1 /*
2  * Copyright (c) 2011-2019, The University of Oxford.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions are met:
7  * 1. Redistributions of source code must retain the above copyright notice,
8  *    this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright notice,
10  *    this list of conditions and the following disclaimer in the documentation
11  *    and/or other materials provided with the distribution.
12  * 3. Neither the name of the University of Oxford nor the names of its
13  *    contributors may be used to endorse or promote products derived from this
14  *    software without specific prior written permission.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
17  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
20  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #ifndef OSKAR_VECTOR_TYPES_H_
30 #define OSKAR_VECTOR_TYPES_H_
31 
32 /**
33  * @file oskar_vector_types.h
34  */
35 
36 #ifdef __CUDACC__
37 /* Include the CUDA vector types header first, if we're compiling with nvcc. */
38 #   include <vector_types.h>
39 #endif
40 
41 /* Memory alignment macros mirroring those used by CUDA. */
42 #if !(defined(__VECTOR_TYPES_H__) || defined(__CUDACC__))
43 #   if defined(__GNUC__)
44 #       define __align__(n) __attribute__((aligned(n)))
45 #   elif defined(_MSC_VER)
46 #       define __align__(n) __declspec(align(n))
47 #   endif
48 #   if defined(__GNUC__) || defined(_WIN64)
49 #       define __builtin_align__(a) __align__(a)
50 #   else
51 #       define __builtin_align__(a)
52 #   endif
53 
54 /**
55  * @brief Two-element structure (single precision).
56  *
57  * @details
58  * Structure used to hold data for a length-2 vector.
59  * This must be compatible with the CUDA float2 type.
60  */
61 struct __builtin_align__(8) float2 { float x, y; };
62 typedef struct float2 float2;
63 
64 /**
65  * @brief Two-element structure (double precision).
66  *
67  * @details
68  * Structure used to hold data for a length-2 vector.
69  * This must be compatible with the CUDA double2 type.
70  */
71 struct __builtin_align__(16) double2 { double x, y; };
72 typedef struct double2 double2;
73 #endif
74 
75 /**
76  * @brief Four-element complex structure (single precision).
77  *
78  * @details
79  * Structure used to hold data for a length-4 single precision complex vector.
80  * When used as a matrix, the elements should be interpreted as:
81  *
82  *   ( a  b )
83  *   ( c  d )
84  */
85 struct __align__(32) float4c { float2 a, b, c, d; };
86 typedef struct float4c float4c;
87 
88 /**
89  * @brief Four-element complex structure (double precision).
90  *
91  * @details
92  * Structure used to hold data for a length-4 double precision complex vector.
93  * When used as a matrix, the elements should be interpreted as:
94  *
95  *   ( a  b )
96  *   ( c  d )
97  */
98 struct __align__(64) double4c { double2 a, b, c, d; };
99 typedef struct double4c double4c;
100 
101 #endif /* include guard */
102