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
61
62
63
64
65
|
/*
VASP modular - vector assembling signal processor / objects for Max/MSP and PD
Copyright (c) 2002 Thomas Grill (xovo@gmx.net)
For information on usage and redistribution, and for a DISCLAIMER OF ALL
WARRANTIES, see the file, "license.txt," in this distribution.
*/
#include "main.h"
#include "ops_arith.h"
#include "opdefs.h"
#include "util.h"
#include <math.h>
template<class T> inline V f_radd(T &v,T a,T b) { v = a+b; }
template<class T> inline V f_rsub(T &v,T a,T b) { v = a-b; }
template<class T> inline V f_rsubr(T &v,T a,T b) { v = b-a; }
template<class T> inline V f_rmul(T &v,T a,T b) { v = a*b; }
template<class T> inline V f_rdiv(T &v,T a,T b) { v = a/b; }
template<class T> inline V f_rdivr(T &v,T a,T b) { v = b/a; }
template<class T> inline V f_rmod(T &v,T a,T b) { v = fmod(a,b); }
BL VecOp::d_add(OpParam &p) { D__rbin(f_radd<S>,p); }
BL VecOp::d_sub(OpParam &p) { D__rbin(f_rsub<S>,p); }
BL VecOp::d_subr(OpParam &p) { D__rbin(f_rsubr<S>,p); }
BL VecOp::d_mul(OpParam &p) { D__rbin(f_rmul<S>,p); }
BL VecOp::d_div(OpParam &p) { D__rbin(f_rdiv<S>,p); }
BL VecOp::d_divr(OpParam &p) { D__rbin(f_rdivr<S>,p); }
BL VecOp::d_mod(OpParam &p) { D__rbin(f_rmod<S>,p); }
VASP_BINARY("vasp.+",add,true,VASP_ARG_R(0),"Adds a value, envelope or vasp")
VASP_BINARY("vasp.-",sub,true,VASP_ARG_R(0),"Subtracts a value, envelope or vasp")
VASP_BINARY("vasp.!-",subr,true,VASP_ARG_R(0),"Reverse subtracts a value, envelope or vasp")
VASP_BINARY("vasp.*",mul,true,VASP_ARG_R(1),"Multiplies with a value, envelope or vasp")
VASP_BINARY("vasp./",div,true,VASP_ARG_R(1),"Divides by a value, envelope or vasp")
VASP_BINARY("vasp.!/",divr,true,VASP_ARG_R(1),"Reverse divides by a value, envelope or vasp")
VASP_BINARY("vasp.%",mod,true,VASP_ARG_R(0),"Calculates the remainder of the division by a value, envelope or vasp")
// -----------------------------------------------------
template<class T> inline V f_rsqr(T &v,T a) { v = a*a; }
template<class T> inline V f_rssqr(T &v,T a) { v = a*fabs(a); }
BL VecOp::d_sqr(OpParam &p) { D__run(f_rsqr<S>,p); }
BL VecOp::d_ssqr(OpParam &p) { d__run(f_rssqr<S>,p); }
VASP_UNARY("vasp.sqr",sqr,true,"Calculates the square")
VASP_UNARY("vasp.ssqr",ssqr,true,"Calculates the square with preservation of the sign")
// -----------------------------------------------------
template<class T> inline V f_rsign(T &v,T a) { v = (a == 0?0:(a < 0?-1.:1.)); }
template<class T> inline V f_rabs(T &v,T a) { v = fabs(a); }
BL VecOp::d_sign(OpParam &p) { D__run(f_rsign<S>,p); }
BL VecOp::d_abs(OpParam &p) { D__run(f_rabs<S>,p); }
VASP_UNARY("vasp.sign",sign,true,"Calculates the sign (signum function)")
VASP_UNARY("vasp.abs",abs,true,"Calulates the absolute value")
|