publication . Conference object . 2019

global optimization of operand transfer fusion in heterogeneous computing

Kessler, Christoph;
Restricted
  • Published: 22 Oct 2019
  • Publisher: ACM Press
Abstract
We consider the problem of minimizing, for a dataflow graph of kernel calls, the overall number of operand data transfers, and thus, the accumulated transfer startup overhead, in heterogeneous systems with non-shared memory. Our approach analyzes the kernel-operand dependence graph and reorders the operand arrays in memory such that transfers and memory allocations of multiple operands adjacent in memory can be merged, saving transfer startup costs and memory allocation overheads.
Subjects
ACM Computing Classification System: Hardware_ARITHMETICANDLOGICSTRUCTURES
free text keywords: Computer science, Global optimization, Dataflow, Program optimization, Distributed memory, Kernel (linear algebra), Operand, Memory management, Parallel computing, Symmetric multiprocessor system
Related Organizations
Funded by
EC| EXA2PRO
Project
EXA2PRO
Enhancing Programmability and boosting Performance Portability for Exascale Computing Systems
  • Funder: European Commission (EC)
  • Project Code: 801015
  • Funding stream: H2020 | RIA
Communities
FET H2020FET HPC: Transition to Exascale Computing
FET H2020FET HPC: Enhancing Programmability and boosting Performance Portability for Exascale Computing Systems
Powered by OpenAIRE Open Research Graph
Any information missing or wrong?Report an Issue
publication . Conference object . 2019

global optimization of operand transfer fusion in heterogeneous computing

Kessler, Christoph;