What about this for a C++ optimisation?

Joern Rennecke amylaar@cygnus.co.uk
Thu Jul 27 09:18:00 GMT 2000


> How about this for a C++ optimisation? I know it's along lines discussed or
> implemented before, but it's not currently supported...

I'd also like this to work for C, where applicable.  I.e. a value that
is passed directly in registers / on the stack when possible.  And pass
it in a register when a scalar would be passed in a register.
For function arguments there is __attribute__ ((transparent_union__)),
but it doesn't work for return values.

It would be nice if you could eventually get the same performance on
i386-linux-gnulibc1 as you have on i386-linuxaout.


More information about the Gcc mailing list