Memory management

Contents

Memory management#

hipError_t hipPointerSetAttribute(const void *value, hipPointer_attribute attribute, hipDeviceptr_t ptr)#

Sets information on the specified pointer.[BETA].

Warning

This API is marked as Beta. While this feature is complete, it can change and might have outstanding issues.

Parameters:
  • value[in] Sets pointer attribute value

  • attribute[in] Attribute to set

  • ptr[in] Pointer to set attributes for

Returns:

hipSuccess, hipErrorInvalidDevice, hipErrorInvalidValue

hipError_t hipPointerGetAttributes(hipPointerAttribute_t *attributes, const void *ptr)#

Returns attributes for the specified pointer.

The output parameter ‘attributes’ has a member named ‘type’ that describes what memory the pointer is associated with, such as device memory, host memory, managed memory, and others. Otherwise, the API cannot handle the pointer and returns hipErrorInvalidValue.

Note

The unrecognized memory type is unsupported to keep the HIP functionality backward compatibility due to hipMemoryType enum values.

Note

The current behavior of this HIP API corresponds to the CUDA API before version 11.0.

Parameters:
  • attributes[out] attributes for the specified pointer

  • ptr[in] pointer to get attributes for

Returns:

hipSuccess, hipErrorInvalidDevice, hipErrorInvalidValue

hipError_t hipPointerGetAttribute(void *data, hipPointer_attribute attribute, hipDeviceptr_t ptr)#

Returns information about the specified pointer.[BETA].

Warning

This API is marked as Beta. While this feature is complete, it can change and might have outstanding issues.

Parameters:
  • data[inout] Returned pointer attribute value

  • attribute[in] Attribute to query for

  • ptr[in] Pointer to get attributes for

Returns:

hipSuccess, hipErrorInvalidDevice, hipErrorInvalidValue

hipError_t hipDrvPointerGetAttributes(unsigned int numAttributes, hipPointer_attribute *attributes, void **data, hipDeviceptr_t ptr)#

Returns information about the specified pointer.[BETA].

Warning

This API is marked as Beta. While this feature is complete, it can change and might have outstanding issues.

Parameters:
  • numAttributes[in] number of attributes to query for

  • attributes[in] attributes to query for

  • data[inout] a two-dimensional containing pointers to memory locations where the result of each attribute query will be written to

  • ptr[in] pointer to get attributes for

Returns:

hipSuccess, hipErrorInvalidDevice, hipErrorInvalidValue

hipError_t hipMalloc(void **ptr, size_t size)#

Allocate memory on the default accelerator.

If size is 0, no memory is allocated, *ptr returns nullptr, and hipSuccess is returned.

Parameters:
  • ptr[out] Pointer to the allocated memory

  • size[in] Requested memory size

Returns:

hipSuccess, hipErrorOutOfMemory, hipErrorInvalidValue (bad context, null *ptr)

hipError_t hipExtMallocWithFlags(void **ptr, size_t sizeBytes, unsigned int flags)#

Allocate memory on the default accelerator.

If requested memory size is 0, no memory is allocated, *ptr returns nullptr, and hipSuccess is returned.

The memory allocation flag should be either hipDeviceMallocDefault, hipDeviceMallocFinegrained, hipDeviceMallocUncached, or hipMallocSignalMemory. If the flag is any other value, the API returns hipErrorInvalidValue.

Parameters:
  • ptr[out] Pointer to the allocated memory

  • sizeBytes[in] Requested memory size

  • flags[in] Type of memory allocation

Returns:

hipSuccess, hipErrorOutOfMemory, hipErrorInvalidValue (bad context, null *ptr)

hipError_t hipHostMalloc(void **ptr, size_t size, unsigned int flags)#

Allocates device accessible page locked (pinned) host memory.

This API allocates pinned host memory which is mapped into the address space of all GPUs in the system, the memory can be accessed directly by the GPU device, and can be read or written with much higher bandwidth than pageable memory obtained with functions such as malloc().

Using the pinned host memory, applications can implement faster data transfers for HostToDevice and DeviceToHost. The runtime tracks the hipHostMalloc allocations and can avoid some of the setup required for regular unpinned memory.

When the memory accesses are infrequent, zero-copy memory can be a good choice, for coherent allocation. GPU can directly access the host memory over the CPU/GPU interconnect, without need to copy the data.

Currently the allocation granularity is 4KB for the API.

Developers need to choose proper allocation flag with consideration of synchronization.

If no input for flags, it will be the default pinned memory allocation on the host.

Parameters:
  • ptr[out] Pointer to the allocated host pinned memory

  • size[in] Requested memory size in bytes If size is 0, no memory is allocated, *ptr returns nullptr, and hipSuccess is returned.

  • flags[in] Type of host memory allocation. See the description of flags in hipSetDeviceFlags.

Returns:

hipSuccess, hipErrorOutOfMemory

hipError_t hipHostAlloc(void **ptr, size_t size, unsigned int flags)#

Allocate device accessible page locked host memory.

If size is 0, no memory is allocated, *ptr returns nullptr, and hipSuccess is returned.

Flags:

Parameters:
  • ptr[out] Pointer to the allocated host pinned memory

  • size[in] Requested memory size in bytes

  • flags[in] Type of host memory allocation see below

Returns:

hipSuccess, hipErrorOutOfMemory, hipErrorInvalidValue

hipError_t hipHostGetDevicePointer(void **devPtr, void *hstPtr, unsigned int flags)#

Get Device pointer from Host Pointer allocated through hipHostMalloc.

Parameters:
  • devPtr[out] Device Pointer mapped to passed host pointer

  • hstPtr[in] Host Pointer allocated through hipHostMalloc

  • flags[in] Flags to be passed for extension

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorOutOfMemory

hipError_t hipHostGetFlags(unsigned int *flagsPtr, void *hostPtr)#

Return flags associated with host pointer.

See also

hipHostMalloc

Parameters:
  • flagsPtr[out] Memory location to store flags

  • hostPtr[in] Host Pointer allocated through hipHostMalloc

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipHostRegister(void *hostPtr, size_t sizeBytes, unsigned int flags)#

Register host memory so it can be accessed from the current device.

Flags:

After registering the memory, use hipHostGetDevicePointer to obtain the mapped device pointer. On many systems, the mapped device pointer will have a different value than the mapped host pointer. Applications must use the device pointer in device code, and the host pointer in host code.

On some systems, registered memory is pinned. On some systems, registered memory may not be actually be pinned but uses OS or hardware facilities to all GPU access to the host memory.

Developers are strongly encouraged to register memory blocks which are aligned to the host cache-line size. (typically 64-bytes but can be obtains from the CPUID instruction).

If registering non-aligned pointers, the application must take care when register pointers from the same cache line on different devices. HIP’s coarse-grained synchronization model does not guarantee correct results if different devices write to different parts of the same cache block - typically one of the writes will “win” and overwrite data from the other registered memory region.

Parameters:
  • hostPtr[out] Pointer to host memory to be registered.

  • sizeBytes[in] Size of the host memory

  • flags[in] See below.

Returns:

hipSuccess, hipErrorOutOfMemory

hipError_t hipHostUnregister(void *hostPtr)#

Un-register host pointer.

See also

hipHostRegister

Parameters:

hostPtr[in] Host pointer previously registered with hipHostRegister

Returns:

Error code

hipError_t hipMallocPitch(void **ptr, size_t *pitch, size_t width, size_t height)#

Allocates at least width (in bytes) * height bytes of linear memory Padding may occur to ensure alighnment requirements are met for the given row The change in width size due to padding will be returned in *pitch. Currently the alignment is set to 128 bytes

If size is 0, no memory is allocated, *ptr returns nullptr, and hipSuccess is returned.

Parameters:
  • ptr[out] Pointer to the allocated device memory

  • pitch[out] Pitch for allocation (in bytes)

  • width[in] Requested pitched allocation width (in bytes)

  • height[in] Requested pitched allocation height

Returns:

Error code

hipError_t hipMemAllocPitch(hipDeviceptr_t *dptr, size_t *pitch, size_t widthInBytes, size_t height, unsigned int elementSizeBytes)#

Allocates at least width (in bytes) * height bytes of linear memory Padding may occur to ensure alighnment requirements are met for the given row The change in width size due to padding will be returned in *pitch. Currently the alignment is set to 128 bytes

If size is 0, no memory is allocated, ptr returns nullptr, and hipSuccess is returned. The intended usage of pitch is as a separate parameter of the allocation, used to compute addresses within the 2D array. Given the row and column of an array element of type T, the address is computed as: T pElement = (T*)((char*)BaseAddress + Row * Pitch) + Column;

Parameters:
  • dptr[out] Pointer to the allocated device memory

  • pitch[out] Pitch for allocation (in bytes)

  • widthInBytes[in] Requested pitched allocation width (in bytes)

  • height[in] Requested pitched allocation height

  • elementSizeBytes[in] The size of element bytes, should be 4, 8 or 16

Returns:

Error code

hipError_t hipFree(void *ptr)#

Free memory allocated by the hcc hip memory allocation API. This API performs an implicit hipDeviceSynchronize() call. If pointer is NULL, the hip runtime is initialized and hipSuccess is returned.

Parameters:

ptr[in] Pointer to memory to be freed

Returns:

hipSuccess

Returns:

hipErrorInvalidDevicePointer (if pointer is invalid, including host pointers allocated with hipHostMalloc)

hipError_t hipMemcpy(void *dst, const void *src, size_t sizeBytes, hipMemcpyKind kind)#

Copy data from src to dst.

It supports memory from host to device, device to host, device to device and host to host The src and dst must not overlap.

For hipMemcpy, the copy is always performed by the current device (set by hipSetDevice). For multi-gpu or peer-to-peer configurations, it is recommended to set the current device to the device where the src data is physically located. For optimal peer-to-peer copies, the copy device must be able to access the src and dst pointers (by calling hipDeviceEnablePeerAccess with copy agent as the current device and src/dest as the peerDevice argument. if this is not done, the hipMemcpy will still work, but will perform the copy using a staging buffer on the host. Calling hipMemcpy with dst and src pointers that do not match the hipMemcpyKind results in undefined behavior.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • kind[in] Kind of transfer

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorUnknown

hipError_t hipMemcpyWithStream(void *dst, const void *src, size_t sizeBytes, hipMemcpyKind kind, hipStream_t stream)#

Memory copy on the stream. It allows single or multiple devices to do memory copy on single or multiple streams.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • kind[in] Kind of transfer

  • stream[in] Valid stream

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorUnknown, hipErrorContextIsDestroyed

hipError_t hipMemcpyHtoD(hipDeviceptr_t dst, void *src, size_t sizeBytes)#

Copy data from Host to Device.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyDtoH(void *dst, hipDeviceptr_t src, size_t sizeBytes)#

Copy data from Device to Host.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyDtoD(hipDeviceptr_t dst, hipDeviceptr_t src, size_t sizeBytes)#

Copy data from Device to Device.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyAtoD(hipDeviceptr_t dstDevice, hipArray_t srcArray, size_t srcOffset, size_t ByteCount)#

Copies from one 1D array to device memory.

Parameters:
  • dstDevice[out] Destination device pointer

  • srcArray[in] Source array

  • srcOffset[in] Offset in bytes of source array

  • ByteCount[in] Size of memory copy in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyDtoA(hipArray_t dstArray, size_t dstOffset, hipDeviceptr_t srcDevice, size_t ByteCount)#

Copies from device memory to a 1D array.

Parameters:
  • dstArray[out] Destination array

  • dstOffset[in] Offset in bytes of destination array

  • srcDevice[in] Source device pointer

  • ByteCount[in] Size of memory copy in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyAtoA(hipArray_t dstArray, size_t dstOffset, hipArray_t srcArray, size_t srcOffset, size_t ByteCount)#

Copies from one 1D array to another.

Parameters:
  • dstArray[out] Destination array

  • dstOffset[in] Offset in bytes of destination array

  • srcArray[in] Source array

  • srcOffset[in] Offset in bytes of source array

  • ByteCount[in] Size of memory copy in bytes

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyHtoDAsync(hipDeviceptr_t dst, void *src, size_t sizeBytes, hipStream_t stream)#

Copy data from Host to Device asynchronously.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyDtoHAsync(void *dst, hipDeviceptr_t src, size_t sizeBytes, hipStream_t stream)#

Copy data from Device to Host asynchronously.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyDtoDAsync(hipDeviceptr_t dst, hipDeviceptr_t src, size_t sizeBytes, hipStream_t stream)#

Copy data from Device to Device asynchronously.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyAtoHAsync(void *dstHost, hipArray_t srcArray, size_t srcOffset, size_t ByteCount, hipStream_t stream)#

Copies from one 1D array to host memory.

Parameters:
  • dstHost[out] Destination pointer

  • srcArray[in] Source array

  • srcOffset[in] Offset in bytes of source array

  • ByteCount[in] Size of memory copy in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipMemcpyHtoAAsync(hipArray_t dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, hipStream_t stream)#

Copies from host memory to a 1D array.

Parameters:
  • dstArray[out] Destination array

  • dstOffset[in] Offset in bytes of destination array

  • srcHost[in] Source host pointer

  • ByteCount[in] Size of memory copy in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue

hipError_t hipModuleGetGlobal(hipDeviceptr_t *dptr, size_t *bytes, hipModule_t hmod, const char *name)#

Returns a global pointer from a module. Returns in *dptr and *bytes the pointer and size of the global of name name located in module hmod. If no variable of that name exists, it returns hipErrorNotFound. Both parameters dptr and bytes are optional. If one of them is NULL, it is ignored and hipSuccess is returned.

Parameters:
  • dptr[out] Returns global device pointer

  • bytes[out] Returns global size in bytes

  • hmod[in] Module to retrieve global from

  • name[in] Name of global to retrieve

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotFound, hipErrorInvalidContext

hipError_t hipGetSymbolAddress(void **devPtr, const void *symbol)#

Gets device pointer associated with symbol on the device.

Parameters:
  • devPtr[out] pointer to the device associated the symbole

  • symbol[in] pointer to the symbole of the device

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipGetSymbolSize(size_t *size, const void *symbol)#

Gets the size of the given symbol on the device.

Parameters:
  • symbol[in] pointer to the device symbole

  • size[out] pointer to the size

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipGetProcAddress(const char *symbol, void **pfn, int hipVersion, uint64_t flags, hipDriverProcAddressQueryResult *symbolStatus)#

Gets the pointer of requested HIP driver function.

Returns hipSuccess if the returned pfn is addressed to the pointer of found driver function.

Parameters:
  • symbol[in] The Symbol name of the driver function to request.

  • pfn[out] Output pointer to the requested driver function.

  • hipVersion[in] The HIP version for the requested driver function symbol. HIP version is defined as 100*version_major + version_minor. For example, in HIP 6.1, the hipversion is 601, for the symbol function “hipGetDeviceProperties”, the specified hipVersion 601 is greater or equal to the version 600, the symbol function will be handle properly as backend compatible function.

  • flags[in] Currently only default flag is suppported.

  • symbolStatus[out] Optional enumeration for returned status of searching for symbol driver function based on the input hipVersion.

Returns:

hipSuccess, hipErrorInvalidValue.

hipError_t hipMemcpyToSymbol(const void *symbol, const void *src, size_t sizeBytes, size_t offset, hipMemcpyKind kind)#

Copies data to the given symbol on the device. Symbol HIP APIs allow a kernel to define a device-side data symbol which can be accessed on the host side. The symbol can be in __constant or device space. Note that the symbol name needs to be encased in the HIP_SYMBOL macro. This also applies to hipMemcpyFromSymbol, hipGetSymbolAddress, and hipGetSymbolSize. For detailed usage, see the memcpyToSymbol example in the HIP Porting Guide.

Parameters:
  • symbol[out] pointer to the device symbole

  • src[in] pointer to the source address

  • sizeBytes[in] size in bytes to copy

  • offset[in] offset in bytes from start of symbole

  • kind[in] type of memory transfer

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemcpyToSymbolAsync(const void *symbol, const void *src, size_t sizeBytes, size_t offset, hipMemcpyKind kind, hipStream_t stream)#

Copies data to the given symbol on the device asynchronously.

Parameters:
  • symbol[out] pointer to the device symbole

  • src[in] pointer to the source address

  • sizeBytes[in] size in bytes to copy

  • offset[in] offset in bytes from start of symbole

  • kind[in] type of memory transfer

  • stream[in] stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemcpyFromSymbol(void *dst, const void *symbol, size_t sizeBytes, size_t offset, hipMemcpyKind kind)#

Copies data from the given symbol on the device.

Parameters:
  • dst[out] Returns pointer to destinition memory address

  • symbol[in] Pointer to the symbole address on the device

  • sizeBytes[in] Size in bytes to copy

  • offset[in] Offset in bytes from the start of symbole

  • kind[in] Type of memory transfer

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemcpyFromSymbolAsync(void *dst, const void *symbol, size_t sizeBytes, size_t offset, hipMemcpyKind kind, hipStream_t stream)#

Copies data from the given symbol on the device asynchronously.

Parameters:
  • dst[out] Returns pointer to destinition memory address

  • symbol[in] pointer to the symbole address on the device

  • sizeBytes[in] size in bytes to copy

  • offset[in] offset in bytes from the start of symbole

  • kind[in] type of memory transfer

  • stream[in] stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemcpyAsync(void *dst, const void *src, size_t sizeBytes, hipMemcpyKind kind, hipStream_t stream)#

Copy data from src to dst asynchronously.

For multi-gpu or peer-to-peer configurations, it is recommended to use a stream which is a attached to the device where the src data is physically located. For optimal peer-to-peer copies, the copy device must be able to access the src and dst pointers (by calling hipDeviceEnablePeerAccess with copy agent as the current device and src/dest as the peerDevice argument. if this is not done, the hipMemcpy will still work, but will perform the copy using a staging buffer on the host.

Warning

If host or dest are not pinned, the memory copy will be performed synchronously. For best performance, use hipHostMalloc to allocate host memory that is transferred asynchronously.

Warning

on HCC hipMemcpyAsync does not support overlapped H2D and D2H copies. For hipMemcpy, the copy is always performed by the device associated with the specified stream.

Parameters:
  • dst[out] Data being copy to

  • src[in] Data being copy from

  • sizeBytes[in] Data size in bytes

  • kind[in] Type of memory transfer

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorUnknown

hipError_t hipMemset(void *dst, int value, size_t sizeBytes)#

Fills the first sizeBytes bytes of the memory area pointed to by dest with the constant byte value value.

Parameters:
  • dst[out] Data being filled

  • value[in] Value to be set

  • sizeBytes[in] Data size in bytes

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetD8(hipDeviceptr_t dest, unsigned char value, size_t count)#

Fills the first sizeBytes bytes of the memory area pointed to by dest with the constant byte value value.

Parameters:
  • dest[out] Data ptr to be filled

  • value[in] Value to be set

  • count[in] Number of values to be set

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetD8Async(hipDeviceptr_t dest, unsigned char value, size_t count, hipStream_t stream)#

Fills the first sizeBytes bytes of the memory area pointed to by dest with the constant byte value value.

hipMemsetD8Async() is asynchronous with respect to the host, so the call may return before the memset is complete. The operation can optionally be associated to a stream by passing a non-zero stream argument. If stream is non-zero, the operation may overlap with operations in other streams.

Parameters:
  • dest[out] Data ptr to be filled

  • value[in] Constant value to be set

  • count[in] Number of values to be set

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetD16(hipDeviceptr_t dest, unsigned short value, size_t count)#

Fills the first sizeBytes bytes of the memory area pointed to by dest with the constant short value value.

Parameters:
  • dest[out] Data ptr to be filled

  • value[in] Constant value to be set

  • count[in] Number of values to be set

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetD16Async(hipDeviceptr_t dest, unsigned short value, size_t count, hipStream_t stream)#

Fills the first sizeBytes bytes of the memory area pointed to by dest with the constant short value value.

hipMemsetD16Async() is asynchronous with respect to the host, so the call may return before the memset is complete. The operation can optionally be associated to a stream by passing a non-zero stream argument. If stream is non-zero, the operation may overlap with operations in other streams.

Parameters:
  • dest[out] Data ptr to be filled

  • value[in] Constant value to be set

  • count[in] Number of values to be set

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetD32(hipDeviceptr_t dest, int value, size_t count)#

Fills the memory area pointed to by dest with the constant integer value for specified number of times.

Parameters:
  • dest[out] Data being filled

  • value[in] Constant value to be set

  • count[in] Number of values to be set

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMemsetAsync(void *dst, int value, size_t sizeBytes, hipStream_t stream)#

Fills the first sizeBytes bytes of the memory area pointed to by dev with the constant byte value value.

hipMemsetAsync() is asynchronous with respect to the host, so the call may return before the memset is complete. The operation can optionally be associated to a stream by passing a non-zero stream argument. If stream is non-zero, the operation may overlap with operations in other streams.

Parameters:
  • dst[out] Pointer to device memory

  • value[in] Value to set for each byte of specified memory

  • sizeBytes[in] Size in bytes to set

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemsetD32Async(hipDeviceptr_t dst, int value, size_t count, hipStream_t stream)#

Fills the memory area pointed to by dev with the constant integer value for specified number of times.

hipMemsetD32Async() is asynchronous with respect to the host, so the call may return before the memset is complete. The operation can optionally be associated to a stream by passing a non-zero stream argument. If stream is non-zero, the operation may overlap with operations in other streams.

Parameters:
  • dst[out] Pointer to device memory

  • value[in] Value to set for each byte of specified memory

  • count[in] Number of values to be set

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemset2D(void *dst, size_t pitch, int value, size_t width, size_t height)#

Fills the memory area pointed to by dst with the constant value.

Parameters:
  • dst[out] Pointer to device memory

  • pitch[in] Data size in bytes

  • value[in] Constant value to be set

  • width[in]

  • height[in]

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemset2DAsync(void *dst, size_t pitch, int value, size_t width, size_t height, hipStream_t stream)#

Fills asynchronously the memory area pointed to by dst with the constant value.

Parameters:
  • dst[in] Pointer to 2D device memory

  • pitch[in] Pitch size in bytes

  • value[in] Value to be set for each byte of specified memory

  • width[in] Width of matrix set columns in bytes

  • height[in] Height of matrix set rows in bytes

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemset3D(hipPitchedPtr pitchedDevPtr, int value, hipExtent extent)#

Fills synchronously the memory area pointed to by pitchedDevPtr with the constant value.

Parameters:
  • pitchedDevPtr[in] Pointer to pitched device memory

  • value[in] Value to set for each byte of specified memory

  • extent[in] Size parameters for width field in bytes in device memory

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemset3DAsync(hipPitchedPtr pitchedDevPtr, int value, hipExtent extent, hipStream_t stream)#

Fills asynchronously the memory area pointed to by pitchedDevPtr with the constant value.

Parameters:
  • pitchedDevPtr[in] Pointer to pitched device memory

  • value[in] Value to set for each byte of specified memory

  • extent[in] Size parameters for width field in bytes in device memory

  • stream[in] Stream identifier

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMemGetInfo(size_t *free, size_t *total)#

Query memory info.

On ROCM, this function gets the actual free memory left on the current device, so supports the cases while running multi-workload (such as multiple processes, multiple threads, and multiple GPUs).

Warning

On Windows, the free memory only accounts for memory allocated by this process and may be optimistic.

Parameters:
  • free[out] Returns free memory on the current device in bytes

  • total[out] Returns total allocatable memory on the current device in bytes

Returns:

hipSuccess, hipErrorInvalidDevice, hipErrorInvalidValue

hipError_t hipMemPtrGetInfo(void *ptr, size_t *size)#

Get allocated memory size via memory pointer.

This function gets the allocated shared virtual memory size from memory pointer.

Parameters:
  • ptr[in] Pointer to allocated memory

  • size[out] Returns the allocated memory size in bytes

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipMallocArray(hipArray_t *array, const hipChannelFormatDesc *desc, size_t width, size_t height, unsigned int flags)#

Allocate an array on the device.

Parameters:
  • array[out] Pointer to allocated array in device memory

  • desc[in] Requested channel format

  • width[in] Requested array allocation width

  • height[in] Requested array allocation height

  • flags[in] Requested properties of allocated array

Returns:

hipSuccess, hipErrorOutOfMemory

hipError_t hipArrayCreate(hipArray_t *pHandle, const HIP_ARRAY_DESCRIPTOR *pAllocateArray)#

Create an array memory pointer on the device.

Parameters:
  • pHandle[out] Pointer to the array memory

  • pAllocateArray[in] Requested array desciptor

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotSupported

hipError_t hipArrayDestroy(hipArray_t array)#

Destroy an array memory pointer on the device.

Parameters:

array[in] Pointer to the array memory

Returns:

hipSuccess, hipErrorInvalidValue

hipError_t hipArray3DCreate(hipArray_t *array, const HIP_ARRAY3D_DESCRIPTOR *pAllocateArray)#

Create a 3D array memory pointer on the device.

Parameters:
  • array[out] Pointer to the 3D array memory

  • pAllocateArray[in] Requested array desciptor

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotSupported

hipError_t hipMalloc3D(hipPitchedPtr *pitchedDevPtr, hipExtent extent)#

Create a 3D memory pointer on the device.

Parameters:
  • pitchedDevPtr[out] Pointer to the 3D memory

  • extent[in] Requested extent

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotSupported

hipError_t hipFreeArray(hipArray_t array)#

Frees an array on the device.

Parameters:

array[in] Pointer to array to free

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorNotInitialized

hipError_t hipMalloc3DArray(hipArray_t *array, const struct hipChannelFormatDesc *desc, struct hipExtent extent, unsigned int flags)#

Allocate an array on the device.

Parameters:
  • array[out] Pointer to allocated array in device memory

  • desc[in] Requested channel format

  • extent[in] Requested array allocation width, height and depth

  • flags[in] Requested properties of allocated array

Returns:

hipSuccess, hipErrorOutOfMemory

hipError_t hipArrayGetInfo(hipChannelFormatDesc *desc, hipExtent *extent, unsigned int *flags, hipArray_t array)#

Gets info about the specified array.

Parameters:
  • desc[out] - Returned array type

  • extent[out] - Returned array shape. 2D arrays will have depth of zero

  • flags[out] - Returned array flags

  • array[in] - The HIP array to get info for

Returns:

hipSuccess, hipErrorInvalidValue hipErrorInvalidHandle

hipError_t hipArrayGetDescriptor(HIP_ARRAY_DESCRIPTOR *pArrayDescriptor, hipArray_t array)#

Gets a 1D or 2D array descriptor.

Parameters:
  • pArrayDescriptor[out] - Returned array descriptor

  • array[in] - Array to get descriptor of

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue hipErrorInvalidHandle

hipError_t hipArray3DGetDescriptor(HIP_ARRAY3D_DESCRIPTOR *pArrayDescriptor, hipArray_t array)#

Gets a 3D array descriptor.

Parameters:
  • pArrayDescriptor[out] - Returned 3D array descriptor

  • array[in] - 3D array to get descriptor of

Returns:

hipSuccess, hipErrorDeinitialized, hipErrorNotInitialized, hipErrorInvalidContext, hipErrorInvalidValue hipErrorInvalidHandle, hipErrorContextIsDestroyed

hipError_t hipMemcpy2D(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, hipMemcpyKind kind)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • dpitch[in] Pitch of destination memory

  • src[in] Source memory address

  • spitch[in] Pitch of source memory

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpyParam2D(const hip_Memcpy2D *pCopy)#

Copies memory for 2D arrays.

Parameters:

pCopy[in] Parameters for the memory copy

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpyParam2DAsync(const hip_Memcpy2D *pCopy, hipStream_t stream)#

Copies memory for 2D arrays.

Parameters:
  • pCopy[in] Parameters for the memory copy

  • stream[in] Stream to use

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, hipMemcpyKind kind, hipStream_t stream)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • dpitch[in] Pitch of destination memory

  • src[in] Source memory address

  • spitch[in] Pitch of source memory

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

  • stream[in] Stream to use

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DToArray(hipArray_t dst, size_t wOffset, size_t hOffset, const void *src, size_t spitch, size_t width, size_t height, hipMemcpyKind kind)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • wOffset[in] Destination starting X offset

  • hOffset[in] Destination starting Y offset

  • src[in] Source memory address

  • spitch[in] Pitch of source memory

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DToArrayAsync(hipArray_t dst, size_t wOffset, size_t hOffset, const void *src, size_t spitch, size_t width, size_t height, hipMemcpyKind kind, hipStream_t stream)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • wOffset[in] Destination starting X offset

  • hOffset[in] Destination starting Y offset

  • src[in] Source memory address

  • spitch[in] Pitch of source memory

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

  • stream[in] Accelerator view which the copy is being enqueued

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DArrayToArray(hipArray_t dst, size_t wOffsetDst, size_t hOffsetDst, hipArray_const_t src, size_t wOffsetSrc, size_t hOffsetSrc, size_t width, size_t height, hipMemcpyKind kind)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • wOffsetDst[in] Destination starting X offset

  • hOffsetDst[in] Destination starting Y offset

  • src[in] Source memory address

  • wOffsetSrc[in] Source starting X offset

  • hOffsetSrc[in] Source starting Y offset (columns in bytes)

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DFromArray(void *dst, size_t dpitch, hipArray_const_t src, size_t wOffset, size_t hOffset, size_t width, size_t height, hipMemcpyKind kind)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • dpitch[in] Pitch of destination memory

  • src[in] Source memory address

  • wOffset[in] Source starting X offset

  • hOffset[in] Source starting Y offset

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy2DFromArrayAsync(void *dst, size_t dpitch, hipArray_const_t src, size_t wOffset, size_t hOffset, size_t width, size_t height, hipMemcpyKind kind, hipStream_t stream)#

Copies data between host and device asynchronously.

Parameters:
  • dst[in] Destination memory address

  • dpitch[in] Pitch of destination memory

  • src[in] Source memory address

  • wOffset[in] Source starting X offset

  • hOffset[in] Source starting Y offset

  • width[in] Width of matrix transfer (columns in bytes)

  • height[in] Height of matrix transfer (rows)

  • kind[in] Type of transfer

  • stream[in] Accelerator view which the copy is being enqueued

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpyAtoH(void *dst, hipArray_t srcArray, size_t srcOffset, size_t count)#

Copies data between host and device.

Parameters:
  • dst[in] Destination memory address

  • srcArray[in] Source array

  • srcOffset[in] Offset in bytes of source array

  • count[in] Size of memory copy in bytes

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpyHtoA(hipArray_t dstArray, size_t dstOffset, const void *srcHost, size_t count)#

Copies data between host and device.

Parameters:
  • dstArray[in] Destination memory address

  • dstOffset[in] Offset in bytes of destination array

  • srcHost[in] Source host pointer

  • count[in] Size of memory copy in bytes

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy3D(const struct hipMemcpy3DParms *p)#

Copies data between host and device.

Parameters:

p[in] 3D memory copy parameters

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipMemcpy3DAsync(const struct hipMemcpy3DParms *p, hipStream_t stream)#

Copies data between host and device asynchronously.

Parameters:
  • p[in] 3D memory copy parameters

  • stream[in] Stream to use

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipDrvMemcpy3D(const HIP_MEMCPY3D *pCopy)#

Copies data between host and device.

Parameters:

pCopy[in] 3D memory copy parameters

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

hipError_t hipDrvMemcpy3DAsync(const HIP_MEMCPY3D *pCopy, hipStream_t stream)#

Copies data between host and device asynchronously.

Parameters:
  • pCopy[in] 3D memory copy parameters

  • stream[in] Stream to use

Returns:

hipSuccess, hipErrorInvalidValue, hipErrorInvalidPitchValue, hipErrorInvalidDevicePointer, hipErrorInvalidMemcpyDirection

template<typename T>
hipError_t hipGetSymbolAddress(void **devPtr, const T &symbol)#

Gets the address of a symbol.

Parameters:
  • devPtr[out] - Returns device pointer associated with symbol.

  • symbol[in] - Device symbol.

Returns:

hipSuccess, hipErrorInvalidValue

template<typename T>
hipError_t hipGetSymbolSize(size_t *size, const T &symbol)#

Gets the size of a symbol.

Parameters:
  • size[out] - Returns the size of a symbol.

  • symbol[in] - Device symbol address.

Returns:

hipSuccess, hipErrorInvalidValue

template<typename T>
hipError_t hipMemcpyToSymbol(const T &symbol, const void *src, size_t sizeBytes, size_t offset, hipMemcpyKind kind)#

Copies data to the given symbol on the device.

Returns:

hipSuccess, hipErrorInvalidMemcpyDirection, hipErrorInvalidValue

template<typename T>
hipError_t hipMemcpyToSymbolAsync(const T &symbol, const void *src, size_t sizeBytes, size_t offset, hipMemcpyKind kind, hipStream_t stream)#

Copies data to the given symbol on the device asynchronously on the stream.

Returns:

hipSuccess, hipErrorInvalidMemcpyDirection, hipErrorInvalidValue

template<typename T>
hipError_t hipMemcpyFromSymbol(void *dst, const T &symbol, size_t sizeBytes, size_t offset, hipMemcpyKind kind)#

Copies data from the given symbol on the device.

Returns:

hipSuccess, hipErrorInvalidMemcpyDirection, hipErrorInvalidValue

template<typename T>
hipError_t hipMemcpyFromSymbolAsync(void *dst, const T &symbol, size_t sizeBytes, size_t offset, hipMemcpyKind kind, hipStream_t stream)#

Copies data from the given symbol on the device asynchronously on the stream.

Returns:

hipSuccess, hipErrorInvalidMemcpyDirection, hipErrorInvalidValue

template<class T>
static inline hipError_t hipMalloc(T **devPtr, size_t size)#

: C++ wrapper for hipMalloc

Perform automatic type conversion to eliminate the need for excessive typecasting (ie void**)

HIP_DISABLE_CPP_FUNCTIONS macro can be defined to suppress these wrappers. It is useful for applications which need to obtain decltypes of HIP runtime APIs.

See also

hipMalloc

template<class T>
static inline hipError_t hipMallocPitch(T **devPtr, size_t *pitch, size_t width, size_t height)#

: C++ wrapper for hipMallocPitch

Perform automatic type conversion to eliminate the need for excessive typecasting (ie void**)

HIP_DISABLE_CPP_FUNCTIONS macro can be defined to suppress these wrappers. It is useful for applications which need to obtain decltypes of HIP runtime APIs.

See also

hipMallocPitch

template<class T>
static inline hipError_t hipHostMalloc(T **ptr, size_t size, unsigned int flags = hipHostMallocDefault)#

: C++ wrapper for hipHostMalloc

Provide an override to automatically typecast the pointer type from void**, and also provide a default for the flags.

HIP_DISABLE_CPP_FUNCTIONS macro can be defined to suppress these wrappers. It is useful for applications which need to obtain decltypes of HIP runtime APIs.

See also

hipHostMalloc

template<class T>
static inline hipError_t hipHostAlloc(T **ptr, size_t size, unsigned int flags = hipHostAllocDefault)#

: C++ wrapper for hipHostAlloc

Provide an override to automatically typecast the pointer type from void**, and also provide a default for the flags.

HIP_DISABLE_CPP_FUNCTIONS macro can be defined to suppress these wrappers. It is useful for applications which need to obtain decltypes of HIP runtime APIs.

See also

hipHostAlloc