Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Implement devices to devices reshard. (#3721)

Browse Source 
* Force clearing device memory before Reshard.
* Remove calculating row_segments for gpu_hist and gpu_sketch.
* Guard against changing device.
This commit is contained in:
trivialfis
2018-09-28 17:40:23 +12:00
committed by Rory Mitchell
parent 0b7fd74138
commit 5a7f7e7d49
11 changed files with 179 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

48
tests/cpp/common/test_host_device_vector.cu
View File

@@ -162,7 +162,7 @@ TEST(HostDeviceVector, TestCopy) {
std::vector<size_t> starts{0, 501};
std::vector<size_t> sizes{501, 500};
SetCudaSetDeviceHandler(SetDevice);
HostDeviceVector<int> v;
{
// a separate scope to ensure that v1 is gone before further checks
@@ -178,6 +178,52 @@ TEST(HostDeviceVector, TestCopy) {
SetCudaSetDeviceHandler(nullptr);
}
// The test is not really useful if n_gpus < 2
TEST(HostDeviceVector, Reshard) {
std::vector<int> h_vec (2345);
for (size_t i = 0; i < h_vec.size(); ++i) {
h_vec[i] = i;
}
HostDeviceVector<int> vec (h_vec);
auto devices = GPUSet::AllVisible();
std::vector<size_t> devices_size (devices.Size());
// From CPU to GPUs.
// Assuming we have > 1 devices.
vec.Reshard(devices);
size_t total_size = 0;
for (size_t i = 0; i < devices.Size(); ++i) {
total_size += vec.DeviceSize(i);
devices_size[i] = vec.DeviceSize(i);
}
ASSERT_EQ(total_size, h_vec.size());
ASSERT_EQ(total_size, vec.Size());
auto h_vec_1 = vec.HostVector();
ASSERT_TRUE(std::equal(h_vec_1.cbegin(), h_vec_1.cend(), h_vec.cbegin()));
vec.Reshard(GPUSet::Empty()); // clear out devices memory
// Shrink down the number of devices.
vec.Reshard(GPUSet::Range(0, 1));
ASSERT_EQ(vec.Size(), h_vec.size());
ASSERT_EQ(vec.DeviceSize(0), h_vec.size());
h_vec_1 = vec.HostVector();
ASSERT_TRUE(std::equal(h_vec_1.cbegin(), h_vec_1.cend(), h_vec.cbegin()));
vec.Reshard(GPUSet::Empty()); // clear out devices memory
// Grow the number of devices.
vec.Reshard(devices);
total_size = 0;
for (size_t i = 0; i < devices.Size(); ++i) {
total_size += vec.DeviceSize(i);
ASSERT_EQ(devices_size[i], vec.DeviceSize(i));
}
ASSERT_EQ(total_size, h_vec.size());
ASSERT_EQ(total_size, vec.Size());
h_vec_1 = vec.HostVector();
ASSERT_TRUE(std::equal(h_vec_1.cbegin(), h_vec_1.cend(), h_vec.cbegin()));
}
TEST(HostDeviceVector, Span) {
HostDeviceVector<float> vec {1.0f, 2.0f, 3.0f, 4.0f};
vec.Reshard(GPUSet{0, 1});