CS Theory and MS tuning

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

CRoberson
Posts: 2055
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

CS Theory and MS tuning

Post by CRoberson »

Ok, saw something on an overclocking forum. It was a list of tweaks
to speed up MS OS. One of them made me say huh?

Basically, it claimed that the OS initially doesn't know to access the
CPU's L2 cache and that you should put the size of the L2 cache
(in megabytes) into a registry location.

From my days in CS classes, I remember that the L1 and L2 caches
are controlled by the HW not the OS. The OS and all other programs
make a request to the CPU to access memory. How the
HW/CPU/memory subsytem accesses the memory is fully abstracted
from the SW including the OS.


So, I looked into the registry and the location is there. I've not tried
it as it sounds rediculous. Here is the quote from the forum.

To make Windows XP use L2 cache

by default windows doesn't recognise your L2 cache on processor and doesn't make use of it, to enable it run regedit and navigate to

HKEY_LOCAL_MACHINE > SYSTEM > CurrentControlSet > Control > Session Manager > MemoryManagment

there is a DWORD called "SecondLevelDataCache" double click it, select Decimal and set the amount of L2 cache you have (e.g. if you have 256KB, then just type 256, and make sure you enter it in Decimal, not Hexadecimal)
Does this make any sense to anybody? I think it is just bull. Or did
MS do something outside the bounds of normal CS theory
(wouldn't be the first time)? Although, such efforts by them that I am
aware of seem more hype than performance.
F. Bluemers
Posts: 868
Joined: Thu Mar 09, 2006 11:21 pm
Location: Nederland

Re: CS Theory and MS tuning

Post by F. Bluemers »

the key does not do anything in windows xp.
It was for old NT systems using procs without cpuid.
most tweaks are a kind of urban computer sories anyway

Best
Fonzy
Bryan Hofmann

Re: CS Theory and MS tuning

Post by Bryan Hofmann »

If you are using a very old orignal pentium (ie old 60/66mhz) CPU this could be used, other wise it does nothing.

http://support.microsoft.com/kb/183063
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: CS Theory and MS tuning

Post by bob »

CRoberson wrote:Ok, saw something on an overclocking forum. It was a list of tweaks
to speed up MS OS. One of them made me say huh?

Basically, it claimed that the OS initially doesn't know to access the
CPU's L2 cache and that you should put the size of the L2 cache
(in megabytes) into a registry location.

From my days in CS classes, I remember that the L1 and L2 caches
are controlled by the HW not the OS. The OS and all other programs
make a request to the CPU to access memory. How the
HW/CPU/memory subsytem accesses the memory is fully abstracted
from the SW including the OS.


So, I looked into the registry and the location is there. I've not tried
it as it sounds rediculous. Here is the quote from the forum.

To make Windows XP use L2 cache

by default windows doesn't recognise your L2 cache on processor and doesn't make use of it, to enable it run regedit and navigate to

HKEY_LOCAL_MACHINE > SYSTEM > CurrentControlSet > Control > Session Manager > MemoryManagment

there is a DWORD called "SecondLevelDataCache" double click it, select Decimal and set the amount of L2 cache you have (e.g. if you have 256KB, then just type 256, and make sure you enter it in Decimal, not Hexadecimal)
Does this make any sense to anybody? I think it is just bull. Or did
MS do something outside the bounds of normal CS theory
(wouldn't be the first time)? Although, such efforts by them that I am
aware of seem more hype than performance.
Makes absolutely no sense, and is complete and utter hogwash. No idea who came up with that lame-brained idea, but it's wrong. The processor uses L1/L2 unless it is turned off in the BIOS which nobody in their right mind would do...