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
|
// ----------------------------------------------------------------------------
// - Open3D: www.open3d.org -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2024 www.open3d.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
// Eigen still applies __host__ __device__ annotation to defaulted functions for
// some classes. This causes compiler errors on Windows + CUDA. Disable CUDA
// support for Eigen until this issue has been fixed upstream.
#define EIGEN_NO_CUDA
#include <vector>
#include "open3d/Macro.h"
#include "open3d/core/Dispatch.h"
#include "open3d/core/Dtype.h"
#include "open3d/core/ParallelFor.h"
#include "open3d/core/Tensor.h"
#include "tests/Tests.h"
#include "tests/core/CoreTest.h"
namespace open3d {
namespace tests {
// CUDA does not allow using extended lambdas in private class scope like in
// googletest's internal test class.
void RunParallelForOn(core::Tensor& tensor) {
int64_t* tensor_data = tensor.GetDataPtr<int64_t>();
core::ParallelFor(
tensor.GetDevice(), tensor.NumElements(),
[=] OPEN3D_HOST_DEVICE(int64_t idx) { tensor_data[idx] = idx; });
}
TEST(ParallelFor, LambdaCUDA) {
const core::Device device("CUDA:0");
const size_t N = 10000000;
core::Tensor tensor({N, 1}, core::Int64, device);
RunParallelForOn(tensor);
core::Tensor tensor_cpu = tensor.To(core::Device("CPU:0"));
for (int64_t i = 0; i < tensor.NumElements(); ++i) {
ASSERT_EQ(tensor_cpu.GetDataPtr<int64_t>()[i], i);
}
}
} // namespace tests
} // namespace open3d
|
