►
From YouTube: 0B Welcome Codee
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Thank
you.
Thank
you,
helen.
It's!
It's
really
a
pleasure
to
be
here,
presenting
cody
one
more
time,
so
it's
really
exciting
to
see
so
many
people
interested
in
understanding,
which
is
the
real,
how
how
can
cody
help
with
er
and
what
is
different
in
kodi,
with
respect
to
other
tools
that
we
typically
use
in
the
the
software
development
workflow.
A
So
we
are
really
helpful
that
we're
really
thankful
that
the
doe
through
nerd
and
ocf
are
trusting.
A
penton
has
been
trusting
us
for
for
these
years.
So
really
hope
that
this
is
the
beginning
of
a
journey
to
for
all
of
this
community
to
start
using
kodi,
and
this
only
starts
so
we
really
as
nurse
as
helen
pointed
out.
We
really
think
any
feedback
that
you
can
provide
about.
A
A
So
essentially,
as
helen
pointed
out
today
and
tomorrow,
we
will
be
doing
the
first
training
sessions
of
a
series
of
events
that
are
planned
for
this
year.
So
in
this
first
two
sessions
we
will
begin
with.
A
Let's
call
it
basic
to
introductory
topics
to
cody
and
how
cause
it
can
help
to
optimize
performance
of
codes,
in
particular
in
general,
and
in
particular,
how
it
can
assist
you
in
creating
gpu
calls
enabled,
with
openmp
and
openecc
fragmas,
to
run
on
permuter,
and
this
will
be
the
first
events
of
a
series
that
will
continue
in
september
october.
The
date
is
still
to
be
defined.
A
We
will
cover
more
advanced
topics
from
gpu
development
technique
and
programming
techniques
from
vectorization
techniques
that
are
also
relevant
and
used
very
useful
in
today's
supercomputers,
like
like
promoter
so
today
and
tomorrow,
we'll
be
covering
what
we
consider
basic
topics
and
intermediate
topics
of
the
usage
of
kodi.
So
we
have
split
and
organized
this.
A
We
have
three
hours
ahead
and
we
plan
to
do
a
30
minute
breaks
in
break
in
the
middle,
so
we
plan
to
start
with
this
part,
one
where
we
will
be
basically
introducing
the
coding
tools
and
what
makes
them
different,
and
essentially,
you
can
summarize
most
of
the
capabilities
available
in
kodi
in
with
three
words
shift
left
performance,
so
we
will
see
and
explore
what
shift
left
means
and
how,
for
the
first
time,
kodi
is
enabling
to
shift
left
performance,
and
this
cannot
be
done
with
software.
Only.
A
This
needs
to
be
done
only
with
collection
and
curation
of
coding,
best
practices
for
performance
and
producing
support,
documentation
that
the
tool
can
refer
to
and
that
the
user
of
the
tool
can
inspect
and
read
to
understand,
which
are
the
issues
in
terms
of
performance.
What
is
the
solution
and
how
to
actually
fix
it?
A
So
in
this
first
part
we
will
cover
essentially
shift
left
performance
with
the
help
of
the
kodi
software
installed
on
ersk,
and
essentially,
you
will
also
see
a
quick
walkthrough
of
how
to
optimize
the
very
well-known,
a
p
computation
in
in
perimeter.
A
The
good
news
of
this
is
that
what
you
will
be
seeing
with
pi
with
the
pi
example,
exactly
the
same
sequence
of
steps
is
what
you
will
need
to
use
in
the
following
parts
of
the
training,
for
instance,
in
the
part
2
also
covered.
Today,
we
will
be
addressing
several
gpu
challenges
and
in
particular
one
of
them
is
how
to
identify
parts
of
the
code
that
can
be
actually
offloaded
to
the
gpu
and
how
to
optimize
the
memory
layout
for
data
transfers,
particularly
to
avoid
to
increase
the
performance
on
tomorrow.
A
We
will
also
see
how
to
avoid
defects
or
incorrect
gpu
code,
but
the
good
news.
What
we
are,
what
kodi
is
expected
to
enable
is
a
repeatable,
systematic,
more
predictable
approach
to
performance,
optimization,
instead
of
relying
on
years
or
decades
of
experience
in
performance,
optimization
and
coming
up
with
an
idea
of
what
can
work
for
a
given
piece
of
code.
Cody
provides
a
systematic
approach
that
you
will
see
so
the
exactly
the
same
sequence
and
process
that
you
will
be
using
for
the
pi
example.
A
Finally,
tomorrow,
the
plan
is
also
to
leave
half
of
the
session,
maybe
one
hour
one
hour
and
a
half
for
you
to
work
on
a
more
realistic
example.
Instead
of
working
with
simple
kernels,
very
well
known
like
pi
or
matmul,
we
have
developed
a
simplification
of
the
well-known
knowledge,
a
choral
benchmark
that,
for
the
sake
of
training,
we
have
simplified
some
parts
of
the
code,
but
other
parts
of
the
code
are
exactly
the
same
pieces
of
code
that
are
key
to
parallelize
and
key
to
optimize
for
gpus
in
the
real
clueless
application.
A
Now,
for
this
more
free
part
of
the
session,
we
also
typically
try
to
encourage
users
to
bring
their
own
calls
and
to
try
to
get
started
with
kodi
one
of
the
things
we
have
been
working
very
very
hard
in
this
last
months
of
last
year.
2021
is
in
streamlining
how
to
get
started
with
kodi,
so
we
have
simplified
a
lot:
how
to
interact
with
build
systems,
how
to
interact
with
compilers.
A
So
that
you
can,
we
want
you
to
be
up
and
running,
producing
the
performance,
optimization
report
of
kodi
as
soon
as
possible,
because
that
is
really
where
the
real
value
begins,
where
you
really
start
to
have
insights
about
what
are
the
issues
in
your
code
and
how
to
fix
them
and
which
are
the
actual
solutions
to
it.
So
tomorrow,
for
those
of
you
that
want
to
work
at
or
try
their
own
codes,
we
and
the
team
will
be
supporting
you
through
that
step
or
after
the
feed.
A
After
the
training
session,
you
will
have
these
office
hours
available
for
you
to
make
an
appointment
with
nerska
and
appento
team
to
try
to
see
to
help
you
how
to
use
kodi
with
your
own
real
application
for
those
that
you
don't
want
to,
or
don't
have
available
a
code
to
work.
We
will
probably
be
providing
this
mk
for
you
to
play
with
the
tour.
I
understand
and
apply
the
concepts.
A
So
we
use
the
simple
examples
of
pi
and
matmul
to
introduce
and
present
labs
oriented
to
problem
solving.
We
will
be
focusing
on
one
particular
challenge
that
is
relevant
and
that
in
the
end,
you
will
need
all
of
those
challenges
to
really
optimize
the
performance
and
port
real
applications
to
to
gpus.
So
here
you
will
see
that
for
pi
we
will
be
addressing
one
single
challenge
for
matt
mool.
A
We
will
be
addressing
three
possible
challenges
and
other
two
challenges
are
left
out
of
the
scope
for
this
quick
training,
but
will
be
addressed
through
problem
solving
and
labs
in
the
second
part
of
the
training
later
in
september
and
october,
and
you
will
see
us
mk,
a
microkernel
that
we
propose
to.
You
essentially
has
something
of
all
of
the
challenges
that
we
expect
to
cover
to
take
you
and
guide
you
through
the
journey
of
optimizing
code
to
gpus
okay.
A
So
this
is
essentially
what
we
wanted
to
to
cover
as
part
of
our
welcome
so
just
keep
in
mind
and
bear
in
mind
this
table
with
the
journey
and
with
the
how
we
will
be
presenting
you
with
example,
calls
of
increasing
complexity
starting
from
simple
ones
like
pi
and
matmul,
but
that
already
have
some
gpu.
You
will
be
solving
some
gpu
challenges
that
you
will
really
need
to
address
in
real
applications
or
in
more
complicated
applications
like
mk,
okay,.