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

use __HIPCC__ for device code

Browse Source
This commit is contained in:
Hui Liu
2024-01-24 11:30:01 -08:00
parent 1e0ccf7b87
commit 069cf1d019
23 changed files with 68 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
include/xgboost/base.h
View File

@@ -58,19 +58,19 @@
/*!
* \brief Tag function as usable by device
*/
#if defined (__CUDA__) || defined(__NVCC__) || defined(__HIP_PLATFORM_AMD__)
#if defined (__CUDA__) || defined(__NVCC__) || defined(__HIPCC__)
#define XGBOOST_DEVICE __host__ __device__
#else
#define XGBOOST_DEVICE
#endif // defined (__CUDA__) || defined(__NVCC__) || defined(__HIP_PLATFORM_AMD__)
#endif // defined (__CUDA__) || defined(__NVCC__) || defined(__HIPCC__)
#if defined(__CUDA__) || defined(__CUDACC__) || defined(__HIP_PLATFORM_AMD__)
#if defined(__CUDA__) || defined(__CUDACC__) || defined(__HIPCC__)
#define XGBOOST_HOST_DEV_INLINE XGBOOST_DEVICE __forceinline__
#define XGBOOST_DEV_INLINE __device__ __forceinline__
#else
#define XGBOOST_HOST_DEV_INLINE
#define XGBOOST_DEV_INLINE
#endif // defined(__CUDA__) || defined(__CUDACC__) || defined(__HIP_PLATFORM_AMD__)
#endif // defined(__CUDA__) || defined(__CUDACC__) || defined(__HIPCC__)
// These check are for Makefile.
#if !defined(XGBOOST_MM_PREFETCH_PRESENT) && !defined(XGBOOST_BUILTIN_PREFETCH_PRESENT)

4
include/xgboost/host_device_vector.h
View File

@@ -58,11 +58,11 @@
namespace xgboost {
#if defined(__CUDACC__) || defined(__HIP_PLATFORM_AMD__)
#if defined(__CUDACC__) || defined(__HIPCC__)
// Sets a function to call instead of cudaSetDevice();
// only added for testing
void SetCudaSetDeviceHandler(void (*handler)(int));
#endif // __CUDACC__ || __HIP_PLATFORM_AMD__
#endif // __CUDACC__ || __HIPCC__
template <typename T> struct HostDeviceVectorImpl;

10
include/xgboost/linalg.h
View File

@@ -30,11 +30,11 @@
// decouple it from xgboost.
#ifndef LINALG_HD
#if defined(__CUDA__) || defined(__NVCC__) || defined(__HIP_PLATFORM_AMD__)
#if defined(__CUDA__) || defined(__NVCC__) || defined(__HIPCC__)
#define LINALG_HD __host__ __device__
#else
#define LINALG_HD
#endif // defined (__CUDA__) || defined(__NVCC__) || defined(__HIP_PLATFORM_AMD__)
#endif // defined (__CUDA__) || defined(__NVCC__) || defined(__HIPCC__)
#endif // LINALG_HD
namespace xgboost::linalg {
@@ -118,7 +118,7 @@ using IndexToTag = std::conditional_t<std::is_integral<RemoveCRType<S>>::value,
template <int32_t n, typename Fn>
LINALG_HD constexpr auto UnrollLoop(Fn fn) {
#if defined(__CUDA_ARCH__) || defined(__HIP_PLATFORM_AMD__)
#if defined(__CUDA_ARCH__) || defined(__HIPCC__)
#pragma unroll n
#endif // defined __CUDA_ARCH__
for (int32_t i = 0; i < n; ++i) {
@@ -136,7 +136,7 @@ int32_t NativePopc(T v) {
inline LINALG_HD int Popc(uint32_t v) {
#if defined(__CUDA_ARCH__)
return __popc(v);
#elif defined(__GNUC__) || defined(__clang__) || defined(__HIP_PLATFORM_AMD__)
#elif defined(__GNUC__) || defined(__clang__) || defined(__HIPCC__)
return __builtin_popcount(v);
#elif defined(_MSC_VER)
return __popcnt(v);
@@ -148,7 +148,7 @@ inline LINALG_HD int Popc(uint32_t v) {
inline LINALG_HD int Popc(uint64_t v) {
#if defined(__CUDA_ARCH__)
return __popcll(v);
#elif defined(__GNUC__) || defined(__clang__) || defined(__HIP_PLATFORM_AMD__)
#elif defined(__GNUC__) || defined(__clang__) || defined(__HIPCC__)
return __builtin_popcountll(v);
#elif defined(_MSC_VER) && defined(_M_X64)
return __popcnt64(v);

6
include/xgboost/span.h
View File

@@ -41,7 +41,7 @@
#if defined(__CUDACC__)
#include <cuda_runtime.h>
#elif defined(__HIP_PLATFORM_AMD__)
#elif defined(__HIPCC__)
#include <hip/hip_runtime.h>
#endif
@@ -106,7 +106,7 @@ namespace common {
#define SPAN_CHECK KERNEL_CHECK
#elif defined(__HIP_PLATFORM_AMD__)
#elif defined(__HIPCC__)
// Usual logging facility is not available inside device code.
#if defined(_MSC_VER)
@@ -157,7 +157,7 @@ namespace common {
#endif // defined(XGBOOST_STRICT_R_MODE)
#endif // __CUDA_ARCH__ || __HIP_PLATFORM_AMD__
#endif // __CUDA_ARCH__ || __HIPCC__
#define SPAN_LT(lhs, rhs) SPAN_CHECK((lhs) < (rhs))