rocm-systems/hipamd/src/hip_device.cpp

/* Copyright (c) 2018 - 2022 Advanced Micro Devices, Inc.

 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal
 in the Software without restriction, including without limitation the rights
 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 copies of the Software, and to permit persons to whom the Software is
 furnished to do so, subject to the following conditions:

 The above copyright notice and this permission notice shall be included in
 all copies or substantial portions of the Software.

 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE. */

#include <hip/hip_runtime.h>

#include "hip_internal.hpp"
#include "hip_mempool_impl.hpp"

#undef hipGetDeviceProperties
#undef hipDeviceProp_t

namespace hip {

// ================================================================================================
hip::Stream* Device::NullStream() {
  if (null_stream_ == nullptr) {
    null_stream_ = new Stream(this, Stream::Priority::Normal, 0, true);
  }

  if (null_stream_ == nullptr) {
    return nullptr;
  }
  // Wait for all active streams before executing commands on the default
  iHipWaitActiveStreams(null_stream_);
  return null_stream_;
}

// ================================================================================================
bool Device::Create() {
  // Create default memory pool
  default_mem_pool_ = new MemoryPool(this);
  if (default_mem_pool_ == nullptr) {
    return false;
  }

  // Create graph memory pool
  graph_mem_pool_ = new MemoryPool(this);
  if (graph_mem_pool_ == nullptr) {
    return false;
  }

  if (!HIP_MEM_POOL_USE_VM) {
    uint64_t max_size = std::numeric_limits<uint64_t>::max();
    // Use maximum value to hold memory, because current implementation doesn't support VM
    // Note: the call for the threshold is always successful
    auto error = graph_mem_pool_->SetAttribute(hipMemPoolAttrReleaseThreshold, &max_size);
  }

  // Current is default pool after device creation
  current_mem_pool_ = default_mem_pool_;
  return true;
}

// ================================================================================================
void Device::AddMemoryPool(MemoryPool* pool) {
  amd::ScopedLock lock(lock_);
  if (auto it = mem_pools_.find(pool); it == mem_pools_.end()) {
    mem_pools_.insert(pool);
  }
}

// ================================================================================================
void Device::RemoveMemoryPool(MemoryPool* pool) {
  amd::ScopedLock lock(lock_);
  if (auto it = mem_pools_.find(pool); it != mem_pools_.end()) {
    mem_pools_.erase(it);
  }
}

// ================================================================================================
bool Device::FreeMemory(amd::Memory* memory, Stream* stream) {
  amd::ScopedLock lock(lock_);
  // Search for memory in the entire list of pools
  for (auto it : mem_pools_) {
    if (it->FreeMemory(memory, stream)) {
      return true;
    }
  }
  return false;
}

// ================================================================================================
void Device::ReleaseFreedMemory(Stream* stream) {
  amd::ScopedLock lock(lock_);
  // Search for memory in the entire list of pools
  for (auto it : mem_pools_) {
    it->ReleaseFreedMemory(stream);
  }
}

// ================================================================================================
void Device::RemoveStreamFromPools(Stream* stream) {
  amd::ScopedLock lock(lock_);
  // Update all pools with the destroyed stream
  for (auto it : mem_pools_) {
    it->RemoveStream(stream);
  }
}

// ================================================================================================
void Device::Reset() {
  {
    amd::ScopedLock lock(lock_);
    auto it = mem_pools_.begin();
    while (it != mem_pools_.end()) {
      auto current = it++;
      (*current)->ReleaseAllMemory();
      delete *current;
    }
    mem_pools_.clear();
  }
  flags_ = hipDeviceScheduleSpin;
  hip::Stream::destroyAllStreams(deviceId_);
  amd::MemObjMap::Purge(devices()[0]);
  Create();
}

// ================================================================================================
Device::~Device() {
  if (default_mem_pool_ != nullptr) {
    default_mem_pool_->release();
  }

  if (graph_mem_pool_ != nullptr) {
    graph_mem_pool_->release();
  }

  if (null_stream_ != nullptr) {
    hip::Stream::Destroy(null_stream_);
  }
}

}  // namespace hip

void ihipDestroyDevice() {
  for (auto deviceHandle : g_devices) {
    delete deviceHandle;
  }
}

hipError_t ihipDeviceGet(hipDevice_t* device, int deviceId) {
  if (device == nullptr) {
    return hipErrorInvalidValue;
  }

  if (deviceId < 0 || static_cast<size_t>(deviceId) >= g_devices.size()) {
    return hipErrorInvalidDevice;
  }

  *device = deviceId;
  return hipSuccess;
}

hipError_t hipDeviceGet(hipDevice_t* device, int deviceId) {
  HIP_INIT_API(hipDeviceGet, device, deviceId);

  HIP_RETURN(ihipDeviceGet(device, deviceId));
}

hipError_t hipDeviceTotalMem(size_t* bytes, hipDevice_t device) {
  HIP_INIT_API(hipDeviceTotalMem, bytes, device);

  if (device < 0 || static_cast<size_t>(device) >= g_devices.size()) {
    HIP_RETURN(hipErrorInvalidDevice);
  }

  if (bytes == nullptr) {
    HIP_RETURN(hipErrorInvalidValue);
  }

  auto* deviceHandle = g_devices[device]->devices()[0];
  const auto& info = deviceHandle->info();

  *bytes = info.globalMemSize_;

  HIP_RETURN(hipSuccess);
}

hipError_t hipDeviceComputeCapability(int* major, int* minor, hipDevice_t device) {
  HIP_INIT_API(hipDeviceComputeCapability, major, minor, device);

  if (device < 0 || static_cast<size_t>(device) >= g_devices.size()) {
    HIP_RETURN(hipErrorInvalidDevice);
  }

  if (major == nullptr || minor == nullptr) {
    HIP_RETURN(hipErrorInvalidValue);
  }

  auto* deviceHandle = g_devices[device]->devices()[0];
  const auto& isa = deviceHandle->isa();
  *major = isa.versionMajor();
  *minor = isa.versionMinor();

  HIP_RETURN(hipSuccess);
}

hipError_t hipDeviceGetCount(int* count) {
  HIP_INIT_API(hipDeviceGetCount, count);

  HIP_RETURN(ihipDeviceGetCount(count));
}

hipError_t ihipDeviceGetCount(int* count) {
  if (count == nullptr) {
    return hipErrorInvalidValue;
  }

  // Get all available devices
  *count = g_devices.size();

  if (*count < 1) {
    return hipErrorNoDevice;
  }

  return hipSuccess;
}

hipError_t hipDeviceGetName(char* name, int len, hipDevice_t device) {
  HIP_INIT_API(hipDeviceGetName, (void*)name, len, device);

  if (device < 0 || static_cast<size_t>(device) >= g_devices.size()) {
    HIP_RETURN(hipErrorInvalidDevice);
  }

  if (name == nullptr || len <= 0) {
    HIP_RETURN(hipErrorInvalidValue);
  }

  auto* deviceHandle = g_devices[device]->devices()[0];
  const auto& info = deviceHandle->info();
  const auto nameLen = ::strlen(info.boardName_);

  // Only copy partial name if size of `dest` is smaller than size of `src` including
  // trailing zero byte
  auto memcpySize = (len <= (nameLen + 1) ? (len - 1) : nameLen);
  ::memcpy(name, info.boardName_, memcpySize);
  name[memcpySize] = '\0';

  HIP_RETURN(hipSuccess);
}

hipError_t hipDeviceGetUuid(hipUUID* uuid, hipDevice_t device) {
  HIP_INIT_API(hipDeviceGetUuid, reinterpret_cast<void*>(uuid), device);

  if (device < 0 || static_cast<size_t>(device) >= g_devices.size()) {
    HIP_RETURN(hipErrorInvalidDevice);
  }

  if (uuid == nullptr) {
    HIP_RETURN(hipErrorInvalidValue);
  }

  auto* deviceHandle = g_devices[device]->devices()[0];
  const auto& info = deviceHandle->info();
  memcpy(uuid->bytes, info.uuid_, sizeof(info.uuid_));

  HIP_RETURN(hipSuccess);
}

hipError_t ihipGetDeviceProperties(hipDeviceProp_tR0600* props, hipDevice_t device) {
  if (props == nullptr) {
    return hipErrorInvalidValue;
  }

  if (unsigned(device) >= g_devices.size()) {
    return hipErrorInvalidDevice;
  }
  auto* deviceHandle = g_devices[device]->devices()[0];

  constexpr auto int32_max = static_cast<uint64_t>(std::numeric_limits<int32_t>::max());
  constexpr auto uint16_max = static_cast<uint64_t>(std::numeric_limits<uint16_t>::max()) + 1;
  hipDeviceProp_tR0600 deviceProps = {0};

  const auto& info = deviceHandle->info();
  const auto& isa = deviceHandle->isa();
  ::strncpy(deviceProps.name, info.boardName_, sizeof(info.boardName_));
  ::strncpy(deviceProps.uuid.bytes, info.uuid_, sizeof(info.uuid_));
  deviceProps.totalGlobalMem = info.globalMemSize_;
  deviceProps.sharedMemPerBlock = info.localMemSizePerCU_;
  deviceProps.sharedMemPerMultiprocessor = info.localMemSizePerCU_ * info.numRTCUs_;
  deviceProps.regsPerBlock = info.availableRegistersPerCU_;
  deviceProps.warpSize = info.wavefrontWidth_;
  deviceProps.maxThreadsPerBlock = info.maxWorkGroupSize_;
  deviceProps.maxThreadsDim[0] = info.maxWorkItemSizes_[0];
  deviceProps.maxThreadsDim[1] = info.maxWorkItemSizes_[1];
  deviceProps.maxThreadsDim[2] = info.maxWorkItemSizes_[2];
  deviceProps.maxGridSize[0] = int32_max;
  deviceProps.maxGridSize[1] = uint16_max;
  deviceProps.maxGridSize[2] = uint16_max;
  deviceProps.clockRate = info.maxEngineClockFrequency_ * 1000;
  deviceProps.memoryClockRate = info.maxMemoryClockFrequency_ * 1000;
  deviceProps.memoryBusWidth = info.globalMemChannels_;
  deviceProps.totalConstMem = std::min(info.maxConstantBufferSize_, int32_max);
  deviceProps.major = isa.versionMajor();
  deviceProps.minor = isa.versionMinor();
  deviceProps.multiProcessorCount = info.maxComputeUnits_;
  deviceProps.l2CacheSize = info.l2CacheSize_;
  deviceProps.maxThreadsPerMultiProcessor = info.maxThreadsPerCU_;
  deviceProps.maxBlocksPerMultiProcessor = int(info.maxThreadsPerCU_ / info.maxWorkGroupSize_);
  deviceProps.computeMode = 0;
  deviceProps.clockInstructionRate = info.timeStampFrequency_;
  deviceProps.arch.hasGlobalInt32Atomics = 1;
  deviceProps.arch.hasGlobalFloatAtomicExch = 1;
  deviceProps.arch.hasSharedInt32Atomics = 1;
  deviceProps.arch.hasSharedFloatAtomicExch = 1;
  deviceProps.arch.hasFloatAtomicAdd = 1;
  deviceProps.arch.hasGlobalInt64Atomics = 1;
  deviceProps.arch.hasSharedInt64Atomics = 1;
  deviceProps.hostNativeAtomicSupported = info.pcie_atomics_ ? 1 : 0;
  deviceProps.arch.hasDoubles = 1;
  deviceProps.arch.hasWarpVote = 1;
  deviceProps.arch.hasWarpBallot = 1;
  deviceProps.arch.hasWarpShuffle = 1;
  deviceProps.arch.hasFunnelShift = 0;
  deviceProps.arch.hasThreadFenceSystem = 1;
  deviceProps.arch.hasSyncThreadsExt = 0;
  deviceProps.arch.hasSurfaceFuncs = 0;
  deviceProps.arch.has3dGrid = 1;
  deviceProps.arch.hasDynamicParallelism = 0;
  deviceProps.concurrentKernels = 1;
  deviceProps.pciDomainID = info.pciDomainID;
  deviceProps.pciBusID = info.deviceTopology_.pcie.bus;
  deviceProps.pciDeviceID = info.deviceTopology_.pcie.device;
  deviceProps.maxSharedMemoryPerMultiProcessor = info.localMemSizePerCU_;
  deviceProps.canMapHostMemory = 1;
  deviceProps.regsPerMultiprocessor = info.availableRegistersPerCU_;
  sprintf(deviceProps.gcnArchName, "%s", isa.targetId());
  deviceProps.cooperativeLaunch = info.cooperativeGroups_;
  deviceProps.cooperativeMultiDeviceLaunch = info.cooperativeMultiDeviceGroups_;

  deviceProps.cooperativeMultiDeviceUnmatchedFunc = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedGridDim = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedBlockDim = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedSharedMem = info.cooperativeMultiDeviceGroups_;

  deviceProps.maxTexture1DLinear =
      std::min(16 * info.imageMaxBufferSize_, int32_max);  // Max pixel size is 16 bytes
  deviceProps.maxTexture1DMipmap = std::min(16 * info.imageMaxBufferSize_, int32_max);
  deviceProps.maxTexture1D = deviceProps.maxSurface1D = std::min(info.image1DMaxWidth_, int32_max);
  deviceProps.maxTexture2D[0] = deviceProps.maxSurface2D[0] =
      std::min(info.image2DMaxWidth_, int32_max);
  deviceProps.maxTexture2D[1] = deviceProps.maxSurface2D[1] =
      std::min(info.image2DMaxHeight_, int32_max);
  deviceProps.maxTexture3D[0] = deviceProps.maxSurface3D[0] =
      std::min(info.image3DMaxWidth_, int32_max);
  deviceProps.maxTexture3D[1] = deviceProps.maxSurface3D[1] =
      std::min(info.image3DMaxHeight_, int32_max);
  deviceProps.maxTexture3D[2] = deviceProps.maxSurface3D[2] =
      std::min(info.image3DMaxDepth_, int32_max);
  deviceProps.maxTexture1DLayered[0] = deviceProps.maxSurface1DLayered[0] =
      std::min(info.image1DAMaxWidth_, int32_max);
  deviceProps.maxTexture1DLayered[1] = deviceProps.maxSurface1DLayered[1] =
      std::min(info.imageMaxArraySize_, int32_max);
  deviceProps.maxTexture2DLayered[0] = deviceProps.maxSurface2DLayered[0] =
      std::min(info.image2DAMaxWidth_[0], int32_max);
  deviceProps.maxTexture2DLayered[1] = deviceProps.maxSurface2DLayered[1] =
      std::min(info.image2DAMaxWidth_[1], int32_max);
  deviceProps.maxTexture2DLayered[2] = deviceProps.maxSurface2DLayered[2] =
      std::min(info.imageMaxArraySize_, int32_max);
  deviceProps.hdpMemFlushCntl = info.hdpMemFlushCntl;
  deviceProps.hdpRegFlushCntl = info.hdpRegFlushCntl;

  deviceProps.memPitch = std::min(info.maxMemAllocSize_, int32_max);
  deviceProps.textureAlignment = deviceProps.surfaceAlignment = info.imageBaseAddressAlignment_;
  deviceProps.texturePitchAlignment = info.imagePitchAlignment_;
  deviceProps.kernelExecTimeoutEnabled = 0;
  deviceProps.ECCEnabled = info.errorCorrectionSupport_ ? 1 : 0;
  deviceProps.isLargeBar = info.largeBar_ ? 1 : 0;
  deviceProps.asicRevision = info.asicRevision_;
  deviceProps.ipcEventSupported = 1;
  deviceProps.streamPrioritiesSupported = 1;
  deviceProps.multiGpuBoardGroupID = info.deviceTopology_.pcie.device;

  // HMM capabilities
  deviceProps.asyncEngineCount = info.numAsyncQueues_;
  deviceProps.deviceOverlap = (info.numAsyncQueues_ > 0) ? 1 : 0;
  deviceProps.unifiedAddressing = info.hmmDirectHostAccess_;
  deviceProps.managedMemory = info.hmmSupported_;
  deviceProps.concurrentManagedAccess = info.hmmSupported_;
  deviceProps.directManagedMemAccessFromHost = info.hmmDirectHostAccess_;
  deviceProps.canUseHostPointerForRegisteredMem = info.hostUnifiedMemory_;
  deviceProps.pageableMemoryAccess = info.hmmCpuMemoryAccessible_;
  deviceProps.hostRegisterSupported = info.hostUnifiedMemory_;
  deviceProps.pageableMemoryAccessUsesHostPageTables = info.hostUnifiedMemory_;

  // Mem pool
  deviceProps.memoryPoolsSupported = HIP_MEM_POOL_SUPPORT;
  deviceProps.memoryPoolSupportedHandleTypes = 0;

  // Caching behavior
  deviceProps.globalL1CacheSupported = 1;
  deviceProps.localL1CacheSupported = 1;
  deviceProps.persistingL2CacheMaxSize = info.l2CacheSize_;
  deviceProps.reservedSharedMemPerBlock = 0;
  deviceProps.sharedMemPerBlockOptin = 0;

  // Unsupported features
  // Single to double precision perf ratio
  deviceProps.singleToDoublePrecisionPerfRatio = 0;
  // Flag hipHostRegisterReadOnly
  deviceProps.hostRegisterReadOnlySupported = 0;
  // Compute preemption
  deviceProps.computePreemptionSupported = 0;
  // Cubemaps
  deviceProps.maxTextureCubemap = 0;
  deviceProps.maxTextureCubemapLayered[0] = 0;
  deviceProps.maxTextureCubemapLayered[1] = 0;
  deviceProps.maxSurfaceCubemap = 0;
  deviceProps.maxSurfaceCubemapLayered[0] = 0;
  deviceProps.maxSurfaceCubemapLayered[1] = 0;
  // Texture gather ops
  deviceProps.maxTexture2DGather[0] = 0;
  deviceProps.maxTexture2DGather[1] = 0;
  // Textures bound to pitch memory
  deviceProps.maxTexture2DLinear[0] = 0;
  deviceProps.maxTexture2DLinear[1] = 0;
  deviceProps.maxTexture2DLinear[2] = 0;
  // Alternate 3D texture
  deviceProps.maxTexture3DAlt[0] = 0;
  deviceProps.maxTexture3DAlt[1] = 0;
  deviceProps.maxTexture3DAlt[2] = 0;
  // access policy
  deviceProps.accessPolicyMaxWindowSize = 0;
  // cluster launch
  deviceProps.clusterLaunch = 0;
  // Mapping HIP array
  deviceProps.deferredMappingHipArraySupported = 0;
  // RDMA options
  deviceProps.gpuDirectRDMASupported = 0;
  deviceProps.gpuDirectRDMAFlushWritesOptions = 0;
  deviceProps.gpuDirectRDMAWritesOrdering = 0;
  *reinterpret_cast<uint32_t*>(&deviceProps.luid[0]) = info.luidLowPart_;
  *reinterpret_cast<uint32_t*>(&deviceProps.luid[sizeof(uint32_t)]) = info.luidHighPart_;
  deviceProps.luidDeviceNodeMask = info.luidDeviceNodeMask_;

  deviceProps.sparseHipArraySupported = 0;
  deviceProps.timelineSemaphoreInteropSupported = 0;
  deviceProps.unifiedFunctionPointers = 0;

  *props = deviceProps;
  return hipSuccess;
}

hipError_t hipGetDevicePropertiesR0600(hipDeviceProp_tR0600* props, hipDevice_t device) {
  HIP_INIT_API(hipGetDevicePropertiesR0600, props, device);

  HIP_RETURN(ihipGetDeviceProperties(props, device));
}

extern "C" typedef struct hipDeviceProp_t {
  char name[256];            ///< Device name.
  size_t totalGlobalMem;     ///< Size of global memory region (in bytes).
  size_t sharedMemPerBlock;  ///< Size of shared memory region (in bytes).
  int regsPerBlock;          ///< Registers per block.
  int warpSize;              ///< Warp size.
  int maxThreadsPerBlock;    ///< Max work items per work group or workgroup max size.
  int maxThreadsDim[3];      ///< Max number of threads in each dimension (XYZ) of a block.
  int maxGridSize[3];        ///< Max grid dimensions (XYZ).
  int clockRate;             ///< Max clock frequency of the multiProcessors in khz.
  int memoryClockRate;       ///< Max global memory clock frequency in khz.
  int memoryBusWidth;        ///< Global memory bus width in bits.
  size_t totalConstMem;      ///< Size of shared memory region (in bytes).
  int major;  ///< Major compute capability.  On HCC, this is an approximation and features may
              ///< differ from CUDA CC.  See the arch feature flags for portable ways to query
              ///< feature caps.
  int minor;  ///< Minor compute capability.  On HCC, this is an approximation and features may
              ///< differ from CUDA CC.  See the arch feature flags for portable ways to query
              ///< feature caps.
  int multiProcessorCount;          ///< Number of multi-processors (compute units).
  int l2CacheSize;                  ///< L2 cache size.
  int maxThreadsPerMultiProcessor;  ///< Maximum resident threads per multi-processor.
  int computeMode;                  ///< Compute mode.
  int clockInstructionRate;  ///< Frequency in khz of the timer used by the device-side "clock*"
                             ///< instructions.  New for HIP.
  hipDeviceArch_t arch;      ///< Architectural feature flags.  New for HIP.
  int concurrentKernels;     ///< Device can possibly execute multiple kernels concurrently.
  int pciDomainID;           ///< PCI Domain ID
  int pciBusID;              ///< PCI Bus ID.
  int pciDeviceID;           ///< PCI Device ID.
  size_t maxSharedMemoryPerMultiProcessor;  ///< Maximum Shared Memory Per Multiprocessor.
  int isMultiGpuBoard;                      ///< 1 if device is on a multi-GPU board, 0 if not.
  int canMapHostMemory;                     ///< Check whether HIP can map host memory
  int gcnArch;                              ///< DEPRECATED: use gcnArchName instead
  char gcnArchName[256];                    ///< AMD GCN Arch Name.
  int integrated;                           ///< APU vs dGPU
  int cooperativeLaunch;                    ///< HIP device supports cooperative launch
  int cooperativeMultiDeviceLaunch;         ///< HIP device supports cooperative launch on multiple
                                            ///< devices
  int maxTexture1DLinear;                   ///< Maximum size for 1D textures bound to linear memory
  int maxTexture1D;                         ///< Maximum number of elements in 1D images
  int maxTexture2D[2];  ///< Maximum dimensions (width, height) of 2D images, in image elements
  int maxTexture3D[3];  ///< Maximum dimensions (width, height, depth) of 3D images, in image
                        ///< elements
  unsigned int* hdpMemFlushCntl;  ///< Addres of HDP_MEM_COHERENCY_FLUSH_CNTL register
  unsigned int* hdpRegFlushCntl;  ///< Addres of HDP_REG_COHERENCY_FLUSH_CNTL register
  size_t memPitch;                ///< Maximum pitch in bytes allowed by memory copies
  size_t textureAlignment;        ///< Alignment requirement for textures
  size_t texturePitchAlignment;   ///< Pitch alignment requirement for texture references bound to
                                  ///< pitched memory
  int kernelExecTimeoutEnabled;   ///< Run time limit for kernels executed on the device
  int ECCEnabled;                 ///< Device has ECC support enabled
  int tccDriver;                  ///< 1:If device is Tesla device using TCC driver, else 0
  int cooperativeMultiDeviceUnmatchedFunc;       ///< HIP device supports cooperative launch on
                                                 ///< multiple
                                                 /// devices with unmatched functions
  int cooperativeMultiDeviceUnmatchedGridDim;    ///< HIP device supports cooperative launch on
                                                 ///< multiple
                                                 /// devices with unmatched grid dimensions
  int cooperativeMultiDeviceUnmatchedBlockDim;   ///< HIP device supports cooperative launch on
                                                 ///< multiple
                                                 /// devices with unmatched block dimensions
  int cooperativeMultiDeviceUnmatchedSharedMem;  ///< HIP device supports cooperative launch on
                                                 ///< multiple
                                                 /// devices with unmatched shared memories
  int isLargeBar;                                ///< 1: if it is a large PCI bar device, else 0
  int asicRevision;                              ///< Revision of the GPU in this device
  int managedMemory;                   ///< Device supports allocating managed memory on this system
  int directManagedMemAccessFromHost;  ///< Host can directly access managed memory on the device
                                       ///< without migration
  int concurrentManagedAccess;  ///< Device can coherently access managed memory concurrently with
                                ///< the CPU
  int pageableMemoryAccess;     ///< Device supports coherently accessing pageable memory
                                ///< without calling hipHostRegister on it
  int pageableMemoryAccessUsesHostPageTables;  ///< Device accesses pageable memory via the host's
                                               ///< page tables
} hipDeviceProp_t;

extern "C" hipError_t hipGetDeviceProperties(hipDeviceProp_t* props, hipDevice_t device) {
  // Removing this API from tracing.
  // This API is now in backwards compatibility mode and is not callable from newly compiled apps.
  HIP_INIT_VOID();

  if (props == nullptr) {
    return hipErrorInvalidValue;
  }

  if (unsigned(device) >= g_devices.size()) {
    return hipErrorInvalidDevice;
  }
  auto* deviceHandle = g_devices[device]->devices()[0];

  constexpr auto int32_max = static_cast<uint64_t>(std::numeric_limits<int32_t>::max());
  constexpr auto uint16_max = static_cast<uint64_t>(std::numeric_limits<uint16_t>::max()) + 1;
  hipDeviceProp_t deviceProps = {0};

  const auto& info = deviceHandle->info();
  const auto& isa = deviceHandle->isa();
  ::strncpy(deviceProps.name, info.boardName_, 128);
  deviceProps.totalGlobalMem = info.globalMemSize_;
  deviceProps.sharedMemPerBlock = info.localMemSizePerCU_;
  deviceProps.regsPerBlock = info.availableRegistersPerCU_;
  deviceProps.warpSize = info.wavefrontWidth_;
  deviceProps.maxThreadsPerBlock = info.maxWorkGroupSize_;
  deviceProps.maxThreadsDim[0] = info.maxWorkItemSizes_[0];
  deviceProps.maxThreadsDim[1] = info.maxWorkItemSizes_[1];
  deviceProps.maxThreadsDim[2] = info.maxWorkItemSizes_[2];
  deviceProps.maxGridSize[0] = int32_max;
  deviceProps.maxGridSize[1] = uint16_max;
  deviceProps.maxGridSize[2] = uint16_max;
  deviceProps.clockRate = info.maxEngineClockFrequency_ * 1000;
  deviceProps.memoryClockRate = info.maxMemoryClockFrequency_ * 1000;
  deviceProps.memoryBusWidth = info.globalMemChannels_;
  deviceProps.totalConstMem = std::min(info.maxConstantBufferSize_, int32_max);
  deviceProps.major = isa.versionMajor();
  deviceProps.minor = isa.versionMinor();
  deviceProps.multiProcessorCount = info.maxComputeUnits_;
  deviceProps.l2CacheSize = info.l2CacheSize_;
  deviceProps.maxThreadsPerMultiProcessor = info.maxThreadsPerCU_;
  deviceProps.computeMode = 0;
  deviceProps.clockInstructionRate = info.timeStampFrequency_;
  deviceProps.arch.hasGlobalInt32Atomics = 1;
  deviceProps.arch.hasGlobalFloatAtomicExch = 1;
  deviceProps.arch.hasSharedInt32Atomics = 1;
  deviceProps.arch.hasSharedFloatAtomicExch = 1;
  deviceProps.arch.hasFloatAtomicAdd = 1;
  deviceProps.arch.hasGlobalInt64Atomics = 1;
  deviceProps.arch.hasSharedInt64Atomics = 1;
  deviceProps.arch.hasDoubles = 1;
  deviceProps.arch.hasWarpVote = 1;
  deviceProps.arch.hasWarpBallot = 1;
  deviceProps.arch.hasWarpShuffle = 1;
  deviceProps.arch.hasFunnelShift = 0;
  deviceProps.arch.hasThreadFenceSystem = 1;
  deviceProps.arch.hasSyncThreadsExt = 0;
  deviceProps.arch.hasSurfaceFuncs = 0;
  deviceProps.arch.has3dGrid = 1;
  deviceProps.arch.hasDynamicParallelism = 0;
  deviceProps.concurrentKernels = 1;
  deviceProps.pciDomainID = info.pciDomainID;
  deviceProps.pciBusID = info.deviceTopology_.pcie.bus;
  deviceProps.pciDeviceID = info.deviceTopology_.pcie.device;
  deviceProps.maxSharedMemoryPerMultiProcessor = info.localMemSizePerCU_;
  deviceProps.canMapHostMemory = 1;
  // FIXME: This should be removed, targets can have character names as well.
  deviceProps.gcnArch = isa.versionMajor() * 100 + isa.versionMinor() * 10 + isa.versionStepping();
  sprintf(deviceProps.gcnArchName, "%s", isa.targetId());
  deviceProps.cooperativeLaunch = info.cooperativeGroups_;
  deviceProps.cooperativeMultiDeviceLaunch = info.cooperativeMultiDeviceGroups_;

  deviceProps.cooperativeMultiDeviceUnmatchedFunc = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedGridDim = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedBlockDim = info.cooperativeMultiDeviceGroups_;
  deviceProps.cooperativeMultiDeviceUnmatchedSharedMem = info.cooperativeMultiDeviceGroups_;

  deviceProps.maxTexture1DLinear =
      std::min(16 * info.imageMaxBufferSize_, int32_max);  // Max pixel size is 16 bytes
  deviceProps.maxTexture1D = std::min(info.image1DMaxWidth_, int32_max);
  deviceProps.maxTexture2D[0] = std::min(info.image2DMaxWidth_, int32_max);
  deviceProps.maxTexture2D[1] = std::min(info.image2DMaxHeight_, int32_max);
  deviceProps.maxTexture3D[0] = std::min(info.image3DMaxWidth_, int32_max);
  deviceProps.maxTexture3D[1] = std::min(info.image3DMaxHeight_, int32_max);
  deviceProps.maxTexture3D[2] = std::min(info.image3DMaxDepth_, int32_max);
  deviceProps.hdpMemFlushCntl = info.hdpMemFlushCntl;
  deviceProps.hdpRegFlushCntl = info.hdpRegFlushCntl;

  deviceProps.memPitch = std::min(info.maxMemAllocSize_, int32_max);
  deviceProps.textureAlignment = info.imageBaseAddressAlignment_;
  deviceProps.texturePitchAlignment = info.imagePitchAlignment_;
  deviceProps.kernelExecTimeoutEnabled = 0;
  deviceProps.ECCEnabled = info.errorCorrectionSupport_ ? 1 : 0;
  deviceProps.isLargeBar = info.largeBar_ ? 1 : 0;
  deviceProps.asicRevision = info.asicRevision_;

  // HMM capabilities
  deviceProps.managedMemory = info.hmmSupported_;
  deviceProps.concurrentManagedAccess = info.hmmSupported_;
  deviceProps.directManagedMemAccessFromHost = info.hmmDirectHostAccess_;
  deviceProps.pageableMemoryAccess = info.hmmCpuMemoryAccessible_;
  deviceProps.pageableMemoryAccessUsesHostPageTables = info.hostUnifiedMemory_;

  *props = deviceProps;
  return hipSuccess;
}