Re: Building newlib SPU port with FSF gcc

[jschopp <jschopp at austin dot ibm dot com>]

Was on vacation, sorry for the slow reply.  Thanks for the pointer.  I'll add it to the 
queue.  To be fair this did all compile with the only available compiler when I left for 
vacation.  Nice to see some of that compiler making it into FSF though.  I will change to 
match.

-Joel

trevor_smigiel@playstation.sony.com wrote:
> Joel,
>
> I tried building the latest newlib with the latest FSF gcc and
> encountered a few compile time errors.
>
> There are a few uses of Altivec style initializers still in the code.

I'll remove them.

> There is a use of "vector" as a keyword which is implemented as a define
> in spu_intrinsics.h, which is not included in that case.  It is better
> to use "__vector".

OK.  I'll do that too.

> Also, there is an issue with uses of size_t, which is defined as a long
> int.  For certain intrinsics, like spu_splats(), a long type is
> considered ambiguous, and the compiler will generate an error.  The
> specification of the intrinsics, could be extended to handle long, and
> I'll raise that as a possiblitiy with the 3C group.  But you might want
> to fix the cases where it occurs anyway.

I think it should be extended.  And I'll fix the casees anyway.

> I've included a patch below which identifies the places which cause
> problems.  

Many thanks.