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
|
These fortran examples (double-precision source files stored in x*.fm4
files) are designed to mimic the results of the equivalent standard C, Tcl,
Java, and Python examples except for x16a.fm4 which is a unique fortran
example which exercises plshade in cmap1 mode. Depending on whether PLplot
has been built with single-precision libraries or double-precision libraries
(the default), these x*.fm4 files are either processed with the simple sed
script double2single.sed to produce the equivalent single-precision x*.f
files or simply copied to a double-precision form of the x*.f files.
The .fm4 suffix is left over from the days when we used m4 processing to
process mixed m4 and fortran-like code into either double- or
single-precision fortran source. The new method of using either cp or a sed
script to convert double precision source code (which follows the standard
below) is much easier to understand and maintain.
N.B. that maintenance consists of making sure no non-standard number or
single-precision construct of any kind gets into the double-precision source
code, the x*.fm4 files. The rules are simple for what is allowed in x*.fm4
files (see double2single.sed for more commentary):
* Always use lower case
* Always use implicit none
* Always use floating-point numbers in standard form consisting of optional
sign, one or more digits, a decimal point, zero or more digits, "d", an
optional sign and one or more digits. 1.d0 and 1.d-10 are OK, but 1.,
1.e0, 1.D0, etc. are NOT OK.
* Always use dble to convert from integer to double precision. float is
archaic and actually returns a single-precision result.
N.B. If/when the x*.fm4 files require more complexity we may have to extend
these rules and adjust double2single.sed accordingly to convert to the
single-precision form. But for now, the rules above are religiously
followed, and double2single.sed is correspondingly simple.
|
