[EM] Support mmap backed ellpack. (dmlc#10602)

- Support resource view in ellpack. - Define the CUDA version of MMAP resource. - Define the CUDA version of malloc resource. - Refactor cuda runtime API wrappers, and add memory access related wrappers. - gather windows macros into a single header.
kirchho · Jul 18, 2024 · 292bb67 · 292bb67
1 parent e9fbce9
commit 292bb67
Show file tree

Hide file tree

Showing 59 changed files with 887 additions and 644 deletions.
diff --git a/R-package/src/Makevars.in b/R-package/src/Makevars.in
@@ -113,6 +113,7 @@ OBJECTS= \
     $(PKGROOT)/src/common/charconv.o \
     $(PKGROOT)/src/common/column_matrix.o \
     $(PKGROOT)/src/common/common.o \
+    $(PKGROOT)/src/common/cuda_rt_utils.o \
     $(PKGROOT)/src/common/error_msg.o \
     $(PKGROOT)/src/common/hist_util.o \
     $(PKGROOT)/src/common/host_device_vector.o \

diff --git a/R-package/src/Makevars.win b/R-package/src/Makevars.win
@@ -113,6 +113,7 @@ OBJECTS= \
     $(PKGROOT)/src/common/charconv.o \
     $(PKGROOT)/src/common/column_matrix.o \
     $(PKGROOT)/src/common/common.o \
+    $(PKGROOT)/src/common/cuda_rt_utils.o \
     $(PKGROOT)/src/common/error_msg.o \
     $(PKGROOT)/src/common/hist_util.o \
     $(PKGROOT)/src/common/host_device_vector.o \

diff --git a/include/xgboost/base.h b/include/xgboost/base.h
@@ -7,6 +7,8 @@
 #define XGBOOST_BASE_H_
 
 #include <dmlc/omp.h>  // for omp_uint, omp_ulong
+// Put the windefs here to guard as many files as possible.
+#include <xgboost/windefs.h>
 
 #include <cstdint>  // for int32_t, uint64_t, int16_t
 #include <ostream>  // for ostream

diff --git a/include/xgboost/collective/poll_utils.h b/include/xgboost/collective/poll_utils.h
@@ -4,13 +4,14 @@
  * \author Tianqi Chen
  */
 #pragma once
-#include "xgboost/collective/result.h"
-#include "xgboost/collective/socket.h"
+#include <xgboost/collective/result.h>
+#include <xgboost/collective/socket.h>
 
 #if defined(_WIN32)
+#include <xgboost/windefs.h>
+// Socket API
 #include <winsock2.h>
 #include <ws2tcpip.h>
-
 #else
 
 #include <arpa/inet.h>

diff --git a/include/xgboost/collective/socket.h b/include/xgboost/collective/socket.h
@@ -1,12 +1,8 @@
 /**
- * Copyright (c) 2022-2024, XGBoost Contributors
+ * Copyright 2022-2024, XGBoost Contributors
  */
 #pragma once
 
-#if !defined(NOMINMAX) && defined(_WIN32)
-#define NOMINMAX
-#endif                   // !defined(NOMINMAX)
-
 #include <cerrno>        // errno, EINTR, EBADF
 #include <climits>       // HOST_NAME_MAX
 #include <cstddef>       // std::size_t
@@ -18,18 +14,12 @@
 
 #if defined(__linux__)
 #include <sys/ioctl.h>  // for TIOCOUTQ, FIONREAD
-#endif  // defined(__linux__)
-
-#if !defined(xgboost_IS_MINGW)
-
-#if defined(__MINGW32__)
-#define xgboost_IS_MINGW 1
-#endif  // defined(__MINGW32__)
-
-#endif  // xgboost_IS_MINGW
+#endif                  // defined(__linux__)
 
 #if defined(_WIN32)
-
+// Guard the include.
+#include <xgboost/windefs.h>
+// Socket API
 #include <winsock2.h>
 #include <ws2tcpip.h>
 
@@ -41,9 +31,9 @@ using in_port_t = std::uint16_t;
 
 #if !defined(xgboost_IS_MINGW)
 using ssize_t = int;
-#endif                    // !xgboost_IS_MINGW()
+#endif  // !xgboost_IS_MINGW()
 
-#else                     // UNIX
+#else  // UNIX
 
 #include <arpa/inet.h>    // inet_ntop
 #include <fcntl.h>        // fcntl, F_GETFL, O_NONBLOCK
@@ -839,7 +829,3 @@ Result INetNToP(H const &host, std::string *p_out) {
 }  // namespace xgboost
 
 #undef xgboost_CHECK_SYS_CALL
-
-#if defined(xgboost_IS_MINGW)
-#undef xgboost_IS_MINGW
-#endif
diff --git a/include/xgboost/windefs.h b/include/xgboost/windefs.h
@@ -0,0 +1,33 @@
+/**
+ * Copyright 2024, XGBoost Contributors
+ *
+ * @brief Macro for Windows.
+ */
+#pragma once
+
+#if !defined(xgboost_IS_WIN)
+
+#if defined(_MSC_VER) || defined(__MINGW32__)
+#define xgboost_IS_WIN 1
+#endif  // defined(_MSC_VER) || defined(__MINGW32__)
+
+#endif  // !defined(xgboost_IS_WIN)
+
+#if defined(xgboost_IS_WIN)
+
+#if !defined(NOMINMAX)
+#define NOMINMAX
+#endif  // !defined(NOMINMAX)
+
+// A macro used inside `windows.h` to avoid conflicts with `winsock2.h`
+#define WIN32_LEAN_AND_MEAN
+
+#if !defined(xgboost_IS_MINGW)
+
+#if defined(__MINGW32__)
+#define xgboost_IS_MINGW 1
+#endif  // defined(__MINGW32__)
+
+#endif  // xgboost_IS_MINGW
+
+#endif  // defined(xgboost_IS_WIN)
diff --git a/src/c_api/c_api_error.h b/src/c_api/c_api_error.h
@@ -7,10 +7,9 @@
 #define XGBOOST_C_API_C_API_ERROR_H_
 
 #include <dmlc/base.h>
-#include <dmlc/logging.h>
 
-#include "c_api_utils.h"
-#include "xgboost/collective/result.h"
+#include "c_api_utils.h"  // for XGBoostAPIGuard
+#include "xgboost/logging.h"
 
 /*! \brief  macro to guard beginning and end section of all functions */
 #ifdef LOG_CAPI_INVOCATION

diff --git a/src/cli_main.cc b/src/cli_main.cc
@@ -4,29 +4,26 @@
  * \brief The command line interface program of xgboost.
  *  This file is not included in dynamic library.
  */
-#if !defined(NOMINMAX) && defined(_WIN32)
-#define NOMINMAX
-#endif  // !defined(NOMINMAX)
-
 #include <dmlc/timer.h>
-
-#include <xgboost/learner.h>
+#include <xgboost/base.h>
 #include <xgboost/data.h>
 #include <xgboost/json.h>
+#include <xgboost/learner.h>
 #include <xgboost/logging.h>
 #include <xgboost/parameter.h>
 
-#include <iomanip>
-#include <ctime>
-#include <string>
 #include <cstdio>
 #include <cstring>
+#include <ctime>
+#include <iomanip>
+#include <string>
 #include <vector>
+
+#include "c_api/c_api_utils.h"
 #include "common/common.h"
 #include "common/config.h"
 #include "common/io.h"
 #include "common/version.h"
-#include "c_api/c_api_utils.h"
 
 namespace xgboost {
 enum CLITask {

diff --git a/src/collective/tracker.cc b/src/collective/tracker.cc
@@ -7,11 +7,10 @@
 #include <sys/socket.h>  // socket, AF_INET6, AF_INET, connect, getsockname
 #endif                   // defined(__unix__) || defined(__APPLE__)
 
-#if !defined(NOMINMAX) && defined(_WIN32)
-#define NOMINMAX
-#endif  // !defined(NOMINMAX)
-
 #if defined(_WIN32)
+// Guard the include
+#include <xgboost/windefs.h>
+// Socket API
 #include <winsock2.h>
 #include <ws2tcpip.h>
 #endif  // defined(_WIN32)

diff --git a/src/common/common.cc b/src/common/common.cc
@@ -1,5 +1,5 @@
 /**
- * Copyright 2015-2023 by Contributors
+ * Copyright 2015-2024, XGBoost Contributors
  */
 #include "common.h"
 
@@ -54,9 +54,4 @@ void EscapeU8(std::string const &string, std::string *p_buffer) {
     }
   }
 }
-
-#if !defined(XGBOOST_USE_CUDA)
-int AllVisibleGPUs() { return 0; }
-#endif  // !defined(XGBOOST_USE_CUDA)
-
 }  // namespace xgboost::common
diff --git a/src/common/common.cu b/src/common/common.cu
@@ -1,29 +1,21 @@
-/*!
- * Copyright 2018-2022 XGBoost contributors
+/**
+ * Copyright 2018-2024, XGBoost contributors
  */
-#include "common.h"
-
-namespace xgboost {
-namespace common {
+#include <thrust/system/cuda/error.h>
+#include <thrust/system_error.h>
 
-void SetDevice(std::int32_t device) {
-  if (device >= 0) {
-    dh::safe_cuda(cudaSetDevice(device));
-  }
-}
+#include "common.h"
 
-int AllVisibleGPUs() {
-  int n_visgpus = 0;
-  try {
-    // When compiled with CUDA but running on CPU only device,
-    // cudaGetDeviceCount will fail.
-    dh::safe_cuda(cudaGetDeviceCount(&n_visgpus));
-  } catch (const dmlc::Error &) {
-    cudaGetLastError();  // reset error.
-    return 0;
+namespace dh {
+void ThrowOnCudaError(cudaError_t code, const char *file, int line) {
+  if (code != cudaSuccess) {
+    std::string f;
+    if (file != nullptr) {
+      f = file;
+    }
+    LOG(FATAL) << thrust::system_error(code, thrust::cuda_category(),
+                                       f + ": " + std::to_string(line))
+                      .what();
   }
-  return n_visgpus;
 }
-
-}  // namespace common
-}  // namespace xgboost
+}  // namespace dh
diff --git a/src/common/common.h b/src/common/common.h
@@ -1,5 +1,5 @@
 /**
- * Copyright 2015-2023 by XGBoost Contributors
+ * Copyright 2015-2024, XGBoost Contributors
  * \file common.h
  * \brief Common utilities
  */
@@ -19,9 +19,8 @@
 #include "xgboost/base.h"     // for XGBOOST_DEVICE
 #include "xgboost/logging.h"  // for LOG, LOG_FATAL, LogMessageFatal
 
+// magic to define functions based on the compiler.
 #if defined(__CUDACC__)
-#include <thrust/system/cuda/error.h>
-#include <thrust/system_error.h>
 
 #define WITH_CUDA() true
 
@@ -31,23 +30,20 @@
 
 #endif  // defined(__CUDACC__)
 
+#if defined(XGBOOST_USE_CUDA)
+#include <cuda_runtime_api.h>
+#endif
+
 namespace dh {
-#if defined(__CUDACC__)
+#if defined(XGBOOST_USE_CUDA)
 /*
- * Error handling  functions
+ * Error handling functions
  */
+void ThrowOnCudaError(cudaError_t code, const char *file, int line);
+
 #define safe_cuda(ans) ThrowOnCudaError((ans), __FILE__, __LINE__)
 
-inline cudaError_t ThrowOnCudaError(cudaError_t code, const char *file,
-                                    int line) {
-  if (code != cudaSuccess) {
-    LOG(FATAL) << thrust::system_error(code, thrust::cuda_category(),
-                                       std::string{file} + ": " +  // NOLINT
-                                       std::to_string(line)).what();
-  }
-  return code;
-}
-#endif  // defined(__CUDACC__)
+#endif  // defined(XGBOOST_USE_CUDA)
 }  // namespace dh
 
 namespace xgboost::common {
@@ -167,8 +163,6 @@ class Range {
   Iterator end_;
 };
 
-int AllVisibleGPUs();
-
 inline void AssertGPUSupport() {
 #ifndef XGBOOST_USE_CUDA
     LOG(FATAL) << "XGBoost version not compiled with GPU support.";
@@ -187,16 +181,6 @@ inline void AssertSYCLSupport() {
 #endif  // XGBOOST_USE_SYCL
 }
 
-void SetDevice(std::int32_t device);
-
-#if !defined(XGBOOST_USE_CUDA)
-inline void SetDevice(std::int32_t device) {
-  if (device >= 0) {
-    AssertGPUSupport();
-  }
-}
-#endif
-
 /**
  * @brief Last index of a group in a CSR style of index pointer.
  */