← 研踪 PaperTrailarXiv:2610.39222 · v1
When Does Heterogeneity Pay Off? Understanding GPU-WSE System Complementarity
首次提交 2026-10-29 10:42
版本更新 2026-10-29 10:42
雷达观测 2026-10-30 11:25
摘要
As AI accelerator architectures become increasingly diverse, integrating different architectures into a unified system is emerging as a new heterogeneous computing paradigm. GPUs and WSEs differ substantially in compute, memory, and communication organization, creating both architectural complementarity and additional overheads, yet how to translate this heterogeneity into practical performance gains remains unclear. To address this gap, we propose a hierarchical benchmarking methodology to characterize GPU--WSE heterogeneous execution and the conditions for exploiting complementarity, together with an analytical performance model that identifies when heterogeneity becomes a bottleneck and when it yields system-level performance gains.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-10-29 10:42 · 最新观测版本
When Does Heterogeneity Pay Off? Understanding GPU-WSE System Complementarity
As AI accelerator architectures become increasingly diverse, integrating different architectures into a unified system is emerging as a new heterogeneous computing paradigm. GPUs and WSEs differ substantially in compute, memory, and communication organization, creating both architectural complementarity and additional overheads, yet how to translate this heterogeneity into practical performance gains remains unclear. To address this gap, we propose a hierarchical benchmarking methodology to characterize GPU--WSE heterogeneous execution and the conditions for exploiting complementarity, together with an analytical performance model that identifies when heterogeneity becomes a bottleneck and when it yields system-level performance gains.
查看此版本解读 →查看此版本原文 ↗