aboutsummaryrefslogtreecommitdiff
path: root/vendor/gmp-6.3.0/mpn/mips64/README
blob: 7ddd0e572c704805a175ef77c2c7b11d104e2c60 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Copyright 1996 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.

or

  * the GNU General Public License as published by the Free Software
    Foundation; either version 2 of the License, or (at your option) any
    later version.

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,
see https://www.gnu.org/licenses/.





This directory contains mpn functions optimized for MIPS3.  Example of
processors that implement MIPS3 are R4000, R4400, R4600, R4700, and R8000.

RELEVANT OPTIMIZATION ISSUES

1. On the R4000 and R4400, branches, both the plain and the "likely" ones,
   take 3 cycles to execute.  (The fastest possible loop will take 4 cycles,
   because of the delay insn.)

   On the R4600, branches takes a single cycle

   On the R8000, branches often take no noticeable cycles, as they are
   executed in a separate function unit..

2. The R4000 and R4400 have a load latency of 4 cycles.

3. On the R4000 and R4400, multiplies take a data-dependent number of
   cycles, contrary to the SGI documentation.  There seem to be 3 or 4
   possible latencies.

4. The R1x000 processors can issue one floating-point operation, two integer
   operations, and one memory operation per cycle.  The FPU has very short
   latencies, while the integer multiply unit is non-pipelined.  We should
   therefore write fp based mpn_Xmul_1.

STATUS

Good...