TITLE: Technologies for patascale supercomputers and Fujitsu's ...

munchsistersAI and Robotics

Oct 17, 2013 (3 years and 7 months ago)

59 views

TITLE:

Technologies for patascale supercomputers a
nd Fujitsu

s
challenges in moving

toward
s

E
xascale computing


Abstract:

T
he pres
entation will highlight technologies introduced in Fujitsu’s massively parallel
supercomputer series and
the
K computer

by cou
rtesy of RIKEN
.


Fujitsu has been developing SPARC based massively parallel supercomputers, FX1 in 2008,
the
K computer in 2
010, a
nd PRIMEHPC FX10 in 2012.

T
he architecture
development

conform
s

to the following
design approach
;

(1)

S
ingle CPU per node design for high memory
bandwidth

(2)

S
tep by step enhancement for massively parallel
calculations

(3)

Preserving

application compatibility

as much as possible

Through the development, Fujitsu invented fundamental technologies for scalabi
lity,
efficiency, and usability
.

The

technologies include

a SIMD extension for HPC applications
(HPC
-
ACE),

automatic
parallelization

support

for multi
-
thread
ed

execution

(VISIMPACT)
,
a
direct network
based on a

6D mesh/torus interconnect (Tofu)
,

and
hardware support for
collective communication

(
via the
Tofu & its dedicated
algorithms
)
.

We evaluated the technologies using real applications and real machines
.

The evaluations
show that
not a few real applications
perform

good efficiency with over 30% of peak
performance. The level of vector machine efficiency is accomplished by a thorough utilization
of software pipelining and sophisticated tuning t
echniques
us
ing

the Performance Analysis
counters

of the CPU
.

In addition, some real applications

show that

the performance of
hybrid
parallel execution
model
outperforms
that of
flat MPI execution model
.

The
compiler support
of
VISIMPACT technology
transforms

the app
lications to
the
hybrid execution model

with

flexible combination of process
es

and threads
.

Fujitsu’s recent activities and technology development toward Exascale computing

will be
also
introduced

briefly
.


CV:

Mr.

Yuji
Oinaga is a head of the Next
Generation Techn
ical Computing Unit
in

Fujitsu
Limited. His work
ing career at

Fujitsu
start
ed

in development of the CPU of Mainframe
computers and vector supercomputers.
F
rom then, he has been working on development of
all Fujitsu supercomputers as a chi
ef architect of the systems.
He
recently

le
d
the K
computer

development project
. He

currently
is
also

working for the next
generation

supercomputer,
Exascale system.

He
graduated from

Department of Physics
,

School of Science
,

Hokkaido University

in 1974.