►
From YouTube: SimPEG meeting May 21st
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
A
C
D
E
D
C
Sort
of
I
think
yeah
that
could
be
a
that
could
be
a
good
solution
and
I
said,
and
especially
if
you,
if
you
want
to
I,
don't
know
it's
like
a
depending
upon
that.
The
way
like
if
you're,
using
binder
or
as
or
whatever
cloth
platform
I
mean
that's
okay,
but
if
you
really
want
to
install
their
own
machine,
yeah
I,
don't
know
I'm,
not
sure.
D
C
C
Doh
I
mean
I,
didn't
know
that,
but
anyway
they
gave
a
talk
which
was
interesting
just
like
cloak
one,
but
but
by
the
way,
what
we
were
talking
with
Fernando
was
having
a
project
between
Stanford
and
and
Berkeley,
having
some
sort
of
like
a
cloud-based
platform
in
terms
of
data
management
and
whatever
running
the
simulation.
So
like
a
Pangea
project,
but
I
think
that's
probably
way
to
go,
and
it
seems
like
a
running
a
if
you
have
some
sort
of
cloud
suppose
we
have
money
to
buy.
A
One
thing
we
don't
need
to
don't
read
number
five
we
work
with.
That
was
like
something
it's
a
sure
if
you
work
with
use
of
the
Joseph
of
the
problem.
So
that's
was
the
problem.
We
sneaked
and
the
thing
is
that
right
for
me
it's
the
problem
was
that
is
on
Windows,
so
he's
using
the
three
compile
and
is
easily
dispatched
from
tip.
A
So
it's
a
free
compiled
version
that
work
with
actually
the
newer
version
of
numpy
and
he
was
trying
to
down
random
file
while
working
with
a
disguised
version
that
was
three
compile
with
the
new
verse.
Let's
suppose
the
world
also
so
pip
has
a
few
precompiled
one
two
I
but
that's
what
we
do
is
up
there.
So
we
have
to
do
that
for
Windows
users.
Otherwise,
is
it
not
of
the
c
compiler
code?
It's
not
working
for
them
and
it's
not
by
default
on
Windows
machine.
A
So
that's
why
we
have
to
have
this
free
compiled
version
we
do
like.
Ideally
we
do
it
with
advair,
but
sometimes
we
break
a
barrier.
So
I
can't
actually
me
doing
it
manually
to
approve
that.
But
right
now
a
viper
is
working,
but
the
like
the
fusion
by
magazine
that
change
was
like
a
pretty
cool
challenge.
I
find
this.
If
I
understand
well
so
like,
like
you,
three
compiled
with
numpy
1.16,
you
cannot
use.
A
A
A
C
You
could
you
could,
like
a
you
know
like
a
virtual
box
like
a
like,
you
can
have
a
Linux
system
on
your
window
machine,
oh
so
that
like
it,
can
print
install
everything
into
VirtualBox
and
then
just
wrap
it
up
everything.
So
we
could
basically
provide
that
as
a
zip
file,
then
you
can
download
it
and
run
it.
So
that's
kind
of
old
style,
but
kind
of
works.
Okay,.
D
E
C
F
Yeah
so
I
think
in
doing
these
tutorials
I
was
just
gonna.
Do
a
little
bit
of
a
walkthrough
I.
Guess
we
could
go
screen
share.
Yeah
yeah
I
was
gonna,
show
you
guys
what
I
had
and
maybe
get
a
few
comments.
The
things
that
I
sort
of
wanted
to
do
was
assume
that
they
don't
know
anything,
don't
don't
assume
they
know
any
of
the
terminology.
I
really
want
the
tutorials
to
be
well
commented.
So
I
spent
a
lot
of
time.
F
Easing
people
into
the
terminology.
What
we're
doing
I
wanted
some
practical
examples,
because
if
we,
if
they're
practical,
then
they
can
sort
of
be
adapted
to
a
general
workflow
and
I
want
to
eventually
touch
as
much
of
the
package
as
I.
Possibly
can
so
yeah
I'll
just
shoot
me
over
here,
but
yeah
I've
got
three
sort
of
sets
of
tutorials
right
now,
so
we'll
start
off
at
mesh
generation
kind
of
a
first
step
for
using
this
package
and
then
a
little
bit
on
the
operators.
F
So
averaging
and
differential
operators
like
what
are
those
some
like
how
we
did,
how
do
we
construct
them?
How
do
we
apply
them
and
then
sort
of
them?
The
last
one
that
I
wanted
to
do
was
sort
of
like
a
like
I
solve
your
own
PDE.
So
now
that
we
know
we
have
all
the
building
blocks
like
making
matches
and
how
do
these
inner
product
matrices
things
work
then,
eventually,
you're
gonna
get
to
a
point
where
you've
got
a
PDE,
it's
got
some
boundary
conditions
and
then
you're
gonna
need
to
construct
it.
F
F
E
F
F
We
have
a
little
bit
of
a
seal
on
like
what
it
actually
is
like
a
little
bit
of
an
introduction,
we're
going
to
demonstrate
how
to
create
some
basic
tree
meshes
in
2d
and
3d.
Some
strategies
for
local
mesh
refinement
how
to
plot
the
tree
meshes.
And
how
do
you
extract
some
properties
like
the
volume
or?
What
is
the
the
number
of
edges
in
X
right,
though,
how
we
create
the
tensor
meshes.
Sorry
how
we
create
these
tree
meshes.
We
first
sort
of
define
the
base
tensor
mesh,
and
then
we
refine
in
certain
areas.
F
We
have
a
few
rules.
The
number
of
base
mesh
cells
means
to
all
the
powers
of
two
so
really
thoroughly
going
through
how
you
would
do
all
this,
so
yeah
I
see
first
import
the
packages,
so
we
have
a
basic
example
where
we're
gonna
take
it.
We're
gonna
make
a
quadtree
mesh
2d
and
we're
gonna
go
and
define
the
highest
level
of
discretization
in
a
rectangular
box.
F
And
so
we've
got
each
step
right.
We
comment
it
so
the
minimum
cell
width
of
the
base
mesh
cells,
the
minimum
cell
with
the
number
of
base
left
base,
base
mesh
cells
in
x
and
y.
Then
we
define
our
name
and
create
the
tree
mesh
our
instance
of
the
tree
mesh
class.
Then
we
go
and
define
based
on
this
box
that
we
end
up
getting
a
plot
like
this.
So
that's
the
most
basic
straightforward
example
that
you
could
possibly
get
now.
F
G
F
Yeah
you
have
to
ship
you'd
have
to
shift
those
functions
over
right
yeah.
So
we
do
actually
talk
about
that
and
I
think
the
tensor
mesh
tutorial
where
we
might,
you
could
actually
put
in
like
your
your
bottom
southwest
corner
as
values.
We
also
say
that
there's
some
flags
like
CN
and
zero,
so
we
actually
talked
about
being
able
to
do
that
as
well
right
and
then
I
think
in
the
the
tensor
mesh
example,
because
it's
the
first
example
we
sort
of
start
with
the
the
general
way
that
you
could
go
and
define
this
mesh.
F
And
then
we
do
an
example
afterward
where
we
say.
Oh
there's
a
couple,
little
shorthand
things
that
you
could
do
if
you
want
to
save
space,
and
you
want
to
get
fancy
so
yeah,
you
can
define
the
bottom
southwest
corner
as
being
numbers
or
you
could
use
those
flags
like
C
and
O
for
shifting
the
mesh
okay.
So
we
do
actually
have
that
in
an
example.
F
So
then
we
have
one
about
extracting
mesh
properties,
so
once
the
mesh
is
created,
you
may
want
to
extract
some
properties.
Here
we
demonstrate
some
of
the
things
that
you
can.
You
can
pull
out
so
yeah.
You
could
extract
what's
the
location
of
the
bottom
west
corner,
because
it's
a
2d
mesh,
total
number
of
mesh
cells,
the
cell
center
locations,
which
cells
lie
on
the
boundary
volume
and
then
now
we're
plotting
the
cell
areas
nice.
F
So,
instead
of
digging
plot
grid,
this
is
now
plot
image
and
then
to
have
a
3d
example
as
well.
So
now
we
do
a
full,
a
full
3d
everything
we've
shown
on
the
page
that
we
can
do
extracting
the
mesh
properties.
The
the
mesh
refinement
is
now
3d
points,
and
then
we
take
a
slice
of
the
3d
and
and
see
that
so
that's
sort
of
the
general
flavor
of
these
men,
tutorials.
F
H
A
D
F
Know
I've
thought
about
doing
that,
but
there
is
many
warnings
and
things
that
have
not
been
addressed
in
auto
generation
of
the
API.
Oh,
it's
it
so
there's
there's
a
lot
of
warnings
and
a
lot
of
errors
still
in
the
build
of
this
website
and
those
need
to
be
I
think
addressed
before
we
start
just
adding
a
whole
bunch
of
new
stuff
and
getting
too
excited
okay.
D
E
F
F
F
So
yeah
numerical
solutions
to
differential
equations
using
finite
volume
method
to
require
discrete
operators.
These
include
averaging
and
differential
operators
kind
of
give
a
little
bit
of
a
definition
of
what
that
actually
means
right.
We
might
need
to
average
a
quantity
that
lives
on
one
part
of
the
mesh
to
another
part
of
the
mesh
or
carry
out
these
differential
operators.
So
we
have
one
for
averaging.
F
So
we're
gonna
go
how
to
construct
and
apply
an
averaging
matrix
talk
about
some
differences
between
averaging
matrices
and
one
two
and
three
D.
What
happens
when
you
average
a
discontinuous
function
and
then
talk
a
little
bit
about
the
transpose
and
I'm?
Definitely
hoping
I
can
hear
a
few
comments
on
that,
because
there's
some
details,
I
think
that
are
important.
F
So
in
for
the
packages,
so
we
have
an
easy
one.
Key
example:
where
we're
just
gonna
compute
a
scalar
quantity
on
the
nodes,
1d
example
we're
going
to
average
it
to
sell
centers
and
then
we're
going
to
compute
that
function
at
centers.
That
is
a
way
to
validate
the
averaging
is,
is
numerically
accurate.
So
this
is
sort
of
us
yeah.
It's
a
bit
cut
off
when
I
got
to
fix
it,
but
Liam
are
sparse.
F
F
F
We
get
matrix
that
looks
like
this,
and
in
3d
we
get
a
matrix
that
looks
like
this
for
a
tensor
mesh,
it
would
be
different
to
was
a
tree
mesh
and
then
we
we
do
some
things
about.
Okay,
what's
the
number
of
cells,
the
number
of
faces,
the
dimensions
of
this
matrix
and
the
number
of
nonzero
elements.
G
F
Kind
of
a
model,
and
then
this
is
what
happens
if
you
use
the
cell
centers
two
phases
operator.
This
is,
if
you
we
didn't,
have
all
of
the
averaging
matrices
and
because
I
made
a
uniform
mesh.
If
I
want
to
go,
what
is
it
if
I
want
to
go?
Centers
two
nodes
I
can
take
the
averaging
matrix
from
nodes
to
centers
and
take
the
transpose,
but
I
think
that
only
works.
If
you
have
a
uniform
cells,
you
start
having
padding
cells
or
a
tree
mesh
I.
Don't
think
you're
allowed
to
do
that.
F
F
F
E
F
C
Insides
should
be
should
be
fine,
I
think,
there's
no
dimensions.
Only
thing
is
to
boundary
like
a
cell
to
node
no
to
center,
or
one
is
not
well-defined,
like
a
cell
to
face
face
to
face
cell
is
well-defined
right,
like
you,
don't
need
any
boundary
condition
that
spells
to
face.
You
need
to
put
some
boundary
condition
and
we
dead.
So
that's
like
that's
the
only
change
yeah.
F
Yeah,
can
you
discuss
that
because
yet
you're
effectively
asking
it
to
go
and
use
values
that
are
outside
of
the
mesh
yeah
or
it
basically
assumes
that
anything
outside
is
zero?
Well
from
what
you're
telling
me
this,
this
transpose
like
if
I
have,
if
I
have
padding
if
I'm
using
a
tree
mesh,
then
that
transpose
that
that
works
or
does
it
work
too,
like?
Is
it
still
a
second
or
we're
averaging
I?
Think
there's
some
if
you're
putting
a.
C
F
D
D
F
We're
gonna
do
some
demonstration
so
how
to
construct
and
apply.
These
talked
a
little
bit
about
the
mapping
and
dimensions
so
you're
gonna
apply
some
of
these
operators
to
maybe
a
scalar
quantity
and
you're
gonna
get
a
vector
quantity
out
of
it.
So
wherever
the,
where
do
things
go
and
some
applications
of
that
transpose
again
so
one
the
example.
F
Then
we
used
basically
a
1d
gradient
operator,
so
we
took
the
derivative
with
respect
to
X
and
that's
what
the
dots
are
and
then
we
took
the
analytic
derivative
at
the
cell
centers
and
compared
it
and
and
we
find
that
it
matches
so
to
try
and
validate
the
operator,
talked
a
little
bit
about
mapping
and
dimensions
and
just
sort
of
like
we
I
didn't
really
figure
out.
Maybe
a
better
way
to
do
this,
but
I
decided
to
just
kind
of
make
the
gradient
divergence
and
curl
operators
in
3d
and
kind
of
plot
them
up.
F
So
you
could
sort
of
see
the
relative
sizes
I
mean
I
could
have
invented
some
functions
and
applied
it,
but
this
just
seemed
like
a
good
way
of
demonstrating
what
these
actually
were
their
sparse
matrices.
These
are
the
dimensions
they
have
and
then
printing
it
out.
So
there
is
some
statements
that
say:
okay,
this
is
a
gradient
that
goes
from
yeah
nodes
to
edges,
and
so
there's
the
number
of
nose.
There's
the
number
of
edges
and
lo
and
behold,
there's
the
dimensions,
the
operator,
here's
the
number
of
nonzero
elements.
F
E
F
F
Then
we
have
a
vector
living
on
the
faces
and
we
take
the
divergence
of
that
and
we
get
this
and
then
we're
gonna
have
a
sort
of
a
rotational
field
at
the
edges
and
we
take
the
curl
of
that,
and
this
is
what
we
get
Excel
centers,
because
it's
a
2d
example
right.
So
if
we're
going
we're
gonna
go
from
from
edges
to
faces.
Well,
that's
just
the
cell
centers
of
TD
right
you're,
taking
the
curl
of
a
function,
that's
dependent!
Why
you're
just
gonna
get
the
set
component!
F
F
C
Like
that's,
how
we're
testing
the
all
the
gradient
operators,
if
you
can,
you
can
take
a
look
at
how
we're
testing
and
there
are
like
testing
examples
so
like
it,
maybe
even
just
for
1d
case.
Rather
you
can
plot
it
out,
but
they're,
showing
one
order.
Test
example
will
be
beneficial.
I
think
you
I
thought.
F
About
it,
but
it's
it's
a
tutorial
more
for
how
to
use
it
so
yeah
that
statement
of
saying
these
are
second-order.
Differential
operators
I
think
is
sufficient.
They
don't
they.
That
would
be
useful
information
to
know,
but
we're
supposed
to
have
made
the
operators
and
they
should
be
correct.
They
just
just
need
to
know
how
to
use
them
and
know
kind
of
a
bit
about
what
they
are.
Okay,.
C
Think
in
the
docks
of
not
sure
actually
that's
a
good
question,
because
I
actually
get
your.
D
C
F
Okay,
top
yeah,
so
in
the
last
part,
is
the
solving
PDEs,
so
yeah
any
comments
you
want
to
give
on
this.
I
have
a
feeling
it's
99%
correct,
but
I
would
like
to
get
that
up
to
100,
okay,
so
yeah.
These
were
sort
of
the
four
steps
that
I
sort
of
thought.
If
you
were
starting
from
scratch,
you'd
have
you
have
your
PDE
you'd
want
to
formulate
the
problem
and
it's
boundary
conditions?
F
I
know
how
everything
works
before
this
and
now
I
need
to
solve
my
particular
problem.
Can
they
go
through
this?
This
kind
of
create
your
own
process
and
be
able
to
kind
of
customize
like
solve
their
particular
situations,
so
I
decided
to
do
Gauss's
law
of
electrostatics
and
then
the
advection
diffusion
equation,
because
the
operators
we
have
kind
of,
let
me
do
this.
F
So
in
all
of
these
examples,
yeah
we're
gonna
state
we've
got
sort
of
a
2d
charge
distribution.
You
can
have
a
couple
of
point
charges
and
then
we're
gonna
go
and
try
and
get
the
electric
potential
and
then
go
and
get
the
electric
fields.
So
we
start
with
Gauss's
law
in
Faraday's
law
and
Faraday's
law,
because
it's
static
will.
Let
us
write
the
electric
fields
in
terms
of
the
scalar
potential
and
then
we
have
our
boundary
condition.
F
F
We
just
found
a
homogeneous
physical
properties,
there's
going
to
be
no
charge
buildup
on
on
boundaries.
So
beyond
the
faces
is
fine,
we're
not
violating
any
like
natural
boundary
conditions
of
the
cells
yeah.
And
so
then
we
do
our
finite
volume
approach
by
a
valley.
Maybe
I'll
evaluate
these
inner
products
and
then
we
can't
rule
out
any
of
the
terms
and
we
end
up
with
a
system.
F
So
we
put
a
little
charge
distribution
at
minus
10
and
win
it
10
zero.
We
create
our
mesh,
create
the
system,
define
our
right
hand,
side
and
in
the
end,
we
get
something.
That's
like
this.
So
we've
got
a
little
positive,
dent
charge,
density
on
the
right
and
one
little
negative,
one
on
the
Left.
We
have
positive
potential
and
your
the
positive
charge,
and
then
we
see
electric
fields
go
from
positive
to
negative.
F
F
F
So
we
picked
sort
of
a
not
really
fancy
version
of
the
infection
diffusion,
so
we've
got
a
non
compressible
fluid
and
this
diffusion
coefficient
or
function
is
just
a
constant.
In
this
case.
We're
gonna
have
a
closed
system,
so
this
variable
P
that
we're
solving,
for
maybe
it's
like
the
concentration
of
salt
in
the
water
or
something
we're
gonna.
Have
the
flux
of
that
on
the
boundary
is
equal
to
zero,
so
nothing
leaves
or
comes
into
the
system.
That's
sort
of
the
natural
boundary
condition
for
this
at.
C
The
F,
hey,
Dad,
yeah
I,
think
the
the
DC
one
there's
a
couple
of
tricks
for
the
weak
form
and
ok,
you
need
to
do
an
integration
by
part.
I
totally
do
that.
Did
you
obviously
did
it
and
then,
like
I,
felt
like
it
that's
a
little
bit
what
you
were
doing
in
in
in
the
code
and
the
and
the
like,
and
the
equation
wasn't
quite
matching
so
in
what
you're
doing
in
the
code
is
more
like
a
finite
difference.
Yeah
yeah!
So
it's
not
like
we're.
Not
gonna
use
this.
C
So
the
whole
idea
about
the
weak
form,
there's
only
one
like
a
differential
operator.
So
here
in
this
case,
is
a
divergence,
so
we're
not
using
quite
gradient.
So
gradient
is
like
a
divergence
transpose
and
some
sort
of
nice
matrix.
So
it
like
a
there's,
there's
a
little
bit
of
detail
about
that
I
can
I
can
help
later
yeah.
That
would
be
good,
yeah,
yeah,
I
think
I'm
Shep,
some
I
have
some
notes,
so
I
think
that'll
be
nicer
because
it's
not
like
I
not
really
want
one
at
the
moment.
F
C
F
F
F
Yeah
yeah
I'm
not
sure
where
it
got
sort
of
buggered,
but
I
mean
I'd,
yeah,
I
didn't
I,
did
take
the
inner
product,
I
used
trig
identities,
and
then
you
get
this.
The
I
did
use
the
divergence.
Theorem
seems
like
pretty
standard
practice
when
doing
this,
and
because
the
flux
leaving
or
coming
into
my
system
is
zero.
I
was
under
the
impression
that
this
term
is
zero.
Well,.
C
F
C
F
F
I
F
F
E
G
F
We
did
the
2d
problem
and
in
the
end
it
looked
like
it
was
pretty
legit.
So
we've
got
a
couple
sources
at
early
times.
It's
all
kind
of
concentrated
around
the
sources
and
yeah.
So
we
have
a.
We
have
a
velocity
field,
that's
going
to
the
bottom
right,
so
you
can
actually
see
that
the
materials
diffusing
and
that
it
wants
to
go
to
the
bottom
right
and
at
the
source.
F
That's
in
the
bottom
right
when
it
hits
the
boundary
I'm,
hoping
you
can
kind
of
see
that
it's
collecting
there,
because
it
can't
leave
because
the
boundary
condition
says
that
there's
no
flux
that
can
leave
or
enter
the
system
so
I
don't
think
visually.
It's
the
best
plot,
but
first
glance
it
looked
like
it
was
doing
what
it
was
supposed
to.
But
are
you
putting
like
a
two
sources?
Yeah,
okay,
yeah
yeah,
well,
one
source
in
the
in
the
top
right
and
then
one
source
in
the
bottom
right.
F
So
one
one
thing
was
just
to
show
we're
getting
diffusion
and
we're
also
getting
this
infection
going
to
the
bottom
right.
And
then
another
thing
I
wanted
to
show
is
that
we
actually
did
invoke
boundary
conditions
and
that
we
are
getting
sort
of
a
collection
of
material
in
the
bottom
right
corner,
because
it
can't
leave.
I
Devon,
if
you
sure,
if
you
show
this
simulation
for
another
ten
or
twenty
seconds
or
so,
would
there
be
more
material
that
collects
bottom
right
works
it
already
a
steady
state
by
five
seconds.
It.
F
F
G
F
F
That's
that's
basically,
it
so
I
mean
I
have
some
stuff
that
maybe
I
want
to
I'd
like
to
talk
to
sake
Lindsay
about
in
terms
of
the
process
of
adding
new
examples,
because
you'd
like
to
encourage
people
to
add
stuff.
But
if
it
breaks
the
generation
of
the
website,
master
is
broken
and
then
any
time
you
go
on
branch
off
of
master,
that's
broken
that
it
leads
to
problems.
These
dissing
problems-
and
you
know,
there's
still
some
basic
functionality
and
discretized
that
hasn't
been
put
in.
F
We
can
put
in
neumann
boundary
conditions
for
tensor
meshes,
but
well,
I,
don't
think
we're
able
to
put
boundary
condition,
anointment
boundary
conditions
for
tree
meshes
net,
or
maybe
in
3d,
even
on
2d,
because
I
wanted
to
do
one
of
these
examples.
I
want
to
not
just
do
all
the
examples
on
on
tensor
meshes
I
wanted
to
mix
it
up
a
bit
so.
E
C
Do
you
define
your
potential,
for
instance?
So
if
you
are
using
an
old
gradient,
a
natural
boundary
condition
is
national
norm,
and
so
that's
what
we're
doing
at
EC
so
we're
basically
using
like
a
solving
model
problem
to
set
up
the
anointment
boundary
condition
for
DC
problem,
for
instance.
So
if
you
have
an
idea
and
like
a
yeah,
I
can
definitely
help.
So
let
me
know
yeah
if
some
kind
of
idea
of
extension
or
what's.
F
Missing
I
guess
I
just
wanted
to
know
we're
designing
this
package
to
be
like
this
is
something
kind
of
more
specifically
tailored
to
us
doing,
e/m
problems
or
was
the
vision
of
this
particular
package
like
a
general
like
a
way
that
we
can
solve.
General
PD
is
using
finite
volume
because
you
know
we
have
a
weave
the
curl
operator
that
goes
from
edges
to
faces,
because
that's
what
we
need
to
solve
our
a.m.
problems,
but
we
haven't
divine
to
find
a
curl
operator
that
goes
from
faces
to
edges.
D
F
F
Yeah
cuz
I'm
I,
understand
that
is
like
natural
I
understand
that
each
differential
operator,
whether
or
not
you're,
going
from
like
notes
to
edges
or
edges
to
notes
or
whatever
it
is.
There
is
kind
of
a
natural
boundary
condition
that
you
can
put
in
each
of
those
using
the
ghost
points,
but
that
that
list
of
all
the
operators
that
map
between
all
of
the
different
places
and
having
all
of
the
boundary
conditions
like
that,
it's
an
incomplete
list
right
now
we
got
everything
we
need
to
solve.
B
F
C
G
G
B
G
G
A
G
G
G
G
D
B
G
I
think
the
important
part
here
is
that
kind
of
go
from
A
to
Z
here
and
that
well,
the
most
important
parts
is
that
you
actually
identify
places
where
you're
a
little
unsure
about
you
know
better
than
like
okay
guys
here
it
is
bring
it
up.
Let's
make
sure
you
kind
of
go
by
on
little
things
so
that
any
event,
okay.
F
I'm
thinking,
yeah
I'm
thinking,
maybe
wet
soggy
was
getting
at
was
we
have
we
have
this
module
called
the
like
inner
product
where
it
basically
will
boil
some
take
something
like
this,
and
it
will
give
you
sort
of
the
mass
matrix
that
you're
supposed
to
get
from
that
to
a
numerically
evaluate
the
the
inner
product
on
on
a
single
cell
and
then
extend
it
to
all
cells.
So
it's
sort
of
I
think
it's
it's
this
step
right
here
that
he's
sort
of
talking
about
and
kind
of
see
where
he's
going
with
it,
and
it.