Parallel GEMM-based convolution for deep learning on multicore RISC-V processors

• Published in: The Journal of supercomputing Vol. 80; no. 9; pp. 12623 - 12643
• Main Authors: Ramírez, Cristian, Castelló, Adrián, Martínez, Héctor, Quintana-Ortí, Enrique S.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2024; Springer Nature B.V
• Subjects: Compilers; Computer Science; Convolution; Deep learning; Interpreters; Microprocessors; Processor Architectures; Programming Languages; RISC; Tiling; Deep learning; Edge processors; Convolutional layers; Performance analysis
• ISSN: 0920-8542; 1573-0484
• DOI: 10.1007/s11227-024-05927-y

We address the efficient implementation of the convolution operator on the GAP8 parallel ultra-low power platform (PULP), a heterogeneous multi-core processor equipped with a fabric controller (FC); a cluster of eight compute cores; and a four-level memory hierarchy with scratchpads instead of conventional, hardware-assisted cache memories. Our solution for this platform transforms the convolution into a general matrix–matrix multiplication ( gemm ) via the lowering approach, demonstrating that it is possible to attain reasonable performance on the GAP8 by carefully adapting techniques such as tiling and loop parallelism, which are mainstream in the multi-threaded, cache-aware realization of gemm .