Open Issues Need Help
View All on GitHubAI Summary: The user needs help integrating a group convolution kernel for the Neon backend into the Arm Compute Library (ACL). The existing ACL 2D convolution kernel declares `num_groups` as unused, indicating a potential missing feature or requiring modification to support group convolutions.
The Compute Library is a set of computer vision and machine learning functions optimised for both Arm CPUs and GPUs using SIMD technologies.
AI Summary: Investigate a reported 7-10% performance degradation in the Winograd Convolution function of the Arm Compute Library (ACL) version 52.2.0 compared to v25.03, specifically on Apple M2 Pro hardware. The task involves analyzing provided benchmark code, comparing performance results, and potentially debugging the ACL implementation to identify the source of the performance regression.
The Compute Library is a set of computer vision and machine learning functions optimised for both Arm CPUs and GPUs using SIMD technologies.