The proposed architecture contains some data processing units and configurable accelerators and on-chip network connecting the heterogeneous cores.
结构中包含数据处理单元、可配置加速部件以及连接各个核的片上网络。
Experimental results also reveal two primary causes which make theoretical speedup hard to achieve: limited DRAM bandwidth and resource contention of on-chip network.
通过实验还证明实测性能难以达到理论预测值的两个主要原因:访存带宽有限和片上网络的资源竞争。
Experimental results also reveal two primary causes which make theoretical speedup hard to achieve: limited DRAM bandwidth and resource contention of on-chip network.
通过实验还证明实测性能难以达到理论预测值的两个主要原因:访存带宽有限和片上网络的资源竞争。
应用推荐