GCC build failed with your patch on 2001-01-09T11:35:00Z.

[David Edelsohn]

>>>>> Richard Henderson writes:

Richard> But it is also not acceptable to slow down a native 32-bit ppc-linux 
Richard> compiler by forcing it to do multi-word arithmetic when it is not
Richard> necessary.

	It's not just slowing down a 32-bit target, it is slowing down a
32x64 cross when the developer does not want it.

	Deciding that GCC must / should use the target width as CONST_INT,
regardless of the host width, is not a decision to be made lightly.
Because cross-compilers always are built with GCC, which does support
64-bit "long long" on 32-bit hosts, we could make that a requirement.
That essentially is what your patch does.  I don't want that decision to
be made implicitly.

	Are you going to allow / require that 64-bit targets only need to
handle CONST_INT and not CONST_DOUBLE?  This line of reasoning can have
far-reaching consequences.

David