Global enum and defines#
Namespaces | |
| namespace | hip_impl |
Typedefs | |
| typedef enum __HIP_NODISCARD hipError_t | hipError_t |
| typedef struct ihipCtx_t * | hipCtx_t |
| typedef struct ihipExecutionCtx_t * | hipExecutionCtx_t |
| typedef struct ihipDevResourceDesc_t * | hipDevResourceDesc_t |
| typedef int | hipDevice_t |
| typedef struct ihipStream_t * | hipStream_t |
| typedef struct ihipModule_t * | hipModule_t |
| typedef struct ihipModuleSymbol_t * | hipFunction_t |
| typedef struct ihipLinkState_t * | hipLinkState_t |
| typedef struct ihipLibrary_t * | hipLibrary_t |
| typedef struct ihipKernel_t * | hipKernel_t |
| typedef struct ihipMemPoolHandle_t * | hipMemPool_t |
| typedef struct ihipEvent_t * | hipEvent_t |
| typedef void * | hipExternalMemory_t |
| typedef void * | hipExternalSemaphore_t |
| typedef struct _hipGraphicsResource | hipGraphicsResource |
| typedef hipGraphicsResource * | hipGraphicsResource_t |
| typedef struct ihipGraph * | hipGraph_t |
| typedef struct hipGraphNode * | hipGraphNode_t |
| typedef struct hipGraphExec * | hipGraphExec_t |
| typedef struct hipUserObject * | hipUserObject_t |
| typedef void(* | hipHostFn_t) (void *userData) |
| typedef struct ihipMemGenericAllocationHandle * | hipMemGenericAllocationHandle_t |
| typedef unsigned int | GLuint |
| typedef unsigned int | GLenum |
Detailed Description
Data Structure Documentation
◆ hipDeviceArch_t
| struct hipDeviceArch_t |
◆ hipUUID
◆ hipDeviceProp_t
| struct hipDeviceProp_t |
hipDeviceProp
| Data Fields | ||
|---|---|---|
| int | accessPolicyMaxWindowSize | Max value of access policy window. |
| hipDeviceArch_t | arch | Architectural feature flags. New for HIP. |
| int | asicRevision | Revision of the GPU in this device. |
| int | asyncEngineCount | Number of async engines. |
| int | canMapHostMemory | Check whether HIP can map host memory. |
| int | canUseHostPointerForRegisteredMem |
Device can access host registered memory with same address as the host |
| int | clockInstructionRate |
Frequency in khz of the timer used by the device-side "clock*" instructions. New for HIP. |
| int | clockRate | Max clock frequency of the multiProcessors in khz. |
| int | clusterLaunch | Device supports cluster launch. |
| int | computeMode | Compute mode. |
| int | computePreemptionSupported | Is compute preemption supported on the device. |
| int | concurrentKernels | Device can possibly execute multiple kernels concurrently. |
| int | concurrentManagedAccess |
Device can coherently access managed memory concurrently with the CPU |
| int | cooperativeLaunch | HIP device supports cooperative launch. |
| int | cooperativeMultiDeviceLaunch |
HIP device supports cooperative launch on multiple devices |
| int | cooperativeMultiDeviceUnmatchedBlockDim |
HIP device supports cooperative launch on multiple devices with unmatched block dimensions |
| 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 | cooperativeMultiDeviceUnmatchedSharedMem |
HIP device supports cooperative launch on multiple devices with unmatched shared memories |
| int | deferredMappingHipArraySupported |
Device supports deferred mapping HIP arrays and HIP mipmapped arrays |
| int | deviceOverlap | Deprecated. Use asyncEngineCount instead. |
| int | directManagedMemAccessFromHost |
Host can directly access managed memory on the device without migration |
| int | ECCEnabled | Device has ECC support enabled. |
| char | gcnArchName[256] | AMD GCN Arch Name. HIP Only. |
| int | globalL1CacheSupported | Indicates globals are cached in L1. |
| unsigned int | gpuDirectRDMAFlushWritesOptions |
Bitmask to be interpreted according to hipFlushGPUDirectRDMAWritesOptions |
| int | gpuDirectRDMASupported | Indicates device support of RDMA APIs. |
| int | gpuDirectRDMAWritesOrdering | value of hipGPUDirectRDMAWritesOrdering |
| unsigned int * | hdpMemFlushCntl | Addres of HDP_MEM_COHERENCY_FLUSH_CNTL register. |
| unsigned int * | hdpRegFlushCntl | Addres of HDP_REG_COHERENCY_FLUSH_CNTL register. |
| int | hipReserved[32] | Reserved for adding new entries for HIP/CUDA. |
| int | hostNativeAtomicSupported | Link between host and device supports native atomics. |
| int | hostRegisterReadOnlySupported |
Device supports using the hipHostRegisterReadOnly flag with hipHostRegistger |
| int | hostRegisterSupported | Device supports hipHostRegister. |
| int | integrated | APU vs dGPU. |
| int | ipcEventSupported | Device supports IPC events. |
| int | isLargeBar | 1: if it is a large PCI bar device, else 0 |
| int | isMultiGpuBoard | 1 if device is on a multi-GPU board, 0 if not. |
| int | kernelExecTimeoutEnabled | Run time limit for kernels executed on the device. |
| int | l2CacheSize | L2 cache size. |
| int | localL1CacheSupported | Locals are cahced in L1. |
| char | luid[8] | 8-byte unique identifier. Only valid on windows |
| unsigned int | luidDeviceNodeMask | LUID node mask. |
| int | major |
Major compute capability version. This indicates the core instruction set of the GPU architecture. For example, a value of 11 would correspond to Navi III (RDNA3). See the arch feature flags for portable ways to query feature caps. |
| int | managedMemory | Device supports allocating managed memory on this system. |
| int | maxBlocksPerMultiProcessor | Max number of blocks on CU. |
| int | maxGridSize[3] | Max grid dimensions (XYZ). |
| size_t | maxSharedMemoryPerMultiProcessor | Maximum Shared Memory Per CU. HIP Only. |
| int | maxSurface1D | Maximum 1D surface size. |
| int | maxSurface1DLayered[2] | Maximum 1D layered surface size. |
| int | maxSurface2D[2] | Maximum 2D surface size. |
| int | maxSurface2DLayered[3] | Maximum 2D layared surface size. |
| int | maxSurface3D[3] | Maximum 3D surface size. |
| int | maxSurfaceCubemap | Maximum cubemap surface size. |
| int | maxSurfaceCubemapLayered[2] | Maximum cubemap layered surface size. |
| int | maxTexture1D | Maximum number of elements in 1D images. |
| int | maxTexture1DLayered[2] | Maximum number of elements in 1D array images. |
| int | maxTexture1DLinear | Maximum size for 1D textures bound to linear memory. |
| int | maxTexture1DMipmap | Maximum 1D mipmap texture size. |
| int | maxTexture2D[2] | Maximum dimensions (width, height) of 2D images, in image elements. |
| int | maxTexture2DGather[2] | Maximum 2D tex dimensions if gather has to be performed. |
| int | maxTexture2DLayered[3] | Maximum number of elements in 2D array images. |
| int | maxTexture2DLinear[3] | Maximum 2D tex dimensions if tex are bound to pitched memory. |
| int | maxTexture2DMipmap[2] | Maximum number of elements in 2D array mipmap of images. |
| int | maxTexture3D[3] |
Maximum dimensions (width, height, depth) of 3D images, in image elements |
| int | maxTexture3DAlt[3] | Maximum alternate 3D texture dims. |
| int | maxTextureCubemap | Maximum cubemap texture dims. |
| int | maxTextureCubemapLayered[2] | Maximum cubemaps layered texture dims. |
| int | maxThreadsDim[3] | Max number of threads in each dimension (XYZ) of a block. |
| int | maxThreadsPerBlock | Max work items per work group or workgroup max size. |
| int | maxThreadsPerMultiProcessor | Maximum resident threads per multi-processor. |
| int | memoryBusWidth | Global memory bus width in bits. |
| int | memoryClockRate | Max global memory clock frequency in khz. |
| int | memoryPoolsSupported | Indicates if device supports hipMallocAsync and hipMemPool APIs. |
| unsigned int | memoryPoolSupportedHandleTypes | Bitmask of handle types support with mempool based IPC. |
| size_t | memPitch |
Maximum pitch in bytes allowed by memory copies pitched memory |
| int | minor |
Minor compute capability version. This indicates a particular configuration, feature set, or variation within the group represented by the major compute capability version. For example, different models within the same major version might have varying levels of support for certain features or optimizations. See the arch feature flags for portable ways to query feature caps. |
| int | multiGpuBoardGroupID | Unique identifier for a group of devices on same multiboard GPU. |
| int | multiProcessorCount |
Number of multi-processors. When the GPU works in Compute Unit (CU) mode, this value equals the number of CUs; when in Workgroup Processor (WGP) mode, this value equels half of CUs, because a single WGP contains two CUs. |
| char | name[256] | Device name. |
| 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 |
| int | pciBusID | PCI Bus ID. |
| int | pciDeviceID | PCI Device ID. |
| int | pciDomainID | PCI Domain ID. |
| int | persistingL2CacheMaxSize | Device's max L2 persisting lines in bytes. |
| int | regsPerBlock | Registers per block. |
| int | regsPerMultiprocessor | registers available per multiprocessor |
| int | reserved[63] | CUDA Reserved. |
| size_t | reservedSharedMemPerBlock | Shared memory reserved by driver per block. |
| size_t | sharedMemPerBlock | Size of shared memory per block (in bytes). |
| size_t | sharedMemPerBlockOptin | Per device m ax shared mem per block usable by special opt in. |
| size_t | sharedMemPerMultiprocessor | Amount of shared memory available per multiprocessor. |
| int | singleToDoublePrecisionPerfRatio | Deprecated. CUDA only. |
| int | sparseHipArraySupported | Indicates if device supports sparse hip arrays. |
| int | streamPrioritiesSupported | Device supports stream priority. |
| size_t | surfaceAlignment | Alignment requirement for surface. |
| int | tccDriver | 1:If device is Tesla device using TCC driver, else 0 |
| size_t | textureAlignment | Alignment requirement for textures. |
| size_t | texturePitchAlignment | Pitch alignment requirement for texture references bound to. |
| int | timelineSemaphoreInteropSupported | Indicates external timeline semaphore support. |
| size_t | totalConstMem |
Size of shared constant memory region on the device (in bytes). |
| size_t | totalGlobalMem | Size of global memory region (in bytes). |
| int | unifiedAddressing | Does device and host share unified address space. |
| int | unifiedFunctionPointers | Indicates device supports unified function pointers. |
| hipUUID | uuid | UUID of a device. |
| int | warpSize | Warp size. |
◆ hipPointerAttribute_t
| struct hipPointerAttribute_t |
Pointer attributes
| Data Fields | ||
|---|---|---|
| unsigned | allocationFlags | |
| int | device | |
| void * | devicePointer | |
| void * | hostPointer | |
| int | isManaged | |
| enum hipMemoryType | type | |
◆ hipDevSmResource
| struct hipDevSmResource |
◆ hipDevWorkqueueConfigResource
| struct hipDevWorkqueueConfigResource |
| Data Fields | ||
|---|---|---|
| int | device | |
| hipDevWorkqueueConfigScope | sharingScope | |
| unsigned int | wqConcurrencyLimit | |
◆ hipDevWorkqueueResource
| struct hipDevWorkqueueResource |
| Data Fields | ||
|---|---|---|
| unsigned char | reserved[HIP_RESOURCE_ABI_BYTES] | |
◆ hipDevResource
| struct hipDevResource |
| Data Fields | ||
|---|---|---|
| union hipDevResource.__unnamed16__ | __unnamed__ | |
| unsigned char | _internal_padding[92] | |
| struct hipDevResource_st * | nextResource | |
| hipDevResourceType | type | |
◆ hipDevSmResourceGroupParams
| struct hipDevSmResourceGroupParams |
◆ hipIpcMemHandle_t
| struct hipIpcMemHandle_t |
| Data Fields | ||
|---|---|---|
| char | reserved[HIP_IPC_HANDLE_SIZE] | |
◆ hipIpcEventHandle_t
| struct hipIpcEventHandle_t |
| Data Fields | ||
|---|---|---|
| char | reserved[HIP_IPC_HANDLE_SIZE] | |
◆ hipMemFabricHandle_t
| struct hipMemFabricHandle_t |
| Data Fields | ||
|---|---|---|
| unsigned char | data[HIP_IPC_HANDLE_SIZE] | |
◆ hipFuncAttributes
| struct hipFuncAttributes |
◆ hipStreamBatchMemOpParams
| union hipStreamBatchMemOpParams |
Union representing batch memory operation parameters for HIP streams.
hipStreamBatchMemOpParams is used to specify the parameters for batch memory operations in a HIP stream. This union supports various operations including waiting for a specific value, writing a value, and different flags for wait conditions.
The union includes fields for different types of operations defined in the enum hipStreamBatchMemOpType:
- hipStreamMemOpWaitValue32: Wait for a 32-bit value.
- hipStreamMemOpWriteValue32: Write a 32-bit value.
- hipStreamMemOpWaitValue64: Wait for a 64-bit value.
- hipStreamMemOpWriteValue64: Write a 64-bit value.
Each operation type includes an address, the value to wait for or write, flags, and an optional alias that is not relevant on AMD GPUs. Flags can be used to specify different wait conditions such as equality, bitwise AND, greater than or equal, and bitwise NOR.
Example usage:
| Data Fields | ||
|---|---|---|
| struct hipStreamMemOpFlushRemoteWritesParams_t | flushRemoteWrites | Currently not supported on AMD. |
| struct hipStreamMemOpMemoryBarrierParams_t | memoryBarrier | Currently not supported on AMD. |
| hipStreamBatchMemOpType | operation | |
| uint64_t | pad[6] | |
| struct hipStreamMemOpWaitValueParams_t | waitValue | |
| struct hipStreamMemOpWriteValueParams_t | writeValue | |
◆ hipBatchMemOpNodeParams
| struct hipBatchMemOpNodeParams |
Structure representing node parameters for batch memory operations in HIP graphs.
hipBatchMemOpNodeParams is used to specify the parameters for batch memory operations in HIP graphs. This struct includes the context to use for the operations, the number of operations, and an array of hipStreamBatchMemOpParams that describe the operations.
The structure includes the following fields:
- ctx: The HIP context to use for the operations.
- count: The number of operations in the paramArray.
- paramArray: A pointer to an array of hipStreamBatchMemOpParams.
- flags: Flags to control the node.
Example usage:
| Data Fields | ||
|---|---|---|
| unsigned int | count | |
| hipCtx_t | ctx | |
| unsigned int | flags | |
| hipStreamBatchMemOpParams * | paramArray | |
◆ hipMemAccessDesc
| struct hipMemAccessDesc |
Memory access descriptor structure is used to specify memory access permissions for a virtual memory region in Virtual Memory Management API. This structure changes read, and write permissions for specific memory regions.
| Data Fields | ||
|---|---|---|
| hipMemAccessFlags | flags | Accessibility flags to set. |
| hipMemLocation | location | Location on which the accessibility has to change. |
◆ hipMemPoolProps
| struct hipMemPoolProps |
Specifies the properties of allocations made from the pool.
| Data Fields | ||
|---|---|---|
| hipMemAllocationType | allocType |
Allocation type. Currently must be specified as hipMemAllocationTypePinned. |
| hipMemAllocationHandleType | handleTypes | Handle types that will be supported by allocations from the pool. |
| hipMemLocation | location | Location where allocations should reside. |
| size_t | maxSize | Maximum pool size. When set to 0, defaults to a system dependent value. |
| unsigned char | reserved[56] | Reserved for future use, must be 0. |
| void * | win32SecurityAttributes |
Windows-specific LPSECURITYATTRIBUTES required when |
◆ hipMemPoolPtrExportData
| struct hipMemPoolPtrExportData |
◆ hipLaunchParams
| struct hipLaunchParams |
struct hipLaunchParams_t
| Data Fields | ||
|---|---|---|
| void ** | args | Arguments. |
| dim3 | blockDim | Block dimensions. |
| void * | func | Device function symbol. |
| dim3 | gridDim | Grid dimensions. |
| size_t | sharedMem | Shared memory. |
| hipStream_t | stream | Stream identifier. |
◆ hipFunctionLaunchParams
| struct hipFunctionLaunchParams |
struct hipFunctionLaunchParams_t
| Data Fields | ||
|---|---|---|
| unsigned int | blockDimX | X dimension of each thread block. |
| unsigned int | blockDimY | Y dimension of each thread block. |
| unsigned int | blockDimZ | Z dimension of each thread block. |
| hipFunction_t | function | Kernel to launch. |
| unsigned int | gridDimX | Width(X) of grid in blocks. |
| unsigned int | gridDimY | Height(Y) of grid in blocks. |
| unsigned int | gridDimZ | Depth(Z) of grid in blocks. |
| hipStream_t | hStream | Stream identifier. |
| void ** | kernelParams | Kernel parameters. |
| unsigned int | sharedMemBytes | Shared memory. |
◆ hipExternalMemoryHandleDesc
| struct hipExternalMemoryHandleDesc |
| Data Fields | ||
|---|---|---|
| unsigned int | flags | |
| union hipExternalMemoryHandleDesc.handle | handle | |
| unsigned int | reserved[16] | |
| unsigned long long | size | |
| hipExternalMemoryHandleType | type | |
◆ hipExternalMemoryBufferDesc
| struct hipExternalMemoryBufferDesc |
◆ hipExternalMemoryMipmappedArrayDesc
| struct hipExternalMemoryMipmappedArrayDesc |
| Data Fields | ||
|---|---|---|
| hipExtent | extent | |
| unsigned int | flags | |
| hipChannelFormatDesc | formatDesc | |
| unsigned int | numLevels | |
| unsigned long long | offset | |
◆ hipExternalSemaphoreHandleDesc
| struct hipExternalSemaphoreHandleDesc |
| Data Fields | ||
|---|---|---|
| unsigned int | flags | |
| union hipExternalSemaphoreHandleDesc.handle | handle | |
| unsigned int | reserved[16] | |
| hipExternalSemaphoreHandleType | type | |
◆ hipExternalSemaphoreSignalParams
| struct hipExternalSemaphoreSignalParams |
| Data Fields | ||
|---|---|---|
| unsigned int | flags | |
| struct hipExternalSemaphoreSignalParams.params | params | |
| unsigned int | reserved[16] | |
◆ hipExternalSemaphoreWaitParams
| struct hipExternalSemaphoreWaitParams |
External semaphore wait parameters, compatible with driver type
| Data Fields | ||
|---|---|---|
| unsigned int | flags | |
| struct hipExternalSemaphoreWaitParams.params | params | |
| unsigned int | reserved[16] | |
◆ hipHostNodeParams
| struct hipHostNodeParams |
| Data Fields | ||
|---|---|---|
| hipHostFn_t | fn | |
| void * | userData | |
◆ hipKernelNodeParams
| struct hipKernelNodeParams |
◆ hipMemsetParams
| struct hipMemsetParams |
◆ hipMemAllocNodeParams
| struct hipMemAllocNodeParams |
| Data Fields | ||
|---|---|---|
| size_t | accessDescCount |
The number of access descriptors. Must not be bigger than the number of GPUs |
| const hipMemAccessDesc * | accessDescs | The number of memory access descriptors. |
| size_t | bytesize | The size of the requested allocation in bytes. |
| void * | dptr | Returned device address of the allocation. |
| hipMemPoolProps | poolProps |
Pool properties, which contain where the location should reside |
◆ hipAccessPolicyWindow
| struct hipAccessPolicyWindow |
| Data Fields | ||
|---|---|---|
| void * | base_ptr | Starting address of the access policy window. |
| hipAccessProperty | hitProp | hipAccessProperty set for hit |
| float | hitRatio | hitRatio specifies percentage of lines assigned hitProp |
| hipAccessProperty | missProp | hipAccessProperty set for miss |
| size_t | num_bytes | Size in bytes of the window policy. |
◆ hipLaunchMemSyncDomainMap
| struct hipLaunchMemSyncDomainMap |
◆ hipExtDynDataPrefetchRegion
| struct hipExtDynDataPrefetchRegion |
Describes one 2D memory region to prefetch into L2 cache.
address must be aligned to the device's L2 cache line size. width is measured in bytes and must be a multiple of the cache line size. height is the number of rows to prefetch. stride is the byte stride between the start of consecutive rows.
◆ hipExtDynDataPrefetchConfig
| struct hipExtDynDataPrefetchConfig |
Configuration for dynamic data prefetch.
Pointed to by a launch attribute via hipLaunchAttributeExtDynDataPrefetch. numRegions must not exceed the device limit queried via hipDeviceAttributeMaxDynDataPrefetchRegions (use hipDeviceGetAttribute).
Cooperative prefetch is always enabled internally; the user only controls the temporal policy.
| Data Fields | ||
|---|---|---|
| unsigned int | numRegions | Number of valid regions (1-max) |
| hipExtDynDataPrefetchRegion | regions[HIP_EXT_DYN_DATA_PREFETCH_MAX_REGIONS] | Prefetch regions. |
| hipExtDynDataPrefetchTemporal | temporal | Cache retention policy for prefetched data. |
◆ hipLaunchAttributeValue
| union hipLaunchAttributeValue |
Launch Attribute Value
| Data Fields | ||
|---|---|---|
| hipAccessPolicyWindow | accessPolicyWindow | Value of launch attribute hipLaunchAttributeAccessPolicyWindow. |
| struct hipLaunchAttributeValue.clusterDim | clusterDim |
Specifies the desired cluster dimensions for a kernel launch. This opaque type is used as the value for the launch attribute hipLaunchAttributeClusterDimension. It defines the dimensions of the compute cluster in terms of blocks, where each field must evenly divide the corresponding grid dimension:
|
| hipClusterSchedulingPolicy | clusterSchedulingPolicyPreference |
Value of launch attribute :: hipLaunchAttributeClusterSchedulingPolicyPreference determines the preferred strategy for distributing blocks within a compute cluster |
| int | cooperative |
Value of launch attribute hipLaunchAttributeCooperative. Indicates whether the kernel is cooperative. |
| const hipExtDynDataPrefetchConfig * | dynDataPrefetch | Value of launch attribute hipLaunchAttributeExtDynDataPrefetch. |
| hipLaunchMemSyncDomain | memSyncDomain | Value of launch attribute hipLaunchAttributeMemSyncDomain. |
| hipLaunchMemSyncDomainMap | memSyncDomainMap | Value of launch attribute hipLaunchAttributeMemSyncDomainMap. |
| char | pad[64] | 64 byte padding |
| int | priority |
Value of launch attribute :: hipLaunchAttributePriority. Execution priority of kernel |
| hipSynchronizationPolicy | syncPolicy |
Value of launch attribute :: hipLaunchAttributeSynchronizationPolicy. Used to work queued up in stream |
◆ hipGraphInstantiateParams
| struct hipGraphInstantiateParams |
Graph Instantiation parameters
| Data Fields | ||
|---|---|---|
| hipGraphNode_t | errNode_out |
The node which caused instantiation to fail, if any |
| unsigned long long | flags |
Instantiation flags |
| hipGraphInstantiateResult | result_out |
Whether instantiation was successful. If it failed, the reason why |
| hipStream_t | uploadStream |
Upload stream |
◆ hipMemAllocationProp
| struct hipMemAllocationProp |
Memory allocation properties
| Data Fields | ||
|---|---|---|
| union hipMemAllocationProp.__unnamed35__ | __unnamed__ | |
| struct hipMemAllocationProp.allocFlags | allocFlags | |
| hipMemLocation | location | Memory location. |
| hipMemAllocationType | type | Memory allocation type. |
| void * | win32HandleMetaData | Metadata for Win32 handles. |
◆ hipExternalSemaphoreSignalNodeParams
| struct hipExternalSemaphoreSignalNodeParams |
External semaphore signal node parameters
| Data Fields | ||
|---|---|---|
| hipExternalSemaphore_t * | extSemArray |
< Array containing external semaphore handles. Array containing parameters of external signal semaphore. |
| unsigned int | numExtSems | |
| const hipExternalSemaphoreSignalParams * | paramsArray | Total number of handles and parameters contained in extSemArray and paramsArray. |
◆ hipExternalSemaphoreWaitNodeParams
| struct hipExternalSemaphoreWaitNodeParams |
External semaphore wait node parameters
| Data Fields | ||
|---|---|---|
| hipExternalSemaphore_t * | extSemArray |
< Array containing external semaphore handles. Array containing parameters of external wait semaphore. |
| unsigned int | numExtSems | |
| const hipExternalSemaphoreWaitParams * | paramsArray | Total number of handles and parameters contained in extSemArray and paramsArray. |
◆ hipArrayMapInfo
| struct hipArrayMapInfo |
Map info for arrays
| Data Fields | ||
|---|---|---|
| unsigned int | deviceBitMask | Device ordinal bit mask. |
| unsigned int | flags | flags for future use, must be zero now. |
| union hipArrayMapInfo.memHandle | memHandle | |
| hipMemHandleType | memHandleType | Memory handle type. |
| hipMemOperationType | memOperationType | Memory operation type. |
| unsigned long long | offset | Offset within the memory. |
| unsigned int | reserved[2] | Reserved for future use, must be zero now. |
| union hipArrayMapInfo.resource | resource | |
| hipResourceType | resourceType | Resource type. |
| union hipArrayMapInfo.subresource | subresource | |
| hipArraySparseSubresourceType | subresourceType | Sparse subresource type. |
◆ hipMemcpyNodeParams
| struct hipMemcpyNodeParams |
Memcpy node params
| Data Fields | ||
|---|---|---|
| hipMemcpy3DParms | copyParams | Params set for the memory copy. |
| int | flags | Must be zero. |
| int | reserved[3] | Must be zero. |
◆ hipChildGraphNodeParams
| struct hipChildGraphNodeParams |
Child graph node params
| Data Fields | ||
|---|---|---|
| hipGraph_t | graph |
Either the child graph to clone into the node, or a handle to the graph possesed by the node used during query |
◆ hipEventWaitNodeParams
| struct hipEventWaitNodeParams |
Event record node params
| Data Fields | ||
|---|---|---|
| hipEvent_t | event | Event to wait on. |
◆ hipEventRecordNodeParams
| struct hipEventRecordNodeParams |
Event record node params
| Data Fields | ||
|---|---|---|
| hipEvent_t | event | The event to be recorded when node executes. |
◆ hipMemFreeNodeParams
◆ hipGraphNodeParams
| struct hipGraphNodeParams |
Params for different graph nodes
| Data Fields | ||
|---|---|---|
| union hipGraphNodeParams.__unnamed43__ | __unnamed__ | |
| int | reserved0[3] | |
| long long | reserved2 | |
| hipGraphNodeType | type | |
◆ hipGraphEdgeData
| struct hipGraphEdgeData |
◆ hipLaunchAttribute
| struct hipLaunchAttribute |
Used to specify custom attributes for launching kernels
| Data Fields | ||
|---|---|---|
| union hipLaunchAttribute.__unnamed45__ | __unnamed__ | |
| hipLaunchAttributeID | id | Identifier of the launch attribute. |
| char | pad[8 - sizeof(hipLaunchAttributeID)] | Padding to align the structure to 8 bytes. |
◆ hipLaunchConfig_t
| struct hipLaunchConfig_t |
HIP extensible launch configuration
| Data Fields | ||
|---|---|---|
| hipLaunchAttribute * | attrs | Attributes list. |
| dim3 | blockDim | Block dimensions. |
| size_t | dynamicSmemBytes | Dynamic shared-memory size per thread block. |
| dim3 | gridDim | Grid dimensions. |
| unsigned int | numAttrs | Number of attributes. |
| hipStream_t | stream | Stream identifier. |
◆ HIP_LAUNCH_CONFIG
| struct HIP_LAUNCH_CONFIG |
HIP driver extensible launch configuration
| Data Fields | ||
|---|---|---|
| hipLaunchAttribute * | attrs | Attribute list. |
| unsigned int | blockDimX | Thread block dimension in X. |
| unsigned int | blockDimY | Thread block dimension in Y. |
| unsigned int | blockDimZ | Thread block dimension in Z. |
| unsigned int | gridDimX | Grid width in blocks. |
| unsigned int | gridDimY | Grid height in blocks. |
| unsigned int | gridDimZ | Grid depth in blocks. |
| hipStream_t | hStream | HIP stream identifier. |
| unsigned int | numAttrs | Number of attributes. |
| unsigned int | sharedMemBytes | Dynamic shared-memory size in bytes per block. |
◆ hipArrayMemoryRequirements
| struct hipArrayMemoryRequirements |
◆ hipDevResource.__unnamed16__
| union hipDevResource.__unnamed16__ |
| Data Fields | ||
|---|---|---|
| unsigned char | _oversize[HIP_RESOURCE_ABI_BYTES] | |
| hipDevSmResource | sm | |
| hipDevWorkqueueResource | wq | |
| hipDevWorkqueueConfigResource | wqConfig | |
◆ hipExternalMemoryHandleDesc.handle
| union hipExternalMemoryHandleDesc.handle |
| Data Fields | ||
|---|---|---|
| int | fd | |
| const void * | nvSciBufObject | |
| struct hipExternalMemoryHandleDesc.handle.win32 | win32 | |
◆ hipExternalMemoryHandleDesc.handle.win32
◆ hipExternalSemaphoreHandleDesc.handle
| union hipExternalSemaphoreHandleDesc.handle |
| Data Fields | ||
|---|---|---|
| int | fd | |
| const void * | NvSciSyncObj | |
| struct hipExternalSemaphoreHandleDesc.handle.win32 | win32 | |
◆ hipExternalSemaphoreHandleDesc.handle.win32
◆ hipExternalSemaphoreSignalParams.params
| struct hipExternalSemaphoreSignalParams.params |
| Data Fields | ||
|---|---|---|
| struct hipExternalSemaphoreSignalParams.params.fence | fence | |
| struct hipExternalSemaphoreSignalParams.params.keyedMutex | keyedMutex | |
| union hipExternalSemaphoreSignalParams.params.nvSciSync | nvSciSync | |
| unsigned int | reserved[12] | |
◆ hipExternalSemaphoreSignalParams.params.fence
◆ hipExternalSemaphoreSignalParams.params.nvSciSync
| union hipExternalSemaphoreSignalParams.params.nvSciSync |
◆ hipExternalSemaphoreSignalParams.params.keyedMutex
◆ hipExternalSemaphoreWaitParams.params
| struct hipExternalSemaphoreWaitParams.params |
| Data Fields | ||
|---|---|---|
| struct hipExternalSemaphoreWaitParams.params.fence | fence | |
| struct hipExternalSemaphoreWaitParams.params.keyedMutex | keyedMutex | |
| union hipExternalSemaphoreWaitParams.params.nvSciSync | nvSciSync | |
| unsigned int | reserved[10] | |
◆ hipExternalSemaphoreWaitParams.params.fence
◆ hipExternalSemaphoreWaitParams.params.nvSciSync
| union hipExternalSemaphoreWaitParams.params.nvSciSync |
◆ hipExternalSemaphoreWaitParams.params.keyedMutex
| struct hipExternalSemaphoreWaitParams.params.keyedMutex |
◆ hipLaunchAttributeValue.clusterDim
| struct hipLaunchAttributeValue.clusterDim |
Specifies the desired cluster dimensions for a kernel launch.
This opaque type is used as the value for the launch attribute hipLaunchAttributeClusterDimension. It defines the dimensions of the compute cluster in terms of blocks, where each field must evenly divide the corresponding grid dimension:
x:Number of blocks along the X-axis.y:Number of blocks along the Y-axis.z:Number of blocks along the Z-axis.
| Data Fields | ||
|---|---|---|
| unsigned int | x | |
| unsigned int | y | |
| unsigned int | z | |
◆ hipMemAllocationProp.__unnamed35__
| union hipMemAllocationProp.__unnamed35__ |
| Data Fields | ||
|---|---|---|
| hipMemAllocationHandleType | requestedHandleType | Requested handle type. |
| hipMemAllocationHandleType | requestedHandleTypes | Requested handle types. |
◆ hipMemAllocationProp.allocFlags
| struct hipMemAllocationProp.allocFlags |
◆ hipArrayMapInfo.resource
| union hipArrayMapInfo.resource |
| Data Fields | ||
|---|---|---|
| hipArray_t | array | |
| hipMipmappedArray | mipmap | |
◆ hipArrayMapInfo.subresource
| union hipArrayMapInfo.subresource |
| Data Fields | ||
|---|---|---|
| struct hipArrayMapInfo.subresource.miptail | miptail | |
| struct hipArrayMapInfo.subresource.sparseLevel | sparseLevel | |
◆ hipArrayMapInfo.subresource.sparseLevel
| struct hipArrayMapInfo.subresource.sparseLevel |
◆ hipArrayMapInfo.subresource.miptail
| struct hipArrayMapInfo.subresource.miptail |
◆ hipArrayMapInfo.memHandle
| union hipArrayMapInfo.memHandle |
| Data Fields | ||
|---|---|---|
| hipMemGenericAllocationHandle_t | memHandle | |
◆ hipGraphNodeParams.__unnamed43__
| union hipGraphNodeParams.__unnamed43__ |
| Data Fields | ||
|---|---|---|
| hipMemAllocNodeParams | alloc | |
| hipEventRecordNodeParams | eventRecord | |
| hipEventWaitNodeParams | eventWait | |
| hipExternalSemaphoreSignalNodeParams | extSemSignal | |
| hipExternalSemaphoreWaitNodeParams | extSemWait | |
| hipMemFreeNodeParams | free | |
| hipChildGraphNodeParams | graph | |
| hipHostNodeParams | host | |
| hipKernelNodeParams | kernel | |
| hipMemcpyNodeParams | memcpy | |
| hipMemsetParams | memset | |
| long long | reserved1[29] | |
◆ hipLaunchAttribute.__unnamed45__
| union hipLaunchAttribute.__unnamed45__ |
| Data Fields | ||
|---|---|---|
| hipLaunchAttributeValue | val | Value associated with the launch attribute. |
| hipLaunchAttributeValue | value | Value associated with the launch attribute. |
Macro Definition Documentation
◆ __dparm
| #define __dparm | ( | x | ) | = x |
◆ __HIP_NODISCARD
| #define __HIP_NODISCARD |
◆ GENERIC_GRID_LAUNCH
| #define GENERIC_GRID_LAUNCH 1 |
◆ HIP_DEPRECATED
| #define HIP_DEPRECATED | ( | msg | ) | __attribute__((deprecated(msg))) |
◆ HIP_DEPRECATED_MSG
| #define HIP_DEPRECATED_MSG |
◆ HIP_EXT_DYN_DATA_PREFETCH_MAX_REGIONS
| #define HIP_EXT_DYN_DATA_PREFETCH_MAX_REGIONS 2 |
Compile-time upper bound for the number of prefetch regions in hipExtDynDataPrefetchConfig. The actual device limit is queried at runtime via hipDeviceAttributeMaxDynDataPrefetchRegions.
◆ HIP_GET_PROC_ADDRESS_DEFAULT
| #define HIP_GET_PROC_ADDRESS_DEFAULT 0x0 |
Default flag. Equivalent to HIP_GET_PROC_ADDRESS_PER_THREAD_DEFAULT_STREAM if compiled with -fgpu-default-stream=per-thread flag or HIP_API_PER_THREAD_DEFAULT_STREAM macro is defined.
◆ HIP_GET_PROC_ADDRESS_LEGACY_STREAM
| #define HIP_GET_PROC_ADDRESS_LEGACY_STREAM 0x1 |
Search for all symbols except the corresponding per-thread versions.
◆ HIP_GET_PROC_ADDRESS_PER_THREAD_DEFAULT_STREAM
| #define HIP_GET_PROC_ADDRESS_PER_THREAD_DEFAULT_STREAM 0x2 |
Search for all symbols including the per-thread versions. If a per-thread version cannot be found, returns the legacy version.
◆ HIP_IPC_HANDLE_SIZE
| #define HIP_IPC_HANDLE_SIZE 64 |
◆ HIP_LAUNCH_PARAM_BUFFER_POINTER
| #define HIP_LAUNCH_PARAM_BUFFER_POINTER ((void*)0x01) |
◆ HIP_LAUNCH_PARAM_BUFFER_SIZE
| #define HIP_LAUNCH_PARAM_BUFFER_SIZE ((void*)0x02) |
◆ HIP_LAUNCH_PARAM_END
| #define HIP_LAUNCH_PARAM_END ((void*)0x03) |
◆ HIP_RESOURCE_ABI_BYTES
| #define HIP_RESOURCE_ABI_BYTES 40 |
◆ hipArrayCubemap
| #define hipArrayCubemap 0x04 |
◆ hipArrayDefault
| #define hipArrayDefault 0x00 |
Default HIP array allocation flag.
◆ hipArrayLayered
| #define hipArrayLayered 0x01 |
◆ hipArraySurfaceLoadStore
| #define hipArraySurfaceLoadStore 0x02 |
◆ hipArrayTextureGather
| #define hipArrayTextureGather 0x08 |
◆ hipChooseDevice
| #define hipChooseDevice hipChooseDeviceR0600 |
◆ hipCooperativeLaunchMultiDeviceNoPostSync
| #define hipCooperativeLaunchMultiDeviceNoPostSync 0x02 |
◆ hipCooperativeLaunchMultiDeviceNoPreSync
| #define hipCooperativeLaunchMultiDeviceNoPreSync 0x01 |
◆ hipCpuDeviceId
| #define hipCpuDeviceId ((int)-1) |
◆ hipDeviceLmemResizeToMax
| #define hipDeviceLmemResizeToMax 0x10 |
◆ hipDeviceMallocContiguous
| #define hipDeviceMallocContiguous 0x4 |
Memory allocated will be contiguous.
◆ hipDeviceMallocDefault
| #define hipDeviceMallocDefault 0x0 |
◆ hipDeviceMallocFinegrained
| #define hipDeviceMallocFinegrained 0x1 |
Memory is allocated in fine grained region of device.
◆ hipDeviceMallocUncached
| #define hipDeviceMallocUncached 0x3 |
Memory allocated will be uncached.
◆ hipDeviceMapHost
| #define hipDeviceMapHost 0x8 |
◆ hipDeviceProp_t
| #define hipDeviceProp_t hipDeviceProp_tR0600 |
◆ hipDeviceScheduleAuto
| #define hipDeviceScheduleAuto 0x0 |
Automatically select between Spin and Yield.
◆ hipDeviceScheduleBlockingSync
| #define hipDeviceScheduleBlockingSync 0x4 |
◆ hipDeviceScheduleMask
| #define hipDeviceScheduleMask 0x7 |
◆ hipDeviceScheduleSpin
| #define hipDeviceScheduleSpin 0x1 |
Dedicate a CPU core to spin-wait. Provides lowest latency, but burns a CPU core and may consume more power.
◆ hipDeviceScheduleYield
| #define hipDeviceScheduleYield 0x2 |
Yield the CPU to the operating system when waiting. May increase latency, but lowers power and is friendlier to other threads in the system.
◆ hipDrvLaunchAttribute
| #define hipDrvLaunchAttribute hipLaunchAttribute |
◆ hipDrvLaunchAttributeCooperative
| #define hipDrvLaunchAttributeCooperative hipLaunchAttributeCooperative |
hip Drv attributes
◆ hipDrvLaunchAttributeID
| #define hipDrvLaunchAttributeID hipLaunchAttributeID |
◆ hipDrvLaunchAttributeValue
| #define hipDrvLaunchAttributeValue hipLaunchAttributeValue |
◆ hipEnableDefault
| #define hipEnableDefault 0x0 |
Default flag. Equivalent to hipEnablePerThreadDefaultStream if compiled with -fgpu-default-stream=per-thread flag or HIP_API_PER_THREAD_DEFAULT_STREAM macro is defined.
◆ hipEnableLegacyStream
| #define hipEnableLegacyStream 0x1 |
Search for all symbols except the corresponding per-thread versions.
◆ hipEnablePerThreadDefaultStream
| #define hipEnablePerThreadDefaultStream 0x2 |
Search for all symbols including the per-thread versions. If a per-thread version cannot be found, returns the legacy version.
◆ hipEventBlockingSync
| #define hipEventBlockingSync 0x1 |
Waiting will yield CPU. Power-friendly and usage-friendly but may increase latency.
◆ hipEventDefault
| #define hipEventDefault 0x0 |
Default flags.
◆ hipEventDisableSystemFence
| #define hipEventDisableSystemFence 0x20000000 |
Disable performing a system scope sequentially consistent memory fence when the event transitions from recording to recorded. This can be used for events that are only being used to measure timing, and do not require the event inspection operations (see hipEventSynchronize, hipEventQuery, and hipEventElapsedTime) to synchronize-with the work on which the recorded event (see hipEventRecord) is waiting. On some AMD GPU devices this can improve the accuracy of timing measurements by avoiding the cost of cache writeback and invalidation, and the performance impact of those actions on the execution of following work.
◆ hipEventDisableTiming
| #define hipEventDisableTiming 0x2 |
Disable event's capability to record timing information. May improve performance.
◆ hipEventInterprocess
| #define hipEventInterprocess 0x4 |
Event can support IPC. hipEventDisableTiming also must be set.
◆ hipEventRecordDefault
| #define hipEventRecordDefault 0x00 |
Default flag.
◆ hipEventRecordExternal
| #define hipEventRecordExternal 0x01 |
Event is captured in the graph as an external event node when performing stream capture.
◆ hipEventReleaseToDevice
| #define hipEventReleaseToDevice 0x40000000 |
Use a device-scope release when recording this event. This flag is useful to obtain more precise timings of commands between events. The flag is a no-op on CUDA platforms.
◆ hipEventReleaseToSystem
| #define hipEventReleaseToSystem 0x80000000 |
Use a system-scope release when recording this event. This flag is useful to make non-coherent host memory visible to the host. The flag is a no-op on CUDA platforms.
◆ hipEventWaitDefault
| #define hipEventWaitDefault 0x00 |
Default flag.
◆ hipEventWaitExternal
| #define hipEventWaitExternal 0x01 |
Wait is captured in the graph as an external event node when performing stream capture.
◆ hipExtAnyOrderLaunch
| #define hipExtAnyOrderLaunch 0x01 |
AnyOrderLaunch of kernels.
◆ hipExternalMemoryDedicated
| #define hipExternalMemoryDedicated 0x1 |
◆ hipExtHostRegisterCoarseGrained
| #define hipExtHostRegisterCoarseGrained 0x8 |
Coarse Grained host memory lock.
◆ hipExtHostRegisterUncached
| #define hipExtHostRegisterUncached 0x80000000 |
Map host memory onto extended fine grained access host memory pool when enabled. It is applicable on AMD devices, except for Navi4X, in Linux only.
◆ hipExtStreamWriteValueDecrement
| #define hipExtStreamWriteValueDecrement 0x1001 |
◆ hipExtStreamWriteValueIncrement
| #define hipExtStreamWriteValueIncrement 0x1000 |
◆ hipGetDeviceProperties
| #define hipGetDeviceProperties hipGetDevicePropertiesR0600 |
◆ hipGraphKernelNodePortDefault
| #define hipGraphKernelNodePortDefault 0 |
This port activates when the kernel has finished executing.
◆ hipGraphKernelNodePortLaunchCompletion
| #define hipGraphKernelNodePortLaunchCompletion 2 |
This port activates when all blocks of the kernel have begun execution.
◆ hipGraphKernelNodePortProgrammatic
| #define hipGraphKernelNodePortProgrammatic 1 |
This port activates when all blocks of the kernel have performed hipTriggerProgrammaticLaunchCompletion() or have terminated. It must be used with edge type hipGraphDependencyTypeProgrammatic.
◆ hipHostAllocDefault
| #define hipHostAllocDefault 0x0 |
Default pinned memory allocation on the host.
◆ hipHostAllocMapped
| #define hipHostAllocMapped 0x2 |
Map the allocation into the address space for the current device. The device pointer can be obtained with hipHostGetDevicePointer.
◆ hipHostAllocPortable
| #define hipHostAllocPortable 0x1 |
Memory is considered allocated by all contexts.
◆ hipHostAllocUncached
| #define hipHostAllocUncached hipHostMallocUncached |
◆ hipHostAllocWriteCombined
| #define hipHostAllocWriteCombined 0x4 |
Allocates the memory as write-combined. On some system configurations, write-combined allocation may be transferred faster across the PCI Express bus, however, could have low read efficiency by most CPUs. It's a good option for data transfer from host to device via mapped pinned memory.
- Note
- This flag is only for CUDA source compatibility but not functional within HIP runtime, because the allocation path is currently not supported on the AMD platform.
◆ hipHostMallocCoherent
| #define hipHostMallocCoherent 0x40000000 |
Allocate coherent memory. Overrides HIP_HOST_COHERENT for specific allocation.
◆ hipHostMallocDefault
| #define hipHostMallocDefault 0x0 |
Default pinned memory allocation on the host.
- Note
- This is the same definition as hipHostAllocPortable.
◆ hipHostMallocMapped
| #define hipHostMallocMapped 0x2 |
Map the allocation into the address space for the current device. The device pointer can be obtained with hipHostGetDevicePointer.
- Note
- This is the same hipHostMallocMapped.
◆ hipHostMallocNonCoherent
| #define hipHostMallocNonCoherent 0x80000000 |
Allocate non-coherent memory. Overrides HIP_HOST_COHERENT for specific allocation.
◆ hipHostMallocNumaUser
| #define hipHostMallocNumaUser 0x20000000 |
Host memory allocation will follow numa policy set by user.
- Note
- This numa allocation flag is applicable on Linux, under development on Windows.
◆ hipHostMallocPortable
| #define hipHostMallocPortable 0x1 |
Memory is considered allocated by all contexts.
- Note
- This is the same definition as hipHostAllocPortable.
◆ hipHostMallocUncached
| #define hipHostMallocUncached 0x10000000 |
Host memory will be forcedly allocated on extended fine grained system memory pool which is with MTYPE_UC.
- Note
- This allocation flag is applicable on AMD devices, except for Navi4X, in Linux only.
◆ hipHostMallocWriteCombined
| #define hipHostMallocWriteCombined 0x4 |
Allocates the memory as write-combined. On some system configurations, write-combined allocation may be transferred faster across the PCI Express bus, however, could have low read efficiency by most CPUs. It's a good option for data transfer from host to device via mapped pinned memory.
- Note
- This flag is the same definition as hipHostAllocWriteCombined which is equivalent to cudaHostAllocWriteCombined. It is only for CUDA source compatibility but not functional within HIP runtime, because the allocation path is currently not supported on the AMD platform.
◆ hipHostRegisterDefault
| #define hipHostRegisterDefault 0x0 |
Memory is Mapped and Portable.
◆ hipHostRegisterIoMemory
| #define hipHostRegisterIoMemory 0x4 |
The passed memory pointer is treated as pointing to some memory-mapped I/O space, e.g. belonging to a third-party PCIe device, and it will be marked as non cache-coherent and contiguous.
◆ hipHostRegisterMapped
| #define hipHostRegisterMapped 0x2 |
Map the allocation into the address space for the current device. The device pointer can be obtained with hipHostGetDevicePointer.
◆ hipHostRegisterPortable
| #define hipHostRegisterPortable 0x1 |
Memory is considered registered by all contexts.
◆ hipHostRegisterReadOnly
| #define hipHostRegisterReadOnly 0x08 |
This flag is ignored On AMD devices.
◆ hipInvalidDeviceId
| #define hipInvalidDeviceId ((int)-2) |
◆ hipIpcMemLazyEnablePeerAccess
| #define hipIpcMemLazyEnablePeerAccess 0x01 |
◆ hipKernelNodeAttributeAccessPolicyWindow
| #define hipKernelNodeAttributeAccessPolicyWindow hipLaunchAttributeAccessPolicyWindow |
◆ hipKernelNodeAttributeCooperative
| #define hipKernelNodeAttributeCooperative hipLaunchAttributeCooperative |
◆ hipKernelNodeAttributePriority
| #define hipKernelNodeAttributePriority hipLaunchAttributePriority |
◆ hipKernelNodeAttrID
| #define hipKernelNodeAttrID hipLaunchAttributeID |
Kernel node attributeID
◆ hipKernelNodeAttrValue
| #define hipKernelNodeAttrValue hipLaunchAttributeValue |
Kernel node attribute value
◆ hipMallocSignalMemory
| #define hipMallocSignalMemory 0x2 |
Memory represents a HSA signal.
◆ hipMemAttachGlobal
| #define hipMemAttachGlobal 0x01 |
Memory can be accessed by any stream on any device
◆ hipMemAttachHost
| #define hipMemAttachHost 0x02 |
Memory cannot be accessed by any stream on any device.
◆ hipMemAttachSingle
| #define hipMemAttachSingle 0x04 |
Memory can only be accessed by a single stream on the associated device.
◆ hipOccupancyDefault
| #define hipOccupancyDefault 0x00 |
◆ hipOccupancyDisableCachingOverride
| #define hipOccupancyDisableCachingOverride 0x01 |
◆ HIPRTC_JIT_CACHE_MODE
| #define HIPRTC_JIT_CACHE_MODE hipJitOptionCacheMode |
CUDA Only Set cache mode.
◆ HIPRTC_JIT_ERROR_LOG_BUFFER
| #define HIPRTC_JIT_ERROR_LOG_BUFFER hipJitOptionErrorLogBuffer |
CUDA Only Pointer to the buffer/*#end#*/ with logged error(s)
◆ HIPRTC_JIT_ERROR_LOG_BUFFER_SIZE_BYTES
| #define HIPRTC_JIT_ERROR_LOG_BUFFER_SIZE_BYTES hipJitOptionErrorLogBufferSizeBytes |
CUDA Only Size of the buffer in/*#end#*/ bytes for logged error(s)
◆ HIPRTC_JIT_FALLBACK_STRATEGY
| #define HIPRTC_JIT_FALLBACK_STRATEGY hipJitOptionFallbackStrategy |
CUDA Only Fallback strategy.
◆ HIPRTC_JIT_FAST_COMPILE
| #define HIPRTC_JIT_FAST_COMPILE hipJitOptionFastCompile |
CUDA Only Set fast compile.
◆ HIPRTC_JIT_FMA
| #define HIPRTC_JIT_FMA hipJitOptionFma |
- Deprecated:
- CUDA Only Enable/*#end#*/ floating-point multiplies and adds/subtracts operations
◆ HIPRTC_JIT_FTZ
| #define HIPRTC_JIT_FTZ hipJitOptionFtz |
- Deprecated:
- CUDA Only Set/*#end#*/ single-precision denormals.
◆ HIPRTC_JIT_GENERATE_DEBUG_INFO
| #define HIPRTC_JIT_GENERATE_DEBUG_INFO hipJitOptionGenerateDebugInfo |
CUDA Only Generate debug information/*#end#*/.
◆ HIPRTC_JIT_GENERATE_LINE_INFO
| #define HIPRTC_JIT_GENERATE_LINE_INFO hipJitOptionGenerateLineInfo |
CUDA Only Generate line number information/*#end#*/.
◆ HIPRTC_JIT_GLOBAL_SYMBOL_ADDRESS
| #define HIPRTC_JIT_GLOBAL_SYMBOL_ADDRESS hipJitOptionGlobalSymbolAddresses |
CUDA Only Array of host addresses to be/*#end#*/ relocated to the device
◆ HIPRTC_JIT_GLOBAL_SYMBOL_COUNT
| #define HIPRTC_JIT_GLOBAL_SYMBOL_COUNT hipJitOptionGlobalSymbolCount |
CUDA Only Number of symbol count./*#end#*/.
◆ HIPRTC_JIT_GLOBAL_SYMBOL_NAMES
| #define HIPRTC_JIT_GLOBAL_SYMBOL_NAMES hipJitOptionGlobalSymbolNames |
CUDA Only Array of device symbol names to be/*#end#*/ relocated to the host
◆ HIPRTC_JIT_INFO_LOG_BUFFER
| #define HIPRTC_JIT_INFO_LOG_BUFFER hipJitOptionInfoLogBuffer |
CUDA Only Pointer to the buffer with/*#end#*/ logged information
◆ HIPRTC_JIT_INFO_LOG_BUFFER_SIZE_BYTES
| #define HIPRTC_JIT_INFO_LOG_BUFFER_SIZE_BYTES hipJitOptionInfoLogBufferSizeBytes |
CUDA Only Size of the buffer/*#end#*/ in bytes for logged info
◆ HIPRTC_JIT_INPUT_CUBIN
| #define HIPRTC_JIT_INPUT_CUBIN hipJitInputCubin |
Cuda only Input Cubin.
◆ HIPRTC_JIT_INPUT_FATBINARY
| #define HIPRTC_JIT_INPUT_FATBINARY hipJitInputFatBinary |
Cuda Only Input FAT Binary.
◆ HIPRTC_JIT_INPUT_LIBRARY
| #define HIPRTC_JIT_INPUT_LIBRARY hipJitInputLibrary |
Cuda Only Archive of Host Objects with embedded device code/*#end#*/.
◆ HIPRTC_JIT_INPUT_LLVM_ARCHIVES_OF_BUNDLED_BITCODE
| #define HIPRTC_JIT_INPUT_LLVM_ARCHIVES_OF_BUNDLED_BITCODE hipJitInputLLVMArchivesOfBundledBitcode |
HIP Only LLVM/*#end#*/ Archives of Bundled Bitcode
◆ HIPRTC_JIT_INPUT_LLVM_BITCODE
| #define HIPRTC_JIT_INPUT_LLVM_BITCODE hipJitInputLLVMBitcode |
HIP Only LLVM Bitcode or IR assembly/*#end#*/.
◆ HIPRTC_JIT_INPUT_LLVM_BUNDLED_BITCODE
| #define HIPRTC_JIT_INPUT_LLVM_BUNDLED_BITCODE hipJitInputLLVMBundledBitcode |
HIP Only LLVM Clang Bundled Code/*#end#*/.
◆ HIPRTC_JIT_INPUT_NVVM
| #define HIPRTC_JIT_INPUT_NVVM hipJitInputNvvm |
- Deprecated:
- CUDA only High Level intermediate code for LTO/*#end#*/
◆ HIPRTC_JIT_INPUT_OBJECT
| #define HIPRTC_JIT_INPUT_OBJECT hipJitInputObject |
Cuda Only Host Object with embedded device code/*#end#*/.
◆ HIPRTC_JIT_INPUT_PTX
| #define HIPRTC_JIT_INPUT_PTX hipJitInputPtx |
Cuda only Input PTX.
◆ HIPRTC_JIT_INPUT_SPIRV
| #define HIPRTC_JIT_INPUT_SPIRV hipJitInputSpirv |
HIP Only SPIRV Code Object.
◆ HIPRTC_JIT_IR_TO_ISA_OPT_COUNT_EXT
| #define HIPRTC_JIT_IR_TO_ISA_OPT_COUNT_EXT hipJitOptionIRtoISAOptCountExt |
HIP Only Count of linker options/*#end#*/ to be passed on to
◆ HIPRTC_JIT_IR_TO_ISA_OPT_EXT
| #define HIPRTC_JIT_IR_TO_ISA_OPT_EXT hipJitOptionIRtoISAOptExt |
HIP Only Linker options to be/*#end#*/ passed on to compiler
◆ HIPRTC_JIT_LOG_VERBOSE
| #define HIPRTC_JIT_LOG_VERBOSE hipJitOptionLogVerbose |
CUDA Only Generate log verbose.
◆ HIPRTC_JIT_LTO
| #define HIPRTC_JIT_LTO hipJitOptionLto |
- Deprecated:
- CUDA Only Enable link-time/*#end#*/ optimization for device code
◆ HIPRTC_JIT_MAX_REGISTERS
| #define HIPRTC_JIT_MAX_REGISTERS hipJitOptionMaxRegisters |
CUDA Only Maximum registers may be used in a/*#end#*/ thread, passed to compiler
◆ HIPRTC_JIT_MAX_THREADS_PER_BLOCK
| #define HIPRTC_JIT_MAX_THREADS_PER_BLOCK hipJitOptionMaxThreadsPerBlock |
CUDA only Maximum number of/*#end#*/ threads in a thread block
◆ HIPRTC_JIT_MIN_CTA_PER_SM
| #define HIPRTC_JIT_MIN_CTA_PER_SM hipJitOptionMinCTAPerSM |
CUDA Only Hints to JIT compiler/*#end#*/ the minimum number of CTAs frin kernel's grid to be mapped to SM
◆ HIPRTC_JIT_NEW_SM3X_OPT
| #define HIPRTC_JIT_NEW_SM3X_OPT hipJitOptionSm3xOpt |
- Deprecated:
- CUDA Only New SM3X option.
◆ HIPRTC_JIT_NUM_INPUT_TYPES
| #define HIPRTC_JIT_NUM_INPUT_TYPES hipJitNumInputTypes |
Count of Input Types.
◆ HIPRTC_JIT_NUM_LEGACY_INPUT_TYPES
| #define HIPRTC_JIT_NUM_LEGACY_INPUT_TYPES hipJitNumLegacyInputTypes |
Count of Legacy Input Types/*#end#*/.
◆ HIPRTC_JIT_NUM_OPTIONS
| #define HIPRTC_JIT_NUM_OPTIONS hipJitOptionNumOptions |
Number of options.
◆ HIPRTC_JIT_OPTIMIZATION_LEVEL
| #define HIPRTC_JIT_OPTIMIZATION_LEVEL hipJitOptionOptimizationLevel |
Value of optimization level for/*#end#*/ generated codes, acceptable options -O0, -O1, -O2, -O3
◆ HIPRTC_JIT_OVERRIDE_DIRECT_VALUES
| #define HIPRTC_JIT_OVERRIDE_DIRECT_VALUES hipJitOptionOverrideDirectiveValues |
CUDA only Override Directive/*#end#*/ Values
◆ HIPRTC_JIT_POSITION_INDEPENDENT_CODE
| #define HIPRTC_JIT_POSITION_INDEPENDENT_CODE hipJitOptionPositionIndependentCode |
CUDA Only Generates/*#end#*/ Position Independent code
◆ HIPRTC_JIT_PREC_DIV
| #define HIPRTC_JIT_PREC_DIV hipJitOptionPrecDiv |
- Deprecated:
- CUDA Only Set/*#end#*/ single-precision floating-point division and reciprocals
◆ HIPRTC_JIT_PREC_SQRT
| #define HIPRTC_JIT_PREC_SQRT hipJitOptionPrecSqrt |
- Deprecated:
- CUDA Only Set/*#end#*/ single-precision floating-point square root
◆ HIPRTC_JIT_TARGET
| #define HIPRTC_JIT_TARGET hipJitOptionTarget |
CUDA Only JIT target.
◆ HIPRTC_JIT_TARGET_FROM_HIPCONTEXT
| #define HIPRTC_JIT_TARGET_FROM_HIPCONTEXT hipJitOptionTargetFromContext |
CUDA Only The target context,/*#end#*/ which is the default
◆ HIPRTC_JIT_THREADS_PER_BLOCK
| #define HIPRTC_JIT_THREADS_PER_BLOCK hipJitOptionThreadsPerBlock |
CUDA Only Number of thread per block/*#end#*/.
◆ HIPRTC_JIT_WALL_TIME
| #define HIPRTC_JIT_WALL_TIME hipJitOptionWallTime |
CUDA Only Value for total wall clock time.
◆ hiprtcJIT_option
| #define hiprtcJIT_option hipJitOption |
hiprtc JIT option
◆ hiprtcJITInputType
| #define hiprtcJITInputType hipJitInputType |
hiprtc JIT input type
◆ hipStreamAttributeAccessPolicyWindow
| #define hipStreamAttributeAccessPolicyWindow hipLaunchAttributeAccessPolicyWindow |
◆ hipStreamAttributeMemSyncDomain
| #define hipStreamAttributeMemSyncDomain hipLaunchAttributeMemSyncDomain |
◆ hipStreamAttributeMemSyncDomainMap
| #define hipStreamAttributeMemSyncDomainMap hipLaunchAttributeMemSyncDomainMap |
◆ hipStreamAttributePriority
| #define hipStreamAttributePriority hipLaunchAttributePriority |
◆ hipStreamAttributeSynchronizationPolicy
| #define hipStreamAttributeSynchronizationPolicy hipLaunchAttributeSynchronizationPolicy |
◆ hipStreamAttrID
| #define hipStreamAttrID hipLaunchAttributeID |
Stream attributes
◆ hipStreamAttrValue
| #define hipStreamAttrValue hipLaunchAttributeValue |
◆ hipStreamDefault
| #define hipStreamDefault 0x00 |
Flags that can be used with hipStreamCreateWithFlags. Default stream creation flags. These are used with hipStreamCreate().
◆ hipStreamLegacy
| #define hipStreamLegacy ((hipStream_t)1) |
◆ hipStreamNonBlocking
| #define hipStreamNonBlocking 0x01 |
Stream does not implicitly synchronize with null stream.
◆ hipStreamPerThread
| #define hipStreamPerThread ((hipStream_t)2) |
Implicit stream per application thread.
◆ hipStreamWaitValueAnd
| #define hipStreamWaitValueAnd 0x2 |
◆ hipStreamWaitValueEq
| #define hipStreamWaitValueEq 0x1 |
◆ hipStreamWaitValueGte
| #define hipStreamWaitValueGte 0x0 |
◆ hipStreamWaitValueNor
| #define hipStreamWaitValueNor 0x3 |
◆ hipStreamWriteValueDefault
| #define hipStreamWriteValueDefault 0x0 |
Typedef Documentation
◆ GLenum
| typedef unsigned int GLenum |
GLenum as uint.
◆ GLuint
| typedef unsigned int GLuint |
GLuint as uint.
◆ hipCtx_t
| typedef struct ihipCtx_t* hipCtx_t |
◆ hipDevice_t
| typedef int hipDevice_t |
◆ hipDevResourceDesc_t
| typedef struct ihipDevResourceDesc_t* hipDevResourceDesc_t |
◆ hipError_t
HIP error type
◆ hipEvent_t
| typedef struct ihipEvent_t* hipEvent_t |
◆ hipExecutionCtx_t
| typedef struct ihipExecutionCtx_t* hipExecutionCtx_t |
◆ hipExternalMemory_t
| typedef void* hipExternalMemory_t |
◆ hipExternalSemaphore_t
| typedef void* hipExternalSemaphore_t |
◆ hipFunction_t
| typedef struct ihipModuleSymbol_t* hipFunction_t |
◆ hipGraph_t
| typedef struct ihipGraph* hipGraph_t |
An opaque value that represents a hip graph
◆ hipGraphExec_t
| typedef struct hipGraphExec* hipGraphExec_t |
An opaque value that represents a hip graph Exec
◆ hipGraphicsResource
| typedef struct _hipGraphicsResource hipGraphicsResource |
◆ hipGraphicsResource_t
| typedef hipGraphicsResource* hipGraphicsResource_t |
◆ hipGraphNode_t
| typedef struct hipGraphNode* hipGraphNode_t |
An opaque value that represents a hip graph node
◆ hipHostFn_t
| typedef void(* hipHostFn_t) (void *userData) |
◆ hipKernel_t
| typedef struct ihipKernel_t* hipKernel_t |
◆ hipLibrary_t
| typedef struct ihipLibrary_t* hipLibrary_t |
◆ hipLinkState_t
| typedef struct ihipLinkState_t* hipLinkState_t |
◆ hipMemGenericAllocationHandle_t
| typedef struct ihipMemGenericAllocationHandle* hipMemGenericAllocationHandle_t |
Generic handle for memory allocation
◆ hipMemPool_t
| typedef struct ihipMemPoolHandle_t* hipMemPool_t |
HIP memory pool
◆ hipModule_t
| typedef struct ihipModule_t* hipModule_t |
◆ hipStream_t
| typedef struct ihipStream_t* hipStream_t |
◆ hipUserObject_t
| typedef struct hipUserObject* hipUserObject_t |
An opaque value that represents a user obj
Enumeration Type Documentation
◆ hipAccessProperty
| enum hipAccessProperty |
Specifies performance hint with hipAccessPolicyWindow
◆ hipArraySparseSubresourceType
◆ hipClusterSchedulingPolicy
◆ hipComputeMode
| enum hipComputeMode |
◆ hipDeviceAttribute_t
| enum hipDeviceAttribute_t |
hipDeviceAttribute_t hipDeviceAttributeUnused number: 5
| Enumerator | |
|---|---|
| hipDeviceAttributeCudaCompatibleBegin | |
| hipDeviceAttributeEccEnabled | Whether ECC support is enabled. |
| hipDeviceAttributeAccessPolicyMaxWindowSize | Cuda only. The maximum size of the window policy in bytes. |
| hipDeviceAttributeAsyncEngineCount | Asynchronous engines number. |
| hipDeviceAttributeCanMapHostMemory | Whether host memory can be mapped into device address space |
| hipDeviceAttributeCanUseHostPointerForRegisteredMem | Device can access host registered memory at the same virtual address as the CPU |
| hipDeviceAttributeClockRate | Peak clock frequency in kilohertz. |
| hipDeviceAttributeComputeMode | Compute mode that device is currently in. |
| hipDeviceAttributeComputePreemptionSupported | Device supports Compute Preemption. |
| hipDeviceAttributeConcurrentKernels | Device can possibly execute multiple kernels concurrently. |
| hipDeviceAttributeConcurrentManagedAccess | Device can coherently access managed memory concurrently with the CPU |
| hipDeviceAttributeCooperativeLaunch | Support cooperative launch. |
| hipDeviceAttributeCooperativeMultiDeviceLaunch | Support cooperative launch on multiple devices |
| hipDeviceAttributeDeviceOverlap | Device can concurrently copy memory and execute a kernel. Deprecated. Use instead asyncEngineCount. |
| hipDeviceAttributeDirectManagedMemAccessFromHost | Host can directly access managed memory on the device without migration |
| hipDeviceAttributeGlobalL1CacheSupported | Device supports caching globals in L1. |
| hipDeviceAttributeHostNativeAtomicSupported | Link between the device and the host supports native atomic operations |
| hipDeviceAttributeIntegrated | Device is integrated GPU. |
| hipDeviceAttributeIsMultiGpuBoard | Multiple GPU devices. |
| hipDeviceAttributeKernelExecTimeout | Run time limit for kernels executed on the device. |
| hipDeviceAttributeL2CacheSize | Size of L2 cache in bytes. 0 if the device doesn't have L2 cache. |
| hipDeviceAttributeLocalL1CacheSupported | caching locals in L1 is supported |
| hipDeviceAttributeLuid | 8-byte locally unique identifier in 8 bytes. Undefined on TCC and non-Windows platforms |
| hipDeviceAttributeLuidDeviceNodeMask | Luid device node mask. Undefined on TCC and non-Windows platforms |
| hipDeviceAttributeComputeCapabilityMajor | Major compute capability version number. |
| hipDeviceAttributeManagedMemory | Device supports allocating managed memory on this system. |
| hipDeviceAttributeMaxBlocksPerMultiProcessor | Max block size per multiprocessor. |
| hipDeviceAttributeMaxBlockDimX | Max block size in width. |
| hipDeviceAttributeMaxBlockDimY | Max block size in height. |
| hipDeviceAttributeMaxBlockDimZ | Max block size in depth. |
| hipDeviceAttributeMaxGridDimX | Max grid size in width. |
| hipDeviceAttributeMaxGridDimY | Max grid size in height. |
| hipDeviceAttributeMaxGridDimZ | Max grid size in depth. |
| hipDeviceAttributeMaxSurface1D | Maximum size of 1D surface. |
| hipDeviceAttributeMaxSurface1DLayered | Cuda only. Maximum dimensions of 1D layered surface. |
| hipDeviceAttributeMaxSurface2D | Maximum dimension (width, height) of 2D surface. |
| hipDeviceAttributeMaxSurface2DLayered | Cuda only. Maximum dimensions of 2D layered surface. |
| hipDeviceAttributeMaxSurface3D | Maximum dimension (width, height, depth) of 3D surface. |
| hipDeviceAttributeMaxSurfaceCubemap | Cuda only. Maximum dimensions of Cubemap surface. |
| hipDeviceAttributeMaxSurfaceCubemapLayered | Cuda only. Maximum dimension of Cubemap layered surface. |
| hipDeviceAttributeMaxTexture1DWidth | Maximum size of 1D texture. |
| hipDeviceAttributeMaxTexture1DLayered | Maximum dimensions of 1D layered texture. |
| hipDeviceAttributeMaxTexture1DLinear | Maximum number of elements allocatable in a 1D linear texture. Use cudaDeviceGetTexture1DLinearMaxWidth() instead on Cuda. |
| hipDeviceAttributeMaxTexture1DMipmap | Maximum size of 1D mipmapped texture. |
| hipDeviceAttributeMaxTexture2DWidth | Maximum dimension width of 2D texture. |
| hipDeviceAttributeMaxTexture2DHeight | Maximum dimension hight of 2D texture. |
| hipDeviceAttributeMaxTexture2DGather | Maximum dimensions of 2D texture if gather operations performed. |
| hipDeviceAttributeMaxTexture2DLayered | Maximum dimensions of 2D layered texture. |
| hipDeviceAttributeMaxTexture2DLinear | Maximum dimensions (width, height, pitch) of 2D textures bound to pitched memory. |
| hipDeviceAttributeMaxTexture2DMipmap | Maximum dimensions of 2D mipmapped texture. |
| hipDeviceAttributeMaxTexture3DWidth | Maximum dimension width of 3D texture. |
| hipDeviceAttributeMaxTexture3DHeight | Maximum dimension height of 3D texture. |
| hipDeviceAttributeMaxTexture3DDepth | Maximum dimension depth of 3D texture. |
| hipDeviceAttributeMaxTexture3DAlt | Maximum dimensions of alternate 3D texture. |
| hipDeviceAttributeMaxTextureCubemap | Maximum dimensions of Cubemap texture. |
| hipDeviceAttributeMaxTextureCubemapLayered | Maximum dimensions of Cubemap layered texture. |
| hipDeviceAttributeMaxThreadsDim | Maximum dimension of a block. |
| hipDeviceAttributeMaxThreadsPerBlock | Maximum number of threads per block. |
| hipDeviceAttributeMaxThreadsPerMultiProcessor | Maximum resident threads per multiprocessor. |
| hipDeviceAttributeMaxPitch | Maximum pitch in bytes allowed by memory copies. |
| hipDeviceAttributeMemoryBusWidth | Global memory bus width in bits. |
| hipDeviceAttributeMemoryClockRate | Peak memory clock frequency in kilohertz. |
| hipDeviceAttributeComputeCapabilityMinor | Minor compute capability version number. |
| hipDeviceAttributeMultiGpuBoardGroupID | Unique ID of device group on the same multi-GPU board |
| hipDeviceAttributeMultiprocessorCount | Number of multi-processors. When the GPU works in Compute Unit (CU) mode, this value equals the number of CUs; when in Workgroup Processor (WGP) mode, this value equels half of CUs, because a single WGP contains two CUs. |
| hipDeviceAttributeUnused1 | Previously hipDeviceAttributeName. |
| hipDeviceAttributePageableMemoryAccess | Device supports coherently accessing pageable memory without calling hipHostRegister on it |
| hipDeviceAttributePageableMemoryAccessUsesHostPageTables | Device accesses pageable memory via the host's page tables |
| hipDeviceAttributePciBusId | PCI Bus ID. |
| hipDeviceAttributePciDeviceId | PCI Device ID. Returns pcie slot id. |
| hipDeviceAttributePciDomainId | PCI Domain Id. |
| hipDeviceAttributePciDomainID | PCI Domain ID, for backward compatibility. |
| hipDeviceAttributePersistingL2CacheMaxSize | Maximum l2 persisting lines capacity in bytes. |
| hipDeviceAttributeMaxRegistersPerBlock | 32-bit registers available to a thread block. This number is shared by all thread blocks simultaneously resident on a multiprocessor. |
| hipDeviceAttributeMaxRegistersPerMultiprocessor | 32-bit registers available per block. |
| hipDeviceAttributeReservedSharedMemPerBlock | Shared memory reserved by CUDA driver per block. |
| hipDeviceAttributeMaxSharedMemoryPerBlock | Maximum shared memory available per block in bytes. |
| hipDeviceAttributeSharedMemPerBlockOptin | Maximum shared memory per block usable by special opt in. |
| hipDeviceAttributeSharedMemPerMultiprocessor | Shared memory available per multiprocessor. |
| hipDeviceAttributeSingleToDoublePrecisionPerfRatio | Cuda only. Performance ratio of single precision to double precision. |
| hipDeviceAttributeStreamPrioritiesSupported | Whether to support stream priorities. |
| hipDeviceAttributeSurfaceAlignment | Alignment requirement for surfaces. |
| hipDeviceAttributeTccDriver | Cuda only. Whether device is a Tesla device using TCC driver. |
| hipDeviceAttributeTextureAlignment | Alignment requirement for textures. |
| hipDeviceAttributeTexturePitchAlignment | Pitch alignment requirement for 2D texture references bound to pitched memory; |
| hipDeviceAttributeTotalConstantMemory | Constant memory size in bytes. |
| hipDeviceAttributeTotalGlobalMem | Global memory available on devicice. |
| hipDeviceAttributeUnifiedAddressing | Cuda only. An unified address space shared with the host. |
| hipDeviceAttributeUnused2 | Previously hipDeviceAttributeUuid. |
| hipDeviceAttributeWarpSize | Warp size in threads. |
| hipDeviceAttributeMemoryPoolsSupported | Device supports HIP Stream Ordered Memory Allocator. |
| hipDeviceAttributeVirtualMemoryManagementSupported | Device supports HIP virtual memory management |
| hipDeviceAttributeHostRegisterSupported | Can device support host memory registration via hipHostRegister |
| hipDeviceAttributeMemoryPoolSupportedHandleTypes | Supported handle mask for HIP Stream Ordered Memory Allocator |
| hipDeviceAttributeHostNumaId | NUMA ID of the cpu node closest to the device, or -1 when NUMA isn't supported |
| hipDeviceAttributeDmaBufSupported | Device supports DMABuf buffer sharing. |
| hipDeviceAttributeGPUDirectRDMAWithHipVMMSupported | GPU Direct RDMA with HIP VMM is supported (requires DMA-Buf and HIP virtual memory management) |
| hipDeviceAttributeHandleTypeFabricSupported | Device supports exporting memory to a fabric handle. |
| hipDeviceAttributeCudaCompatibleEnd | |
| hipDeviceAttributeAmdSpecificBegin | |
| hipDeviceAttributeClockInstructionRate | Frequency in khz of the timer used by the device-side "clock*" |
| hipDeviceAttributeUnused3 | Previously hipDeviceAttributeArch. |
| hipDeviceAttributeMaxSharedMemoryPerMultiprocessor | Maximum Shared Memory PerMultiprocessor. |
| hipDeviceAttributeUnused4 | Previously hipDeviceAttributeGcnArch. |
| hipDeviceAttributeUnused5 | Previously hipDeviceAttributeGcnArchName. |
| hipDeviceAttributeHdpMemFlushCntl | Address of the HDP_MEM_COHERENCY_FLUSH_CNTL register. |
| hipDeviceAttributeHdpRegFlushCntl | Address of the HDP_REG_COHERENCY_FLUSH_CNTL register. |
| hipDeviceAttributeCooperativeMultiDeviceUnmatchedFunc | Supports cooperative launch on multiple devices with unmatched functions |
| hipDeviceAttributeCooperativeMultiDeviceUnmatchedGridDim | Supports cooperative launch on multiple devices with unmatched grid dimensions |
| hipDeviceAttributeCooperativeMultiDeviceUnmatchedBlockDim | Supports cooperative launch on multiple devices with unmatched block dimensions |
| hipDeviceAttributeCooperativeMultiDeviceUnmatchedSharedMem | Supports cooperative launch on multiple devices with unmatched shared memories |
| hipDeviceAttributeIsLargeBar | Whether it is LargeBar. |
| hipDeviceAttributeAsicRevision | Revision of the GPU in this device. |
| hipDeviceAttributeCanUseStreamWaitValue | '1' if Device supports hipStreamWaitValue32() and hipStreamWaitValue64(), '0' otherwise. |
| hipDeviceAttributeImageSupport | '1' if Device supports image, '0' otherwise. |
| hipDeviceAttributePhysicalMultiProcessorCount | All available physical compute units for the device |
| hipDeviceAttributeFineGrainSupport | '1' if Device supports fine grain, '0' otherwise |
| hipDeviceAttributeWallClockRate | Constant frequency of wall clock in kilohertz. |
| hipDeviceAttributeNumberOfXccs | Number of XCC(s) in the device’s current compute partition(XCP). An XCC is AMD’s fundamental GPU compute tile. Devices like MI300X include multiple XCCs per package. Value varies by boot‑time partition mode(compute or memory). Partition modes can be set using amd‑smi. |
| hipDeviceAttributeMaxAvailableVgprsPerThread | Max number of available (directly or indirectly addressable) VGPRs per thread in DWORDs. |
| hipDeviceAttributePciChipId | GPU Manufacturer device id. |
| hipDeviceAttributeExpertSchedMode | '1' if Device supports expert scheduling mode, '0' otherwise. |
| hipDeviceAttributeMaxDynDataPrefetchRegions | Maximum number of dynamic data prefetch regions per kernel launch (0 if unsupported). |
| hipDeviceAttributeAmdSpecificEnd | |
| hipDeviceAttributeVendorSpecificBegin | |
◆ hipDeviceP2PAttr
| enum hipDeviceP2PAttr |
◆ hipDevResourceType
| enum hipDevResourceType |
◆ hipDevSmResourceGroup_flags
◆ hipDevSmResourceSplitByCount_flags
◆ hipDevWorkqueueConfigScope
◆ hipDriverEntryPointQueryResult
◆ hipDriverProcAddressQueryResult
◆ hipExtDynDataPrefetchTemporal
◆ hipExternalMemoryHandleType
◆ hipExternalSemaphoreHandleType
◆ hipFlushGPUDirectRDMAWritesOptions
◆ hipFuncAttribute
| enum hipFuncAttribute |
- Warning
- On AMD devices and some Nvidia devices, these hints and controls are ignored.
◆ hipFuncCache_t
| enum hipFuncCache_t |
- Warning
- On AMD devices and some Nvidia devices, these hints and controls are ignored.
◆ hipGLDeviceList
| enum hipGLDeviceList |
◆ hipGPUDirectRDMAWritesOrdering
◆ hipGraphDebugDotFlags
| Enumerator | |
|---|---|
| hipGraphDebugDotFlagsVerbose | Output all debug data as if every debug flag is enabled |
| hipGraphDebugDotFlagsKernelNodeParams | Adds hipKernelNodeParams to output |
| hipGraphDebugDotFlagsMemcpyNodeParams | Adds hipMemcpy3DParms to output |
| hipGraphDebugDotFlagsMemsetNodeParams | Adds hipMemsetParams to output |
| hipGraphDebugDotFlagsHostNodeParams | Adds hipHostNodeParams to output |
| hipGraphDebugDotFlagsEventNodeParams | Adds hipEvent_t handle from record and wait nodes to output |
| hipGraphDebugDotFlagsExtSemasSignalNodeParams | Adds hipExternalSemaphoreSignalNodeParams values to output |
| hipGraphDebugDotFlagsExtSemasWaitNodeParams | Adds hipExternalSemaphoreWaitNodeParams to output |
| hipGraphDebugDotFlagsKernelNodeAttributes | Adds hipKernelNodeAttrID values to output |
| hipGraphDebugDotFlagsHandles | Adds node handles and every kernel function handle to output |
◆ hipGraphDependencyType
◆ hipGraphExecUpdateResult
Graph execution update result
◆ hipGraphicsRegisterFlags
HIP Access falgs for Interop resources.
◆ hipGraphInstantiateFlags
◆ hipGraphInstantiateResult
hipGraphInstantiateWithParams results
◆ hipGraphMemAttributeType
◆ hipGraphNodeType
| enum hipGraphNodeType |
hipGraphNodeType
◆ hipLaunchAttributeID
| enum hipLaunchAttributeID |
Launch Attribute ID
◆ hipLaunchMemSyncDomain
◆ hipLimit_t
| enum hipLimit_t |
hipLimit
- Note
- In HIP device limit-related APIs, any input limit value other than those defined in the enum is treated as "UnsupportedLimit" by default.
◆ hipMemAccessFlags
| enum hipMemAccessFlags |
◆ hipMemAllocationGranularity_flags
◆ hipMemAllocationHandleType
Flags for specifying handle types for memory pool allocations
◆ hipMemAllocationType
| enum hipMemAllocationType |
◆ hipMemHandleType
| enum hipMemHandleType |
◆ hipMemOperationType
| enum hipMemOperationType |
◆ hipMemoryAdvise
| enum hipMemoryAdvise |
HIP Memory Advise values
- Note
- This memory advise enumeration is used on Linux, not Windows.
◆ hipMemoryType
| enum hipMemoryType |
hipMemoryType (for pointer attributes)
- Note
- hipMemoryType enum values are combination of cudaMemoryType and cuMemoryType and AMD specific enum values.
◆ hipMemPoolAttr
| enum hipMemPoolAttr |
HIP memory pool attributes
◆ hipMemRangeAttribute
| enum hipMemRangeAttribute |
HIP range attributes
| Enumerator | |
|---|---|
| hipMemRangeAttributeReadMostly | Whether the range will mostly be read and only occassionally be written to |
| hipMemRangeAttributePreferredLocation | The preferred location of the range. |
| hipMemRangeAttributeAccessedBy | Memory range has hipMemAdviseSetAccessedBy set for the specified device |
| hipMemRangeAttributeLastPrefetchLocation | The last location to where the range was prefetched |
| hipMemRangeAttributeCoherencyMode | Returns coherency mode hipMemRangeCoherencyMode for the range |
◆ hipMemRangeCoherencyMode
HIP Coherency Mode
◆ hipMemRangeFlags
| enum hipMemRangeFlags |
◆ hipMemRangeHandleType
◆ hiprtcResult
| enum hiprtcResult |
hiprtc error code
◆ hipSharedMemConfig
| enum hipSharedMemConfig |
- Warning
- On AMD devices and some Nvidia devices, these hints and controls are ignored.
◆ hipStreamBatchMemOpType
◆ hipStreamCaptureMode
| enum hipStreamCaptureMode |
◆ hipStreamCaptureStatus
◆ hipStreamUpdateCaptureDependenciesFlags
◆ hipSynchronizationPolicy
Stream Synchronization Policy. Can be set with hipStreamSetAttribute
◆ hipUserObjectFlags
| enum hipUserObjectFlags |