cues to self
Locked
 
|
All,
Frank sent me the following off line and gave me permission to post it. I dont know how its actually done on modern computers but I can guess how it would have been done on computers as they were when I learned assembly language programming (late 60s early 70s). You have a register that increments a count every time an instruction is executed. You have an interrupt every second (there is a clock). When that interrupt occurs you calculate the record the number of instructions executed since the last such interrupt (there are methods for keeping track of hierarchies of interrupts). You plot that number on a graph (scaled) or print it in a report or whatever. It doesnt matter if it is accurate to the nth decimal place since the user (or whoever) just wants a rough estimate of the performance of the CPU. By the way, the CPU does this calculation. Thats why I say its like Flor doing her home work. Periodically she just counts (by looking at the paper) how many problems she had done since the last time she checked. Etc. My thoughts are: So, the cue to how I am performing NOW, is how I was performing in the 5 nanoseconds just after you asked me? But not really performing, though, is it? I realize that we are coming up against the concept of VALIDITY of a measure, here. To what extent is "how many calculations did you perform in the last 10 nanoseconds'"a valid measure of performance.? Frank, can you help me out with this validity thing? Nick Nicholas Thompson nickthompson at earthlink.net http://home.earthlink.net/~nickthompson ----- Original Message ----- From: Frank Wimberly To: nickthompson at earthlink.net Sent: 11/29/2005 11:11:27 AM Subject: RE: [FRIAM] Is Robert a Deterministic System? Nick, Frank --- Nicholas Thompson nickthompson at earthlink.net http://home.earthlink.net/~nickthompson -------------- next part -------------- An HTML attachment was scrubbed... URL: http://redfish.com/pipermail/friam_redfish.com/attachments/20051129/fde4e6ed/attachment.htm |
|
|
I think the idea is that it's an approximation that is more or less valid
and that's "good enough for government work". To get an exact measure of the instantaneous performance is probably not possible nor would it be needed for practical puposes. Frank --- Frank C. Wimberly 140 Calle Ojo Feliz Santa Fe, NM 87505 (505) 995-8715 or (505) 670-9918 (cell) wimberly3 at earthlink.net or wimberly at andrew.cmu.edu or wimberly at cal.berkeley.edu -----Original Message----- From: [hidden email] [mailto:[hidden email]] On Behalf Of Nicholas Thompson Sent: Tuesday, November 29, 2005 12:06 PM To: friam at redfish.com Subject: [FRIAM] cues to self All, Frank sent me the following off line and gave me permission to post it. I don't know how it's actually done on modern computers but I can guess how it would have been done on computers as they were when I learned assembly language programming (late 60's early 70's). You have a register that increments a count every time an instruction is executed. You have an interrupt every second (there is a clock). When that interrupt occurs you calculate the record the number of instructions executed since the last such interrupt (there are methods for keeping track of hierarchies of interrupts). You plot that number on a graph (scaled) or print it in a report or whatever. It doesn't matter if it is accurate to the nth decimal place since the user (or whoever) just wants a rough estimate of the performance of the CPU. By the way, the CPU does this calculation. That's why I say it's like Flor doing her home work. Periodically she just counts (by looking at the paper) how many problems she had done since the last time she checked. Etc. My thoughts are: So, the cue to how I am performing NOW, is how I was performing in the 5 nanoseconds just after you asked me? But not really performing, though, is it? I realize that we are coming up against the concept of VALIDITY of a measure, here. To what extent is "how many calculations did you perform in the last 10 nanoseconds'"a valid measure of performance.? Frank, can you help me out with this validity thing? Nick Nicholas Thompson nickthompson at earthlink.net http://home.earthlink.net/~nickthompson ----- Original Message ----- From: Frank Wimberly <mailto:[hidden email]> To: nickthompson at earthlink.net Sent: 11/29/2005 11:11:27 AM Subject: RE: [FRIAM] Is Robert a Deterministic System? Nick, Frank --- Nicholas Thompson nickthompson at earthlink.net http://home.earthlink.net/~nickthompson -------------- next part -------------- An HTML attachment was scrubbed... URL: http://redfish.com/pipermail/friam_redfish.com/attachments/20051129/daa4744b/attachment-0001.htm |
«
Return to Friam
|
1 view|%1 views
| Free forum by Nabble | Edit this page |