MFEM  v4.5.1
Finite element discretization library
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Pages
simd.hpp
Go to the documentation of this file.
1 // Copyright (c) 2010-2022, Lawrence Livermore National Security, LLC. Produced
2 // at the Lawrence Livermore National Laboratory. All Rights reserved. See files
3 // LICENSE and NOTICE for details. LLNL-CODE-806117.
4 //
5 // This file is part of the MFEM library. For more information and source code
6 // availability visit https://mfem.org.
7 //
8 // MFEM is free software; you can redistribute it and/or modify it under the
9 // terms of the BSD-3 license. We welcome feedback and contributions, see file
10 // CONTRIBUTING.md for details.
11 
12 #ifndef MFEM_SIMD_HPP
13 #define MFEM_SIMD_HPP
14 
15 #include "../config/tconfig.hpp"
16 
17 // --- AutoSIMD + specializations with intrinsics
18 #include "simd/auto.hpp"
19 #ifdef MFEM_USE_SIMD
20 #if defined(__aarch64__) && defined(__ARM_FEATURE_SVE)
21 #include "simd/sve.hpp"
22 #elif defined(__VSX__)
23 #include "simd/vsx.hpp"
24 #elif defined (__bgq__)
25 #include "simd/qpx.hpp"
26 #elif defined(__x86_64__) || defined(_M_X64) || defined(_M_IX86)
27 #include "simd/x86.hpp"
28 #elif !defined(_MSC_VER)
29 #warning Unknown SIMD architecture
30 #else
31 #pragma message("warning: Unknown SIMD architecture")
32 #endif
33 #endif
34 
35 // MFEM_SIMD_BYTES is the default SIMD size used by MFEM, see e.g. class
36 // TBilinearForm and the default traits class AutoSIMDTraits.
37 // MFEM_ALIGN_BYTES determines the padding used in TVector when its 'align'
38 // template parameter is set to true -- it ensures that the size of such TVector
39 // types is a multiple of MFEM_ALIGN_BYTES. MFEM_ALIGN_BYTES must be a multiple
40 // of MFEM_SIMD_BYTES.
41 #if !defined(MFEM_USE_SIMD)
42 #define MFEM_SIMD_BYTES 8
43 #define MFEM_ALIGN_BYTES 32
44 #elif defined(__AVX512F__)
45 #define MFEM_SIMD_BYTES 64
46 #define MFEM_ALIGN_BYTES 64
47 #elif defined(__aarch64__) && defined(__ARM_FEATURE_SVE)
48 #define MFEM_SIMD_BYTES 64
49 #define MFEM_ALIGN_BYTES 64
50 #elif defined(__AVX__) || defined(__VECTOR4DOUBLE__)
51 #define MFEM_SIMD_BYTES 32
52 #define MFEM_ALIGN_BYTES 32
53 #elif defined(__SSE2__) || defined(__VSX__)
54 #define MFEM_SIMD_BYTES 16
55 #define MFEM_ALIGN_BYTES 32
56 #else
57 #define MFEM_SIMD_BYTES 8
58 #define MFEM_ALIGN_BYTES 32
59 #endif
60 
61 // derived macros
62 #define MFEM_ROUNDUP(val,base) ((((val)+(base)-1)/(base))*(base))
63 #define MFEM_ALIGN_SIZE(size,type) \
64  MFEM_ROUNDUP(size,(MFEM_ALIGN_BYTES)/sizeof(type))
65 
66 namespace mfem
67 {
68 
69 template<typename complex_t, typename real_t>
71 {
72  static const int block_size = MFEM_TEMPLATE_BLOCK_SIZE;
73 
74  // Alignment for arrays of vcomplex_t and vreal_t
75  static const int align_bytes = MFEM_SIMD_BYTES;
76 
77  static const int batch_size = 1;
78 
79  static const int simd_size = MFEM_SIMD_BYTES/sizeof(real_t);
80 
84 };
85 
86 template<typename complex_t, typename real_t>
88 {
89  static const int block_size = MFEM_TEMPLATE_BLOCK_SIZE;
90 
91  // Alignment for arrays of vcomplex_t and vreal_t
92  static const int align_bytes = sizeof(real_t);
93 
94  static const int batch_size = 1;
95 
96  static const int simd_size = 1;
97 
101 };
102 
103 } // mfem namespace
104 
105 #endif // MFEM_SIMD_HPP
static const int simd_size
Definition: simd.hpp:96
AutoSIMD< complex_t, simd_size, align_bytes > vcomplex_t
Definition: simd.hpp:98
static const int batch_size
Definition: simd.hpp:94
static const int align_bytes
Definition: simd.hpp:75
AutoSIMD< real_t, simd_size, align_bytes > vreal_t
Definition: simd.hpp:99
static const int batch_size
Definition: simd.hpp:77
static const int block_size
Definition: simd.hpp:72
AutoSIMD< int, simd_size, simd_size *sizeof(int)> vint_t
Definition: simd.hpp:100
AutoSIMD< int, simd_size, simd_size *sizeof(int)> vint_t
Definition: simd.hpp:83
static const int block_size
Definition: simd.hpp:89
static const int simd_size
Definition: simd.hpp:79
AutoSIMD< real_t, simd_size, MFEM_SIMD_BYTES > vreal_t
Definition: simd.hpp:82
static const int align_bytes
Definition: simd.hpp:92
AutoSIMD< complex_t, simd_size, MFEM_SIMD_BYTES > vcomplex_t
Definition: simd.hpp:81