Re: Donkey's main point as per Microsoft.
- From: "f0dder" <f0dder_nospam@xxxxxxxxxxxxxxxx>
- Date: Mon, 5 Sep 2005 05:49:02 +0200
hutch-- wrote:
If you bother to actually look at ntoskrnl.exe in a disassembler, rather
than divulge in malt-induced fantasies from a mere EXPORTS listing, you will
see that (on a uniprocessor HAL), many of the spinlock functions are null
stubs that simply RET, and there are no cross references to them. If you
understood what they were used for, you would know why. The ones that aren't
NULL stubs do more work than just spinning, like manipulating the IRQL.
On the other hand, KeWaitForMutexObject (which is the same as
KeWaitForSingleObject, just another name for it) has 269 cross references,
and KeWaitForMultipleObjects has 9.
> When a system may have any number of pending exit conditions to
> determine and noting the fundamental Von Neuman architecture of all
> current x86 machines, it must test the exit conditions one after
> another in sequence. Now you can try and hang a newspeak name to this
> action but at its most basic, it is polling the list which means it is
> USING PROCESSOR CYCLES.
There's a difference between polling a set of events constantly (like
your GetExitCodeProcess), and only testing the set of events once an
event is actually triggered. This has already been explained multiple
times, so I won't bother doing it again.
> [...] we now get the
> generalisation that ANY FUNCTION THAT DOES NOT RETURN IMMEDIATELY is a
> "blocking" function [...]
No. But whether a function is blocking or not will be evident from a
disassembly (if it isn't already evident from the performance of the
functions, as well as by plain common sense).
> It would have
> to be obvious that Microsoft did not consult the "Gospel according to
> f0dder" and the more recent "Epistle of Donkey" when they succesfully
> wrote a sequence of Windows operating systems that continue to sell
> day after day, polling technology and all.
It would have to be obvious that microsoft depend HEAVILY on the wait
state functions, if you read the MICROSOFT PRESS published book
"Inside Windows 2000", which has a foreword by the chief designer of
the windows NT architecture, and is authored by two respected and
knowledgable system-level tinkerers, one of which runs sysinternals.com .
.
- Follow-Ups:
- Re: Donkey's main point as per Microsoft.
- From: hutch--
- Re: Donkey's main point as per Microsoft.
- From: f0dder
- Re: Donkey's main point as per Microsoft.
- References:
- Donkey's main point as per Microsoft.
- From: hutch--
- Donkey's main point as per Microsoft.
- Prev by Date: Donkey's main point as per Microsoft.
- Next by Date: Re: f0dder's Fabulous Wait States.
- Previous by thread: Donkey's main point as per Microsoft.
- Next by thread: Re: Donkey's main point as per Microsoft.
- Index(es):
Relevant Pages
|
