Close

Presentation

Hans Meuer Award Finalists
Research Paper
:
Hans Meuer Award Finalist 2: Compiler-assisted Source-to-Source Skeletonization of Application Models for System Simulation
Event Type
Hans Meuer Award Finalists
Research Paper
Passes
Tags
Exascale Systems
MPI
Networks
Performance Analysis and Optimization
TimeMonday, June 25th4:50pm - 5:30pm
LocationPanorama 1
DescriptionPerformance modeling of network and memory systems through simulation requires application endpoint models that inject traffic into the simulation models. Today, endpoint models for system-scale studies consist mainly of post-mortem trace replay, but these off-line simulations may lack flexibility and scalability. On-line simulations running so-called skeleton applications run reduced versions of an application that generate the same or similar traffic to the full application. These skeleton apps have advantages for flexibility and scalability, but they often must be custom written for the simulator itself. Auto-skeletonization of existing application source code via compiler tools would provide endpoint models with minimal development effort. These source-to-source transformations have been only narrowly explored. We introduce a pragma language and corresponding Clang-driven source-to-source compiler that performs auto-skeletonization based on provided pragma annotations. We describe the compiler toolchain, validate the generated skeletons, and show scalability of the generated simulation models beyond 100K endpoints for example MPI applications. Overall, we assert that our proposed auto-skeletonization approach and the flexibility of skeletons can be an essential building block towards realizing balanced exascale interconnect designs.