Mahdi Nazm Bojnordi

Hoda Roodaki, Mahdi Nazm Bojnordi

The ever-increasing demand for 3D modeling in the emerging immersive applications has made point clouds an essential class of data for 3D image and video processing. Tree based structures are commonly used for representing point clouds where pointers are used to realize the connection between nodes. Tree-based structures significantly suffer from irregular access patterns for large point clouds. Memory access indirection in such structures is disruptive to bandwidth efficiency and performance. In this paper, we propose a point cloud representation format based on compressed geometric arrays (CGA). Then, we examine new methods for point cloud processing based on CGA. The proposed format enables a higher bandwidth efficiency via eliminating memory access indirections (i.e., pointer chasing at the nodes of tree) thereby improving the efficiency of point cloud processing. Our experimental results show that using CGA for point cloud operations achieves 1328x speed up, 1321x better bandwidth utilization, and 54% reduction in the volume of transferred data as compared to the state-of-the-art tree-based format from point cloud library (PCL).

Geng Yuan, Payman Behnam, Zhengang Li et al.

Recent works demonstrated the promise of using resistive random access memory (ReRAM) as an emerging technology to perform inherently parallel analog domain in-situ matrix-vector multiplication -- the intensive and key computation in DNNs. With weights stored in the ReRAM crossbar cells as conductance, when the input vector is applied to word lines, the matrix-vector multiplication results can be generated as the current in bit lines. A key problem is that the weight can be either positive or negative, but the in-situ computation assumes all cells on each crossbar column with the same sign. The current architectures either use two ReRAM crossbars for positive and negative weights, or add an offset to weights so that all values become positive. Neither solution is ideal: they either double the cost of crossbars, or incur extra offset circuity. To better solve this problem, this paper proposes FORMS, a fine-grained ReRAM-based DNN accelerator with polarized weights. Instead of trying to represent the positive/negative weights, our key design principle is to enforce exactly what is assumed in the in-situ computation -- ensuring that all weights in the same column of a crossbar have the same sign. It naturally avoids the cost of an additional crossbar. Such weights can be nicely generated using alternating direction method of multipliers (ADMM) regularized optimization, which can exactly enforce certain patterns in DNN weights. To achieve high accuracy, we propose to use fine-grained sub-array columns, which provide a unique opportunity for input zero-skipping, significantly avoiding unnecessary computations. It also makes the hardware much easier to implement. Putting all together, with the same optimized models, FORMS achieves significant throughput improvement and speed up in frame per second over ISAAC with similar area cost.