A message-driven mapping framework for VGG-19 inference on the MAVeC accelerator is claimed to generate over 97% of messages on-chip and sustain 88-92% SiteO utilization in simulation.
Immediately after programming , A_MULS compute messages are multicast to all C-0 SiteOs (entry 2), each carrying an image data and the convolution-pattern bits
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Hardware-Aware Data and Instruction Mapping for AI Tasks: Balancing Parallelism, I/O and Memory Tradeoffs
A message-driven mapping framework for VGG-19 inference on the MAVeC accelerator is claimed to generate over 97% of messages on-chip and sustain 88-92% SiteO utilization in simulation.