Apache OpenOffice (AOO) Bugzilla – Issue 24561
adding a private member function void VtableFactory::flushCode( unsigned char * begin, unsigned char * end);
Last modified: 2007-06-18 10:41:53 UTC
Hi Stephan, As requested, I have created an issue to ask you to modify the multiple inheritance bridge code to offer a private member function void VtableFactory::flushCode( unsigned char * begin, unsigned char * end); to be implemented by each platform, called from the shared code once for each generated vtable. I have modified the flush_range code on my copy of the bridge code for gcc3_linux_powerpc to batch the instruction cache store and invalidate instructions across a contiguous region of memory and then do only one "expensive" sync and isync at the end. This will be a big performance win for smp systems and systems with fast processors that piepeline starvation really impacts harshly. Once you have the new hook in place, I will implement this for PPC Linux and MacOSX. Thanks, Kevin
accepted
set target milestone
sb->khendricks: Function has been added on cws_src680_sb15: bridges/inc/bridges/cpp_uno/shared/vtablefactory.hxx 1.2.4.1 bridges/source/cpp_uno/shared/vtablefactory.cxx 1.2.4.1 Please add definitions of that function to gcc3_linux_powerpc and gcc3_macosx_powerpc.
Hi Stephabn, Thanks, I will do so for both PPC Linux and MacOSX. and commit it back to your tree after test builds of m24 complete. Kevin
Filed issue 25819 to fix (Solaris) SPARC bridge to also flush instruction cache (once it is clear how to do so).
verified
do we have a spec for this?
integrated in SRC680m32
How is the status on this NEW when the resolution is FIXED?
=> set this task to teh correct state 'resolved/fixed' => please check, if the fix is integrated and set it to 'verified' and close this task
This issue has been "RESOLVED" for quite a time, I assume I can close it ...