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

RMM integration plugin (#5873)

Browse Source 
* [CI] Add RMM as an optional dependency

* Replace caching allocator with pool allocator from RMM

* Revert "Replace caching allocator with pool allocator from RMM"

This reverts commit e15845d4e72e890c2babe31a988b26503a7d9038.

* Use rmm::mr::get_default_resource()

* Try setting default resource (doesn't work yet)

* Allocate pool_mr in the heap

* Prevent leaking pool_mr handle

* Separate EXPECT_DEATH() in separate test suite suffixed DeathTest

* Turn off death tests for RMM

* Address reviewer's feedback

* Prevent leaking of cuda_mr

* Fix Jenkinsfile syntax

* Remove unnecessary function in Jenkinsfile

* [CI] Install NCCL into RMM container

* Run Python tests

* Try building with RMM, CUDA 10.0

* Do not use RMM for CUDA 10.0 target

* Actually test for test_rmm flag

* Fix TestPythonGPU

* Use CNMeM allocator, since pool allocator doesn't yet support multiGPU

* Use 10.0 container to build RMM-enabled XGBoost

* Revert "Use 10.0 container to build RMM-enabled XGBoost"

This reverts commit 789021fa31112e25b683aef39fff375403060141.

* Fix Jenkinsfile

* [CI] Assign larger /dev/shm to NCCL

* Use 10.2 artifact to run multi-GPU Python tests

* Add CUDA 10.0 -> 11.0 cross-version test; remove CUDA 10.0 target

* Rename Conda env rmm_test -> gpu_test

* Use env var to opt into CNMeM pool for C++ tests

* Use identical CUDA version for RMM builds and tests

* Use Pytest fixtures to enable RMM pool in Python tests

* Move RMM to plugin/CMakeLists.txt; use PLUGIN_RMM

* Use per-device MR; use command arg in gtest

* Set CMake prefix path to use Conda env

* Use 0.15 nightly version of RMM

* Remove unnecessary header

* Fix a unit test when cudf is missing

* Add RMM demos

* Remove print()

* Use HostDeviceVector in GPU predictor

* Simplify pytest setup; use LocalCUDACluster fixture

* Address reviewers' commments

Co-authored-by: Hyunsu Cho <chohyu01@cs.wasshington.edu>
This commit is contained in:
Philip Hyunsu Cho
2020-08-12 01:26:02 -07:00
committed by GitHub
parent c3ea3b7e37
commit 9adb812a0a
26 changed files with 508 additions and 140 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
src/common/device_helpers.cuh
View File

@@ -36,7 +36,12 @@
#ifdef XGBOOST_USE_NCCL
#include "nccl.h"
#endif
#endif // XGBOOST_USE_NCCL
#if defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
#include "rmm/mr/device/per_device_resource.hpp"
#include "rmm/mr/device/thrust_allocator_adaptor.hpp"
#endif // defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
#if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 600 || defined(__clang__)
@@ -370,12 +375,21 @@ inline void DebugSyncDevice(std::string file="", int32_t line = -1) {
}
namespace detail {
#if defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
template <typename T>
using XGBBaseDeviceAllocator = rmm::mr::thrust_allocator<T>;
#else // defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
template <typename T>
using XGBBaseDeviceAllocator = thrust::device_malloc_allocator<T>;
#endif // defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
/**
* \brief Default memory allocator, uses cudaMalloc/Free and logs allocations if verbose.
*/
template <class T>
struct XGBDefaultDeviceAllocatorImpl : thrust::device_malloc_allocator<T> {
using SuperT = thrust::device_malloc_allocator<T>;
struct XGBDefaultDeviceAllocatorImpl : XGBBaseDeviceAllocator<T> {
using SuperT = XGBBaseDeviceAllocator<T>;
using pointer = thrust::device_ptr<T>; // NOLINT
template<typename U>
struct rebind // NOLINT
@@ -391,10 +405,15 @@ struct XGBDefaultDeviceAllocatorImpl : thrust::device_malloc_allocator<T> {
GlobalMemoryLogger().RegisterDeallocation(ptr.get(), n * sizeof(T));
return SuperT::deallocate(ptr, n);
}
#if defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
XGBDefaultDeviceAllocatorImpl()
: SuperT(rmm::mr::get_current_device_resource(), cudaStream_t{0}) {}
#endif // defined(XGBOOST_USE_RMM) && XGBOOST_USE_RMM == 1
};
/**
* \brief Caching memory allocator, uses cub::CachingDeviceAllocator as a back-end and logs allocations if verbose. Does not initialise memory on construction.
* \brief Caching memory allocator, uses cub::CachingDeviceAllocator as a back-end and logs
* allocations if verbose. Does not initialise memory on construction.
*/
template <class T>
struct XGBCachingDeviceAllocatorImpl : thrust::device_malloc_allocator<T> {

2
src/common/host_device_vector.cu
View File

@@ -11,6 +11,7 @@
#include "xgboost/data.h"
#include "xgboost/host_device_vector.h"
#include "xgboost/tree_model.h"
#include "device_helpers.cuh"
namespace xgboost {
@@ -402,6 +403,7 @@ template class HostDeviceVector<FeatureType>;
template class HostDeviceVector<Entry>;
template class HostDeviceVector<uint64_t>; // bst_row_t
template class HostDeviceVector<uint32_t>; // bst_feature_t
template class HostDeviceVector<RegTree::Node>;
#if defined(__APPLE__)
/*