►
From YouTube: 02 Practical PI
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
Okay,
so
these
are
the
use
cases
that
we
use
so
now
what
we
propose
is
we
are
have
provided
you
with
attack,
to
explain
or
to
guide
you
step
by
step
to
the
process
of
optimizing
pi
that
you
have
seen
in
the
live
demo.
So
essentially,
we
are
suggesting
that
you
produce
an
open,
acc
and
openmp
versions
for
gpu
and
that
you
build
it
and
run
it
on
per
motor
with
a
problem
size
that
is
big
enough
to
see
some
seconds
of
execution.
A
Essentially
the
pi
example.
If
we
go
back
to
the
one
of
the
first
slides
we
presented
one
hour
ago,
is
addressing
one
of
the
challenges
that
is
see
how
we
can
use
kodi
to
identify
opportunities
in
the
code,
essentially
loops
that
can
be
offloaded
to
the
gpu,
so
the
code
card
runs
correctly
and
faster
on
the
gpu
okay,
you
have
also
seen
the
code.
This
is
the
source
code,
a
simple
example
to
illustrate
this
use
case.
A
You
have
also
seen
use
cases
one
and
two,
how
to
produce
the
report
number
one
and
how
to
use
the
actions
report
to
actually
annotate
your
code
with
directives
and
benchmark
that
code
on
kodi
of
sorry
on
permuter,
and
here
we
also
provide
you
in
the
deck,
with
all
the
steps
you
need
to
do
so.
Please
copy
the
lab
example.
The
example
codes
in
your
home
account
whole
directory
enter
into
the
directory
about
the
p
lab
and
follow
these
steps.
You
just
need
to
copy
and
paste
the
commands.
A
You
will
see
in
this
terminal
and
you
should
be
able
to
reproduce
the
same
output
that
you
are
seeing
in
the
screen.
So
we
suggest
that
you
start
with
the
pw
report.
With
the
entry
level
higher
level
report
and
note,
multiplying
and
of
loading
are
disabled
by
default.
This
has
a
usability
reason.
We
have
seen
users
trying
to
use
kodi
on
their
own
laptops,
for
instance,
trying
to
exploit
offloading
capabilities,
but
not
being
aware
of
how
this
this
operating
system
and
the
computer
needs
to
be
set
up
with
a
gcc
or
a
clan
compiler.
A
That
is,
that
supports
openmp,
open
sec
of
loading
capabilities.
So
without
being
aware
of
this,
they
blamed
on
kodi
not
working
properly,
but
in
reality
because
it
was
working
properly,
but
the
setup
of
the
computer
was
not
correctly
set
for
the
type
of
capabilities
the
user
wanted
to
explore.
So
in
order
to
help
the
user
the
programmer
to
be
aware
to
be
conscious
of
what
needs
to
be
prepared
in
the
applique
in
the
system
to
use
different
types
of
capabilities,
we
decided
to
disable
material
offloading.
A
So
you
need
to
explicitly
enable
those
reporting
capabilities
in
kodi,
and
how
can
you
do
that
with
the
include
tax
flag
in
pw
report?
Okay,
this
is
a
way
of
filtering
the
output
of
the
actions
reported
by
the
tool.
Here
we
are
requesting.
Please
report,
all
all
the
actions
everything
you
find
in
my
code.
A
Step
number
three
is
increase
the
level
of
reporting
abs.
Sorry,
where
you
will
see
this
level
of
complexity
in
real
cause,
you
can
have
tens
or
hundreds
of
loops.
So
what
you
really
want
to
start
is
with
low
hanging
fruit.
You
want
to
start
with
the
hottest
loops
that
consume
most
of
the
runtime.
The
hottest
spots
and
from
those
loops
can
be
one
or
can
be
10
loops.
A
You
want
to
start
with
those
that
have
a
low
complexity
in
terms
of
performance,
optimization
and
wait
or
defer
to
later,
using
or
addressing
those
that
has
a
high
complexity
during
the
labs
today
and
tomorrow
we
can
provide
you
more
details
of
how
low
medium
or
high
difficulty
is
actually
determined
by
cody.
But
this
is
not
the
purpose
of
this
first
lab
next
step
produce.
A
The
actions
report
increase
the
level
of
reporting
to
see
all
the
details
about
a
particular
action
copy
and
paste
the
suggested
commands
to
produce
gpu
enable
code
using
openmp
gpu
enable
call
using
openhcc
and
finally,
in
the
labs.
We
also
provide
you
the
scripts,
how
to
launch
and
how
to
build
and
run
the
different,
the
sequential
version,
as
well
as
the
optimized
versions
that
you
have
produced
in
previous
steps,
and
here
we
provide
you
with
default
reservation
kodi
day.
A
One
is
the
full
reservation
for
today,
but
please
check
double
check
the
welcome
slide
from
presented
by
helen,
because
maybe
you
need
to
use
additional
flags
to
the
reservation.
You
also
need
to
use
the
entrain
to
group
for
the
reservation,
so
please
double
check
in
the
in
the
launch
script
that
you
are
using
the
appropriate
options
to
use
the
nodes
reserved
for
the
training
today
and
tomorrow.
A
A
A
You
are
working
on
the
code
and
you
need
to
guarantee
that
it
runs
first
correctly
and
second,
faster
and
final
remark:
don't
forget
to
take
a
look
at
the
kodi.com
knowledge
database,
it's
public,
it's
open!
It
is
extended
to
provide
you
intended
to
provide
you
with
the
knowledge
that
the
experts
have
in
mind
when
they
are
optimizing,
their
own
codes,
so
that
you
can
learn
from
the
experts
or
you
can
even
being
an
expert,
recycle
yourself
and
learn
more
in
other
areas
of
expertise
that
you
have
not
addressed
in
your
career.
For
some
reason.
B
So
yeah
check
out
the
nurse
welcome,
slides
the
read
me
in
the
exercise
directory
and
just
do
the
lab2
right
now,
so
you
can
do
pw
all
these
commands
and
do
everything
on
the
command
line
and
then,
at
the
end,
after
you
generate
the
openmp
and
open
acc
code
as
batch
and
that
launch
dot
sh
script
so
that
it
will
launch
your
compile
and
run
on
the
compute
on
the
gpu
compute
nodes
also
we'll
be
using
the
reservation.
It's
already
added
included
in
the
launch
script.
B
So
once
you
copy
over
the
exercise
from
the
global
cfs
directory
and
and
the
best
way
to
do.
This
is
also
follow
the
slide
2
deck.
It
has
step-by-step
examples
and
command
lines
on
multiple
slides,
so
follow.
Those
will
be
very
helpful.