binary compatibility ABI (was: Re: [boehm-gc] Import 6.3 alpha 1)

Andrew Haley aph@redhat.com
Thu Aug 14 16:27:00 GMT 2003


Jeff Sturm writes:
 > On Thu, 14 Aug 2003, Andrew Haley wrote:
 > > Jeff Sturm writes:
 > >  >
 > >  > One of my sore points is the performance gap between static and
 > >  > position-independent code with the current gcj compiler.  For example:
 > >  > reading a static field requires one memory load without -fPIC, but two
 > >  > loads with -fPIC (to obtain the memory address from the GOT offset).
 > >
 > > Short of overwriting the instruction that does the fetch, how would you
 > > do that?
 > 
 > That wasn't really such a good example.
 > 
 > Consider:
 > 
 > static int a, b;
 > ...
 > return a + b;
 > 
 > Does this require 2, 3 or 4 memory loads?
 > 
 > (I believe 2 is the answer for non-PIC, 4 for PIC.  If we take advantage
 > of the fact that static fields are at a fixed offset from the class
 > record, it can be done in three, while preserving position independence.)

Unless a and b are in the exact same class as the 'return a + b'
statment, the positions of a and b relative to the class record may
change.  I guess this isn't true if the whole application is compiled
and we give up on binary compatibility.

Andrew.



More information about the Java mailing list