1 //====--- OMPGridValues.h - Language-specific address spaces --*- C++ -*-====//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 ///
9 /// \file
10 /// \brief Provides definitions for Target specific Grid Values
11 ///
12 //===----------------------------------------------------------------------===//
13 
14 #ifndef LLVM_FRONTEND_OPENMP_OMPGRIDVALUES_H
15 #define LLVM_FRONTEND_OPENMP_OMPGRIDVALUES_H
16 
17 namespace llvm {
18 
19 namespace omp {
20 
21 /// \brief Defines various target-specific GPU grid values that must be
22 ///        consistent between host RTL (plugin), device RTL, and clang.
23 ///        We can change grid values for a "fat" binary so that different
24 ///        passes get the correct values when generating code for a
25 ///        multi-target binary. Both amdgcn and nvptx values are stored in
26 ///        this file. In the future, should there be differences between GPUs
27 ///        of the same architecture, then simply make a different array and
28 ///        use the new array name.
29 ///
30 /// Example usage in clang:
31 ///   const unsigned slot_size =
32 ///   ctx.GetTargetInfo().getGridValue().GV_Warp_Size;
33 ///
34 /// Example usage in libomptarget/deviceRTLs:
35 ///   #include "llvm/Frontend/OpenMP/OMPGridValues.h"
36 ///   #ifdef __AMDGPU__
37 ///     #define GRIDVAL AMDGPUGridValues
38 ///   #else
39 ///     #define GRIDVAL NVPTXGridValues
40 ///   #endif
41 ///   ... Then use this reference for GV_Warp_Size in the deviceRTL source.
42 ///   llvm::omp::GRIDVAL().GV_Warp_Size
43 ///
44 /// Example usage in libomptarget hsa plugin:
45 ///   #include "llvm/Frontend/OpenMP/OMPGridValues.h"
46 ///   #define GRIDVAL AMDGPUGridValues
47 ///   ... Then use this reference to access GV_Warp_Size in the hsa plugin.
48 ///   llvm::omp::GRIDVAL().GV_Warp_Size
49 ///
50 /// Example usage in libomptarget cuda plugin:
51 ///    #include "llvm/Frontend/OpenMP/OMPGridValues.h"
52 ///    #define GRIDVAL NVPTXGridValues
53 ///   ... Then use this reference to access GV_Warp_Size in the cuda plugin.
54 ///    llvm::omp::GRIDVAL().GV_Warp_Size
55 ///
56 
57 struct GV {
58   /// The size reserved for data in a shared memory slot.
59   const unsigned GV_Slot_Size;
60   /// The default value of maximum number of threads in a worker warp.
61   const unsigned GV_Warp_Size;
62 
63   constexpr unsigned warpSlotSize() const {
64     return GV_Warp_Size * GV_Slot_Size;
65   }
66 
67   /// the maximum number of teams.
68   const unsigned GV_Max_Teams;
69   // An alternative to the heavy data sharing infrastructure that uses global
70   // memory is one that uses device __shared__ memory.  The amount of such space
71   // (in bytes) reserved by the OpenMP runtime is noted here.
72   const unsigned GV_SimpleBufferSize;
73   // The absolute maximum team size for a working group
74   const unsigned GV_Max_WG_Size;
75   // The default maximum team size for a working group
76   const unsigned GV_Default_WG_Size;
77 
78   constexpr unsigned maxWarpNumber() const {
79     return GV_Max_WG_Size / GV_Warp_Size;
80   }
81 };
82 
83 /// For AMDGPU GPUs
84 static constexpr GV AMDGPUGridValues64 = {
85     256,  // GV_Slot_Size
86     64,   // GV_Warp_Size
87     128,  // GV_Max_Teams
88     896,  // GV_SimpleBufferSize
89     1024, // GV_Max_WG_Size,
90     256,  // GV_Default_WG_Size
91 };
92 
93 static constexpr GV AMDGPUGridValues32 = {
94     256,  // GV_Slot_Size
95     32,   // GV_Warp_Size
96     128,  // GV_Max_Teams
97     896,  // GV_SimpleBufferSize
98     1024, // GV_Max_WG_Size,
99     256,  // GV_Default_WG_Size
100 };
101 
102 template <unsigned wavesize> constexpr const GV &getAMDGPUGridValues() {
103   static_assert(wavesize == 32 || wavesize == 64, "");
104   return wavesize == 32 ? AMDGPUGridValues32 : AMDGPUGridValues64;
105 }
106 
107 /// For Nvidia GPUs
108 static constexpr GV NVPTXGridValues = {
109     256,  // GV_Slot_Size
110     32,   // GV_Warp_Size
111     1024, // GV_Max_Teams
112     896,  // GV_SimpleBufferSize
113     1024, // GV_Max_WG_Size
114     128,  // GV_Default_WG_Size
115 };
116 
117 } // namespace omp
118 } // namespace llvm
119 
120 #endif // LLVM_FRONTEND_OPENMP_OMPGRIDVALUES_H
121