| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 22
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
sekerob, anyone,
----------------------------------------does anyone have cpu schedular configs sussed out for ht affinity of 2+ process names that should run on even/odd cpu id's ? not hard to do in thoery like the goal: whats needed to force the maximum instruction spread is a) find out instruction usage profiles (use binutils nm, and sort the libc math routines) per executable b) any 2 processes rank toward even (int), and odd (float) when the schedular decides how to pair up ht cores. part b is what i don't have a 2-line shell script snippet to offer. cpu schedular tools and kernel mods exist but i have no idea to what degree [Edit 1 times, last edit by Former Member at Dec 6, 2010 3:32:11 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
a) above from the boinc queue dir: to look at the modules linked into the executables running in the project dirs.
ranking an educated guess like modules starting with gl* are likely FPU access, and so on, gives us a less than 100% random guess of relative comparison of functionality/instructions in the non-stripped executables. file -zr `find -type f `|grep ELF |tr -s ' '|cut -f1 -d: |xargs nm |
||
|
|
|