1TTOL(3) LAM NETWORK LIBRARY TTOL(3)
2
6 ttol, ltot, mttol, mltot - LAM data representation translation suite
7
9 #include <portable.h>
10 #include <t_types.h>
11 void ttoli4 (int4 *pvar1, int4 *pvar2);
13 void ltoti4 (int4 *pvar1, int4 *pvar2);
14 void ttoli2 (int2 *pvar1, int2 *pvar2);
15 void ltoti2 (int2 *pvar1, int2 *pvar2);
16 void ttolf4 (float4 *pvar1, float4 *pvar2);
17 void ltotf4 (float4 *pvar1, float4 *pvar2);
18 void ttolf8 (float8 *pvar1, float8 *pvar2);
19 void ltotf8 (float8 *pvar1, float8 *pvar2);
20 void mttoli4 (int4 *array, int num);
22 void mltoti4 (int4 *array, int num);
23 void mttoli2 (int2 *array, int num);
24 void mltoti2 (int2 *array, int num);
25 void mttolf4 (float4 *array, int num);
26 void mltotf4 (float4 *array, int num);
27 void mttolf8 (float8 *array, int num);
28 void mltotf8 (float8 *array, int num);
29
31 subroutine F4LTOT (lvar, tvar)
32 subroutine F4TTOL (tvar, lvar)
33 real*4 lvar, tvar
34 subroutine F8LTOT (lvar, tvar)
36 subroutine F8TTOL (tvar, lvar)
37 real*8 lvar, tvar
38 subroutine I4LTOT (lvar, tvar)
40 subroutine I4TTOL (tvar, lvar)
41 integer*4 lvar, tvar
42 subroutine I2LTOT (lvar, tvar)
44 subroutine I2TTOL (tvar, lvar)
45 integer*2 tvar, lvar
46 subroutine F4MLTT (array, num)
48 subroutine F4MTTL (array, num)
49 real*4 array(*)
50 integer num
51 subroutine F8MLTT (array, num)
53 subroutine F8MTTL (array, num)
54 real*8 array(*)
55 integer num
56 subroutine I4MLTT (array, num)
58 subroutine I4MTTL (array, num)
59 integer*4 array(*)
60 integer num
61 subroutine I2MLTT (array, num)
63 subroutine I2MTTL (array, num)
64 integer*2 array(*)
65 integer num
66
68 The number and order of bytes in a word may differ between nodes in any
69 LAM network. The representation of floating point numbers may also
70 vary. To transfer message and file data correctly, both sender and
71 receiver in a LAM network must agree on the amount of data being
72 exchanged and its representation. A simple solution to these two prob‐
73 lems is to define a standard representation and require senders and
74 receivers to convert their data from/to the local representation
75 to/from the standard format. Conversion is not necessary if the commu‐
76 nicating processes are on the same node, but it keeps the code porta‐
77 ble.
78 The LAM header file <portable.h> defines fixed size C types for each
80 supported architecture. Fortran already has the ability to declare
81 variable precision.
82 int4 4-byte wide signed integer
84 uint4 4-byte wide unsigned integer
86 int2 2-byte wide signed integer
88 uint2 2-byte wide unsigned integer
90 float4 4-byte wide floating point number
92 float8 8-byte wide floating point number
94 Programmers wishing to write portable LAM code that can run on hetero‐
96 geneous architectures are urged to use these data types for all vari‐
97 ables that are communicated across node boundaries. To gain computing
98 speed, programmers may wish to transfer the communicated data to the
99 appropriate native data type on the local CPU in order to benefit from
100 its faster access of word-sized variables. Following this approach
101 solves the data size problem.
102 The LAM header file <t_types.h> defines functions that convert the
104 order of bytes between the local representation, regardless of what
105 that is, and a standard (LAM) representation. The functions are meant
106 to be used in conjunction with the data types defined in <portable.h>.
107 If no conversion is necessary, the functions will have no effect.
108 These functions are implemented as macros and should not be accessed
109 through pointers to functions.
110 When sending messages between heterogeneous nodes, both the message
112 body (i.e. what is in the nh_msg field of the network message struc‐
113 ture; see nsend(2)) and the message data pouch (i.e. what is in the
114 nh_data field of the network message structure) must be converted.
115 Other fields in the message envelope are automatically converted when
116 the message is passed. Data conversion functions should be used by
117 both the sending and receiving processes. Data should be converted
118 from local to LAM representation before sending, and from LAM to local
119 representation after receiving.
120 As an alternative to the byte-order translation functions, the program‐
122 mer can set various bits in the nh_flags field to convert the nh_msg
123 and/or the nh_data data in other than the default manner (see
124 nsend(2)). The default assumption is that nh_data is an array of 8
125 int4 integers and that nh_msg contains raw bytes.
126 File data should be converted from local to LAM representation before
128 writing, and from LAM to local representation after reading.
129 The conversion functions are:
131 ttoli4() Convert an int4 from LAM to local representation.
133 ltoti4() Convert an int4 from local to LAM representation.
135 ttoli2() Convert an int2 from LAM to local representation.
137 ltoti2() Convert an int2 from local to LAM representation.
139 ttolf4() Convert a float4 from LAM to local representation.
141 ltotf4() Convert a float4 from local to LAM representation.
143 ttolf8() Convert a float8 from LAM to local representation.
145 ltotf8() Convert a float8 from local to LAM representation.
147 mttoli4() Convert an array of int4s from LAM to local representation.
149 mltoti4() Convert an array of int4s from local to LAM representation.
151 mttoli2() Convert an array of int2s from LAM to local representation.
153 mltoti2() Convert an array of int2s from local to LAM representation.
155 mttolf4() Convert an array of float4s from LAM to local representation.
157 mltotf4() Convert an array of float4s from local to LAM representation.
159 mttolf8() Convert an array of float8s from LAM to local representation.
161 mltotf8() Convert an array of float8s from local to LAM representation.
163 The single element conversion functions accept two arguments:
165 pvar1 pointer to a variable of the required type holding the origi‐
167 nal data to be converted
168 pvar2 pointer to a variable of the required type where the result‐
170 ing converted data is returned
171 The original data is not modified. The two pointers may be the same if
173 the conversion is to be done "in place". The multiple element conver‐
174 sion functions accept two arguments:
175 array pointer to the base of an array of data elements of the
177 required type
178 num the number of data elements in the array
180 Before the call, the array holds the data to be converted. After the
182 function returns, the array holds the converted data. The array con‐
183 version functions can only convert data "in place".
184 Incompatible Word Lengths
186 Some CPUs do not support all data type sizes. As an example, Cray
187 machines can access 4-byte and 8-byte integers and 8-byte floating
188 point numbers, but do not access 2-byte integers or 4-byte floating
189 point numbers. Until another solution is offered, when such machines
190 are used in a LAM network it is advised to constrain the data type
191 choices to those available on all CPUs (i.e. to follow the least common
192 denominator approach).
193
195 nsend(2)
196LAM 7.1.2 March, 2006 TTOL(3)