Upload
philippa-parsons
View
216
Download
0
Embed Size (px)
DESCRIPTION
Outline –Charm++ designed for portability between shared and distributed memory –Optimizing multicore charm K-neighbor and its description and performance What optimizations were carried out –Abstractions: Basic: shared object space, Readonly data Plain global variables: still work.. More on disciplined use of these later Nodegroups Passing pointers to shared data structures, including sections of arrays. –Readonly, write-exclusive: permissions by convention or “capability” 1/26/2016Charm++ Workshop3
Citation preview
Programming an SMP Desktop using Charm++
Laxmikant (Sanjay) Kale
http://charm.cs.uiuc.eduParallel Programming LaboratoryDepartment of Computer Science
University of Illinois at Urbana ChampaignSupported in part by IACAT
Prologue• I will present an abbreviated version of the planed
talk– We are running late..– Also, I realized that what I really intended to present, with code
examples, will need an hour long talk..– We will write that in a report later (may be put it in charm
documentation)
05/03/23 Charm++ Workshop 2
Outline– Charm++ designed for portability between shared and distributed
memory– Optimizing multicore charm
• K-neighbor and its description and performance• What optimizations were carried out
– Abstractions:• Basic: shared object space, Readonly data• Plain global variables: still work.. More on disciplined use of
these later• Nodegroups• Passing pointers to shared data structures, including sections
of arrays.– Readonly, write-exclusive: permissions by convention or “capability”
05/03/23 Charm++ Workshop 3
Optimizing SMP implementation of Charm++• Changed memory allocator
– to avoid acquiring a lock per memory allocation
• Reduced the granularity of critical region• Used thread local storage (__thread) to avoid false
sharing• Use memory fence instead of lock for pcqueue• Reduce lock contention by using a separate msg
queue for every other core on the same node• Simplify the data structure of pcqueue
– Assumes queuesize is adequately large
05/03/23 Charm++ Workshop 4
Results on SMP Performance• Improvement on K-Neighbor Test (8 cores, Mar’2009)
Results on SMP Performance• Improvement on K-Neighbor Test (24 cores, Mar’2009)
Results on SMP Performance• Improvement on K-Neighbor Test (16 cores, Apr’2009)
0 64 128 256 512 1024 2048 40960.00
200.00
400.00
600.00
800.00
1000.00
1200.00
1400.00
1600.00
1800.00
2000.00 kNeighbor test on a Power 5 node (15 elements, k=3)
non-smp smpmsg size (bytes)
aver
age
itera
tion
time
(us)
05/03/23 Charm++ Workshop 8
We evaluated many of our applications to test and
demonstrate the efficacy of the optimized SMP
runtime
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160
0.2
0.4
0.6
0.8
1
1.2
Number of processors
Effi
cien
cyJacobi 2D stencil computation on Power 5
(8000x8000 matrix size)
05/03/23 Charm++ Workshop 10
ChaNGa: Barnes-Hut based
production astronomy code
05/03/23 Charm++ Workshop 11
ChaNGa: Barnes-Hut based
production astronomy code
NAMD Scaling with Optimization
0
0.5
1
1.5
2Time/ste
p (
s)
Number of Cores
Multicore vs net
multicorenon-smp
multicore 1.715 0.6 0.303 0.152 0.0826non-smp 1.719 0.631 0.321 0.173 0.11
1 3 6 12 24
NAMD apoa1 running on upcrc
05/03/23 Charm++ Workshop 13
Summary of constructs that use shared memory in
Charm++
Basic Mechanisms• Chares and Chare array constitute a “shared
object space”– Analogous to shared address space
• Readonly globals– Initialized in main::main or any method called from it
synchronously
• Shared global variables
05/03/23 Charm++ Workshop 14
More powerful mechanisms• Node groups• Passing pointers to shared data structures,
including sections of arrays.– Readonly, write-permission
05/03/23 Charm++ Workshop 15
16
Node Groups• Node Groups - a collection of objects (chares)
– Exactly one representative on each node• Ideally suited for system libraries on SMP
– Similar to arrays:• Broadcasts, reductions, indexing
– But not completely like arrays:• Non-migratable; one per node
Conditional packing
• Pass data structure between chares– Pass pointer (dest. within the node)– PUP the entire structure (dest. outside the node)
• Who owns the data and frees it?– Data structure must inherit from CkConditional
• Reference counted
• A data structure can contain info about an array section– Useful in cases like in-place sorting (e.g. quicksort)
Sharing Data and Conditional packing• Pointers can be sent in “messages”, but they are
packed to underlying data structures when going across nodes– (feature in chare kernel since 1989 or so!)
• Data structure being shared should be encapsulated, with a read or write “capability”– If I give you write access, I promise not to modify it, read it, or
grant access to someone else– If I give you a read access, I promise not to change it until you
are done
05/03/23 Charm++ Workshop 18
Disciplined Sharing• My pet idea: shared arrays with restricted modes
– Readonly, write-exclusive, accumulate, and “owner-computes”– Modes can change at well-defined global synch points– Captures a large fraction of uses of shared arrays
05/03/23 Charm++ Workshop 19