Why Does "java_global_trees" Exist?
Bryce McKinlay
mckinlay@redhat.com
Tue Jun 15 15:18:00 GMT 2004
Ranjit Mathew wrote:
>On Mon, 14 Jun 2004 16:45:51 -0400, Bryce McKinlay <mckinlay@redhat.com> wrote:
>[...]
>
>
>>Yeah, this is a leftover from earlier incarnations of GCC's garbage
>>collector. These days we can use GTY markers to easily register roots,
>>so the java_global_trees stuff could be removed. I've found it annoying
>>in the past too, so I vote to get rid of it.
>>
>>
>
>Actually *everyone* (but Ada/Fortran) seems to have "global_trees"
>structures - I've found "global_trees", "c_global_trees",
>"cp_global_trees" and "objc_global_trees", apart from
>"java_global_trees".
>
>Maybe it's better this way to reduce the roots set
>exposed to the GCC garbage collector.
>
>
Yeah, I suppose if they were all separate entries rather than a big
array, the root set data would be larger since each entry in the root
set array generated by gengtype carries type information as well as the
pointer. But, whether that translates into a measureable performance
penalty is hard to say.
Regards
Bryce
More information about the Java
mailing list