This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: libstdc++/4150: catastrophic performance decrease in C++ code
>> Done.
> Great. Thanks again for all your efforts.
Yes, thank you Jason. I can confirm no regressions with
i386-*-freebsd4.5. I updated docs/html/17_intro/porting.texi to
reflect the removal of _GLIBCPP_AVOID_FSEEK.
I can also confirm that adding ``std::ios::sync_with_stdio(false);''
now speeds up the cin run of the test case to be the same as the run
with a newly opened file. There used to be no difference in this
environment.
> Now, I believe this fixes the current 'high' priority GNATS bugs, both
> 4150 and 5396. If so, can you put them into feedback or close them?
> It would be appreciated.
I just reread 4150 and 5396 (and tested both with truss and time) and
agree that Jason got them. I can confirm that the profile of the
system calls now looks as expected.
> Loren, I think you had a patch on top of Jason's. Can you give status
> please?
I am looking at it again against what Jason committed today. I will
post something ASAP.
> I'll start working on the intractivity bug, represented by 5820.
I can report that Jason's patch did not affect this test case on
i386-*-freebsd4.5 (where it already worked as expected by the PR
submitter and myself).
Regards,
Loren