CUTLASS 2.0 (#62)
CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
This commit is contained in:
Andrew Kerr
@@ -0,0 +1,61 @@
|
||||
/***************************************************************************************************
|
||||
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without modification, are permitted
|
||||
* provided that the following conditions are met:
|
||||
* * Redistributions of source code must retain the above copyright notice, this list of
|
||||
* conditions and the following disclaimer.
|
||||
* * 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.
|
||||
* * Neither the name of the NVIDIA CORPORATION 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 NVIDIA CORPORATION 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 TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*
|
||||
**************************************************************************************************/
|
||||
|
||||
#pragma once
|
||||
#pragma warning (disable : 4068 ) /* disable unknown pragma warnings for vistual studio */
|
||||
|
||||
#pragma diag_suppress boolean_controlling_expr_is_constant
|
||||
#include <gtest/gtest.h>
|
||||
#pragma diag_warning boolean_controlling_expr_is_constant
|
||||
#pragma warning( disable : 4503)
|
||||
|
||||
/////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
/// Sets flags for Unit test
|
||||
void FilterArchitecture();
|
||||
|
||||
/////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
// active test macro
|
||||
#define CUTLASS_TEST_LEVEL_ACTIVE(LEVEL,NAME_STATIC,NAME_DYNAMIC,...) \
|
||||
TEST(NAME_STATIC,L##LEVEL##_##NAME_DYNAMIC) __VA_ARGS__
|
||||
|
||||
// disabled test macro
|
||||
#define CUTLASS_TEST_LEVEL_DISABLED(LEVEL,NAME_STATIC,NAME_DYNAMIC,...) \
|
||||
TEST(NAME_STATIC,DISABLED_L##LEVEL##_##NAME_DYNAMIC) {}
|
||||
|
||||
#if CUTLASS_TEST_LEVEL == 0
|
||||
#define CUTLASS_TEST_L0(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(0,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L1(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_DISABLED(1,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L2(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_DISABLED(2,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#elif CUTLASS_TEST_LEVEL == 1
|
||||
#define CUTLASS_TEST_L0(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(0,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L1(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(1,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L2(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_DISABLED(2,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#else
|
||||
#define CUTLASS_TEST_L0(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(0,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L1(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(1,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#define CUTLASS_TEST_L2(NAME_STATIC,NAME_DYNAMIC,...) CUTLASS_TEST_LEVEL_ACTIVE(2,NAME_STATIC,NAME_DYNAMIC,__VA_ARGS__)
|
||||
#endif
|
||||
@@ -0,0 +1,107 @@
|
||||
/***************************************************************************************************
|
||||
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without modification, are permitted
|
||||
* provided that the following conditions are met:
|
||||
* * Redistributions of source code must retain the above copyright notice, this list of
|
||||
* conditions and the following disclaimer.
|
||||
* * 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.
|
||||
* * Neither the name of the NVIDIA CORPORATION 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 NVIDIA CORPORATION 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 TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*
|
||||
**************************************************************************************************/
|
||||
|
||||
#include <cuda_runtime_api.h>
|
||||
|
||||
#include "cutlass_unit_test.h"
|
||||
|
||||
/////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
/// Sets flags for Unit test
|
||||
void FilterArchitecture() {
|
||||
// Default flags can be overwritten by --gtest_filter from commandline
|
||||
cudaError_t err;
|
||||
|
||||
int cudaDeviceId;
|
||||
err = cudaGetDevice(&cudaDeviceId);
|
||||
if (cudaSuccess != err) {
|
||||
std::cerr << "*** Error: Could not detect active GPU device ID"
|
||||
<< " [" << cudaGetErrorString(err) << "]" << std::endl;
|
||||
exit(1);
|
||||
}
|
||||
|
||||
cudaDeviceProp deviceProperties;
|
||||
err = cudaGetDeviceProperties(&deviceProperties, cudaDeviceId);
|
||||
if (cudaSuccess != err) {
|
||||
std::cerr << "*** Error: Could not get device properties for GPU " << cudaDeviceId << " ["
|
||||
<< cudaGetErrorString(err) << "]" << std::endl;
|
||||
exit(1);
|
||||
}
|
||||
|
||||
int deviceMajorMinor = deviceProperties.major * 10 + deviceProperties.minor;
|
||||
int const kMaxDevice = 999;
|
||||
|
||||
// Defines text filters for each GEMM kernel based on minimum supported compute capability
|
||||
struct {
|
||||
|
||||
/// Unit test filter string
|
||||
char const *filter;
|
||||
|
||||
/// Minimum compute capability for the kernels in the named test
|
||||
int min_compute_capability;
|
||||
|
||||
/// Maximum compute capability for which the kernels are enabled
|
||||
int max_compute_capability;
|
||||
|
||||
/// If true, architecture is assumed to be silicon
|
||||
bool silicon;
|
||||
|
||||
}
|
||||
test_filters[] = {
|
||||
{ "SM50*", 50, kMaxDevice, true},
|
||||
{ "SM60*", 60, kMaxDevice, true},
|
||||
{ "SM61*", 61, kMaxDevice, true},
|
||||
{ "SM70*", 70, 75, true},
|
||||
{ "SM75*", 75, kMaxDevice, true},
|
||||
{ 0, 0, false }
|
||||
};
|
||||
|
||||
bool running_on_silicon = false;
|
||||
for (int i = 0; test_filters[i].filter; ++i) {
|
||||
if (deviceMajorMinor == test_filters[i].min_compute_capability) {
|
||||
running_on_silicon = test_filters[i].silicon;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// Set negative test filters
|
||||
std::stringstream ss;
|
||||
ss << "-";
|
||||
for (int i = 0, j = 0; test_filters[i].filter; ++i) {
|
||||
|
||||
if (!running_on_silicon && deviceMajorMinor != test_filters[i].min_compute_capability) {
|
||||
ss << (j++ ? ":" : "") << test_filters[i].filter;
|
||||
}
|
||||
else if (deviceMajorMinor < test_filters[i].min_compute_capability ||
|
||||
deviceMajorMinor > test_filters[i].max_compute_capability) {
|
||||
|
||||
ss << (j++ ? ":" : "") << test_filters[i].filter;
|
||||
}
|
||||
}
|
||||
|
||||
::testing::GTEST_FLAG(filter) = ss.str();
|
||||
}
|
||||
|
||||
/////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
Reference in New Issue
Block a user