Workshops Archive

To SIMD Or Not To SIMD: Comparing RVV vs. CGRA SoCs For Non-Linear Functions In AI


Workshop: Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026)

Authors: Sameer Deshmukh, Naoya Mizuki, and Takumi Honda (Fujitsu Limited)

Abstract: Non-linear functions are an important class of operations in modern large language models, yet chip vendors have so far relied on traditional vector or scalar units for them, whose performance has not kept pace with the matrix multiplication units (MMUs) now found in high-performance and edge devices from almost every vendor. Reconfigurable units such as FPGAs and CGRAs have been proposed, but never systematically compared against traditional units backed by MMUs. We build two SoCs, one with an RVV vector processor and one with a elastic CGRA, each coupled with an MMU and driven by a single-core CPU. For FFNs, we find the CGRA shows maximum speedup in clock cycles of 7.12x over RVV for the activation of simple point-wise functions like SiLU, whereas RVV shows up to 9.8x over the CGRA on complex functions such as softmax; over the enclosing feed-forward block the corresponding maxima are 3.16x and 4.24x.


Back to Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026) Archive Listing

Back to Full Workshop Archive Listing