From d924f2963b0be25431bc6d56c9b7947ba098b360 Mon Sep 17 00:00:00 2001 From: Carl Pearson Date: Fri, 13 Mar 2020 17:31:55 -0600 Subject: [PATCH] stencil paper revision --- content/publication/2020522_pearson_iwapt/index.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/content/publication/2020522_pearson_iwapt/index.md b/content/publication/2020522_pearson_iwapt/index.md index ea336c2..acb8970 100644 --- a/content/publication/2020522_pearson_iwapt/index.md +++ b/content/publication/2020522_pearson_iwapt/index.md @@ -26,9 +26,10 @@ abstract = """ High-performance distributed computing systems increasingly feature nodes that have multiple CPU sockets and multiple GPUs. The communication bandwidth between these components is non-uniform. Furthermore, these systems can expose different communication capabilities between these components. -For communication-heavy applications, optimally using these capabilities is challenging and essential for performance. -This work presents approaches for automatic data placement and communication implementation for 3D stencil codes on multi-GPU nodes with non-homogeneous communication performance and capabilities. -Benchmarking results in the Summit system show that choices in placement can result in a 20% improvement in single-node exchange, and communication specialization canyield a further 6x improvement in exchange time in a single node, and a 16% improvement at 1536 GPUs""" +For communication-heavy applications, optimally using these capabilities is challenging and essential for performance. +Bespoke codes with optimized communication may be non-portable across run-time/software/hardware configurations, and existing stencil frameworks neglect optimized communication. +This work presents node-aware approaches for automatic data placement and communication implementation for 3D stencil codes on multi-GPU nodes with non-homogeneous communication performance and capabilities. +Benchmarking results in the Summit system show that choices in placement can result in a 20% improvement in single-node exchange, and communication specialization can yield a further 6x improvement in exchange time in a single node, and a 16% improvement at 1536 GPUs.""" abstract_short = ""