Activate Cache in Processors *confussed*

I was just wondering:

If the cache (L1&L2) of a processor are completely invisible for the
programmer, being totally controled by hardware, why does Windows (not
sure about linux or other O.S) give the option to activate these and
even more, why do you get a better performance.

Can the Cache be activated de activated? Is it just a tweak?

TIA

Can the Cache be activated de activated? Is it just a tweak?

Atleast from many BIOS'es it can be. There are instructions in x86 (486
isa),sse,sse2,sse3 to affect cache in various ways, you can find more
information from Intel's references for various of their products at
developer.intel.com, they have online reference manuals for their x86
implementations, etc. you can also check www.sandpile.org (sp?) and other
sites.

There are instructions to invalidate cache, read ahead, etc.etc. you must
check feature flags of your processor what instructions are supported and
then look up from the references what the instructions are supposed to do.
This isn't a particularly small topic to cover in a single Usenet post.

By default, CPU cache is enabled and Windows doesn't have an option to
disable it altogether. It has an useless option to reduce its size for
Pentium-II. BIOS may allow you to disable cache for troubleshooting
purposes.

"Bulblight" wrote in message
om...
I was just wondering:

If the cache (L1&L2) of a processor are completely invisible for the
programmer, being totally controled by hardware, why does Windows (not
sure about linux or other O.S) give the option to activate these and
even more, why do you get a better performance.

Can the Cache be activated de activated? Is it just a tweak?

TIA

"Alexander Grigoriev" wrote in message hlink.net...
By default, CPU cache is enabled and Windows doesn't have an option to
disable it altogether. It has an useless option to reduce its size for
Pentium-II. BIOS may allow you to disable cache for troubleshooting
purposes.


So then, Windows can actually change the set of instructions it
requieres to do a certain job depending whether you have the caché
activated of not?

Can you give an example of a x86 instruction that extensively and
explicitly uses cache?

Thanks in advance!

Cheers

On 25 Nov 2003 03:24:43 -0800, (Bulblight) wrote:

I was just wondering:

If the cache (L1&L2) of a processor are completely invisible for the
programmer, being totally controled by hardware, why does Windows (not
sure about linux or other O.S) give the option to activate these and
even more, why do you get a better performance.

Which version of Windows and where in it do you see an option to "activate"
a cache? I'm not sure that the "cache" you are talking about is the L1 or
L2 hardware cache.

Can the Cache be activated de activated? Is it just a tweak?

The cache(s) can be deactivated in BIOS Setup and *could* be deactivated by
a prog, e.g. one which wants to slow the CPU speed down to run legacy code.
If you deactivate both hardware caches, you'll see more difference than a
"tweak" - your CPU will run at at the speed of a 386 or thereabouts.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Can you give an example of a x86 instruction that extensively and
explicitly uses cache?

486:
wbinvd

SSE:
maskmovq
movntq
movntps
prefetch(h)
sfence

SSE2:
clflush
lfence
mfence
pause
maskmovdqu
movntpd
movntdq
movnti

... those are some, I posted my cheating list to my website:

www.liimatta.org/misc/wogsimd.txt

Visit developer.intel.com for references what each instruction does. I only
use that list as "cheat", when I don't remember the precise instruction
mnemonics. SSE3 adds more instructions but I haven't obviously updated my
cheat list yet to cover those (I will when I get the CPU, speaking of which,
that's why I chose P4 originally over Athlon, to be able to program for
SSE2, not because AthlonXP (?) at the time was few % faster than a P4

Anyways, I gave you URLs to in other post already, what was wrong with them?
Didn't find anything?

SSE2, not because AthlonXP (?) at the time was few % faster than a P4

A quick mental lookup gives impression that it was Athlon TB which was the
alternative back then, I could remember still wrong so correct me if feel
inclined. I've got a bad memory and even worse when placing two different
things simultaneously on a timeline. Especially since I never been a into
Athlons to begin with. Cheap, sure, so what.

(Bulblight) wrote in
om:

I was just wondering:

If the cache (L1&L2) of a processor are completely invisible for
the programmer, being totally controled by hardware, why does
Windows (not sure about linux or other O.S) give the option to
activate these and even more, why do you get a better
performance.

Can the Cache be activated de activated? Is it just a tweak?

nooo noo noo totally wrong, You obviously are talking about an
option to specify a cache size so the windoze can addopt his most
time consuming procedures to make adventage of this cache .. but
then again windoze has no such a function there is only a CPU
select function (you can force windoze box to recheck CPU at boot)

There are HDD caches, RAM caches for file/icon/picture operations,
but those have nothing to do with L1/L2.

Pozdrawiam.
--
RusH //
http://kiti.pulse.pdi.net/qv30/ -- heckme
Pent-up passive-aggressive dork alert! Whoop! Whoop!
Whoop! Whoop! Boy, you're really lighting up this alarm here!