as a model... From what I've read those were "private" pictures, and I interpret that as pictures of the person, not the model. > (not to mention wife of Johnny Hallyday, no?) Daughter, afaik. >...is a public figure, and as such does not necessarily "own" her own image. Maybe it's just me, but I think one person can have many images; She's got a professionnal image (owned by her model agency ?), but that doesn't deny her from having the right for privacy, that is the right to protect her intimate image.
The case is not very clear, but again, I don't think it's that _specific_ case that is really important.
Yeah, right. Not to mention off topic messages:) Estelle (nonstanding what you think of her) has the *right* to sue to protect her "image". And someone *has* to be responsible for content. Just imagine another case, child porn or whatever. Do you think it should be allowed, since none is responsible ? Even libertarians among us should say no. It seems that the problem is that you could open an account on the server anonymously. So since justice couldn't sue the poster (who is the real guilty man in this history if you ask me), they sue the hosting site. Morality: now in France, you can open free sites, people can say what they want as far as they take their responsabilities, but you cannot open anonymous free sites if you don't plan de revue content before it goes online. (And 400 000 FF is a huge ammount of money for a free site; I understand the motive, but it's one or two orders of magnitude too much imho; Valentin doesn't look like a bad guy after all, just a bit too naive).
Slashdot Mirror
...As if Sun control over Java protected it from forking...
I think recent kernels allow 4 controllers "out of the box", and can be made to support 8 with very little work...
>If you have that many devices, frankly:
>GET YOURSELF A SCSI SETUP! STOP THE EIDE INSANITY!
Agreed. I used to have more than 2 EIDE controllers, but it was 'cause I had an ATAPI streamer that didn't cooperate very well with any other device, and that had to install with its own dumb controller as a last resort... :(
>IDE I/O is very CPU intensive, you'll find it to be faster to go SCSI.
That used to be true, but DMA modes aren't CPU intensive anymore. SCSI still has advantages over IDE (being able to hang lots of devices out of a single controller, as you mentionned, is one of them), but CPU usage advantage is a thing of the past.
Fine plan. No, go back to reality and Chaos Theory... Boom! :)
>But remember, if your mind is _too_ open, then everything falls out. (Or any trash can fall in).
Or other people may get in and look around. That'd be more "on topic" :)
Actually TrueType fonts should be "enhanced" with "hints" that allow them to be rendered on low res devices and still be readable.
I guess the problem is that very few fonts designers take the time to add those "hints", and that all but very standard fonts are ugly at some resolutions.
And I really don't see how it could be otherwise; Starting a native *nix version (no more megs of "win32 compatilibity libs") from scratch would cost too much of development time, for a low priority platform...
IE is fast because it is integrated (you pay the loading time when the GUI start, where it is less noticable), so if you want IE, you'll have to use Windows or be very disappointed by performance. IE was not designed for portability, mind you.
Personally, I'm don't feel a huge need to replace Netscape bugs with Microsoft ones on my desktop. IE parsers must be horrible cludges (ever tried an URL with "§=" as a substring ? just an example of what shouldn't happen). I hate how IE promotes the use of tables with fixed sizes everywhere. Can anyone write pages "best viewed in 680x800" so it looks good on my browser ?). And so on.
No thanks.
fyi, cdrom.com box *is* a heavyly tweaked system (userland for sure, kernel almost certainly).
>relatively smoothly, except for the fact that all
>the packages were just listed by a single letter,
>with no explanations, and being a Linux novice,
>"D" did not mean anything to me.
It's not the problem of being a novice, it's a problem with reading the docs. Last time I installed a Slackware distro ('95), it came with a small "book" that explained, among things, what each package set was about (D is development tools, iirc).
I must add that Slackware is not the easiest distro to start with, its motto is KISS (Keep It Simple Stupid), so you have to have some hacking abilities to enjoy it.
[Taken from linux-kernel mailing-list]
;)
:-)
To: Juergen Schmidt
cc: linux-kernel@vger.rutgers.edu
Subject: Re: Bad apache perfomance wtih linux SMP
References:
From: Andi Kleen
Date: 18 May 1999 15:02:08 +0200
In-Reply-To: Juergen Schmidt's message of "Tue, 18 May 1999 05:23:42 GMT"
Message-ID:
Juergen Schmidt writes:
> Do you have any ideas, what's happening there?
> Or even better, how to fix this?
One culprit is most likely that the data copy for TCP sending runs completely
serialized. This can be fixed by doing replacing the
skb->csum = csum_and_copy_from_user(from,
skb_put(skb, copy), copy, 0, &err);
in tcp.c:tcp_do_sendmsg with
unlock_kernel();
skb->csum = csum_and_copy_from_user(from,
skb_put(skb, copy), copy, 0, &err);
lock_kernel();
The patch does not violate any locking requirements in the kernel, because
the kerne lock could have been dropped at any time anyways when the copy_from_user
slept to swap a page in.
(I'm not sure if running a published benchmark with such a patch is fair though.
On the other hand Microsoft did some many hidden changes in their service packs
that probably everything is allowed
Another problem is that Linux 2.2 per default uses only 1GB of memory. This can
be tuned by changing the PAGE_OFFSET constant in include/asm/page.h and
arch/i386/vmlinux.lds from 0xc0000000 to 0x80000000 or so and recompiling
(the tradeoff is that that it limits the per process virtual memory to ~1.8GB,
but increases the overall physical memory that can be mapped).
[...]
Message-ID:
Date: Tue, 18 May 1999 19:44:25 +0200
From: Juergen Schmidt
To: Andi Kleen
CC: linux-kernel@vger.rutgers.edu
Subject: Re: Bad apache perfomance wtih linux SMP
References:
Andi Kleen wrote:
> One culprit is most likely that the data copy for TCP sending runs completely
> serialized. This can be fixed by doing replacing the
>
> skb->csum = csum_and_copy_from_user(from,
> skb_put(skb, copy), copy, 0, &err);
>
> in tcp.c:tcp_do_sendmsg with
>
> unlock_kernel();
> skb->csum = csum_and_copy_from_user(from,
> skb_put(skb, copy), copy, 0, &err);
> lock_kernel();
Bingo !!!
This raised performance from 270 rps to 802 rps when 64 clients were
pulling a 4k HTML-page. Single CPU perfomance lies by 890 rps -- but the
new numbers are just from a very short run. (BTW: NT/IIS on 4 CPUs
deliver 840 rps
Or with apaches ab:
ab -c 8 -t 120 127.0.0.1/4k.html
produces:
2.2.8 4 CPUs: 350.95
2.2.8 4 CPUs with patch: 1334.19
2.2.8 no SMP: 1540.22
[...]
If a two lines patch fixed 95% of the problem, I suppose a full de-serialization of the tcp stack (and vfs/filesystems ? please) would allow Linux to beat NT on that test...
This is a much more complex task, but that kind of news help me keeping the faith in Linux kernel developers...
Kernel hackers rocks !!
Not to mention Hans Reiser's fs called (guess what) Reiserfs.
It wouldn't have journaling at first, but it is planned.
Journaling is not about not loosing data (that would requires no-delay write to disk, etc), but about being able to return to a consistent state fast. You may loose several seconds of updates (or more), but you are then guarantee to have all your files as they were supposed to be (from system and applications point-of-vue) at that time, not a "hopefully rightly fixes fs" like what most *fsck can do.
I know, my libc is getting old, but it works :)
Netscape scrolling repaint is plain trash too. :(
And ImageMagic repaint is all but optimized... Try moving a window in "OpaqueMove" mode over it...
Or is it an X problem ? Off topic anyway.
threads are much less problematic than select() actually.
Tried with a PII 350 + 192Mb + Voodoo2
CPU usage was between 45% and 60% most of the time, with peaks at 100% when entering rooms/enabling new visual effects (like "quake power" effect).
q3 also used 85Mb during play, so it may be playable with a voodoo2 and enough ram on your box.
nsisucks.com ?
You missed this interview with Linus some time ago where he said (sort of late answer to Tanenbaum's "Linux is obsolete"), mainly, that microkernels were the panacea for researchers only, but that in real world you could have most of microkernel's benefits without the overhead with a monolithic kernel and good code design...
So "All those hackers [examples of lately caught hand-in-sack people there] are dumb" does not prove anything, nor is helpful in evaluating the damage done by crackers.
I read it as a "nice" way to downplay the involvment of Word macros in industrial piracy ;(
All the crackers that have been caught have been caught.
Stanislav Meduna posted a workaround on the linux-kernel mailing list the 23rd of January, look for the "nspipepatch".
It solved 95% of my problems (Communicator still "freezes" at times 'till I move the mouse, but it occurs much less often now; It still randomly crashes in java/javascript stuff, though).
> Hmmm... I thought that at some point Intel made
> it possible to do 36 bit memory addressing on the
> i686 processors. This would allow access to 8 GB
> of memory. Of course I could totaly be off my
> rocker.
Yes, but that involves gross hacks, thanks to
yet-another-f*cked-up-design-from-youknowwho.
NT mm engineers screamed "yuck" loudly, but will
add support for that because they will be paid for
the job.
Linus (and "official kernel team") won't mangle
Linux mm to support this. 3rd parties may, but
anyone needing such amounts of memory should use
clean 64 bit architectures anyway.
From what I've read those were "private" pictures, and I interpret that as pictures of the person, not the model.
> (not to mention wife of Johnny Hallyday, no?)
Daughter, afaik.
>
Maybe it's just me, but I think one person can have many images; She's got a professionnal image (owned by her model agency ?), but that doesn't deny her from having the right for privacy, that is the right to protect her intimate image.
The case is not very clear, but again, I don't think it's that _specific_ case that is really important.
Are you trying to say that people flocked on Windows because MS-DOS was so bad ?
Well, you're right.
Yeah, right. Not to mention off topic messages :)
Estelle (nonstanding what you think of her) has the *right* to sue to protect her "image".
And someone *has* to be responsible for content. Just imagine another case, child porn or whatever. Do you think it should be allowed, since none is responsible ? Even libertarians among us should say no.
It seems that the problem is that you could open an account on the server anonymously. So since justice couldn't sue the poster (who is the real guilty man in this history if you ask me), they sue the hosting site.
Morality: now in France, you can open free sites, people can say what they want as far as they take their responsabilities, but you cannot open anonymous free sites if you don't plan de revue content before it goes online.
(And 400 000 FF is a huge ammount of money for a free site; I understand the motive, but it's one or two orders of magnitude too much imho; Valentin doesn't look like a bad guy after all, just a bit too naive).