PvaAllocator#

Functions#

NVCVStatus nvcvAllocatorConstructPva(NVCVAllocatorHandle *handle)

Constructs an allocator instance in the given storage.

NVCVStatus nvcvImageCalcRequirementsPva(int32_t width, int32_t height, NVCVImageFormat format, int32_t baseAddrAlignment, int32_t rowAddrAlignment, NVCVImageRequirements *reqs)

Calculates the resource requirements needed to create a tensor with given shape.

NVCVStatus nvcvTensorCalcRequirementsPva(int32_t rank, const int64_t *shape, NVCVDataType dtype, NVCVTensorLayout layout, int32_t baseAddrAlignment, int32_t rowAddrAlignment, NVCVTensorRequirements *reqs)

Calculates the resource requirements needed to create a tensor with given shape.

Functions#

NVCVStatus nvcvAllocatorConstructPva(NVCVAllocatorHandle *handle)#

Constructs an allocator instance in the given storage.

When not needed anymore, the allocator instance must be destroyed by nvcvAllocatorDecRef function.

Parameters:

handle – [out] Where new instance handle will be written to.

  • Must not be NULL.

Return values:

  • NVCV_ERROR_INVALID_ARGUMENT – Some argument is outside its valid range.

  • NVCV_ERROR_OUT_OF_MEMORY – Not enough memory to create the allocator.

  • NVCV_SUCCESS – Allocator created successfully.

NVCVStatus nvcvImageCalcRequirementsPva(
int32_t width,
int32_t height,
NVCVImageFormat format,
int32_t baseAddrAlignment,
int32_t rowAddrAlignment,
NVCVImageRequirements *reqs,
)#

Calculates the resource requirements needed to create a tensor with given shape.

Parameters:

  • width, height – [in] Image dimensions.

    • Width and height must be > 0.

  • format – [in] Image format.

    • Must not be NVCV_IMAGE_FORMAT_NONE.

  • baseAddrAlignment – [in] Alignment, in bytes, of the requested memory buffer. Currently, only a value of 0 is supported, which means buffers will be allocated using the default PVA memory alignment.

  • rowAddrAlignment – [in] Alignment, in bytes, of each image’s row address. In other words, (rowAddr - baseAddr) % rowAddrAlignment == 0. If 0, use a default suitable for optimized memory access. The used alignment is at least the given value. Pass 1 for fully packed rows, i.e., no padding at the end of each row.

    • If different from 0, it must be a power-of-two.

  • reqs – [out] Where the image requirements will be written to.

    • Must not be NULL.

Return values:

  • NVCV_ERROR_INVALID_ARGUMENT – Some parameter is outside valid range.

  • NVCV_SUCCESS – Operation executed successfully.

NVCVStatus nvcvTensorCalcRequirementsPva(
int32_t rank,
const int64_t *shape,
NVCVDataType dtype,
NVCVTensorLayout layout,
int32_t baseAddrAlignment,
int32_t rowAddrAlignment,
NVCVTensorRequirements *reqs,
)#

Calculates the resource requirements needed to create a tensor with given shape.

Parameters:

  • rank – [in] Rank of the tensor (its number of dimensions).

  • shape – [in] Pointer to array with tensor shape. It must contain at least ‘rank’ elements.

  • dtype – [in] Type of tensor’s elements.

  • layout – [in] Tensor layout. Pass NVCV_TENSOR_NONE is layout is not available.

    • Layout rank must be rank.

  • baseAddrAlignment – [in] Alignment, in bytes, of the requested memory buffer. Currently, only a value of 0 is supported, which means buffers will be allocated using the default PVA memory alignment.

  • rowAddrAlignment – [in] Alignment, in bytes, of the start of each second-to-last dimension address with respect to the base address. In other words, (rowAddr - baseAddr) % rowAddrAlignment == 0. If 0, use a default suitable for optimized memory access. The used alignment is at least the given value. Pass 1 for creation of fully packed tensors, i.e., no padding between dimensions.

    • If different from 0, it must be a power-of-two.

  • reqs – [out] Where the tensor requirements will be written to.

    • Must not be NULL.

Return values:

  • NVCV_ERROR_INVALID_ARGUMENT – Some parameter is outside valid range.

  • NVCV_SUCCESS – Operation executed successfully.