Files
ceres-solver/internal/ceres/context_impl.cc
T

221 lines
7.6 KiB
C++
Raw Normal View History

2018-02-22 10:28:39 -08:00
// Ceres Solver - A fast non-linear least squares minimizer
2023-09-19 15:29:34 -07:00
// Copyright 2023 Google Inc. All rights reserved.
2018-02-22 10:28:39 -08:00
// http://ceres-solver.org/
//
// 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 Google Inc. 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 OWNER 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.
//
// Author: vitus@google.com (Michael Vitus)
#include "ceres/context_impl.h"
2022-02-14 20:56:30 -06:00
#include <string>
2022-03-12 15:22:19 -08:00
#include "ceres/internal/config.h"
2022-08-17 19:19:11 -05:00
#include "ceres/stringprintf.h"
2022-08-07 17:14:32 -05:00
#include "ceres/wall_time.h"
2022-03-12 15:22:19 -08:00
2022-02-14 20:56:30 -06:00
#ifndef CERES_NO_CUDA
#include "cublas_v2.h"
#include "cuda_runtime.h"
2022-02-14 20:56:30 -06:00
#include "cusolverDn.h"
#endif // CERES_NO_CUDA
2022-04-21 17:41:10 -07:00
namespace ceres::internal {
2018-02-22 10:28:39 -08:00
2022-02-09 20:10:26 +01:00
ContextImpl::ContextImpl() = default;
2022-02-14 20:56:30 -06:00
#ifndef CERES_NO_CUDA
2022-08-07 17:14:32 -05:00
void ContextImpl::TearDown() {
if (cusolver_handle_ != nullptr) {
cusolverDnDestroy(cusolver_handle_);
cusolver_handle_ = nullptr;
}
if (cublas_handle_ != nullptr) {
cublasDestroy(cublas_handle_);
cublas_handle_ = nullptr;
}
2022-09-28 17:35:40 -07:00
if (cusparse_handle_ != nullptr) {
2022-08-07 17:14:32 -05:00
cusparseDestroy(cusparse_handle_);
cusparse_handle_ = nullptr;
}
for (auto& s : streams_) {
if (s != nullptr) {
cudaStreamDestroy(s);
s = nullptr;
}
2022-08-07 17:14:32 -05:00
}
2022-08-13 16:20:05 -05:00
is_cuda_initialized_ = false;
2022-08-07 17:14:32 -05:00
}
2022-08-17 19:19:11 -05:00
std::string ContextImpl::CudaConfigAsString() const {
return ceres::internal::StringPrintf(
"======================= CUDA Device Properties ======================\n"
2023-09-22 15:25:58 +00:00
"Cuda version : %d.%d\n"
"Device ID : %d\n"
"Device name : %s\n"
"Total GPU memory : %6.f MiB\n"
"GPU memory available : %6.f MiB\n"
"Compute capability : %d.%d\n"
"Warp size : %d\n"
"Max threads per block : %d\n"
"Max threads per dim : %d %d %d\n"
"Max grid size : %d %d %d\n"
"Multiprocessor count : %d\n"
"cudaMallocAsync supported : %s\n"
2022-08-17 19:19:11 -05:00
"====================================================================",
cuda_version_major_,
cuda_version_minor_,
gpu_device_id_in_use_,
gpu_device_properties_.name,
gpu_device_properties_.totalGlobalMem / 1024.0 / 1024.0,
GpuMemoryAvailable() / 1024.0 / 1024.0,
gpu_device_properties_.major,
gpu_device_properties_.minor,
gpu_device_properties_.warpSize,
gpu_device_properties_.maxThreadsPerBlock,
gpu_device_properties_.maxThreadsDim[0],
gpu_device_properties_.maxThreadsDim[1],
gpu_device_properties_.maxThreadsDim[2],
gpu_device_properties_.maxGridSize[0],
gpu_device_properties_.maxGridSize[1],
gpu_device_properties_.maxGridSize[2],
2023-09-22 15:25:58 +00:00
gpu_device_properties_.multiProcessorCount,
gpu_device_properties_.memoryPoolsSupported ? "Yes" : "No");
2022-08-17 19:19:11 -05:00
}
size_t ContextImpl::GpuMemoryAvailable() const {
size_t free, total;
cudaMemGetInfo(&free, &total);
return free;
}
bool ContextImpl::InitCuda(std::string* message) {
2022-08-13 16:20:05 -05:00
if (is_cuda_initialized_) {
2022-02-14 20:56:30 -06:00
return true;
}
2022-08-17 19:19:11 -05:00
CHECK_EQ(cudaGetDevice(&gpu_device_id_in_use_), cudaSuccess);
int cuda_version;
CHECK_EQ(cudaRuntimeGetVersion(&cuda_version), cudaSuccess);
cuda_version_major_ = cuda_version / 1000;
cuda_version_minor_ = (cuda_version % 1000) / 10;
2022-09-01 17:00:34 -07:00
CHECK_EQ(
cudaGetDeviceProperties(&gpu_device_properties_, gpu_device_id_in_use_),
cudaSuccess);
2022-08-17 19:19:11 -05:00
VLOG(3) << "\n" << CudaConfigAsString();
2022-08-07 17:14:32 -05:00
EventLogger event_logger("InitCuda");
2022-02-14 20:56:30 -06:00
if (cublasCreate(&cublas_handle_) != CUBLAS_STATUS_SUCCESS) {
2022-09-01 17:00:34 -07:00
*message =
"CUDA initialization failed because cuBLAS::cublasCreate failed.";
2022-02-14 20:56:30 -06:00
cublas_handle_ = nullptr;
return false;
}
2022-08-07 17:14:32 -05:00
event_logger.AddEvent("cublasCreate");
2022-02-14 20:56:30 -06:00
if (cusolverDnCreate(&cusolver_handle_) != CUSOLVER_STATUS_SUCCESS) {
2022-09-01 17:00:34 -07:00
*message =
"CUDA initialization failed because cuSolverDN::cusolverDnCreate "
"failed.";
2022-08-07 17:14:32 -05:00
TearDown();
return false;
}
event_logger.AddEvent("cusolverDnCreate");
if (cusparseCreate(&cusparse_handle_) != CUSPARSE_STATUS_SUCCESS) {
2022-09-01 17:00:34 -07:00
*message =
"CUDA initialization failed because cuSPARSE::cusparseCreate failed.";
2022-08-07 17:14:32 -05:00
TearDown();
2022-02-14 20:56:30 -06:00
return false;
}
2022-08-07 17:14:32 -05:00
event_logger.AddEvent("cusparseCreate");
for (auto& s : streams_) {
if (cudaStreamCreateWithFlags(&s, cudaStreamNonBlocking) != cudaSuccess) {
*message =
"CUDA initialization failed because CUDA::cudaStreamCreateWithFlags "
"failed.";
TearDown();
return false;
}
2022-02-14 20:56:30 -06:00
}
2022-08-07 17:14:32 -05:00
event_logger.AddEvent("cudaStreamCreateWithFlags");
if (cusolverDnSetStream(cusolver_handle_, DefaultStream()) !=
CUSOLVER_STATUS_SUCCESS ||
cublasSetStream(cublas_handle_, DefaultStream()) !=
CUBLAS_STATUS_SUCCESS ||
cusparseSetStream(cusparse_handle_, DefaultStream()) !=
CUSPARSE_STATUS_SUCCESS) {
2022-08-17 19:19:11 -05:00
*message = "CUDA initialization failed because SetStream failed.";
2022-08-07 17:14:32 -05:00
TearDown();
2022-02-14 20:56:30 -06:00
return false;
}
2022-08-07 17:14:32 -05:00
event_logger.AddEvent("SetStream");
2022-08-13 16:20:05 -05:00
is_cuda_initialized_ = true;
2022-02-14 20:56:30 -06:00
return true;
}
2023-09-22 15:25:58 +00:00
void* ContextImpl::CudaMalloc(size_t size, cudaStream_t stream) const {
void* data = nullptr;
// Stream-ordered alloaction API is available since CUDA 11.4, but might be
// not implemented by particular device
#if CUDART_VERSION < 11040
#warning \
"Stream-ordered allocations are unavailable, consider updating CUDA toolkit to version 11.4+"
CHECK_EQ(cudaSuccess, cudaMalloc(&data, size));
#else
if (gpu_device_properties_.memoryPoolsSupported) {
CHECK_EQ(cudaSuccess, cudaMallocAsync(&data, size, stream));
} else {
CHECK_EQ(cudaSuccess, cudaMalloc(&data, size));
}
#endif
return data;
}
void ContextImpl::CudaFree(void* data, cudaStream_t stream) const {
// Stream-ordered alloaction API is available since CUDA 11.4, but might be
// not implemented by particular device
#if CUDART_VERSION < 11040
#warning \
"Stream-ordered allocations are unavailable, consider updating CUDA toolkit to version 11.4+"
CHECK_EQ(cudaSuccess, cudaFree(data));
#else
if (gpu_device_properties_.memoryPoolsSupported) {
CHECK_EQ(cudaSuccess, cudaFreeAsync(data, stream));
} else {
CHECK_EQ(cudaSuccess, cudaFree(data));
}
#endif
}
2022-02-14 20:56:30 -06:00
#endif // CERES_NO_CUDA
ContextImpl::~ContextImpl() {
#ifndef CERES_NO_CUDA
2022-08-07 17:14:32 -05:00
TearDown();
2022-02-14 20:56:30 -06:00
#endif // CERES_NO_CUDA
}
2022-11-26 16:02:28 -08:00
2018-02-22 10:28:39 -08:00
void ContextImpl::EnsureMinimumThreads(int num_threads) {
thread_pool.Resize(num_threads);
}
2022-11-26 16:02:28 -08:00
2022-04-21 17:41:10 -07:00
} // namespace ceres::internal