forked from NVIDIA/cutlass
-
Notifications
You must be signed in to change notification settings - Fork 21
/
cutlass.h
169 lines (151 loc) · 6.5 KB
/
cutlass.h
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
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
/***************************************************************************************************
* Copyright (c) 2017 - 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: BSD-3-Clause
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice, this
* list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* 3. Neither the name of the copyright holder nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
**************************************************************************************************/
/*! \file
\brief Basic include for CUTLASS.
*/
#pragma once
#include "cutlass/arch/synclog.hpp"
#include "cutlass/detail/helper_macros.hpp"
#include <cutlass/gpu_generics.h>
////////////////////////////////////////////////////////////////////////////////////////////////////
namespace cutlass {
/// Status code returned by CUTLASS operations
enum class Status {
kSuccess, ///< Operation was successful.
kErrorMisalignedOperand, ///< operands fail alignment requirements.
kErrorInvalidDataType, ///< DataType fails requirement.
kErrorInvalidLayout, ///< Layout fails alignment requirement.
kErrorInvalidProblem, ///< Specified problem size is not supported by operator.
kErrorNotSupported, ///< Operation is not supported on current device.
kErrorWorkspaceNull, ///< The given workspace is null when it is required to be non-null.
kErrorInternal, ///< An error within CUTLASS occurred.
kErrorArchMismatch, ///< CUTLASS runs on a device that it was not compiled for.
kErrorInsufficientDriver, ///< CUTLASS runs with a driver that is too old.
kErrorMemoryAllocation, ///< Kernel launch failed due to insufficient device memory.
kInvalid ///< Status is unspecified.
};
/// Convert cutlass status to status strings
CUTLASS_HOST_DEVICE
static char const* cutlassGetStatusString(cutlass::Status status) {
switch (status) {
case cutlass::Status::kSuccess:
return "Success";
case cutlass::Status::kErrorMisalignedOperand:
return "Error Misaligned Operand";
case cutlass::Status::kErrorInvalidDataType:
return "Error Invalid Data Type";
case cutlass::Status::kErrorInvalidLayout:
return "Error Invalid Layout";
case cutlass::Status::kErrorInvalidProblem:
return "Error Invalid Problem";
case cutlass::Status::kErrorNotSupported:
return "Error Not Supported";
case cutlass::Status::kErrorWorkspaceNull:
return "Error Workspace Null";
case cutlass::Status::kErrorInternal:
return "Error Internal";
case cutlass::Status::kErrorInsufficientDriver:
return "Error Insufficient Driver";
case cutlass::Status::kErrorArchMismatch:
return "Error Architecture Mismatch";
case cutlass::Status::kErrorMemoryAllocation:
return "Error Memory Allocation failed";
case cutlass::Status::kInvalid: break;
}
return "Invalid status";
}
////////////////////////////////////////////////////////////////////////////////////////////////////
/// Helper function to return true when called by thread 0 of threadblock 0.
CUTLASS_HOST_DEVICE bool thread0() {
#if defined(__CUDA_ARCH__)
return (!threadIdx.x && !threadIdx.y && !threadIdx.z) && (!blockIdx.x && !blockIdx.y && !blockIdx.z);
#elif defined(__SYCL_DEVICE_ONLY__)
return (!syclcompat::global_id::x() && !syclcompat::global_id::y() && !syclcompat::global_id::z());
#else
return false;
#endif
}
/// Returns a lane index in the warp. The threads in warp may not be convergent
CUTLASS_DEVICE
int canonical_lane_idx() {
#if defined(__CUDA_ARCH__)
return threadIdx.x % NumThreadsPerWarp;
#else
return 0;
#endif
}
/// Returns a warp-uniform value indicating the canonical warp index of the calling threads.
/// Threads within the warp must be converged.
CUTLASS_DEVICE
int canonical_warp_idx_sync() {
#if defined(__CUDA_ARCH__)
return __shfl_sync(0xffffffff, threadIdx.x / NumThreadsPerWarp, 0);
#else
return 0;
#endif
}
/// Returns a warp index in the CTA. The threads in warp may not be convergent
/// As it doesn't sync the warp, it faster and allows forward progress
CUTLASS_DEVICE
int canonical_warp_idx() {
#if defined(__CUDA_ARCH__)
return threadIdx.x / NumThreadsPerWarp;
#else
return 0;
#endif
}
/// Returns a warp-uniform value indicating the canonical warp group index of the calling threads.
/// Threads within the warp must be converged.
CUTLASS_DEVICE
int canonical_warp_group_idx() {
#if defined(__CUDA_ARCH__)
return __shfl_sync(0xffffffff, threadIdx.x / NumThreadsPerWarpGroup, 0);
#else
return 0;
#endif
}
#if defined(SYCL_INTEL_TARGET)
CUTLASS_DEVICE
auto get_sub_group_id() {
return sycl::ext::oneapi::this_work_item::get_nd_item<3>()
.get_sub_group()
.get_group_id()[0];
}
CUTLASS_DEVICE
auto get_sub_group_local_id() {
return sycl::ext::oneapi::this_work_item::get_nd_item<3>()
.get_sub_group()
.get_local_id()[0];
}
#endif
////////////////////////////////////////////////////////////////////////////////////////////////////
} // namespace cutlass
////////////////////////////////////////////////////////////////////////////////////////////////////