This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On Wed, Jul 8, 2009 at 6:46 PM, Jeff Law<law@redhat.com> wrote:I've reviewed the thread and I must admit I'm still somewhat confused -- I still don't see how movement of a clobber can extend a pseudo's lifetime, except in cases where movement of the clobber to an earlier point allows the scheduler to then move a following assignment to the same pseudo to an earlier point in the stream.
After a clobber there should be no useful value in a pseudo until a later set of the pseudo, so how does hoisting the clobber extend a pseudo's lifetime? ISTM that hoisting the clobber to the earliest possible point in the insn stream should be decreasing lifetimes.
What am I missing?
Please see the "Gcc 3.1 performance regressions with respect to 2.95.3" thread from March 2002, including:
http://gcc.gnu.org/ml/gcc-patches/2002-03/msg01763.html
Some of this may have been fixed, but GCC was emitting naked CLOBBERS with open-coded no-conflict blocks and other entertaining ways.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |