Levels of Abstraction: The History of Custom MOS Design Steve - - PowerPoint PPT Presentation

levels of abstraction the history of custom mos design
SMART_READER_LITE
LIVE PREVIEW

Levels of Abstraction: The History of Custom MOS Design Steve - - PowerPoint PPT Presentation

TRILOBYTE 1 SYSTEMS Levels of Abstraction: The History of Custom MOS Design Steve Golson Trilobyte Systems Phone: +1.978.369.9669 Email: sgolson@trilobyte.com Web: http://www.trilobyte.com 2 1968 -- The beginning Hand-drawn schematics


slide-1
SLIDE 1

1

TRILOBYTE SYSTEMS

Levels of Abstraction: The History of Custom MOS Design

Steve Golson Trilobyte Systems Phone: +1.978.369.9669 Email: sgolson@trilobyte.com Web: http://www.trilobyte.com

slide-2
SLIDE 2

2

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

1968 -- The beginning

  • Hand-drawn schematics
  • Standard cells -- new library for each product line

~6 kinds of gates, ~4-5 speed ranges

  • Gate sizing by hand using CT curves (load vs delay)
  • Place & route by hand on mylar
  • TTL breadboard for verification
  • Transistor-level timing analysis using SPICE-like programs

Designer thinks about: Gates, transistors, breadboard, and maybe library

slide-3
SLIDE 3

3

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Early 1970s -- some improvements

+ Module-level design (multiple designers working on one chip) + Software simulation (cycle-based) + On-chip bus design (trade wires for time) + Pitch-matched layout (datapath, PLA, ROM, RAM) + MOS-specific circuit techniques (dynamic logic)

Designer thinks about: Gates, netlist simulation, floorplan, transistors

slide-4
SLIDE 4

4

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Early 1980s -- LSI Logic et al.

  • Hand-drawn schematics
  • Hand-typed netlist

+ Automatic place and route

  • Netlist simulation

+ Gate-level timing analysis

Designer thinks about: Gates, netlist simulation

slide-5
SLIDE 5

5

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Mid 1980s -- some improvements

+ Schematic capture + Automatic netlist generation

  • Automatic place and route
  • Netlist simulation
  • Gate-level timing analysis

Designer thinks about: Gates, netlist simulation

slide-6
SLIDE 6

6

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Early 1990s -- Verilog and synthesis

+ RTL Verilog + RTL simulation + Synthesized netlist

  • Netlist simulation
  • Automatic (?) place and route
  • Gate-level timing analysis

Designer thinks about: RTL code, RTL simulation, gates, netlist simulation

slide-7
SLIDE 7

7

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Late 1990s -- Behavioral synthesis

+ Behavioral Verilog, some generated by graphical HLDA tools + Behavioral Verilog and HLDA simulation

  • RTL Verilog, some automatically generated by HLDA tool
  • RTL simulation
  • Synthesized netlist
  • Netlist simulation
  • Automatic (???) place and route
  • Gate-level timing analysis

+ FPGA breadboard

slide-8
SLIDE 8

8

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Late 1990s -- Behavioral synthesis

Designer thinks about: Behavioral code, RTL code, gates, floorplan, behavioral simulation, RTL simulation, netlist simulation, breadboard

slide-9
SLIDE 9

9

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

2001 and beyond the infinite…

+ System design language (C++, Superlog, SystemC, etc.) + System design language simulation + System design language synthesis

  • RTL Verilog
  • RTL simulation
  • Floorplan
  • RTL synthesized netlist & placement (physical synthesis)
  • Netlist simulation
  • Gate-level timing analysis
slide-10
SLIDE 10

10

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

2001 and beyond the infinite…

Designer thinks about: System code, RTL code, gates, floorplan, placement, system simulation, RTL simulation, netlist simulation, breadboard

slide-11
SLIDE 11

11

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Levels of abstraction today…

2 4 6 8 10 12 1970 1980 1990 2000 2010 levels of abstraction

slide-12
SLIDE 12

12

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Levels of abstraction tomorrow??

2 4 6 8 10 12 1970 1980 1990 2000 2010 levels of abstraction

slide-13
SLIDE 13

13

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

What designers need… What designers need is to have abstraction levels removed, not added

slide-14
SLIDE 14

14

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

What designers need… is RTL signoff

Designer thinks about: System code, RTL code, System simulation, RTL simulation Vendor thinks about: everything else

slide-15
SLIDE 15

15

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Levels of abstraction: RTL signoff

2 4 6 8 10 12 1970 1980 1990 2000 2010 levels of abstraction

slide-16
SLIDE 16

16

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

What designers need…

  • Physical synthesis

+ Improve quality of synthesis + Improve quality of place & route + No more gate tweaks and layout mods + No longer worry about gate-level timing

  • Automatic verification of RTL vs netlist

+ No longer worry about netlist simulation

  • RTL analysis tools for timing, power, area
  • Smarter IP lawyers
slide-17
SLIDE 17

17

Steve Golson -- Trilobyte Systems -- Levels of Abstraction: The History of Custom MOS Design

Things to consider about new EDA tools

  • How do they fit into my existing design flow?
  • Revolutionary or evolutionary?
  • What problems do they solve?
  • What new problems might they introduce?

Higher-level abstractions place you further away from an intuitive feel for hardware and timing: “How fast is it?” Just because you can do something, doesn’t mean it is worth doing