Re: GMP compiled under MacOS/C++ wrap

[Torbjorn Granlund]

Hans Aberg <address@hidden> writes:

  At 13:02 +0200 2001/04/24, Torbjorn Granlund wrote:
  >It is a stretch to call the G4 a 128-bit CPU.  It has SIMD
  >instructions operation on 128-bit registers, but it is still a 32-bit
  >CPU.
  
  The new "titanium" power book has a G4 with a "velocity engine" which is
  claimed to be 128-bit vectored into 4 32-bit words.

Sure, there are 128-bit SIMD instructions as I said.

  The memory bus size also seemed to be 64 or 128 bit, by some speed
  computations I made.
  
I don't understand how you can tell the memory bus width from speed
tests.  In fact, that should be impossible.

  But I have been unable to find document giving the exact details.
  
The G4 manuals are available from Motorola's literature centre.  I
don't have the exact URL, but I know they are there since I have
downloaded them.

  >It is a different question what limb size we should use in GMP.  We
  >need a C type that matches a limb, and that limits the choices.  I
  >strongly doubt usign anything but 32 bits would be a god idea.
  
  Is this necessary if you write assembler instructions for the basic math
  operations?
  
Yes.

  >  - mpq_swap, mpf_swap are mentioned in the manual, but are not in the gmp.h
  >  header.
  >
  >They seem to be there now.
  
  Is there a later version than <ftp://ftp.gnu.org/pub/gnu/gmp/gmp-3.0.tar.gz>?
  
Yes, 3.1.1 has been out several months.

  This is a good idea, but GCC is not available under pre-MacOS X, except by
  using Tenon Mach/UNIX, which I did a few years ago. That old GCC could not
  compile my C++ code, so I switched to CodeWarrior because of that.

The Mac OS 10.0.1 release come with the GNU C++ compiler.  It lives in
usr/libexec/gcc/darwin/ppc/2.95.2/cc1plus.

  >  - In gmp.h, I noticed the lines:
  >    /* Really use `defined (__STDC__)' here; we want it to be true for Sun
  >C */
  >    #if defined (__STDC__) || defined (__cplusplus)
  >  Note that under C++ (cf. C.1.9:1 in the C++ standard), __STDC__
  >  implementation defined.
  >
  >What is your point here?
  
  My impression from reading this comment was that you had put in
  "defined(__cplusplus)" just in order to make it work for Sun C/C++ (but I
  perhaps read it wrong). But it is normal to not have __STDC__ defined under
  C++.
  
Sure, that's why the test on __cplusplus is there.  

-- 
Torbjörn