[PATCH] New port of Cell SPU

Mark Mitchell mark@codesourcery.com
Mon Nov 20 20:53:00 GMT 2006


trevor_smigiel@playstation.sony.com wrote:
> This fourth version includes an update to a newer revision of trunk
> (118894) and corrected ChangeLog.

The SC has already appointed David Edelsohn and Andrew Pinski as
maintainers of this port.  So, if they're happy, this patch can go in.

I did give the patch a quick review, and it looks fine to me.  I don't
claim to be an expert on GCC back-ends, but I didn't see anything that
looks like it would be harmful to any other port, which seems like the
most important criterion.

However, I found the use of "static inline" in the C++ intrinsics
surprising.  Is there a reason to prefer "static inline" over just plain
inline?  In the event that the functions are not in fact inlined,
wouldn't you want only one copy?  Is the concern about namespace
pollution, in that the if not static, these functions would collide with
other (non-SPU) functions with the same name?  In that case, wouldn't it
be better to put the functions in namespace __gnu::spu, and then use a
using-directive to bring them into global scope, for users that include
this file?  This isn't a blocking issue since the current declarations
are at worst pessimizing, but I'd be interested in the answer.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713



More information about the Gcc-patches mailing list