Home | History | Annotate | Line # | Download | only in sparc32
README revision 1.1 
      1 Copyright 1996, 2001 Free Software Foundation, Inc.
      2 
      3 This file is part of the GNU MP Library.
      4 
      5 The GNU MP Library is free software; you can redistribute it and/or modify
      6 it under the terms of the GNU Lesser General Public License as published by
      7 the Free Software Foundation; either version 3 of the License, or (at your
      8 option) any later version.
      9 
     10 The GNU MP Library is distributed in the hope that it will be useful, but
     11 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
     12 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
     13 License for more details.
     14 
     15 You should have received a copy of the GNU Lesser General Public License
     16 along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
     17 
     18 
     19 
     20 
     21 
     22 This directory contains mpn functions for various SPARC chips.  Code that
     23 runs only on version 8 SPARC implementations, is in the v8 subdirectory.
     24 
     25 RELEVANT OPTIMIZATION ISSUES
     26 
     27   Load and Store timing
     28 
     29 On most early SPARC implementations, the ST instructions takes multiple
     30 cycles, while a STD takes just a single cycle more than an ST.  For the CPUs
     31 in SPARCstation I and II, the times are 3 and 4 cycles, respectively.
     32 Therefore, combining two ST instructions into a STD when possible is a
     33 significant optimization.
     34 
     35 Later SPARC implementations have single cycle ST.
     36 
     37 For SuperSPARC, we can perform just one memory instruction per cycle, even
     38 if up to two integer instructions can be executed in its pipeline.  For
     39 programs that perform so many memory operations that there are not enough
     40 non-memory operations to issue in parallel with all memory operations, using
     41 LDD and STD when possible helps.
     42 
     43 UltraSPARC-1/2 has very slow integer multiplication.  In the v9 subdirectory,
     44 we therefore use floating-point multiplication.
     45 
     46 STATUS
     47 
     48 1. On a SuperSPARC, mpn_lshift and mpn_rshift run at 3 cycles/limb, or 2.5
     49    cycles/limb asymptotically.  We could optimize speed for special counts
     50    by using ADDXCC.
     51 
     52 2. On a SuperSPARC, mpn_add_n and mpn_sub_n runs at 2.5 cycles/limb, or 2
     53    cycles/limb asymptotically.
     54 
     55 3. mpn_mul_1 runs at what is believed to be optimal speed.
     56 
     57 4. On SuperSPARC, mpn_addmul_1 and mpn_submul_1 could both be improved by a
     58    cycle by avoiding one of the add instructions.  See a29k/addmul_1.
     59 
     60 The speed of the code for other SPARC implementations is uncertain.
     61