►
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
Yeah,
cool,
okay,
so
I'm
Mike,
Levy
I'm,
a
software
engineer
in
the
oceanography
section
at
incar
honestly,
giving
a
status
report
on
adding
marble
initially
to
the
CSM
implementation
mom.
But
but
eventually
we
hope
to
feel
this
back
to
the
Standalone
lawn
as
well.
I've
had
lots
of
help
from
lots
of
people
that
have
named
blow
and
I
somehow
managed
to
leave
alpair
off
of
this
list,
but
but
he
also
provided
lots
of
input
along
the
way.
A
So
this
is
a
slide
from
2017
just
an
overview
on
what
marble
is.
It's
a
software
library
at
biochemistry
tracers,
Ocean
Models.
It's
got
a
modular
design,
so
it
can
be
it's
independent
of
an
egcm.
Everything
is
done,
column
by
column.
So
so
we
don't
have
horizontal
operators.
A
The
project
started
in
the
fall
of
2014,
with
the
goal
of
sharing
the
top
speed
GC
code
with
the
amps
ocean,
funded
by
a
doe
Grant,
just
in
case
you're
curious
about
the
timing
that
the
CSM
acne
split
I
first
got
wind
of
it
in
Spring
of
2014.
So
this
was
this
was
after
after
that
split
was
was
known
to
be
becoming,
but
I
think
MSO
was
kind
of
seen
as
the
front
runner
for
for
csm-3
for
the
ocean
model,
and
then
we
had
this
goal
in
an
Ideal
World.
A
So
so
the
marble
design
principles
were
open.
Development
on
GitHub
and
I've
provided
a
link
to
our
GitHub
repository
when
we
use
it.
So
we
use
Fortran
for
all
of
our
computation,
and
since
this
is
supported
by
this
is
part
of
CSM.
We
need
to
support
a
wide
range
of
compilers,
so
I
think
Intel,
PGI,
g4train
and
nag
are
the
four
that
we
test
with
we're
using
python
for
for
tools
just
the
scripts.
A
Well,
we
provided
a
standalone
driver
for
for
testing
and
that's
all
managed
by
my
python
emotes
for
train
code.
We
have
tools
for
generating
the
parameter,
settings,
files
and
things
like
that.
That's
all
it's
all
python
based
and
we
also
use
GitHub
to
host
our
documentation.
So
there's
a
link,
our
GitHub,
I
o
page
everything
we
do
should
be
column,
wise
and,
and
we
have
some
requirements
that
we
impose
on
the
GCM.
A
So
so
the
GCM
will
do
all
the
I
o
for
us
all,
the
parallelization,
the
time
stepping
there's
some
data
that
marble
needs
from
one
time
step
to
another
and
and
the
GCM
is
used
to
say
that
for
us,
I
won't
go
into
detail
about
it,
but
but
we
call
it
saved.
State
and
I
might
mention
it
in
a
later
slide
and
then
there
are
some
cases
where
we
do
need
horizontal
operators
and
the
GCM
needs
to
do
that
so
right
now.
The
only
example
is.
If
we
want
variable.
A
We
we
use
consistent
units
in
Marble,
we're
currently
using
CGS,
because
this
code
originated
in
in
pop,
which
is
a
which
uses
CGS,
but
what
we
do
have
an
issue
to
get
open
to
look
into
switching
to
MKS
and
then
this
this
picture
I
think
Matt
long
put
this
in
the
initial
proposal
for
for
marble
back
in
in
2013
or
2014.,
and
it's
it's
kind
of
a
good
overview
of
of
what
we're
we're
doing
so
so
marble
is
the
BGC
model.
A
It's
this
box
down
here
and
then
the
base
model
just
is
the
ocean
model,
and
then
we
introduced
a
a
driver
layer
that
that
belongs
to
the
ocean
model
like
it's
ocean
model
code
and
it
kind
of
does
the
translation
between
Marvel
and
and
the
ocean
model.
So,
for
this
talk
pretty
much
everything
I'm
going
to
talk
about
is
in
the
BGC
driver
box
and
then
I
did
have
to
make
small
changes
in
the
in
Mom's
Tracer
flow
control
to
call
the
driver
code.
But
everything
else
is
in
its
own
module.
A
So
I
guess
Keith
mentioned
this
very
briefly.
In
his
talk
about
the
marble
threads,
it's
32,
IOD
chemistry,
tracers,
there's
an
optional
carbon
isotope
module
that
would
add
14
more
tracers,
it
computes
surface
fluxes
such
as
air,
cco2
flux,
chlorophyll
concentration
and
computes,
Source,
sync
terms
for
the
tracers,
although
the
GCM
needs
to
supply
them
and
then
there's
the
Standalone
marble
for
for
some
small
test
cases
and
that's
separate
from
the
code,
that's
run
by
by
the
gcms
in
space
Oh
Yeah,
a
small
driver
for
for
testing
the
shared
code.
A
So
Marble
Runs
in
in
kind
of
three
three
stages
of
an
initialization,
the
set
parameters
and
Define,
tracers
and
Diagnostics.
Then
every
time
step
we
compute
surface
fluxes
and
interior
tendencies
from
the
sourcing
terms
and
after
after
each
of
those
computation
stages,
we
return
a
set
of
Diagnostics.
So
there's
Diagnostics
related
to
the
surface,
flux,
computation
and
Diagnostics,
related
to
The,
Source,
sync
computation
and
then
at
the
end,
we
we
shut
down.
We
clean
up
memory
space
and
make
sure
that
you
see
them
is
received
the
final
one.
A
So
this
is
the
kind
of
the
short.
How
do
we
add
Marvel
to
mom
I've
broken
it
up
into
two
steps:
the
the
infrastructure
updates
in
and
the
Fortran
updates
on
the
infrastructure
updates.
A
It
was
Israeli,
I,
guess
CSM
mom,
but
we
need
to
create
a
concept
that
will
turn
on
the
BC
tracers
by
default.
You
need
to
interact
with
marble,
Python
scripts,
to
determine
parameter
settings
and
and
what
diagnostic
kits
output
I
introduced.
A
new
parameter
in
the
Tracer
flow
control
use
marble
tracers.
So
we
only
add
the
tracers.
A
If
this
is
true,
and
by
default
it's
it's
false
unless
you're
running
with
one
of
the
concepts
that
turns
it
on
at
the
request
of
gfdl
I,
I
added
a
compile
time,
macro
unders
are
used,
marble,
tracers
and-
and
we
only
build
if
we
compile
with
this
with
this
macro,
so
we
don't
distribute
marble
with
with
Mom
but
it'll
be
brought
in
automatically
in
the
CSM
version
of
mom
and
it's
available
freely
on
GitHub
for
the
Standalone
mod
and
then
we'll
we'll
need
to
update
the
mod
diagnostic
package
to
to
handle
the
marble
output
on
the
Fortran
side.
A
These
are
kind
of
the
the
big
steps
in
my
mind,
it's
not
all-inclusive,
but
we
need
to
read
the
marble
parameters
from
from
just
because
that's
how
Marvel
wants
to
be
configured,
we
we
need
to
call
a
marbles
initialization
routine,
and
we
do
that
from
call
Tracer
register.
A
A
Right
now,
it's
only
passively
affecting
all
the
marble
tracers
we're
not
applying
the
surface
flux,
door,
Source
sync
terms
and
marble
Diagnostics
can
be
added
to
the
output,
but
but
it's
not
in
a
user-friendly
way
right
now,
it's
very
clergy
and
obviously
that'll
get
cleaned
up
as
as
this
work
progresses
and-
and
we
do
write
marble
log
messages
into
the
CSM
blog
files,
so
just
just
to
have
a
picture
to
break
up
with
the
blocks
of
text
on
the
left
is
the
initial
nitrate
field
at
the
surface
that
we
read
in
the
non-initial
condition
file,
then,
after
a
month
of
being
passively
infected
around
it's
different.
A
A
So
so
you'll
need
to
a
little
bit
more
details.
I
think
all
of
my
testing
is
in
the
CSM
framework.
That's
the
primary
focus
for
the
initial
implementation
and
then
once
this
is
once
it's
working
in
CSM
it'll,
hopefully
be
easier
to
to
go
back
and
apply
it
outside
of
the
CSM
framework,
I'm
going
to
avoid
showing
any
any
code
Snippets.
A
But
all
the
code
is
available
in
GitHub
for
the
CSM
implementation,
so
so
there's
mom
six
source
code,
then
there's
also
the
mom
interface
that
sits
between
CSM
and
Mom
and
I
needed
to
update
both
of
both
of
those
layers
to
include
marble.
A
I'm
I'm,
not
sure
if
these
will
be
the
tags
in
csm2
or
if
there's
gonna
be
one
more
set
of
updates
and
then
I'm
trying
to
to
color
code.
The
progress
on
the
following
slides,
so
things
that
are
highlighted
in
green,
are
done
in
yellow
they're,
in
progress
in
red
they're
not
yet
begun,
and
if
the
colors
look
too
close
together,
there's
also
the
icons
the
check
mark,
the
construction
or
a
not
yet
done.
A
I'm
pretty
new
to
PowerPoint
I
couldn't
figure
out
how
to
change
the
bullet
points
themselves
to
just
depended
those
on
the
end,
all
right.
So
on
the
infrastructure
side,
we
build
marble,
and
this
is
done
through
csm's
case
that
build
script
and
it
which
calls
mom
interfaces,
build
web
script
and
that's
done
use.
Marble.
Tracers
is
in
the
Imam
parameter
list.
A
The
the
default
value
is
false
unless
you
specify
an
rcbgc
config
that
is
not
none
and
so
I
created
the
c1850
mom
Eco
comset
in
CSM
that
sets
ocean
BDC
config
to
latest,
which
uses
the
most
up-to-date
tunings
that
we
have
in
progress
building
mom
with
with
the
used
marble
tracers
macro.
A
So
right
now,
my
my
seam
sandbox
just
has
an
update
to
include
this
flag
on
a
really
old
mom,
but
on
my
list
of
things
to
do
eventually
will
be
to
talk
to
our
parent,
also
the
same
developers
about
what
Behavior
we
actually
want.
So
I
I
could
see,
keeping
it
to
that
CSM
always
builds
mom
with
marble
and
keeps
marble
as
a
runtime
configuration.
A
You
can
turn
it
on
or
off
at
will,
but
maybe
we
want
to
mimic
kind
of
Pops,
dough
time
configuration
and
and
have
a
flag
that
turns
the
Dual
macro
on
or
off.
Instead,
that's
that's
a
discussion
for
for
down
the
road.
It's
not
not
too
important
right
now,
also
also
in
progress
or
providing
initial
condition
and
surface
forcing
fields.
A
So
there's
a
a
notebook
in
the
in
the
marble
forcing
repository
that
that
we
use
to
generate
the
initial
condition
files
some
about
Fields
come
from
World
ocean
Atlas.
Some
come
from
glad
app.
Everything
else
comes
from
a
csm21
pi
control
run
for
steam
up
six.
So
so
here's
surface
nitrate
on
the
native
I
think
I
think
this
was
the
world
ocean
atmospherd
and
then
here
it
is
interpolated
onto
the
the
mom
grid.
A
Again
and
I
guess
I
should
have
mentioned
this.
This
is
all
using
the
the
two-thirds
degree
grid
that
we've
been
developing
CSM
CSM
mom
with
some
surface,
forcing
Fields
will
eventually
need
to
be
written
from
files,
but
for
those
I'm
currently
hard-coding
reasonable
values.
Instead
of
that,
instead
of
creating
files
and
figuring
out
how
to
read
them
in,
there
are
some
fields
that
come
from
the
model
and
those
are
being
populated
correctly,
like
temperature
and
salinity
are,
are
Paxton
from
from
Mom
six.
A
Also
in
progress
is
tying
into
Marvel's,
generate
settings
file
script
to
provide
runtime
parameters,
so
the
CSM
build
nml
script
will
generate
a
marble
in
file,
but
first
there's
there's
no
way
to
modify
the
values
that
get
written
into
that
and
secondly,
it
doesn't
matter
if
you
can
modify
those
values
because
there's
no
Fortran
code
to
read
the
file
in
and
pass
it
on
to
Marble.
So
so
that's
I
guess
maybe
it's
a
stretch
to
say,
though
that's
in
progress
in
in,
but
that's
where
that
seems
I
haven't
done.
A
Any
work
to
tie
marbles
generate
Diagnostics
file
script,
which
can
be
used
to
provide
a
list
of
diagnostic
output
into
the
system
so
far,
I've
just
hard-coded
specific
tracers
and
one
diagnostic
into
the
diag
table.
Yeah
I'm
a
file
and
then
as
as
I,
develop
I'll,
continue
to
add
Diagnostics
by
hand
as
as
we
need
it
and
I
won't
try
to
connect
to
the
marble
infrastructure
until
we're
actually
happy
with
where
the
Fortran
driver
is
and
then
I
haven't
done.
A
On
the
Fortran
status,
it
kind
of
follows
nicely
and
that
how
does
Marble
Run
slide?
That
I
previously
shown
there's
some
stuff
done
in
initialization
and
some
work
to
be
done.
I'm,
mostly
focusing
on
the
per
time
step,
work
right
now
and
and
the
next
slide
I'll
go
into
a
little
bit
more
detail
and
and
I
guess,
I
think
I
mentioned
so
far,
but
in
Source
Trace
here.
A
So
when
you,
when
you
turn
on
when
you
set
that
used
marble,
tracer's
flag
to
true
mom
6
adds
32
tracers
to
to
it's
list
and
it
initializes
them
and
it
backs
them
around
the
hooks
between
the
marble,
Diagnostics
and
Mom.
Six
are
there
on
the
Fortran
side,
so
every
diagnostic,
but
Marvel
can
return,
is
reported
to
Mom
and
Tracer
flow
control
in
it.
A
So
in
theory,
you
can
add
any
diagnostic
to
the
diag
table
and
and
it'll
be
spit
out
in
in
practice,
we're
still
not
Computing,
all
of
them
so
I'm
not
making
all
the
calls
into
marble
that
I
need
to
make
so
in
in
progress.
Computing
the
surface
fluxes
are
are
getting
there.
All
the
calls
to
Marble
are
in
place.
A
Some
forcing
fields
are
properly
Filled
from
from
memory.
There
are
other
forcing
fields
that
need
to
be
read
from
disk
or
retrieved
from
the
coupler
and
I
still
need
to
tell
Mom
what
to
do.
With
the
surface.
Fluxes
I
mean
I
feel
like
it's
a
little
bit
cheating
to
have
that
just
be
one
bullet
point,
because
that's
a
lot
of
work,
but
but
it's
not
not
there
yet,
and
then
a
performance
tweak.
Marble
can
compute
flexes
for
multiple
columns
simultaneously,
but
for
like
a
simple
first
implementation,
I'm
just
calling
it
column,
icon.
A
Also
in
progress,
though
so
I've
been
allocating
memory
for
for
things
that
that
mom
needs
to
hold
on
to
for
marble
kind
of
as
we
go,
and
so
so
there's
some
steps
that
that
I
haven't
gotten
to
you
yet
since
I
have
an
allocated
memory
for
them.
Yet,
as
I
mentioned
before,
where
we
can
write
all
the
diagnostics
from
the
surface
from
Computing
the
surface
fluxes,
but
the
interior
tendency
computation
is
not
called,
and
so
those
Diagnostics
are
being
accumulated
and
the
marble
log
is
being
written
to
the
csm.log
file.
A
But
I
don't
think
it's
actually
the
right
place
for
that.
So
at
some
point,
I'll
move
it
to
a
more
appropriate
output
file
and
then
still
to
do
is
reading
the
marble
win
file
to
get
the
marble
tuning
the
the
parameter
settings
and
passes
by
some
marble
I'm
starting
to
call
into
your
tendency
compute.
It
should
be
similar
to
the
column
sequence,
for
what
we've
done
with
the
surface
fluxes
depending
the
vertical
coordinate,
will
be
a
little
bit.
A
Tricky
and
that'll
actually
require
an
update
on
the
marble
side
as
well,
because
right
now
marbles
used
to
pop
just
setting
the
vertical
coordinate
for
for
every
column
during
initialization
and
then
running
it
running
it
be
so
so
handling
the
vertical
coordinate
that
changes
in
space
and
time
will
will
require
an
update
to
the
marble
system
just
to
have
an
interface
to
do
that.
A
So
so
again
the
same
disclaimer
about
all
the
output.
But
this
isn't
scientifically
useful,
but
it's
kind
of
a
proof
of
concept.
But
this
is
the
CO2
piston
velocity
diagnostic
that
gets
computed
when
we
compute
the
surface
fluxes
and
it
was
accumulated
over
over
a
month
because
that's
been
the
length
of
the
test
runs
that
I've
done,
and
it
was
nice
to
see
spatial
variants
in
in
this
plot.
But
when
I
was
first
running,
I
was
setting.
A
A
So,
just
to
summarize,
the
marble
driver
is
still
a
work
in
progress.
Most
of
this
infrastructure
is
in
place
for
CSM,
for
the
CSM
mom
Mom
plus
marble
without
CSM
will
be
supported
and
I'll
do
that
work,
but
but
it's
it's,
it
hasn't
been
started
yet.