Design of MPI Passive Target Synchronization for a Non-Cache- - PowerPoint PPT Presentation

Design of MPI Passive Target Synchronization for a Non-Cache- Coherent Many-Core Processor 27th PARS Workshop, Hagen, Germany, May 5 2017 Steffen Christgau , Bettina Schnor Operating Systems and Distributed Systems Institute for Computer Science University of Potsdam, Germany

Motivation: Distributed Hash Table (DHT) • hash table as cache for computational results in MPI application PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 1 / 14

Motivation: Distributed Hash Table (DHT) • hash table as cache for computational results in MPI application • large amount of data → distribute across processes → DHT PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 1 / 14

Motivation: Distributed Hash Table (DHT) • hash table as cache for computational results in MPI application • large amount of data → distribute across processes → DHT local local local DHT DHT part DHT part DHT part ... rank n − 1 rank 0 rank 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 1 / 14

Motivation: Distributed Hash Table (DHT) • hash table as cache for computational results in MPI application • large amount of data → distribute across processes → DHT local local local DHT DHT part DHT part DHT part ... rank n − 1 rank 0 rank 1 • accessing distributed data: hash function returns arbitrary process and address difficult to program with two-sided message passing MPI passive target one-sided communication to the rescue synchronization required PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 1 / 14

Motivation: nCC Systems • Future many-cores may not provide (global) cache coherence. PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 2 / 14

Motivation: nCC Systems • Future many-cores may not provide (global) cache coherence. Intel Knights Landing: coherent multi-socket systems not feasible https://www.extremetech.com/wp-content/uploads/2016/04/KnightsLanding.png PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 2 / 14

Motivation: nCC Systems • Future many-cores may not provide (global) cache coherence. Intel Knights Landing: coherent multi-socket systems not feasible HPE "The Machine", EuroServer: coherence islands https://regmedia.co.uk/2016/11/22/the_machine_universal_memory_pool_access.jpg PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 2 / 14

Research Platform • nCC many-core research system: Intel SCC 48 Pentium cores with L1/2 caches no HW cache coherence L2$ Core MC 2 MC 3 MIU MPB L2$ Core R Tile MC 0 MC 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 3 / 14

Research Platform • nCC many-core research system: Intel SCC 48 Pentium cores with L1/2 caches no HW cache coherence L2$ Core MC 2 MC 3 MIU MPB L2$ Core R Tile MC 0 MC 1 • This talk: design of synchronization on nCC platform. PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 3 / 14

Agenda MPI Passive Target One-Sided Communication Design for Passive Target Synchronization on the SCC Data Structures and Algorithms Data Placement Outlook and Future Work PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 4 / 14

MPI One-Sided Communication • process memory exposed via windows process ’ address space local DHT part local DHT part local DHT part DHT rank 0 rank 1 ... rank n − 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows process ’ address space local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 • key concept : only one communication partner issues communication operations PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 • key concept : only one communication partner issues communication operations origin processes issue communication operations PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 • key concept : only one communication partner issues communication operations origin processes issue communication operations target processes are addressed by operations PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 • key concept : only one communication partner issues communication operations origin processes issue communication operations target processes are addressed by operations typical RMA operations: PUT, GET, . . . PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI One-Sided Communication • process memory exposed via windows • access to windows with window object (handle) process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) rank 0 rank 1 ... rank n − 1 • key concept : only one communication partner issues communication operations origin processes issue communication operations target processes are addressed by operations typical RMA operations: PUT, GET, . . . explicit synchronization required PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 5 / 14

MPI Passive Target Synchronization • locks as means for synchronization, used by origins only • no participation of targets in synchronization (passive targets) PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 6 / 14

MPI Passive Target Synchronization • locks as means for synchronization, used by origins only • no participation of targets in synchronization (passive targets) • usage similar to shared memory locks WIN_LOCK(win, rank, ...) 1. acquire lock for target window PUT(win, rank, ...) 2. perform operations WIN_UNLOCK(win, rank) 3. release lock PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 6 / 14

MPI Passive Target Synchronization • locks as means for synchronization, used by origins only • no participation of targets in synchronization (passive targets) • usage similar to shared memory locks WIN_LOCK(win, rank, ...) 1. acquire lock for target window PUT(win, rank, ...) 2. perform operations WIN_UNLOCK(win, rank) 3. release lock MPI de fi nes two lock types: shared concurrent accesses on target window allowed exclusive prevent concurrent accesses on same target window PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 6 / 14

Distributed Hash Table with MPI OSC process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) ... rank n − 1 rank 0 rank 1 PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 7 / 14

Distributed Hash Table with MPI OSC process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) ... rank n − 1 rank 0 rank 1 DHT_read LOCK(window_obj, target, SHARED) GET(window_obj, target, &data) UNLOCK(window_obj, target) PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 7 / 14

Distributed Hash Table with MPI OSC process ’ address space window object window object window object local DHT part local DHT part local DHT part DHT (window) (window) (window) ... rank n − 1 rank 0 rank 1 DHT_read DHT_write LOCK(window_obj, target, SHARED) LOCK(window_obj, target, EXCLUSIVE) GET(window_obj, target, &data) PUT(window_obj, target, data) UNLOCK(window_obj, target) UNLOCK(window_obj, target) PARS 2017 S. Christgau (U Potsdam): MPI Passive Target Synchronization 7 / 14