Back in May was the announcement by AMD of ROCm-DS as a new toolkit geared for real-world data science problems with various helpers to accelerate data processing on Instinct accelerators. AMD today is complementing ROCm-DS by announcing ROCm-LS and hipCIM.

ROCm-LS is a new toolkit from AMD for helping to accelerate life science workloads on Instinct accelerators. ROCm-LS is debuting in "early access" form for soliciting early developer feedback. ROCm-LS is described by AMD as:

"The early access release of ROCm-LS enables you to experiment with accelerating your life science workloads, such as digital pathology, automated medical image analysis, and feature extraction and enhancement in large TIFF files on AMD Instinct GPUs. Join us in exploring this tantalizing glimpse into the future capabilities of ROCm-LS, setting the stage for the next evolution in life science computing."

As part of ROCm-LS, AMD also announced the hipCIM library as part of it. The hipCIM library is for computer vision and image processing for multi-dimensional image datasets.

The hipCIM code is in fact a port of NVIDIA's cuCIM library from their RAPIDS ecosystem. With AMD's hipCIM they are maintaining API compatibility with cuCIM to help those porting their software from the CUDA world.

More details on the new ROCm-LS toolkit and hipCIM via the [1]AMD ROCm blog . The code for ROCm-LS/hipCIM is being hosted under the new [2]ROCm-LS organization on GitHub .



[1] https://rocm.blogs.amd.com/software-tools-optimization/rocm-ls-intro/README.html

[2] https://github.com/ROCm-LS