| |
| |
| |
|
Page: 1 2 3 4 5
Comments:
<0> generally somewhere around 1% or 2% is where most kernels are
<0> ie, 64 bytes per 4K page, or something
<0> seems high, but sometimes you gotta do what you gotta do
<1> I'm sure I can shave it down a bit
<2> hi
<1> that's on a 32 bit system right? If I'm working on a 64 bit system, would that be any higher since the pointers are twice as big?
<3> call me stupid, but is ther actually any point to 64 bit except for bigger address space?
<1> you can work with bigger numbers, more data at once
<1> good programs will use the registers efficiently and be able to keep more data in the cpu at once, so you don't have to hit up memory as much
<3> ah, so it actually does improve performance
<3> and isn't just a way to make everyone buy new computers
<4> x86_64 also increases number of registers [not just their size]
<4> though this is x86-specific
<3> more registers is good
<3> 4 GP regs is far too few
<4> 7 GP
<4> unless you
<4> 're one of those crazy people like me who don't need stack pointer sometimes
<3> what's with this can't include mp3 playing software in linux distros any more thing
<3> is the gstreamer mp3 plugin illegal or something?
<1> fraunhauffer labs holds patents
<1> but there's some exception for non-profit stuff... but still it's not kosher since it's not open source
<1> or something like that.
<3> ah
<5> i like about> 30 gp
<5> + a movalbe register-frame
<3> ronny: what CPU has that?
<1> ppc, I think.
<3> nice
<5> hmm
<3> the only no x86 chips i've done ***embly on are MIPS r5000 and sparc
<5> and sparcs got even more afair
<3> don't rember much about them
<4> movable register-frame? sparc it is
<1> what was the r5000 in?
<4> also, there's ia64, but it has 128 registers
<3> ps2
<4> and LOADS of fun stuff :)
<5> sparcs got 64 gp regs 32 x 32 bit and 32x 64 bit
<4> sparc has 32 GPRS, one of which is fixed to 0
<5> mwk: ia64 is kinda neat - but amd was man and made another x86 mess
<4> 32-bit each on 32-bit sparcen, 64-bit each on 64-bit sparcen
<5> oh - wait
<4> then, there are floating point registers, 32-bit each
<5> amn - im on the fpu spec
<4> there are 32 of those on 32-bit sparcen, 64 of them on 64-bit sparcen
<5> nice they got 16 128 bit reg for the fpu
<4> plus, there are control registers and some special ones
<4> 16*128, 32*64, or 32*32, depending on instruction.
<3> so a core duo 2 would be ia64?
<4> though 128-bit fp is actually software-supported only iirc
<1> no, ia64 is itanium
<4> no, core duo 2 is x86_64
<4> ia64 == itanium
<4> aka itanic
<3> does itanium actually exist yet or is it still being developed?
<4> it exists. sort of.
<1> they're onto itanium2
<1> it totally exists
<4> itanium2 is already done. they're onto itanium3.
<3> I haven't seen any itanimum machines in the computer shops
<1> they're giving it another go around? You'd think hp/intel had learned their lesson...
<4> it exists, but noone gives a **** because it's non-x86 :(
<1> it's mostly targeted towards HPC
<4> indeed.
<3> oh right so you can't run windows or linux and stuff on it?
<4> linux is ported to ia64
<1> linux runs on a toaster
<3> o.O toaster linux
<4> windows... some version runs on ia64 IIRC
<4> yeah, and on dead badgers
<1> yeah, there's a windows nt port or something.
<1> I forget
<3> not that compute cluster edition?
<3> that costs more than the hardware it runs on?
<4> but it's some highendish special version
<1> yeah, I dunno if ms even supports it anymore
<4> anyway, /me heads off to bed. night.
<1> urh... hmmm.... looks like I was a bit off... accounting takes up 0.3% worst case on x86
<1> I'm happy
<3> yay for things working
<1> what sort of work have you done on the ps2?
<0> the console or the port?
<6> console
<6> :P
<1> oh, sorry, that was at serishema
<3> sp1nm0nkey: used it as a vector processor
<0> GLOOP I SAY
<1> http://www.robmonroe.net/upload/2007/1/24/elephantintheway.jpg
<3> lmao
<0> also I have a itanium box
<0> first gen workstation. it's pretty nifty
<3> does the elephant run linux?
<1> yes.
<7> what the hell?
<3> just a random thought
<3> someone said linux runs on a toaster
<0> that's the old netbsd trick
<3> so i was wondering if it ran on elephatns too
<0> linux folks need to come up with a different example
<3> what running netbsd on a toaster?
<7> (my wth was at the elephant)
<7> I mean, the drawing
<3> looks like someone, a bored someone's homework to me
<1> looks more like a test to me
<7> yup, there's a grade, gotta be a test
<3> ah
<3> i won't try following people on #fedora's suggestions again
<8> \join
<1> geist: When you said 1-2%, is that including page tables/directories?
<0> depends on how many virtual pages you have
<0> but i dont count those, since you have no control over that (for the most part)
<1> mmm, okay, so looks like I'm pretty much spot on now, It's 1.9% on 64 bit, 0.97% on 32 bit worst case
<1> geist: On a typical extent filesystem, what are the chances of being to expand a closed file contiguously?
<1> *being able to
<1> or when you close the file do you push the new extent onto the bottom of the free extent list so you're less likely to allocate an extent that's directly after a newly closed file?
<9> hi...i want to know...wich OS you are using to osdev??
<10> PoopOS
<10> err linux
<1> linux
<9> but you test your OS in the development machine or in anothe machine??
<1> VMs are heavenly
<9> ok
<1> I try and do as much development outside of the OS's codebase, since debugging inside a running operating system can be a bit difficult
<4> 'd better test on real machines too, since VMs ignore many problems
<4> eg. qemu implementation of PAE that ignores reserved bits, while real CPU chokes dead on them
<9> I read in the Mega Tokyo FAQ about hard disk crashing
<4> about whatnow?
<9> and i'm afraid now to test and happen errors with my HD
<1> yes, you can ruin your partition table/mbr/other partitions if you're testing block device driver/fs code
<4> so don't. your problem.
<1> test inside a VM, and if you want to do that sort of testing on a real box, replace your drive with a CF->ide converter and flash your OS onto a CF card
<4> why not test on semi-expendable machine that you can restore from backup quickly?
<4> or test from floppy/usb-stick with HD removed, if you're paranoid
<9> xD
<9> ok
<1> 400 MHz pII boxen are so cheap on ebay
<9> I have a PII 400
<9> but I have only one monitor
<1> are you using it for anything?
<1> find a KVM or use a serial cable
<4> so? why would you ever want to read from two monitors at once?
<9> I will buy a serial cable
<0> good idea
<4> yea, that's another good thing
<0> using a serial port to debug is a really good idea for initial development
<0> since once you get the port set up (a few outs) you can write a char with a single out, regardless of address space, even from ***embly
<0> and on other arches, it's typically more or less the same, though the uart is usually memory mapped
<11> howdy
<1> geist: Huh? even if you're in pmode the address is the same?
Return to
#osdev
or
Go to some related
logs:
unable to find the GNOME libraries
#asm
fputcontent
#ubuntu
3gpwiz gentoo
grub x975 msi
Linux aplication install SANE tar
setup l2tp fedora core 5
sed + extract part of string
elhaz mp3.
|
|
