Copyright 2003, 2004, 2006, 2008 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP Library is free software; you can redistribute it and/or modify
it under the terms of either:
* the GNU Lesser General Public License as published by the Free
Software Foundation; either version 3 of the License, or (at your
option) any later version.
* the GNU General Public License as published by the Free Software
Foundation; either version 2 of the License, or (at your option) any
or both in parallel, as here.
The GNU MP Library is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received copies of the GNU General Public License and the
GNU Lesser General Public License along with the GNU MP Library. If not,
AMD64 MPN SUBROUTINES
This directory contains mpn functions for AMD64 chips. It is also useful
for 64-bit Pentiums, and "Core 2".
RELEVANT OPTIMIZATION ISSUES
The Opteron and Athlon64 can sustain up to 3 instructions per cycle, but in
practice that is only possible for integer instructions. But almost any
three integer instructions can issue simultaneously, including any 3 ALU
operations, including shifts. Up to two memory operations can issue each
Scheduling typically requires that load-use instructions are split into
separate load and use instructions. That requires more decode resources,
and it is rarely a win. Opteron/Athlon64 have deep out-of-order core.
Optimizing for 64-bit Pentium4 is probably a waste of time, as the most
critical instructions are very poorly implemented here. Perhaps we could
save a cycle or two, but the most common loops now run at between 10 and 22
cycles, so a saved cycle isn't too exciting.
The new spin of the venerable P6 core, the "Core 2" is much better than the
Pentium4 for the GMP loops. Its integer pipeline is somewhat similar to to
the Opteron/Athlon64 pipeline, except that the GMP favourites ADC/SBB and
MUL are slower. Furthermore, an INC/DEC followed by ADC/SBB incur a
pipeline stall of around 10 cycles. The default mpn_add_n and mpn_sub_n
code suffers badly from the stall. The code in the core2 subdirectory uses
the almost forgotten instruction JRCXZ for loop control, and updates the
induction variable using LEA.
"System V Application Binary Interface AMD64 Architecture Processor
Supplement", draft version 0.99, December 2007.