Jihad needs scientists
in [Design]
|
From: MassiveProng on 26 Feb 2007 02:48 On Mon, 26 Feb 2007 02:51:35 +0000 (UTC), kensmith(a)green.rahul.net (Ken Smith) Gave us: >Oh no there's that post again. The tree structure has gotten cross >linked. I do hope the incremental and transaction log are good. It would >be a shame to lose all this valuable stuff. Your little pricks and prods will go unheard, I'm sure. That horse had blinders on a long time ago.
From: MassiveProng on 26 Feb 2007 02:51 On Mon, 26 Feb 2007 03:02:50 +0000 (UTC), kensmith(a)green.rahul.net (Ken Smith) Gave us: >I doubt it is at its end. I would not, on the other hand, be surprised to >see it replaced with a better one. The science is still evolving. Yes. ReiserFS is one hell of a good premise, even if it were to not last. The beat will go on... On a different note: I like XOSL boot manager too, but it hasn't been maintained for years. Maybe I should re-write it for today! It sure is a cool boot manager.
From: jmfbahciv on 26 Feb 2007 07:00 In article <ersfac$ui3$1(a)blue.rahul.net>, kensmith(a)green.rahul.net (Ken Smith) wrote: >In article <ers25a$8qk_002(a)s1016.apx1.sbo.ma.dialup.rcn.com>, > <jmfbahciv(a)aol.com> wrote: >>In article <erpp2s$c02$4(a)blue.rahul.net>, >> kensmith(a)green.rahul.net (Ken Smith) wrote: >[......] >>>>>If a CPU chip needs 1 hour to do a an add instruction, you can't make it >>>>>go faster by anything you code. Like I said it sets the upper limit on >>>>>the performance. >>>> >>>>Sigh! If an ADD takes an hour and the computation has to be done >>>>in less time, then you don't use the ADD instruction. You do >>>>the addition by hand. >>> >>>In other words: You need another CPU to do the operation. >> >>Not at all. You can arithmetic by hand. > >You will say anything to avoid admitting that you missed the point when I >said that the hardware is what sets the upper limit on the performance. I am addressing your comment. > >> >>> No amount of >>>fancy code on a machine that takes an hour per instruction will fix it. >>> >>>This is what I have been trying to explain to you about the hardware >>>setting the upper limit on performance. >> >>Sigh! The IBM 1620 had no arithmetic instructions. Arithmetic >>was done "by hand" by looking up table entries. > >..... and that set a limit on the performance didn't it. No. The architecture could not do arithmetic. Software increased the performance beyond the hardware's capabilities. /BAH
From: jmfbahciv on 26 Feb 2007 07:05 In article <45E1CD23.26249F55(a)hotmail.com>, Eeyore <rabbitsfriendsandrelations(a)hotmail.com> wrote: > > >Ken Smith wrote: > >> <jmfbahciv(a)aol.com> wrote: >> > kensmith(a)green.rahul.net (Ken Smith) wrote: >> >> <jmfbahciv(a)aol.com> wrote: >> >> >> >>>VM isn't swapping. VM allows the OS to manage smaller chunks >> >>>of memory rather than segments. >> >> >> >>That is completely and totally worng. "Virtual memory" means quite >> >>literally "memory that is not real". >> > >> >No. It is memory whose addressing is larger than available physical >> >memory. >> >> No, not only the addressing appears larger. The total memory appears to >> be more. Merely allowing an address space that is larger is merely >> address translation. You only get into virtual memory when it appears the >> programs as though the machine has more memory than there is physical RAM. >> This is exactly what I was telling you when I directed you to how the word >> "virtual" is defined. > >To the processor itself the VM should be transparent. It should 'look' and >behave like acres of RAM. A good example of where the such a task should be >offloaded from the CPU itself. You can also build an EXE, or set of EXEs, on a machine that has a smaller physical memory. Being able to do this was a watermark in the biz. /BAH
From: jmfbahciv on 26 Feb 2007 07:10
In article <ersjj1$ui3$9(a)blue.rahul.net>, kensmith(a)green.rahul.net (Ken Smith) wrote: >In article <45E1CD23.26249F55(a)hotmail.com>, >Eeyore <rabbitsfriendsandrelations(a)hotmail.com> wrote: >[....] >>> No, not only the addressing appears larger. The total memory appears to >>> be more. Merely allowing an address space that is larger is merely >>> address translation. You only get into virtual memory when it appears the >>> programs as though the machine has more memory than there is physical RAM. >>> This is exactly what I was telling you when I directed you to how the word >>> "virtual" is defined. >> >>To the processor itself the VM should be transparent. It should 'look' and >>behave like acres of RAM. A good example of where the such a task should be >>offloaded from the CPU itself. > >No, that isn't done. VM systems are also usually multitaskers. You could >create one that isn't but the rule is that they are. Here's how it the >operation breaks down in a multitask environment. > >- Running Task A >- Task A does a page fault on the real memory >- OS gets an interrupt >- Perhaps some checking is done here >- OS looks for the page to swap out Swap out from where? If the CPU architecture has write-through cache you don't have to move the contents of the page you need to remove in order to fetch the page that Task A needs from memory. >- Complex issue of priority on swapping skipped here. >- OS marks the outgoing page to be not usable >- OS starts swap actions going >- OS looks for a task that can run now >- OS remembers some stuff about task priorities >- OS switches to new context >- Task B runs >- Swap action completes >- OS gets interrupt >- OS marks the new page as ready to go >- OS checks the task priority information >- OS maybe switches tasks >- Task A or B runs depending on what OS decided. > > >This way, a lower priority task can do useful stuff while we wait for the >pages to swap. Priorities are usually set based on hardware at the level you're talking about. /BAH |