|
From: Frank C. <fra...@gm...> - 2011-09-16 01:35:51
|
Hi, We recently built and linked in the google perftools libtcmalloc custom memory allocator into a prototype software tool we are researching at a basic research level. We found that on WIndows and Linux, the speed of this research software tool sped by 350% when we linked in libtcmalloc. Now, we are trying to analyze why this fortuitous speed improvement using kcachegrind -- callgrind, How do find out which tcmalloc functions are responsible for the speed improvement using the kcachegrind -- callgrind profiler? Thank you. |
|
From: Dan K. <da...@ke...> - 2011-09-16 02:09:05
|
On Thu, Sep 15, 2011 at 6:35 PM, Frank Chang <fra...@gm...> wrote: > Hi, We recently built and linked in the google perftools libtcmalloc custom > memory allocator into a prototype software tool we are researching at a > basic research level. We found that on WIndows and Linux, the speed of this > research software tool sped by 350% when we linked in libtcmalloc. Now, we > are trying to analyze why this fortuitous speed improvement using > kcachegrind -- callgrind, How do find out which tcmalloc functions are > responsible for the speed improvement using the kcachegrind -- callgrind > profiler? Thank you. Before you go there, have you read http://goog-perftools.sourceforge.net/doc/tcmalloc.html ? |
|
From: Frank C. <fra...@gm...> - 2011-09-16 13:08:13
|
Dan. I read the kcachegrind manual last night and early this morning and I think we figured out how to visualize the tcmalloc source code hit counts generated by callgrind. Thank you. On Thu, Sep 15, 2011 at 7:08 PM, Dan Kegel <da...@ke...> wrote: > On Thu, Sep 15, 2011 at 6:35 PM, Frank Chang <fra...@gm...> > wrote: > > Hi, We recently built and linked in the google perftools libtcmalloc > custom > > memory allocator into a prototype software tool we are researching at a > > basic research level. We found that on WIndows and Linux, the speed of > this > > research software tool sped by 350% when we linked in libtcmalloc. Now, > we > > are trying to analyze why this fortuitous speed improvement using > > kcachegrind -- callgrind, How do find out which tcmalloc functions are > > responsible for the speed improvement using the kcachegrind -- callgrind > > profiler? Thank you. > > Before you go there, have you read > http://goog-perftools.sourceforge.net/doc/tcmalloc.html ? > |