Merge

group Merge

Functions

NVCVStatus cvcudaMergeCreate(NVCVOperatorHandle *handle)

Creates a stateless Merge operator.

Parameters:

handle – [out] Non-null destination for the operator handle.

Return values:

  • NVCV_SUCCESS – Operator successfully created.

  • NVCV_ERROR_INVALID_ARGUMENT – Handle is null.

  • NVCV_ERROR_OUT_OF_MEMORY – Insufficient memory to create the operator.

NVCVStatus cvcudaMergeSubmit(NVCVOperatorHandle handle, cudaStream_t stream, NVCVTensorBatchHandle in, NVCVTensorHandle out)

Copies input tensor c to output channel c, preserving every element’s bits. TensorBatch entries represent channels, not samples. Each input has an explicit channel dimension of one. Repeated read-only inputs are permitted, including merging the same grayscale tensor three times. C=1 still copies data. There is no blending, color conversion, metadata interpretation or alpha processing. This entry point operates on tensors; see cvcudaMergeVarShapeSubmit for image batches. Submission is asynchronous; all tensor storage must remain alive until completion.

Limitations:

Tensor input (each band): Data Layout: [kNHWC, kHWC, kNCHW, kCHW] Channels: [1]

Data Type

Allowed

8bit Unsigned

Yes

8bit Signed

Yes

16bit Unsigned

Yes

16bit Signed

Yes

32bit Unsigned

Yes

32bit Signed

Yes

16bit Float

Yes

32bit Float

Yes

64bit Float

Yes

Tensor output: Data Layout: [kNHWC, kHWC, kNCHW, kCHW] Channels: [1, 2, 3, 4] Data Type: same scalar type as input; vector/packed/complex types are not supported.

All dimensions must be positive. All inputs must have identical shape, layout and dtype. Output rank, sample count, height, width and dtype must match the inputs; its layout may differ. The number of inputs must equal the output channel count. CUDA-accessible, pitch-linear tensors are required. Base addresses and strides must be aligned to the scalar element size, and all strides must be positive. Each non-singleton dimension’s stride must cover its inner dimensions’ byte span in layout order; singleton strides need not. Row/sample padding is allowed. Input/output byte spans must not overlap, including padding inside those spans. Padding is not written. No resizing or broadcasting occurs. Floating-point and multichannel non-U8 support extend Pillow’s pixel-copy semantics.

Parameters:

  • handle – [in] Merge operator handle, not null.

  • stream – [in] CUDA stream, including the default stream.

  • in – [in] Prepopulated batch of source band tensors, not null.

  • out – [out] Destination tensor, not null.

Return values:

  • NVCV_SUCCESS – Work successfully submitted.

  • NVCV_ERROR_INVALID_ARGUMENT – Invalid metadata, unsupported storage, or overlapping buffers.

NVCVStatus cvcudaMergeVarShapeSubmit(NVCVOperatorHandle handle, cudaStream_t stream, const NVCVImageBatchHandle *in, int32_t numInputs, NVCVImageBatchHandle out)

Copies image n in input batch c to physical storage channel c of output image n. All element bits are preserved, including floating-point payloads. C=1 still copies data; output images are not views. Channel order is the component order within an interleaved pixel or plane order for planar images. Color models, swizzles and alpha metadata are not interpreted; no conversion or alpha processing occurs. Repeated read-only inputs are permitted, including the same band batch in multiple positions. Submission is asynchronous. All batches, images, image/batch descriptors and backing storage must remain alive and must not be mutated by the caller until completion. The host array of batch handles need only survive this call.

Limitations:

ImageBatchVarShape input (each band): Data Layout: one single-channel plane. Channels: [1]

Data Type

Allowed

8bit Unsigned

Yes

8bit Signed

Yes

16bit Unsigned

Yes

16bit Signed

Yes

32bit Unsigned

Yes

32bit Signed

Yes

16bit Float

Yes

32bit Float

Yes

64bit Float

Yes

ImageBatchVarShape output: Data Layout: one interleaved plane, or C full-resolution single-channel planes. Channels: [1, 2, 3, 4] Data Type: same scalar type as input.

Each batch must be nonempty and have a uniform image format. Every batch has the same number N of images; numInputs equals the output channel count C. Semantic format tags may differ across band batches if their scalar dtype matches. Image sizes must be positive and match at each corresponding index across batches, but may vary across indices. Pitches and allocations may vary across images and planes. Plane data kinds and packings must describe contiguous, byte-aligned, equal-width components of exactly one of the listed scalar types, without intra-pixel padding. Subsampled, sub-byte, unequal-component and hybrid plane packings are not supported. Extra-channel metadata must have NVCVExtraChannelInfo.numChannels == 0. CUDA-accessible, pitch-linear storage is required; CUDA arrays are not supported. Plane bases and row pitches must be aligned to the scalar size, and each positive row pitch must cover a full row. Every write/read and write/write overlap across the entire submission is rejected, including across image indices, bands and planes and padding inside their byte spans. Padding is not written. No resizing or broadcasting occurs.

Parameters:

  • handle – [in] Merge operator handle, not null.

  • stream – [in] CUDA stream, including the default stream.

  • in – [in] Host array of numInputs nonempty source band batch handles, none null.

  • numInputs – [in] Number of band batches, equal to C (1-4).

  • out – [out] Prepopulated destination image batch, not null.

Return values:

  • NVCV_SUCCESS – Work successfully submitted.

  • NVCV_ERROR_INVALID_ARGUMENT – Invalid metadata, unsupported storage, or overlapping buffers.