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] Fix PR 13987, compile time slow down on powerpc-apple-darwin


Segher Boessenkool <segher@kernel.crashing.org> writes:

> > We are looking for an integer which is fast even when targetting a
> > 64-bit machine.  I think your suggestion of HOST_FAST_INT is
> > reasonable.  When targetting a 32-bit machine, HOST_FAST_INT and
> > HOST_WIDE_INT would normally be the same.  When targetting a 64-bit
> > machine, HOST_WIDE_INT must be 64 bits, but HOST_FAST_INT might be 32
> > bits.
> 
> So  (unsigned) long  will always work, you think?  Or does it need
> a platform-specific override?  I can only think of 8- or 16-bit
> host machines wanting that; and I don't really care about the speed
> of compiling _on_ such a machine ;-)

My sense is that it should have a platform specific override not to
make it smaller, but to make it larger.  On a machine with 64-bit
registers, we want to use a 64-bit type, because the bitmap code will
(presumably) run faster.  Although many of those systems will have a
64-bit long, some will still use a 32-bit long.  So I think we should
have the capacity for an override to speed up the compiler in such a
case.

Ian


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