QUDA  v1.1.0
A library for QCD on GPUs
fallback.h
Go to the documentation of this file.
1 /*
2 Copyright (c) 2013, NVIDIA Corporation
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  * Redistributions of source code must retain the above copyright
8  notice, this list of conditions and the following disclaimer.
9  * Redistributions in binary form must reproduce the above copyright
10  notice, this list of conditions and the following disclaimer in the
11  documentation and/or other materials provided with the distribution.
12  * Neither the name of the <organization> nor the
13  names of its contributors may be used to endorse or promote products
14  derived from this software without specific prior written permission.
15 
16 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
17 ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19 DISCLAIMED. IN NO EVENT SHALL <COPYRIGHT HOLDER> BE LIABLE FOR ANY
20 DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
22 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
23 ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27 
28 #pragma once
29 #include <trove/utility.h>
30 #include <trove/detail/dismember.h>
31 
32 namespace trove {
33 namespace detail {
34 
35 template<int s, typename T>
38  __device__
39  static array<U, s> impl(const U* src) {
40  return array<U, s>(*src,
42  }
43  __device__
44  static array<U, s> impl(const T* src) {
45  return impl((U*)src);
46  }
47 };
48 
49 template<typename T>
50 struct divergent_loader<1, T> {
52  __device__
53  static array<U, 1> impl(const U* src) {
54  return array<U, 1>(*src);
55  }
56  __device__
57  static array<U, 1> impl(const T* src) {
58  return impl((U*)src);
59  }
60 };
61 
62 template<typename T>
63 struct use_divergent {
64  static const bool value = (sizeof(T) % 4) == 0;
65 };
66 
67 template<typename T>
68 __device__
69 typename enable_if<use_divergent<T>::value, T>::type
70 divergent_load(const T* src) {
71  typedef typename detail::dismember_type<T>::type U;
73  u_store loaded =
75  src);
76  return detail::fuse<T>(loaded);
77 }
78 
79 template<typename T>
80 __device__
81 typename enable_if<!use_divergent<T>::value, T>::type
82 divergent_load(const T* src) {
83  return *src;
84 }
85 
86 template<int s, typename T>
89  __device__
90  static void impl(const array<U, s>& data, U* dest) {
91  *dest = data.head;
93  }
94  __device__
95  static void impl(const array<U, s>& data, const T* dest) {
96  return impl(data, (U*)dest);
97  }
98 };
99 
100 template<typename T>
101 struct divergent_storer<1, T> {
103  __device__
104  static void impl(const array<U, 1>& data, U* dest) {
105  *dest = data.head;
106  }
107  __device__
108  static void impl(const array<U, 1>& data, const T* dest) {
109  return impl(data, (U*)dest);
110  }
111 };
112 
113 template<typename T>
114 __device__
115 typename enable_if<use_divergent<T>::value>::type
116 divergent_store(const T& data, T* dest) {
117  typedef typename detail::dismember_type<T>::type U;
119  u_store lysed = detail::lyse<U>(data);
121  lysed, dest);
122 }
123 
124 template<typename T>
125 __device__
127 divergent_store(const T& data, T* dest) {
128  *dest = data;
129 }
130 
131 
132 }
133 }
Definition: alias.h:4
__device__ enable_if< use_divergent< T >::value, T >::type divergent_load(const T *src)
Definition: fallback.h:70
__device__ enable_if< use_divergent< T >::value >::type divergent_store(const T &data, T *dest)
Definition: fallback.h:116
Definition: aos.h:38
head_type head
Definition: array.h:38
tail_type tail
Definition: array.h:39
static __device__ array< U, 1 > impl(const U *src)
Definition: fallback.h:53
static __device__ array< U, 1 > impl(const T *src)
Definition: fallback.h:57
detail::dismember_type< T >::type U
Definition: fallback.h:51
static __device__ array< U, s > impl(const U *src)
Definition: fallback.h:39
detail::dismember_type< T >::type U
Definition: fallback.h:37
static __device__ array< U, s > impl(const T *src)
Definition: fallback.h:44
static __device__ void impl(const array< U, 1 > &data, U *dest)
Definition: fallback.h:104
static __device__ void impl(const array< U, 1 > &data, const T *dest)
Definition: fallback.h:108
detail::dismember_type< T >::type U
Definition: fallback.h:102
static __device__ void impl(const array< U, s > &data, U *dest)
Definition: fallback.h:90
detail::dismember_type< T >::type U
Definition: fallback.h:88
static __device__ void impl(const array< U, s > &data, const T *dest)
Definition: fallback.h:95
static const bool value
Definition: fallback.h:64