►
From YouTube: Analog Workgroup Meeting May 3, 2022
Description
AWG Meeting with presentation of programmatic analog design by Marko Kusunen and Mirjana Videnovic-Misic
A
A
B
A
That
is
well,
I
underst.
I
understand
that,
and
I
certainly
appreciate
that
I
know
I
know
instructing
is
a
lot
of
work.
A
F
F
D
D
D
Yeah
yeah,
you
know
because
it
is,
it
is
kind
of.
Let
me
see,
download,
because
I'm.
D
F
I
A
We'll
wait
a
couple
minutes
because
I
do
expect
more
folks
to
join
us
and
we
did
marco
mariano.
I
did.
I
did
get
a
number
of
folks
interested
in
the
meeting
from
the
post
thing
I
did
on
linkedin
under
chip's
line,
so
I
thought
that
was
good.
Okay,
cool.
D
D
Yeah
yeah,
it's
actually
interesting.
More
people
are
kind
of
finding
us.
I
think
this
linkedin
posts
help.
D
So
I
mean
more
people
who
are
doing
similar
things
with
generators.
I'm
kind
of
optimistic
that,
especially
with
this
initiative
that
we're
having
right
now
and
marco
and
me,
are
part
of
it
christoph
also
that
will
hopefully
break
the
ice.
You
know
and
create
the
critical
mass
of
people
yeah.
A
I
think
I
think
you're
getting
to
that
point
and,
like
you
know,
like
I
said
I
was
surprised,
pleasantly
surprised.
You
know
at
the
interest
that
I
got
on
linkedin.
You
know
which
thanks
marco
for
suggesting
you
know
trying
to
get
some
additional
socialization
for
this.
So
I
think
that
I
think
that's
great.
A
A
You
know,
including
including
different
collaboration,
project
ideas
and
also
looking
at
improving
automation
in
the
analog
space,
which
of
course,
is
always
a
challenging
arena,
and
today
we're
gonna
have
an
update
on
programmatic,
analog,
ic
design
that
marco
cassinin
and
mariana
is
gonna
update
us
a
lot
of
most
is
gonna
update
us
on
today
and
the
work
that
they
have
been
championing
and
helped
to
try
to
move
the
community
forward
on
that
area.
So
with
that,
why
don't
I
turn
over
to
you,
marco
and
mariana.
D
G
G
Yeah,
okay,
good
there
we
go
yeah,
so
I'm
gonna
give
up
in
the
group
mosaic
back.
This
is
like
the
main
main
group
for
this:
our
fork
of
berkeley,
animal
generator
and,
like
template
environment,
to
use
it
use
it
in
a
relatively
efficient
manner.
G
There
is
one
main
project
under
this
group,
which
is
called
virtual,
so
template,
and
there
is
also
automatic,
continuous
integration,
build
of
a
documentation
of
this
project
containing
online
documentation
of
all
the
sub
modules
or
the
main
sub
modules
in
it,
and
I
briefly
show
you
the
documentation.
This
is
automatically
built
every
time
we
push
to
the
latest
development
branch,
so
it
has
certain
some
introduction
introductory
test
text
and
describes
all
the
modules
what
we
have
there.
G
So
as
this
is
open
source
project,
what
we
decided
with
miriana
at
quite
early
phase
is
that
we
should
have
a
kind
of
well-structured
development
plan
or
release
schedule
for
this.
So
I
wrote
a
brief
developer's
guide
for
this
environment
and
we
decided
that
we
have
do
two
release
dates
per
year,
fourth
or
fourth
and
tenth
of
10th.
They
are
easy
to
remember
and
then
some
instructions
about
branching
and
how
to
do
merge,
requests
for
for
these
releases
and
so
forth.
But
I
think
this
is
kind
of
the
backbone
of
the
development.
G
So
if
anyone
of
you
tries
the
environment,
find
something
that
needs
to
be
improved,
feel
free
to
file
in
a
merge
request,
and
we
or
someone
will
check
that
out
and
and
if
it's
okay,
we
will
merge
it.
If,
if
there
are
some
feedback,
then
we
will
provide
the
feedback
and
eventually
we
will
converse
to
the
better
environment.
G
G
G
The
aim
of
this
template
is
that
we
provide
provide
a
set
of
shell
scripts
that
you
can
easily
use
to
take
this
environment
into
use.
We
have
pip
installation
script
for
to
install
all
the
python
dependencies.
You
need
to
run
this
this
environment.
We
have
idiot
sub
modules
that
initializes
all
the
github
modules
in
this
project.
So
you
don't
need
to
worry
about
what
to
initialize
and
how
it
it
will
do
it
for
you.
G
We
have
readme
file,
of
course,
licensing,
and
then
we
have
two
source
me
scripts.
One
is
for
c
shell
and
derivatives,
and
one
is
for
bash,
and
basically
this
source
will
script.
Miriana
will
later
on,
show
you
how
it
works,
but
it
forces
you
to
set
things
up
correctly.
So
basically
you
start
in
all
of
it.
G
Let's
just
briefly
show
what
it
has.
So
it
has
a
configure
script.
You
run
dodge
class
configure
and
make
gen,
and
it
will
generate
the
layout
and
schematic
of
this
inverter
for
you.
So
it's
a
really
a
minimum
minimum
working
example
that
just
shows
that
your
environment
has
been
set
up
correctly
and
it
works,
and
you
can
start
extending
this
inverted
generator
or
you
can
generate.
G
We
have
a
javascript
that
you
can
use
to
generate
a
new
generator
template,
which
is
actually
it
has
a
functionality
of
inverter,
but
it's
easy
to
start
modifying
it
for
your
needs,
so
hey
marco.
Can
I
ask
a
question.
A
Sure,
yes
yeah,
I'm
just
curious.
Does
this?
Does
this
create
the
the
primitive
schematic
and
layout
in
open
access,
or
is
it
through
hooks
into
virtuoso,
which
then,
of
course
eventually
goes
to
open
access,
but
I'm
just
kind
of
curious
how
the
mechanics
of
that
work.
G
G
Yeah,
so
the
this
is
using
using
virtual,
so
scaling
interface
to
virtue
also,
but.
G
So
if,
if
I
go
now
outside
the
inverted
gen,
we
use
these
back
framework
bag
templates,
they
are
actually
the
back
framework
back
to
frameworks
from
uc
berkeley.
So
they
support
open
access
through
some
interface
layer
but
psybag
oa.
G
But
you
can't
build
that
without
access
to
these
open
access,
c
headers-
and
we
don't
so-
we
haven't
tried
that.
But
in
theory.
G
G
Through
through
what
we
have
here
back
to
environment
settings,
it's
basically
a
collection
of
collection
of
scripts
that
are
common
common
for
all
the
back
environments,
so
we
stored
them
in
the
ones
up
module
in
order
to
be
able
to
propagate
the
changes
effectively
to
any
of
the
projects
using
this
framework,
so
you
can
just
update
the
sub
module.
G
If
you,
if
you
need
to
get
updates
for
those
those
setting
files,
then
we
have
back
to
framework
that
I
said
already,
it's
just
a
fork
from
using
berkeley
back
to
back
to
templates.
This
is
basically
a
sub
module
that
contains
analog
base
and
certain
examples,
also
from
ford
from
a
music
virtual
github
site
and
back
to
technology
template.
This
is
new
production.
G
It's
a
collection
of
virtual
sorry,
abstract,
abstract,
python
classes.
That
tells
you
all
the
properties
that
are
needed
to
define
pros
as
parameters
for
this
environment.
So
let
me
go
back
to
this
release
candidate,
so
I
can
show
you
the
latest
version
of
this.
So
if
I
go
through
this
back
to
technology
templates,
it
contains.
G
Okay,
if
I
go
back
first
so
basically,
this
template
is
something
that
you
can
fork
and
start
building
your
technology
definition
on
this
template.
There
are
instructions
how
to
do
this,
so
if
I
show
you
the
main
class,
it
tells
you
that
okay,
to
use
this
as
a
back
technology
definition
class
comment
the
following
lines
and
then
uncomment
the
following
lines
and
so
forth.
So
you
can
convert
that
this
from
actual
template
to
actual
definition
and
then
what
you
then
need
to
do.
G
You
need
to
figure
out
the
process
parameters
and
those
are
collected
mainly
these
sub
modules,
layer
parameters
in
most
parameters,
resistant
parameters
and
via
parameters,
and
these
are
just
abstract
properties
that
define
the
properties
that
we
use
in
the
technology.
So,
as
these
are
now
properties,
instead
of
like
entries
in
a
yaml
file,
we
can
use
dog
strings
to
document
the
purpose
of
these
properties,
which
is
relatively
handy.
Unfortunately,
for
many
of
these
parameters,
we
actually
don't
know
their
actual
function
or
what
how
they
really
affect
your
layout.
Some
of
them
are
clear.
G
G
So
you
can
also
use
that
as
a
example.
How
how
to
set
is
to
set
this
up.
But
the
reason
why
we
use
this
is
that
when
we
intron
you
introduce
a
new
parameter,
we
can
introduce
it
as
an
abstract
property
and
then
your
actual
technology
definition
inherits
this
class,
this
template
class
and
because
this
abstract
method,
it
forces
you
to
redefine
this.
So
in
your
process,
definition,
which
I
show
next,
you
must
define
values
for
these
parameters,
and
so
I
hope
that
the
idea
of
this
technology
template
is
now
kind
of
clear,
clear.
G
G
It
has
a
property
process
config
that
has
all
these
process
process
variables
or
process
parameters
store
stored.
Indeed,
and
values
of
those
parameters
are
defined,
for
example,
in
this
layer
parameters,
the
pi
which
inherits
the
layer
parameters
template.
So
this
template
tells
what
you
need
to
define,
and
here
we
have
defined
all
the
parameters
that
we
must
overload
in
this
class.
So,
for
example,
this
says
that
the
name
of
the
technology
library
is
sirius
ffmpt
layout
units
are
one
micron
layout
resolution
is
one
thousandth
of
micron,
meaning
one
nanometer
somewhere
yeah,
okay,.
A
I
started
to
trouble
you
with
another
question,
but
now
this
is,
this
is
good.
You
know
the
underlying.
A
Understanding
required
here,
of
course,
is
being
able.
You
know,
basically
understanding
how
to
do
abstract
based
programming
or
you
know,
in
other
words,
class-based
programming
like
you
know
what
was
originally
seen
in
c,
plus
plus
or
an
objective
c.
You
know
years
back.
I
just
was
curious
relative
to
what
are
you
guys,
thoughts
in
terms
of
you
know,
layout
or
circuit
design,
engineers,
understanding
this
and
taking
it
forward?
A
G
This
is
very
good
question,
I
would
say
I'm
48
and
I
was
able
to
do
it,
but
okay
circuit
designers,.
G
I
would
say
that
if
you
think
of
larger
bigger
organization
where
your
server
designers
remain
as
certain
designers,
they
still
need
to
know
the
basics
of
object-oriented
programming
configuring.
This
framework,
configuring
these
processes
in
a
larger
organization.
I
would
leave
those
for
like
technology,
bdk
supper
personal
personnel,
which
should
be
relatively
fluent
with
python,
but
it's.
B
G
It's
it's
possible
to
write
the
generators
without
advanced
knowledge
in
programming.
It's
also
possible
to
maintain
these
processes
without
any,
but
very
special
information
about
programming.
Just
basic
concepts
are
enough,
but
if
you
are
strictly
in
analog
design
strictly
schematic
entry
strictly
in
manual
layout,
there
is
certain
learning
curve
ahead.
J
Okay,
yeah,
may
I
comment
here
somewhat.
Yes,
what
I've
seen
is
that
young
design
engineers
circuit
designers,
they
are
open
for
doing
this,
not
all
of
them,
but
but
many
or
some
of
them.
What
is
more,
tricky
is
to
convince
layout
people
because
they
like
to
do
graphic,
use,
graphic
user
interfaces,
and
it's
it's
obvious
that
doing
layout.
This
is
a
very
graphical
thing,
so
these
people
are
are
more
reluctant
to
switch
and
there's
a
good
reason.
D
May
I
also
comment
here:
rob
yeah.
What
I
believe
is
that
you
know
with
our
joint
effort
and
especially
open
source
community.
We
need
to
create
a
new
new,
let's
say
breed
of
engineers.
J
D
And
you
know
where
automation,
reuse
goes
hand
in
hand
with
ic
design,
and
there
is
no
question
about
it.
So
how
can
we
do
it
only
by
you
know
putting
something
to
get
and
making
this
onboarding
process
for
people
who
wants
to
join,
and
you
know
for
the
new
curriculum
etc,
making
it
easier
and,
like
kristoff
said.
D
I
also
seen
you
know
in
our
organization
very
experienced
designers,
they're,
very
reluctant.
That
is
like
mild
way
to
say,
but
I
believe
that
you
know
the
young
students
just
went
from
the
university.
They
are
much
more
kind
of
susceptible
to
new
methods
and
and
basically
they
should
be
at
this
stage.
Maybe
our
target
group
yeah
thanks.
A
I
appreciate
both
of
the
comments
now
I
agree
with
that.
I
was
just.
I
was
not
trying
to
start
an
insurrection
here.
I
guess
with
my
question,
but
mostly
just
you
know,
basically
just
based
on
my
own
experience.
I
worked
with
analog
folks
over
the
years
and
you
know
I
think,
in
terms
of
you
know
as
boris,
and
I
have
chatted
many
and
I've
chatted.
A
You
know
chatting
with
borah
as
well
and
others
right
that
we
certainly
need
to
look
at
improving
the
overall
productivity
and
reproducibility
of
results,
and
automation
certainly
helps
doing
that
and
also
aids
in
terms
of
technology.
Migration
as
well-
and
I
think
embracing
more
modern
concepts
in
terms
of
how
we
can
do
things
in
design
is,
is
important
and
it
also
will
help
help
us
attract
newer
students
right
and
enabling
different
ways
of
thinking.
So
thank
you.
D
Rob
I
I
would
like
to
say
maybe
only
after
marco
finishes
and
we
finish
with
this
demonstration
about
the
plans.
You
know
after
this
let's
say:
first
first
effort
that
we
put
in,
but
let's
marco
continue,
maybe
with
with.
F
Okay,
I
don't
know
I
do
have
a
couple
questions
for,
for
you
guys,
so
I
thought
the
post
from
borah
who's
using
backtree
to
do
sorry
dc.
So
how
is
this
framework
different
from
factory
and
what
is
the
differences?
If
you
may
help
me
understand.
G
Yes,
this
is
based
on
back
two,
which
is
like
pre-predecessor
of
factory,
and
the
reason
we're
using
this
is
that
with
both
me
and
miriam,
we
have
experience
on
using
back
too,
but
more
experience
of
using
back
too,
and
I
to
my
understanding
battery
is
actually
not
as
open
as
this
is
so
I
would
say
that
yeah.
This
is
it's
easier
to
in
the
future.
If
seem
feasible,
it's
easier
to
migrate
from
this
to
back
three
but
start
developing
the
needed
needed
skills
and
start
cleaning
up
this
environment
with
back
too.
F
I
see
so,
are
you
saying
back
to
it's
just
relying
on
some
go
ahead
mirrors.
D
So
I
think
then
you
can
correct
me,
but
you
know
when
I
was
comparing.
I
never
worked
with
back
three.
There
are
like
a
couple
differences.
The
major
is
that
back
three
does
not
use,
let's
say
skill
and,
like
let's
say
skill
procedures,
to
execute
something
you
know
in
vertozo,
but
does
it
directly
under
the
hood,
with
c
plus,
plus,
okay
and.
F
D
So
I
think
if
you,
if
you
see
like
just
compare,
start
underscore
bag.l
in
back
two
and
back
three
and
you
can
see
that
you
know
a
lot
of
things
are
just
gone.
F
D
And
so
that,
and
that
exactly
part
cannot
be
because
of
this.
Si2
limitations
cannot
be,
let's
say,
open
source
source
cannot
be
shared,
and
I
think
you
know
that
is
the
major
obstacle
of
spreading.
You
know
back
three
in
performance
but
boris,
saying
like
it's
10
times,
you
know
fasting
simulation,
which
is
expected
and
execution
which
is
also
expected
and-
and
you
know,
and
they
kind
of
figured
out
that
back
to
has
a
limitation
when
they
have
been
designing
bigger
socs.
But
you
know,
maybe
then
you
can
add
something.
F
K
My
audio
is
not
working.
Can
you
guys
hear
me
now?
Yes,
yeah
what
I
said
it
is
more
or
less
right,
the
at
least
as
far
as
why
you
wouldn't
want
to
make
that
upgrade.
Is
it
primarily
jumps
from
python
code
that
manipulates
virtuoso
to
python
code
that
manipulates,
open
access
and
spectre
directly
I'd,
say
that's
half
of
like
the
major
goals
of
the
change.
K
The
other
half
is
kind
of
production
of
all
of
the
other
stuff
that
goes
with
the
analog
ip,
so
production
of
verilog
models,
production
of
timing
models
and
the
like-
and
I
would
say
much
of
the
work-
went
into
that
and
yeah
a
lot
of
the
shifts
to
directly
using
that
oa
c
plus
plus
api
was
based
on
having
you
know
sufficiently
large
generators
that
going
through
that
python.
This
skill
loop
really
did
turn
into
like
many
hour
or
even
many
day
run
times.
D
Yeah,
if
I
may
add
just
one
more
thing,
you
know,
for
instance,
in
back
two
we
use
analog
base
and
there
is
digital
base,
which
is
kind
of
you
know
additional
methods
a
bit.
You
know
more
flexibility,
but
if
I
remember
correctly,
baxley
uses
mass
space,
so
it
was
like
iteration
and
and
change
you
know,
and
they
in
my
opinion
they
figure
it
out
and
as
they
go,
they
change.
D
You
know
the
methods
trying
to
how
to
say
optimize
the
the
scripting
of
the
layout,
because
at
the
end
of
the
day,
scripting
of
the
layout
is
extremely
manual
process,
and
I
will
say
something
about
it
at
the
end
not
directly
to
this
yeah.
K
Yeah,
I
would
yeah,
I
would
say,
that's
probably
even
more
true
than
you
said:
there
are
more
kind
of
programming
model
changes
to
how
layout
gets
constructed
and
such
and
just
reaching
back
to.
How
can
you
convince
people
to
write
a
layout
as
a
program
rather
than
you
know,
draw
it
in
a
gooey?
K
I
I
don't
think
that
the
answer
of
they
can
code.
Therefore,
they
can
code,
a
layout
really
cuts
it,
and
we
have
plenty
of
experience
of
that
attitude
really
failing.
K
I
think
it
really
does
boil
down
to
just
what
is
the
mental
model
and
what
is
kind
of
the
interface?
What
are
the
ideas
that
the
library
injects
to
help
someone?
You
know
program
one
of
these
layouts
and
there
are
a
lot
of
the
opposite.
How
detailed
there
are
they
are
versus
how
accessible
they
are,
and
I
think
this
has
generally
drawn
a
very
detailed
line
that
you
know
it's
had.
A
pretty
niche
set
of
people
be
able
to
latch
on
to.
F
Cool,
so
so,
basically
it
so
it
is
a
way
of
drawing
layout
using
you
know,
coding
or
software,
and
do
you
guys
are
you
guys
planning
to
have
like
a
good
way
to
see
inter
interactively
how
your
code
is
changing
your
layout
or
or
do
you
have
to
compile
and
then
go
check
the
gds
and
see
what's
happening
like
what
is
the
process.
G
We
show
the
process
in
the
end
of
this
presentation.
Okay,
it
shows
itself
so,
okay,
yeah.
These
are
valid
remarks
of
over
my
comparison
of
backstreet
and
bacteria
and
they're
highly
evaluated.
G
On
my
side
side,
I
can
say
that
I
tried
to
construct
something
that
gives
lowest
possible
threshold
to
get
your
hands
on
in
generating
layouts
and
schematics
through
python,
because
that's
already
quite
a
huge
step
for
at
least
our
students
in
the
university
that
I'm
working
and
for
that
reason
also,
we
we
have.
G
We
have
kept
simulations
and
even
drc
and
lvs
checking
out
of
the
scope
at
the
moment.
So
we
know
that
in
every
organization
you
want
to
do
it
in
your
own
way.
So
what
we
provide
is
the
way
to
programmatically
control,
your
schematics
and
layouts.
If
you
want
to
simulate
them,
you
can
generate,
of
course,
the
netlist
and
positive
simulation
environment,
but
it's
kind
of
up
to
you
how
you
want
to
do
it.
For
example,
in
other
universities,
we
use
something
that
is
called
the
sidekick,
which
is
a
python
interface
for
various
analog
simulations.
G
G
G
So
the
idea
is
that
every
generator
is
a
git
sub
module
inside
this
project.
This
is
just
a
project
template,
but
if
you
fork
this
and
push
it
to
the
other
repository
pocket
becomes
your
project
and
you
can
add
generators
as
a
git
sub
module,
so
they
will
believe
they
will
live
here.
In
parallel
with
this
university
and
in
this
main
project
we
have
a
configure
file.
G
When
you
run
it,
it
will
create
a
make
recipe
for
all
of
your
generators.
So
if
I
want
to
run
all
the
generators
inside
this
project,
I'll
just
write
make
gen
and
it
will
go
through
all
the
generators
from
them
one
by
one.
G
If
I
want
to
run
only
one
generator,
I
can
give
make
gen
inverted
gen
and
it
will
run
the
generator
also
for
only
for
inverted
j,
so
this
is
kind
of
a
top
level
control
of
generators.
I
don't
take
at
this
point
any
any
stand
on
how
you
pass
the
parameters
to
those
generators
you
can
down
that
in.
You
can
do
that
in
various
ways.
G
G
When
you
run
the
configure,
it
will
create
a
make
file
and
then,
from
that
make
file,
you
can
write,
run,
make
gen
and
it
will
run
the
generator,
and
this
idea
of
using
the
same
structure
over
and
over
is
copied
from
from
kind
of
linux
environments.
Everyone
who
has
installed
linux
programs
from
source.
You
can
do
that
without
reading,
even
the
readme,
because,
most
often
it
goes
that
you
write
configure
it,
make
it
make
install
and
it
will
install
the
software
for
you.
G
If
that
doesn't
work,
then
you
need
to
read
actually
the
readme.
So
if
you
keep
the
structure
constant
for
all
certain
policy,
you
don't
need
to
document
it.
There
is
no
variation
if
the
generator
runs
in
the
same
way
and
also
the
structure
of
the
generator
all
of
the
generator
is
to
say
we
have
templates.
G
So
the
name
of
this
circuit
that
we
are
generating
is
inverter.
The
name
of
this
generator
is
inverted
gen
and
it
has
a
directory
with
the
same
name.
So
all
the
python
leads
here
and
the
templates
schematic
templates
for
this
inverter
they
live
here
called
inverted
templates.
If
you
have
op
amp,
this
would
be
op
amp
templates.
G
G
You
can,
if
you
want,
create
a
method
so
that
you
can
pass
these
properties,
for
example,
from
the
yaml
file,
or
you
can
use
this
inverter
inside
another
generator
and
pass
these
parameters
from
the
higher
level
or
higher
level
of
your
design
to
the
lower
level.
So
you
can
automate
this
parameter
fast
as
you
wish.
This
is
just
python.
G
We
can,
of
course,
in
the
further
development,
also
figure
out
ways
how
to
automate
this
parameter
provocation
throughout
the
hierarchy,
but
what
we
wanted
is
an
environment
where
you
can
use
doc
strings
to
document
your
actual
parameters
and
what
they
do
and
in
every
generator.
You
have
this
main
guard
which
actually
defines
what
you
do
so
the
generation
process
has
now
been
condensed
to
the
three
lines
of
god.
G
And
but
this
is
the
basic
structure
structure
and
this
is
enabled
with
one
helper
class,
which
is
in.
G
So
basically,
we
created
a
class
called
back
design
and
it
has
a
init
method
on
certain
methods
that
are
executed
when
you
actually
instantiate
this
generator
and
basically
what
it
does.
It
imports
the
schematic
from
the
schematic
templates
to
virtue,
also
and
back
then
imports
certain
python
definition
of
your
schematic
to
certain
locations.
So
it
replaces
that
important,
empty
file
with
your
actual
schematic
generator
so
that
you
can
run
it
from
your
generated
location,
but
this
is
kind
of
a
mess
under
the
hood.
G
G
Okay,
so
I
have
gone
all
the
structure
of
this
virtuous
template
now.
Do
you
still
have,
or
do
you
have
more
interesting
questions
that
I
could
probably
answer
or
elaborate
the
structure
for
you?
If,
if
not,
then
it's
miriam
is
trying
to
show
actually
how
this
thing
works.
D
G
Questions,
I
also
I
I
may
apologize
that
I
know
that
it's
not
the
most
interesting
thing
to
go
through
a
hierarchy
of
a
project
that
has
git
sub
modules,
but
anyway,
I
think
this
is
the
easiest
way
to
explain
what
is
there
and
how
they
are
organized
and
how
what
is
their
relations
and
so
forth.
So
then,
the
next
step,
if
you
want
to
try
this,
is
just
to
pull
on
pull
and
try
to
run
it
in
your
organization's
environment.
A
Now,
mark
I
appreciate
you
going
through
the
details
of
the
repository
and
you
know
showing
relative.
You
know
relative
to
a
you
know
a
simple
example
such
as
an
inverter,
how
this
is
put
together,
orchestrated.
G
A
You
know
one
common
theme
I
keep
hearing
here
is
that
you
know
it
seems
that
cadence
is
a
bottleneck
for
us
in
some
ways
in
terms
of
furthering
open
source
development.
C
F
Is
this
process
agnostic
right,
like
can
you?
How
are
you
loading,
a
new
pdk
or
how
are
you
managing
yeah.
G
This
is
a
good
question.
What
I
did
in
our
university.
I've
tested
this
with
two
different
process
configurations
and
you
pull
in
this
template.
You
do
the
trick
for
the
source
me
file
so
that
it
takes
your
virtual
environment
into
use.
That's
of
course,
what
organizes
specific,
and
then
you
replace
this
back
technology
definition
with
the
back
technology
definition
that
you
want
to
use
and
the
generator
runs
without
any
problems.
So.
F
Yeah,
how
does
the
definition
look
like
right,
like
do
you
take
in
the
drc
rules
or
like?
How
is
how
is
it
gonna
switch
from
a
finfet
to
sky
water
130,
or
something
like
that.
G
That
is
a
good
question.
Currently
you
need
to
define
all
these
parameters
manually
or
let's
say
I
don't.
I
don't
have
any
way
to
instruct
you
how
to
figure
out
this
number.
I
can
just
tell
you
what
it
means
because
of
the
like
mds
of
the
processes.
If
you
want
to
set
this
up,
let's
say
tsmc28,
I
probably
have
the
setup
for
the
process.
I
can't
give
it
to
you
got
it.
I
can't.
I
can't
tell
you
any
detail
about
that
process.
F
Okay,
so
another
question,
so
let's
say:
let's
assume
I
have
a
platform
where
I
have
all
these
setup
files
for
different
technologies.
So
then,
if
I
port
an
inverter,
it's
gonna
follow
up
the
the
same
rules
or
like.
Is
it
gonna,
have
a
standard,
sl
dimension
or
is
it
just
gonna
be
an
inverter
functionality.
G
It
will
it
will
follow
the
rules
given
in
these
technology
definition
files,
all
the
dimensions
given
in
the
in
the
inverted
gen
generator.
They
are
relative,
so
basically
we
we
give
them
as
multiples
of
routing
reach,
which
is
process
dependent
or
multiples
of
minimum
channel
length,
which
is
also
like
just.
D
D
Okay,
so
start
from
this
return
template,
I
don't
have,
let's
say,
a
flexible
working
environment.
So
basically
my
it
provided
me
something
like
you
know
this
terminal
where
I
can
just
clone.
You
know.
So.
Basically,
what
am
I
I'm
gonna
do?
Is
you
know?
I'm
just
gonna
get
this
return
template
and
then
you
know
pop
it
in.
B
Yeah,
just
a
brief
note,
just
not
to
be
the
cadence
police
but
like
if
the
repository
has
any
log
files
that
have
time
in
him
that
could
constitute
benchmarking
and
that's
completely
illegal
according
to
academic
licensing,
of
course,
but
just
fy,
I'm
sure
everyone
knows
that,
but
just
throw
it
out
there.
I
D
You,
okay,
so
I'm
gonna
go
out
and
I'm
gonna
copy.
I
D
D
So
here
you
can
see,
you
know
all
the
files
that-
and
let's
say
someone
just
said
mark
already
so
because
you
know
this
is
cds,
ffmpt
technology
and
we
have
certain
scripts
in
bash
to
enable
it
what
I
do.
I
actually
have
a
environment
already
scripts
prepared,
but
I
make
them
you
know
so
for.
I
D
D
D
D
D
But
the
problem
is
I'm
not
here
where
I
can.
I
can
okay.
G
Then
I
think
I
think
this
branch
is
you,
you
have
the
versions,
you
can
still
run
it.
Let's
see,
let's.
I
G
D
D
I
E
F
No
is
this
in
28
or
sky
water
120.
F
F
I
Okay,
let's
see
so,
let
me
see,
I
need
to
now
try
to
read.
I
D
I
D
D
I
I
I
Okay,
so
it's
csh.
E
A
E
Not
a
it's.
D
So
basically,
you
know
before
me
executing
this
before
yeah,
it's
ending,
finished,
implemented
schematic.
So
before
me
executing
all
this
yeah,
you
just
had
inverted
templates,
and
that
is
basically
what
marco
said
it
is
in
this
specific.
When
you
see
you
know
you
have
these
inverter
templates,
but
basically
the
the
core
of
the
generator,
which
means
layout
description
in
python
and
schematic
description
in
python
and
sizing.
D
Sizing
is,
is
kind
of
here
where
we
have
also
the
switch
to
distinguish
between
finfet
and
and
balcimos,
because
there
is
a
different
sizing
for
wits,
one
as
float
the
other.
It
definitely
has
to
be
integer,
and
you
can
see
here
that
in
inverter
generated.
D
I
D
What
dummies
I
mean
very
typical
and
same
as
for
for
back
and
symbol,
also
which
you
can
see
here
so
at
this
point
you
know
you
can
clearly
see
we
demonstrated
the
basic
functionality
of
bag.
Like
marco
said,
we
didn't
want
to
go
into
drc
lbs,
because
every
every
let's
say
institution
has
its
own
customization.
D
The
main
effort
was
spent
basically
to
improve
the
documentation
and
separate
the
information
such
that
it
is
no
longer
just
in
one
yaml
file.
Everything
put
together
but
to
have
some
logical,
let's
say,
logical,
connected
pieces.
I
would
like
to
go.
Take
you
there.
D
So
marco
already
explained
so
the
idea
you
know
the
next
idea
will
be,
let's
say,
layer,
parameters.
What
we
give
to
mosaic
group
members
is
basically
kind
of
a
homework
so
that
they
clone,
execute
and
come
up.
You
know
with
the
problems,
but
also
to
check
these
specific
parameters
and
say
that
you
know
if
they
have
something
you
have
some
better
description
and
mark.
We
should
be
telling
them
to
do
that
into
and
back
to.
I
think
technology
definite
technology
templates,
because
the
documentation
is
generated
from
these
specific.
D
Let's
say
this
is
specific
files,
so
that
is
one
thing
that
I
believe
we
need
to
do
kind
of
to
make
if
we
cannot
share,
share
the
information
technology,
information
specific
and
protect
it,
what
we
can
share
is
description
such
that
people
can
easier,
find
and
debug.
You
know
find
information
in
the
technology
that
are
using.
There
is
also
a
silicon,
austria
labs.
We
have
some
work
going
on
about
generic
technology,
where
these
data
should
be
in
future
kind
of
semi-automa,
semiautomatic
or
optimized.
D
You
know
the
input
will
be
optimized,
but
that
is
not
the
current
status.
What
I
wanted
to
mention
here
also
is
at
the
moment,
if
you
see
here,
we
just
have
let's
say
vertegen,
and
it
is
basically
just
one
of
hopefully
many
generators
that
we
will
have
so
at
the
moment
excel.
I'm
I'm
finishing.
Basically,
I
cleaned
these
templates
and
adapted
for
tsmc
28
nanometer
technology.
D
My
goal
is
basically
to
change
the
project,
adapt
the
project,
basically
more
or
less
everything
is
set.
I
need
to
change
us
back
to
technology
definition
and
then
I
will
share
it
with
people.
Institutions
that
we
have
nda
with
bwc
is
one
of
them
effica
and
feel
like
it's
one
of
them
and
athenian
is
also
one
of
them.
After
that,
my
plan
is
because
we
had
a
tape
out.
D
We
had
a
tape
out,
let's
see,
if
I
can
move
this
yeah,
we
had
a
taped
out
last
may
so
this
basically
back
amplifier
final
was
submitted
and
shared.
It's
basically
open
source,
and
it
contains
a
bunch
of
smaller
generators
that
we
will
at
least
some
of
them.
A
Yeah
we're
just
a
couple
minutes
past
the
top
of
the
hour.
So
if
there's
any
closing
questions.
F
D
D
Jada
will
come
in
august
and
he's
already
working
on
enabling
his
generators
with
magic.
So
we
will
see
if
it's
possible
to
you
know
we
need
to
start
replacing
step
by
step.
So
if
it's
possible
to
introduce
magic.
D
This
moment
I
need
to
see
also
with
infinion,
because
on
their
plate
is
basically
framework,
the
the
the
the
roof.
You
know
above
this
specific
return,
so
template.
D
How
can
you
know
will
some
hooks
be
provided
such
that
we
can
hook
up
some
open
source
tools
and
that
sense,
I
already
before
contacted
some
of
the
the
let's
say,
authors
or
of
the
two
open
source
tools
and
they're
asking
if
they
are
willing
to
contribute
and
in
a
sense
there
is
a
good
will,
but
we
need
to
you
know,
provide
a
hook
such
that
they
they
clearly
understand
what
needs
to
be
done.
D
Yeah,
I'm
sure
it
it
does.
You
just
need
to
see
with
him.
You
know
he
has,
I
think
two
students
or
three
two
students,
two
students
working
on
this,
like
sky
water,
magic
trying
to
make
it
workable.
But
from
what
I
know,
it's
still
not
kind
of
you
know
clean.
I.
F
D
Maybe
it
will
be
good
to
ping
him
and
then
see
with
him
when
he's
ready
to
give
a
presentation
here,
yeah.
A
D
A
Oh,
thank
you
yeah.
I
appreciate
both
of
you
coming
today
and
you
know
updating
us
on
where
you
are
and
all
the
work
that's
ongoing
there.
It
looks
looks
quite
interesting
and
you
know
appreciate
all
the
effort
there.
D
A
You
know
I
just
there
was
actually
one
final
question
from
my
side:
how
many
I'm
just
curious,
how
many
different
the
members
you
have
participating
in
this
in
the
mosaic
group,
I
guess,
is
what
I
would
call
it
right.
D
D
So,
but
if
you
see
yeah,
if
you
see
let's
say
a
mosaic
and
then
group
information,
you
know
there
are
a
lot
of
people
actually
just
hooked
in
that
we
invited
so
some
of
them
are,
let's
say,
thomas
is
actively
working
and
he
he
also
in
involved
in
in
the
whole
story
people
or
like
the
engineers
who
will
be
working
on
the
framework.
D
You
know
some
people,
you
might
also
know-
or
you
know
they-
they
are,
let's
say
they're
here
as
a
more
as
a
guest
to
observe
the
situation,
but
what
I'm
also
planning
from
our
sides?
We,
for
instance
yesterday
we
have
two
new
members
that
next
to
joining
next
month,
to
use
basically
this
vertozo
template
as
as
as
the
ground
where
we
can.
Basically,
you
know
as
the
initial
steps
initial,
where
many
people
who
go
in
this
arena
can
make
initial
steps
together,
because
it
is,
you
can
see,
plug
and
play.
D
If
you
don't
miss
the
branch.
It
is
very
easy
to
execute
sorry,
it
is
very
easy
to
execute
and
it's
self-explanatory.
It
has
quite
good
documentation.
The
technology
information
will
be.
You
know
much
much
clearer
that
this
the
current
situation
will
back
to
and
it
will
have
more
available
generators.
D
You
know
to
plug
it
with,
and
it
will
give
this
like.
You
know
the
strategy,
how
to
do
hierarchical
approach,
so
what
I'm
seeing
is
basic
things
definitely
will
be
solved
and
in
july
we're
we're
putting
one
more
colleague,
one
of
these
two
that
I
mentioned
he
will
be
working
on
the
small
layout
generators
because,
in
my
opinion,
it
is
extremely
important
to
raise
the
level
of
abstraction
in
layout.
You
know,
without
that
people
will
be
very
much
reluctant
to
script.
D
A
No,
that
makes
sense
well,
thank
you
so
much
mariana
and
marco
for
coming
today
and
describing
the
work
that
you've
done.
I
really
appreciate
it
looks
like
a
great
progress.