 |  |  |
| |
Online users/last 15m
16 Guests, 1 User
Alex Kan
22 Members/last 24hAlex Kan, Gecko_R7, Josef W. Segur, Simon, s52d, Geek@Play, hiamps, j_groothu, Pam Heinze, ajs, setala, Arnulf, Devaster, untraceable, Metod, S56RKO, Crunch3r, seti_britta, Bryon Leigh Hatch_(for Carl Sagan), Kna, Vyper, Nebby, BluesilvergreenErektile Dysfunktion,
| |
 | |  |
|
Pages: [1] 
|
 |
|
 Author
|
Topic: Anyway I can see how much of my CPU cache is in use? (Read 994 times)
|
|
pepperammi
|
Sorry. Probably a silly question. I'm just curious like I am at times of how much of my system memory is in use.
Is there anyway to see how much of my CPU cache is in use? I've just upgraded my CPU to a D945 from a D830 and it has 2x2Mb cache.
|
|
|
|
|
Logged
|
|
|
|
|
Pepo
|
Probably continuously the whole, as the recent PCs' OSes won't fit in  Maybe you could find out about the rate of cache misses, if this is interesting to you. Peter |
|
|
|
|
Logged
|
|
|
|
|
pepperammi
|
Probably continuously the whole, as the recent PCs' OSes won't fit in Maybe you could find out about the rate of cache misses, if this is interesting to you. Peter Oh right. Well could be interesting. What's a cache miss anyway?  |
|
|
|
|
Logged
|
|
|
|
|
Pepo
|
What's a cache miss anyway? Simplified story: when "processor core" needs to crunch some data or execute some code, it looks into "its memory pools (a.k.a. cache)". If the data/code is there, it's called a "cache hit". If not, it must be delivered to the cache (instances) from main memory first, this case is called "cache miss". Short story long: other can explain much better then me, see http://en.wikipedia.org/wiki/CPU_cache Peter |
|
|
|
|
Logged
|
|
|
|
|
|
|
citroja
|
I was wondering the same thing...thanks for the info
-citroja
|
|
|
|
|
Logged
|
|
|
|
|
BenHer
|
Modern chips (pentium 4 and later) have counters inside of them that could be used to calculate this...for a range of time, and using the proper tools.
There are many monitoring counters inside of modern CPUs that count all kinds of esoteric things. # of floating point instructions that are "retired", # of times cache misses occured (ie memory was directly read from), and so on.
There are programmer tools from both Intel and AMD that can be used to setup these timers, run a program for a specified length of time, and then read these counters.
I guess the most important thing to really do would be to specify when to start counting: the beginning of seti run of a full WU?, only for a section (function) of the seti code?
and for how long: Full WU run?, after WU is read from data file into RAM?
and under what conditions: Only for certain angle ranges?
|
|
|
|
|
Logged
|
|
|
|
|
Pages: [1] 
|
|
|
|
Quote!
Every solution breeds new problems.- Murphy's Law
|
|
