Processor info

try searching goolge for cpuid and assembly

You have to use the RDTSC instruction, which will return the number of processor cycles executed so far. I don''t remember if RDTSC is guaranteed to be returend in terms of individual processor cycles or some multiple of them (some other architectures with similar features don''t guarantee a 1:1 mapping), but for all X86 processors I''ve tried it on so far, it does.

The timestamp counter is a 64-bit number. You''ll have to do something like set up a timer for 1 second, read the TSC, wait until 1 second has elapsed, read the TSC again, and subtract the first value from it to get the number of cycles per second.

It won''t be a totally accurate figure, but it will be relatively close.

----
Bart

That's what i tried first still, the OS has to find out somehow, right?

Is the PMC easier to use? RDTSC capabilities are returnes in CPUID, but I don't think RDPMC is.

I thought this was an easy question. How is everyone keeping accurate time? GetTickCount()? How does that function know how long a milisecond is?

********


A Problem Worthy of Attack
Proves It's Worth by Fighting Back

[edited by - walkingcarcass on February 26, 2003 3:12:45 PM]

there is a flipcode cotd that does exactly what you want. if you haven''t found it by the time i get home i''ll post a link. i''m at work trying to type on a pda''s onscreen keyboard right now.. not much fun.

500x3

-eldee
;another space monkey;
[ Forced Evolution Studios ]

Do NOT let Dr. Mario touch your genitals. He is not a real doctor!

here it is

-eldee
;another space monkey;
[ Forced Evolution Studios ]

Do NOT let Dr. Mario touch your genitals. He is not a real doctor!

http://www.gamedev.net/community/forums/topic.asp?topic_id=135484

---

:::: [ Triple Buffer V2.0 ] ::::

Thenks, eldee, but it just uses API calls.

QueryPerformanceFrequency and QueryPerformanceCounter can be used under Windows fine.

I expect QueryPerformanceCounter is related to the RDPMC instruction, but how do I get the PMC frequency directly from hardware?

********


A Problem Worthy of Attack
Proves It''s Worth by Fighting Back