LIBXSMM GEMM Microkernel

该数据集包含了针对深度学习任务优化的GEMM（通用矩阵乘法）微核的性能指标和实现，特别关注卷积操作和全连接层。此外，该数据集用于比较PolyDL系统与诸如Intel oneDNN和AutoTVM等最先进的库的性能。数据集涵盖了不同配置和优化下的性能数据，其任务是对深度学习原语和算子融合的性能进行评估。

This dataset contains performance metrics and implementations of GEMM (General Matrix Multiplication) microkernels optimized for deep learning tasks, with a particular focus on convolutional operations and fully connected layers. Furthermore, it is used to compare the performance of the PolyDL system against state-of-the-art libraries such as Intel oneDNN and AutoTVM. The dataset covers performance data across diverse configurations and optimization settings, with the objective of evaluating the performance of deep learning primitives and operator fusion.