Re: why "only" -O optimization for building gcc?

[Nick Ing-Simmons <nik at tiuk dot ti dot com>]

Jonathan Thornburg <jthorn@galileo.thp.univie.ac.at> writes:
>The gcc build instructions   install/build.html  say (in 2.95.2)
>> <p>If you want to save additional space during the bootstrap and in
>> the final installation as well, you can build the compiler binaries
>> without debugging information with "make CFLAGS='-O' LIBCFLAGS='-g
>> -O2' LIBCXXFLAGS='-g -O2 -fno-implicit-templates' bootstrap".  This will save
>> roughly 40% of disk space both for the bootstrap and the final
>
>Is there some deep reason why the instructions don't suggest using
>higher optimization for compiling gcc itself (and also the library), eg
>   make CFLAGS='-O9' LIBCFLAGS='-g -O9' \
>        LIBCXXFLAGS='-g -O9 -fno-implicit-templates' \
>        bootstrap
>This obviously makes for a slower and more memory-intensive build, but
>I'd expect it to yield a faster and maybe smaller gcc binary.  Is there
>a hidden gotcha here?

Well the only higher -O level that does anything is -O3, and IRRC all 
that does is inline functions, so that would make code bigger and 
may not make it faster as cache and paging issues mean inlining is not 
always a win.

>
>If not, perhaps the build instructions should say something about higher
>optimization levels?
-- 
Nick Ing-Simmons <nik@tiuk.ti.com>
Via, but not speaking for: Texas Instruments Ltd.