Re: [4.0 PATCH]: backport "unlocked" stdio patches

[Mark Mitchell <mark at codesourcery dot com>]

Kaveh R. Ghazi wrote:
> I've accumulated the various "unlocked" stdio patches from mainline
> and backported them to the 4.0 branch.
>
> We're still accepting safe compile-time speedup patches, since
> compile-time falls under regressions, right?
>
> Anyway this patch accumulates the following:
>
> 1.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg00453.html
> 2.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg01128.html
> 3.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg01133.html
> 4.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg01527.html
> 5.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg01795.html
> 6.  http://gcc.gnu.org/ml/gcc-patches/2005-04/msg02265.html
>
> All but number 6 are already on mainline showing no problems, the last
> patch awaits mainline approval however.
>
> When I install the patch below on the 4.0 branch, I see various
> speedups on i686-unknown-linux-gnu.  E.g. running genrecog, which does
> a lot of i/o relative to anything else, is about 14% faster.
> Compiling insn-recog.c at -g -O0 -S is about 2.7% faster, and
> c-common.c is 2.3% faster.  Speedups compiling other .c files at -g
> -O0 -S vary but all show improvements in the low single digits,
> usually 1% to 2%.
>
> Bootstrapped against 4.0 on x86_64-unknown-linux-gnu, no regressions
>
> Okay to install in 4.0?

Patch 6 is OK for mainline and the entire set is OK for 4.0.

However, please fix the formatting in Patch 6.  Instead of:

unlock_stream (stdin); unlock_stream (stdout); unlock_stream (stderr);

put those calls on separate lines, or, even better, create an 
unlock_std_streams() function.  That change is preapproved.

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304