Re: MIPS64 soft-float: missing definition for __floatdidf and friends

[Mark Mitchell <mark at codesourcery dot com>]

Roger Sayle wrote:
> Hi Daniel,
> 
> On Wed, 10 May 2006, Daniel Jacobowitz wrote:
>> libgcc2.c is already compiled multiple times... you could easily key
>> off those labels to accomplish what you want, if you preferred the
>> single file approach.
> 
> Indeed, if we wanted to, most of the compiler could be placed in a
> single source file and keyed off of L_reload, L_combine, L_fold_const,
> etc.. with the appropriate #ifdef's.  Clearly, the question is when
> this is reasonable, and at what level of granularity.  I suspect the
> "cohesion" and "coherence" of routines in libgcc, could already argue
> for further splitting.

I think that the current libgcc organization suffers from the faults you
(Roger) have noted; the way the include files and such are set up seems
needlessly baroque.  However, I think it's also reasonable to say we
should avoid duplicating code.  So, I'd suggest that we look for a
middle path, perhaps along the lines that GLIBC uses for soft-fp, where
macros are combined in various ways, or, as previously suggested, in
which a single file is compiled multiple times with different macro
definitions.

(Parenthetically, and probably totally off-topic for this dicussion,
I've been a bit surprised that other architectures haven't enabled the
GLIBC soft-fp code, as I'd imagine the same kinds of improvements Joseph
obtained for PowerPC might well apply.)

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