Cutting edge and innovative AI hardware research from China.

Looks like Amerikan sanctions are driving a new wave of innovation in China.

" This work addresses that gap by introducing the Ten- sor Manipulation Unit (TMU): a reconfigurable, near-memory hardware block designed to execute data-movement-intensive (DMI) operators efficiently. TMU manipulates long datastreams in a memory-to-memory fashion using a RISC-inspired execution model and a unified addressing abstraction, enabling broad support for both coarse- and fine-grained tensor transformations.

The proposed architecture integrates TMU alongside a TPU within a high-throughput AI SoC, leveraging double buffering and output forwarding to improve pipeline utilization. Fab- ricated in SMIC 40 nm technology, the TMU occupies only 0.019 mm2 while supporting over 10 representative TM operators. Benchmarking shows that TMU alone achieves up to 1413.43× and 8.54× operator-level latency reduction over ARM A72 and NVIDIA Jetson TX2, respectively.

When integrated with the in- house TPU, the complete system achieves a 34.6% reduction in end-to-end inference latency, demonstrating the effectiveness and scalability of reconfigurable tensor manipulation in modern AI SoCs."