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]

Re: PATCH - Minor changes to support 64bit int/32bit ABI for PPC



On Sunday, November 9, 2003, at 06:28 PM, David Edelsohn wrote:


Geoff Keating writes:

Geoff> I don't think it's unreasonable to make changes of this nature to GCC.
Geoff> GCC becomes significantly slower when you change H_W_I to be 64 bits.


When Alexandre Oliva patched hwint.h to assume 64-bit HWI for
ports that defined MAX_BITS_PER_WORD to 64, it was decided that this would
be required for portability and consistent behavior. Otherwise one sees
different behavior depending on the host platform and one may not even be
able to cross-target reliably from a true 64-bit platform.


If Apple wants to make this change for speed, that's fine, but
Apple needs to maintain it locally. It is not the FSF GCC policy to
support that mode of operation or to assure that it works. I don't want
to go down the path of making changes that incorrectly give people the
impression this should work because that invites bug reports for future
failures this mode may expose. This is not a supported mode of operation
and imposes a new, implied maintenance burden on FSF GCC developers if
accepted.


David


I see several places in rs6000.c of FSF branch where code is controlled by
HOST_BITS_PER_WIDE_INT == 32. So, above policy must have
been enacted post those changes.


- Fariborz


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]