from high-level algorithms to asml automated digital ......1. asml business drivers: quality and...
TRANSCRIPT
Public
John van Tol
May 21st, 2019 - Eindhoven
From High-Level Algorithms to ASML Automated Digital Design Flows
Competence Leader Embedded Logic and Firmware (ELF) Development
Public
11-03-2019
Slide 2
Public
Key Takeaways
1. ASML business drivers: Quality and Time-to-Market, require an efficient workflow for getting smart algorithms into high speed digital hardware (FPGAs)
2. Adoption of Model-Based Design into the ASML Digital Design Flow achieves this with automatic generation of HDL source code and test bench code
3. Model-Based Design reduces the gap between system architect and design engineer
Public
Chips are everywhere!In 2015, 230 billion ICs were made, 30 for every man, woman and child on the planet
Data: WSTS
0
50
100
150
200
250
199
1
199
2
199
3
199
4
199
5
199
6
199
7
199
8
199
9
200
0
200
1
200
2
200
3
200
4
200
5
200
6
200
7
200
8
200
9
201
0
201
1
2012
201
3
201
4
201
5
IC units, in billions
Public
07-04-2018
Slide 3
Public
We develop and build lithography systems for manufacturing those IC’s
150 tons of precision equipmentPublic
07-04-2019
Slide 4
works with single nano-meter precision: Grass grows at a speed of 33 nano-meter per second
Public
2018
1000 $, 512 MB + GB flash, 200 g,< 1W
07-04-2019
Slide 5
Public
Moore’s Law means doing “More with Less”
Cray Super Computer
1976
30M$, 8 Mb, 5500 kg, 150 kW
Public
Slide 6
Public
The industry is driven by Moore’s law, which continues, through ideas and value creation
Lo
g S
ca
le S
yste
m P
erf
orm
an
ce
/Wa
tt
2008 2018 2028
>2X/2.4 years
2X/2.4 years
¹Lisa Su, AMD, “Immersive era in consumer computing”, IEDM, dec 2017
²Gordon Moore, “Progress in digital integrated Electronics” International; Electronic Device Meeting,, IEEE, 1975, p p 11-13
ASML Market Research
0.001
0.01
0.1
1
Relative cost per function
This is what we’re doing it for!
07-04-2019
Public
07-04-2019
Slide 7
Public
Our systems enable the faster, better, cheaper chips of tomorrow
0.000
0.001
0.010
0.100
1.000
1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023TBD
Re
lati
ve
Co
st
pe
r P
ixe
l
NXE:3400B
Res. 13nm
300mm 125wph NXT:1950i
Res 38nm,
300mm 190wph
XT:1400
Resolution 65nm
300mm 145wph
High NA EUV
Res. <8nm
300mm 185wph
AT:850
Res. 110nm
300mm 102wph
PAS 5500/60
Res. 450nm
200mm 48wph
PAS 2500/10
Res. 900nm,
150mm 66wph
Source: ASML Product specs, Market Research
Public
ASML makes the machines for making IC’s2018-07-25
Public
Slide 8
Public
Electronic Development within D&E ASML2018-07-25
Public
Slide 9
EDEV Wilton USA
EDEV San DiegoUSA
EDEV VeldhovenThe Netherland
Embedded
Logic
Firmware
Public
Goals and Challenges
27-11-2018
Slide 10
Public
07-04-2019
Slide 11
Public
Challenges
Algorithm
complexity
Performance
need
Algorithms in
Software
Algorithms in
Embedded Hardware
(FPGA)
Trend
Implementation
complexity
Increased
performance
Need:
High level design
and implementation
workflow for FPGAs
Public
07-04-2019
Slide 12
Public
Goals
1. One ASML common method of developing algorithms in FPGAs
2. Make modeling, simulation and code generation accessible for ELF engineer
3. Reduce the gap between system architect and ELF engineer
Public
How we do it
27-11-2018
Slide 13
Public
ASML ELF Way-of-Working (WoW)
HDL3rd
party
IP
GDBSimulink
HDL
Integration
GDB = Generic Design Block, HDL = Hardware Description Language
Integration
Verification
Minimal impact on
existing WoW
Public
07-04-2019
Slide 16
Public
ASML ELF Way-of-Working, elaborated
SimulinkSimulation Model
GenerateRTL Code
Generate Verification
Items
Generate Block
Level TB
Block Level Simulation
Public
ASML workflow to create algorithm subsystem, an example
Public
11-03-2019
Slide 18
Public
Automated Build and Test
1. Buildfiles are used to automate the various tasks involved in the FPGA design flow
2. The “Build process” makes extensive use of (GNU) Make
DesignCompilation
and Simulation
System generation
(PlatformDesigner, XPS, Vivado)
Synthesis and place and route
Public
Achievements and Outlook
27-11-2018
Slide 19
Public
07-04-2019
Slide 20
Public
Using HDL Coder in ELF projects
• Established workflow within the entire ASML organization. • IT team (IT4E) maintains development software and Linux servers for any ELF project• Standard use of HDL Code Generation in ELF Design Flow (DDM2/Buildfiles) • Integrated with workflow for communication protocols (GDB), integration tool (QSYS/Platform
designer/Vivado Block design), hardware interfaces (VHDL)
• Improved development of Digital Signal Processing (DSP) functions (vs. handwritten RTL and/or Testbench)
• Easier to make changes in a late stage of the project without compromising quality• Especially suited for complex DSP functions
• Future work: automatic FPGA-In-the-Loop (FIL)
Public
11-03-2019
Slide 21
Public
Key Takeaways
1. ASML business drivers: Quality and Time-to-Market, require an efficient workflow for getting smart algorithms into high speed digital hardware (FPGAs)
2. Adoption of Model-Based Design into the ASML Digital Design Flow achieves this with automatic generation of HDL source code and test bench code
3. Model-Based Design reduces the gap between system architect and design engineer
Public