Releases · ROCm/rpp (original) (raw)

rocm-7.2.4

RPP code for ROCm 7.2 did not change. The library was rebuilt for the updated ROCm 7.2 stack.

rpp 2.2.1 for ROCm 7.2.3

RPP code for ROCm 7.2.3 did not change. The library was rebuilt for the updated ROCm 7.2.3 stack.

rpp 2.2.1 for ROCm 7.2.2

RPP code for ROCm 7.2.2 did not change. The library was rebuilt for the updated ROCm 7.2.2 stack.

rpp 2.2.1 for ROCm 7.2.1

RPP code for ROCm 7.2.1 did not change. The library was rebuilt for the updated ROCm 7.2.1 stack.

rpp 2.1.0 for ROCm 7.2.0

RPP code for ROCm 7.2.0 did not change. The library was rebuilt for the updated ROCm 7.2.0 stack.

rpp 2.1.0 for ROCm 7.1.1

RPP code for ROCm 7.1.1 did not change. The library was rebuilt for the updated ROCm 7.1.1 stack.

RPP 2.1.0 for ROCm 7.1.0

Added

• Solarize augmentation for HOST and HIP
• Hue and Saturation adjustment augmentations for HOST and HIP
• Find RPP - cmake module
• Posterize augmentation for HOST and HIP

Changed

• HALF - Fix half.hpp path updates
• Box filter - padding updates

Removed

• Packaging - Remove Meta Package dependency for HIP
• SLES 15 SP6 support

Resolved issues

• Test Suite - Fixes for accuracy
• HIP Backend - Check return status warning fixes
• Bugfix - HIP vector types init

rpp 2.0.0 for ROCm 7.0.2

RPP code for ROCm 7.0.2 did not change. The library was rebuilt for the updated ROCm 7.0.2 stack.

rpp 2.0.0 for ROCm 7.0.1

RPP code for ROCm 7.0.1 did not change. The library was rebuilt for the updated ROCm 7.0.1 stack.

RPP 2.0.0 for ROCm 7.0.0

Added

• Bitwise NOT, Bitwise AND, Bitwise OR augmentations on HOST (CPU) and HIP backends. (#520)
• Tensor Concat augmentation on HOST (CPU) and HIP backends. (#530)
• JPEG Compression Distortion augmentation on HIP backend. (#538)
• log1p, defined as log (1 + x), tensor augmentation support on HOST (CPU) and HIP backends.
• JPEG Compression Distortion augmentation on HOST (CPU) backend. (#531)

Changed

• All handle creation and destruction APIs have been consolidated to rppCreate(), for handle initialization, and rppDestroy(), for handle destruction (#513)
• RPP function category "logical_operations" more appropriately renamed to "bitwise_operations". (#520)
• TurboJPEG package installation enabled for RPP Test Suite with sudo apt-get install libturbojpeg0-dev. Instructions updated in utilities/test_suite/README.md. (#518)
• Changed API of swap_channels augmentation to be called channel_permute, which now accepts one new argument, "permutationTensor" (pointer to a unsigned int tensor) that provides the permutation order to swap the RGB channels of each input image in the batch in any order. (#547)
  • Old API - RppStatus rppt_swap_channels_host(RppPtr_t srcPtr, RpptDescPtr srcDescPtr, RppPtr_t dstPtr, RpptDescPtr dstDescPtr, rppHandle_t rppHandle);
  • New API - RppStatus rppt_channel_permute_host(RppPtr_t srcPtr, RpptDescPtr srcDescPtr, RppPtr_t dstPtr, RpptDescPtr dstDescPtr, Rpp32u *permutationTensor , rppHandle_t rppHandle);

Removed

• Older versions of RPP handle creation inlcuding rppCreateWithBatchSize(), rppCreateWithStream(), and rppCreateWithStreamAndBatchSize() are now removed and replaced with rppCreate().
• Older versions of RPP handle destruction API including rppDestroyGPU() and rppDestroyHost() are now removed and replaced with rppDestroy().

Resolved issues

• Test package - debian packages will install required dependencies