Extending the accel dialect in AXI4MLIR with direct DMA-mapped allocation eliminates a staging copy and reduces main memory data movement by up to 2x on matrix multiplication accelerators.
Available: https://arxiv.org/abs/2402.19184
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Defeat the Heap: Zero-Copy Data Movement in AXI4MLIR
Extending the accel dialect in AXI4MLIR with direct DMA-mapped allocation eliminates a staging copy and reduces main memory data movement by up to 2x on matrix multiplication accelerators.