FPGA based efficient on-chip memory for image processing algorithms

Deepa, P.; Vasanthanayaki, C.

doi:10.1016/j.mejo.2012.05.001

Cited by 5 publications

(2 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To optimze the performance of matrix multiplication, we employ the Z-Morton memory layout [8], which has been widely studied for the Cache oblivious algorithms on multithreaded CPUs [8,12] and image processing on FPGAs [4]. This memory layout increases both spatial and temporal locality of memory accesses of matrix multiplication and arithmetic operations [8].…”

Section: Matrix Multiplications and Memory Access Patternsmentioning

confidence: 99%

See 1 more Smart Citation

Sparse Winograd Convolutional Neural Networks on Small-scale Systolic Arrays

Shi

Gao

et al. 2019

Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays

View full text Add to dashboard Cite

The reconfigurability, energy-efficiency, and massive parallelism on FPGAs make them one of the best choices for implementing efficient deep learning accelerators. However, state-of-art implementations seldom consider the balance between high throughput of computation power and the ability of the memory subsystem to support it. In this paper, we implement an accelerator on FPGA by combining the sparse Winograd convolution, clusters of small-scale systolic arrays, and a tailored memory layout design. We also provide an analytical model analysis for the general Winograd convolution algorithm as a design reference. Experimental results on VGG16 show that it achieves very high computational resource utilization, 20× ∼ 30× energy efficiency, and more than 5× speedup compared with the dense implementation.

show abstract

Section: Matrix Multiplications and Memory Access Patternsmentioning

confidence: 99%

“…Z-Morton memory layout for both dense and sparse matrix[4,8]: (a) the translation from logical layout to physical layout, (b) the block-based compressed coordinates (BCOO, l × l block and l = 4 for our design) for pruned Winograd weights Divide and Conquer Matrix Multiplication1: function recursive-matmult(A, B, C) 2: n = A.rows 3:…”

mentioning

confidence: 99%