This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PR tree-optimization/37709: debug info at -g3 and more prunning of scope block structures
On Sun, Feb 22, 2009 at 9:04 PM, Jan Hubicka <hubicka@ucw.cz> wrote:
>> On Sun, Feb 22, 2009 at 7:52 PM, Jan Hubicka <hubicka@ucw.cz> wrote:
>> > Hi,
>> > just for some extra data, tramp3d goes 1.83GB->1.73GB garbage...
>>
>> Which is probably due to
>>
>> - /* Removing declarations before inlining is going to affect
>> - DECL_UID that in turn is going to affect hashtables and
>> - code generation. */
>> || !cfun->after_inlining)
>>
>> change. Why does this comment no longer apply? Previously
>> !cfun->after_inlining
>> forced all blocks used. Otherwise the patch looks sane. Note there
>> was an unrelated
>> config/i386/i386.c change in the patch.
>
> I actually attached not cleaned up version of the patch... here is
> proper file.. (with changes you mentioned and also with the commented
> out if (optimized) test removed)
So you still remove dead blocks if !cfun->after_inlining. Why is this ok now?
Richard.