Sujet : Re: How To Speed Startup Of Microsoft Office? Have It Running All The Time!
De : nospam (at) *nospam* needed.invalid (Paul)
Groupes : comp.os.linux.advocacy alt.comp.os.windows-11Date : 29. Mar 2025, 11:25:55
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <vs8hrl$13q7p$1@dont-email.me>
References : 1 2 3 4 5 6 7 8
User-Agent : Ratcatcher/2.0.0.25 (Windows/20130802)
On Sat, 3/29/2025 2:49 AM, Lawrence D'Oliveiro wrote:
On Sat, 29 Mar 2025 01:08:30 -0400, Paul wrote:
I don't know if the message is getting through yet,
but Windows has every feature Linux has. Why do you think they hired
7000 developers ?
Mainly to get in each other’s way, judging from their (lack of)
productivity.
Windows has System Read cache and System Write cache.
It has System Read cache in Win2K. System Write cache came later.
>
The System Read cache is like every other implementation.
Doesn’t seem to be. The usage of RAM in the cache seems to interfere with
regular applications’ use of RAM, in a way that doesn’t happen under
Linux.
It seems to be implemented as some sort of unified cache.
https://learn.microsoft.com/en-us/windows/win32/fileio/file-caching125516 MB 22.333757 sec # Malloc test after a reboot
125405 34.227482 # Malloc after System Read Cache full (system read cache gets purged by this...)
125415 26.925629 # Subsequent runs of malloc64.exe, system read cache empty
125419 25.530308
125370 24.816107
125321 25.015528
There isn't zero overhead on the memory management.
*******
I did the same test on Linux.
The top command has columns like
TOTAL FREE USED BUFF/CACHE
and the first quantity equals the sum of the other three.
This implies, at least from a book-keeping perspective, it
works the same as Windows.
The malloc.c code differs between the Windows and Linux versions,
in the time measurement code. Not in the malloc and memory filling parts.
And this is the result of testing on Linux.
120144 MB 43.035967 sec # Malloc test after a reboot
120144 MB 43.139267 sec # Malloc after System Read Cache full (system read cache gets purged by this...)
It's true that the behavior looks better, but... Linux is slower.
During compile, the same optimization level is used in both environments.
The reason the memory amount is lower, is the malloc-linux program
has to be stopped before the OOM-killer is triggered. The same code
on Windows, simply exits when the last malloc call fails to allocate
memory.
Paul