330d8983d2
In a nutshell, this moves our libomptarget code to populate the offload subproject. With this commit, users need to enable the new LLVM/Offload subproject as a runtime in their cmake configuration. No further changes are expected for downstream code. Tests and other components still depend on OpenMP and have also not been renamed. The results below are for a build in which OpenMP and Offload are enabled runtimes. In addition to the pure `git mv`, we needed to adjust some CMake files. Nothing is intended to change semantics. ``` ninja check-offload ``` Works with the X86 and AMDGPU offload tests ``` ninja check-openmp ``` Still works but doesn't build offload tests anymore. ``` ls install/lib ``` Shows all expected libraries, incl. - `libomptarget.devicertl.a` - `libomptarget-nvptx-sm_90.bc` - `libomptarget.rtl.amdgpu.so` -> `libomptarget.rtl.amdgpu.so.18git` - `libomptarget.so` -> `libomptarget.so.18git` Fixes: https://github.com/llvm/llvm-project/issues/75124 --------- Co-authored-by: Saiyedul Islam <Saiyedul.Islam@amd.com>
113 lines
3.5 KiB
C++
113 lines
3.5 KiB
C++
//===--------- Mapping.h - OpenMP device runtime mapping helpers -- C++ -*-===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#ifndef OMPTARGET_MAPPING_H
|
|
#define OMPTARGET_MAPPING_H
|
|
|
|
#include "Types.h"
|
|
|
|
namespace ompx {
|
|
|
|
namespace mapping {
|
|
|
|
enum {
|
|
DIM_X = 0,
|
|
DIM_Y = 1,
|
|
DIM_Z = 2,
|
|
};
|
|
|
|
#pragma omp begin declare target device_type(nohost)
|
|
|
|
inline constexpr uint32_t MaxThreadsPerTeam = 1024;
|
|
|
|
#pragma omp end declare target
|
|
|
|
/// Initialize the mapping machinery.
|
|
void init(bool IsSPMD);
|
|
|
|
/// Return true if the kernel is executed in SPMD mode.
|
|
bool isSPMDMode();
|
|
|
|
/// Return true if the kernel is executed in generic mode.
|
|
bool isGenericMode();
|
|
|
|
/// Return true if the executing thread is the main thread in generic mode.
|
|
/// These functions will lookup state and it is required that that is OK for the
|
|
/// thread and location. See also `isInitialThreadInLevel0` for a stateless
|
|
/// alternative for certain situations, e.g. during initialization.
|
|
bool isMainThreadInGenericMode();
|
|
bool isMainThreadInGenericMode(bool IsSPMD);
|
|
|
|
/// Return true if this thread is the initial thread in parallel level 0.
|
|
///
|
|
/// The thread for which this returns true should be used for single threaded
|
|
/// initialization tasks. We pick a special thread to ensure there are no
|
|
/// races between the initialization and the first read of initialized state.
|
|
bool isInitialThreadInLevel0(bool IsSPMD);
|
|
|
|
/// Return true if the executing thread has the lowest Id of the active threads
|
|
/// in the warp.
|
|
bool isLeaderInWarp();
|
|
|
|
/// Return a mask describing all active threads in the warp.
|
|
LaneMaskTy activemask();
|
|
|
|
/// Return a mask describing all threads with a smaller Id in the warp.
|
|
LaneMaskTy lanemaskLT();
|
|
|
|
/// Return a mask describing all threads with a larget Id in the warp.
|
|
LaneMaskTy lanemaskGT();
|
|
|
|
/// Return the thread Id in the warp, in [0, getWarpSize()).
|
|
uint32_t getThreadIdInWarp();
|
|
|
|
/// Return the warp size, thus number of threads in the warp.
|
|
uint32_t getWarpSize();
|
|
|
|
/// Return the warp id in the block, in [0, getNumberOfWarpsInBlock()]
|
|
uint32_t getWarpIdInBlock();
|
|
|
|
/// Return the number of warps in the block.
|
|
uint32_t getNumberOfWarpsInBlock();
|
|
|
|
/// Return the thread Id in the block, in [0, getNumberOfThreadsInBlock(Dim)).
|
|
uint32_t getThreadIdInBlock(int32_t Dim = DIM_X);
|
|
|
|
/// Return the block size, thus number of threads in the block.
|
|
uint32_t getNumberOfThreadsInBlock(int32_t Dim = DIM_X);
|
|
|
|
/// Return the block Id in the kernel, in [0, getNumberOfBlocksInKernel(Dim)).
|
|
uint32_t getBlockIdInKernel(int32_t Dim = DIM_X);
|
|
|
|
/// Return the number of blocks in the kernel.
|
|
uint32_t getNumberOfBlocksInKernel(int32_t Dim = DIM_X);
|
|
|
|
/// Return the kernel size, thus number of threads in the kernel.
|
|
uint32_t getNumberOfThreadsInKernel();
|
|
|
|
/// Return the maximal number of threads in the block usable for a team (=
|
|
/// parallel region).
|
|
///
|
|
/// Note: The version taking \p IsSPMD mode explicitly can be used during the
|
|
/// initialization of the target region, that is before `mapping::isSPMDMode()`
|
|
/// can be called by any thread other than the main one.
|
|
uint32_t getMaxTeamThreads();
|
|
uint32_t getMaxTeamThreads(bool IsSPMD);
|
|
|
|
/// Return the number of processing elements on the device.
|
|
uint32_t getNumberOfProcessorElements();
|
|
|
|
} // namespace mapping
|
|
|
|
} // namespace ompx
|
|
|
|
#endif
|