[Patch] Finally implement Matt's "recipe"
Paolo Carlini
pcarlini@suse.de
Fri Nov 5 22:17:00 GMT 2004
Hi,
the below finally *tries* to implement Matt's indications here:
http://gcc.gnu.org/ml/libstdc++/2004-04/msg00118.html
Basically, I have applied in this new context what we used to do once
upon a time in 3.3 for basic_filebuf::_M_underflow_common: I hope the
choice of an additional istream.cc looks ok to everyone.
Regtests are Ok on x86-linux and now I'm going to thoroughly check the
patch on other platforms and verify that the performance are also
still ok.
Matt, everyone, is this the way we want to go? Indeed, in this context
seems easy and neat progressively adding additional optimizations, as
originally envisaged in the thread above ;)
I'll wait for comments at least until monday.
Thanks,
Paolo.
///////////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CL_matt
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20041105/d81751ce/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_matt
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20041105/d81751ce/attachment-0001.ksh>
More information about the Libstdc++
mailing list