Ricap documentation memcpy changes#537
Ricap documentation memcpy changes#537r-abishek merged 6 commits intor-abishek:ar/device_memcpy_removalfrom
Conversation
There was a problem hiding this comment.
Pull request overview
This PR updates RICAP (Random Image Crop And Patch) implementation by removing unnecessary hipMemcpy operations and correcting documentation across multiple API header files. The main changes involve directly using the permutationTensor parameter that is already in HIP memory instead of copying it to scratch buffer, along with widespread documentation corrections to accurately reflect memory location requirements for GPU functions.
Key Changes
- Removed redundant
hipMemcpyoperations in RICAP GPU implementation and replaced stack-allocated array withhipHostMallocin test code - Renamed
permutedIndicesTensortopermutationTensorfor consistency across RICAP API - Updated documentation for GPU functions to specify tensors are in "pinned/HIP memory" instead of "pinned/HOST memory"
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| utilities/test_suite/HIP/Tensor_image_hip.cpp | Replaced stack-allocated array with hipHostMalloc for permutationTensor |
| src/modules/tensor/rppt_tensor_effects_augmentations.cpp | Removed hipMemcpy operations and scratch buffer usage, passing permutationTensor directly to GPU kernels |
| api/rppt_tensor_effects_augmentations.h | Renamed parameter from permutedIndicesTensor to permutationTensor and updated memory location documentation |
| api/rppt_tensor_geometric_augmentations.h | Updated documentation to specify HIP memory for GPU function parameters |
| api/rppt_tensor_filter_augmentations.h | Updated documentation to specify HIP memory for GPU function parameters |
| api/rppt_tensor_color_augmentations.h | Updated documentation to specify HIP memory for GPU function parameters |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| * \param [in] noiseProbailityTensor noiseProbaility values to decide if a destination pixel is a noise-pixel, or equal to source (1D tensor in pinned memory, of size batchSize with 0 <= noiseProbailityTensor[i] <= 1 for each image in batch) | ||
| * \param [in] saltProbailityTensor saltProbaility values to decide if a given destination noise-pixel is salt or pepper (1D tensor in pinned memory, of size batchSize with 0 <= saltProbailityTensor[i] <= 1 for each image in batch) |
There was a problem hiding this comment.
Corrected spelling of 'Probaility' to 'Probability' in parameter names and descriptions.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
* Jenkins - CodeCov * Use rpp codecov token as cred * Jenkins - Install OpenMP --------- Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com>
* Jenkins - CodeCov * Use rpp codecov token as cred * Jenkins - Install OpenMP --------- Co-authored-by: Sam Wu <22262939+samjwu@users.noreply.github.com>
Documentation changes in RICAP and memcpy changes for RICAP