CS 150 – Fall 2005 - Lec #27: FPGA Evolution – 1 trend toward higher levels of integration Evolution of Implementation Technologies zDiscrete devices:

CS 150 – Fall 2005 - Lec #27: FPGA Evolution – 1

trend toward

higher levels

of integration

Evolution of Implementation Technologies

Discrete devices: relays, transistors (1940s-50s) Discrete logic gates (1950s-60s) Integrated circuits (1960s-70s)

e.g. TTL packages: Data Book for 100’s of different parts Map your circuit to the Data Book parts

Gate Arrays (IBM 1970s) “Custom” integrated circuit chips Design using a library (like TTL) Transistors are already on the chip Place and route software puts the chip together automatically + Large circuits on a chip + Automatic design tools (no tedious custom layout) - Only good if you want 1000’s of parts


Gate Array Technology (IBM - 1970s) Simple logic gates

Use transistors toimplement combinationaland sequential logic

Interconnect Wires to connect inputs and

outputs to logic blocks

I/O blocks Special blocks at periphery

for external connections

Add wires to make connections Done when chip is fabed

“mask-programmable” Construct any circuit


Programmable Logic

Disadvantages of the Data Book method Constrained to parts in the Data Book Parts are necessarily small and standard Need to stock many different parts

Programmable logic Use a single chip (or a small number of chips) Program it for the circuit you want No reason for the circuit to be small


Programmable Logic Technologies Fuse and anti-fuse

Fuse makes or breaks link between two wires Typical connections are 50-300 ohm One-time programmable (testing before programming?) Very high density

EPROM and EEPROM High power consumption Typical connections are 2K-4K ohm Fairly high density

RAM-based Memory bit controls a switch that connects/disconnects two

wires Typical connections are .5K-1K ohm Can be programmed and re-programmed in the circuit Low density


Programmable Logic Program a connection

Connect two wires Set a bit to 0 or 1

Regular structures for two-level logic (1960s-70s) All rely on two-level logic minimization PROM connections - permanent EPROM connections - erase with UV light EEPROM connections - erase electrically PROMs

Program connections in the _____________ plane PLAs

Program the connections in the ____________ plane PALs

Program the connections in the ____________ plane


PAL Logic Building Block Programmable AND gates Fixed OR/NOR gate Flipflop/Registered Output Feedback to Array Tri-state Output


XOR PALs

Useful for comparator logic, arithmetic sums, etc. Use of XOR gates can dramatically reduce the number of

AND plane inputs needed to realize certain functions


XOR PAL

And/Or/XOR Logic

Feedback

Registered Outputs

Tri-State Outputs


Another Variation: Synchronous vs. Asynchronous Outputs

DQ

DQ

DQ

Q0

Q1

Open

Com

Seq

Seq

CLK

N

D

Reset


Making Large Programmable Logic Circuits

Alternative 1 : “CPLD” Put a lot of PLDS on a chip Add wires between them whose connections can be

programmed Use fuse/EEPROM technology

Alternative 2: “FPGA” Emulate gate array technology Hence Field Programmable Gate Array You need:

A way to implement logic gates A way to connect them together


Field-Programmable Gate Arrays

PALs, PLAs = 10s – 100s Gate Equivalents

Field Programmable Gate Arrays = FPGAs Altera MAX Family Actel Programmable Gate Array Xilinx Logical Cell Array

1000s - 100000(s) of Gate Equivalents!


Field-Programmable Gate Arrays Logic blocks

To implement combinationaland sequential logic

Interconnect Wires to connect inputs and

outputs to logic blocks

I/O blocks Special logic blocks at

periphery of device forexternal connections

Key questions: How to make logic blocks programmable? How to connect the wires? After the chip has been fabbed


Tradeoffs in FPGAs Logic block - how are functions implemented: fixed

functions (manipulate inputs) or programmable? Support complex functions, need fewer blocks, but they are

bigger so less of them on chip Support simple functions, need more blocks, but they are

smaller so more of them on chip

Interconnect How are logic blocks arranged? How many wires will be needed between them? Are wires evenly distributed across chip? Programmability slows wires down – are some wires specialized

to long distances? How many inputs/outputs must be routed to/from each logic

block? What utilization are we willing to accept? 50%? 20%? 90%?


Clk MUX

Output MUXQ

F/B MUX

Invert Control

AND ARRAY

CLK

pad

8 Product TermAND-OR Array

+Programmable

MUX's

Programmable polarity

I/O Pin

Seq. LogicBlock

Programmable feedback

Altera EPLD (Erasable Programmable Logic Devices)

Historical Perspective PALs: same technology as programmed once bipolar PROM EPLDs: CMOS erasable programmable ROM (EPROM) erased by UV

light

Altera building block = MACROCELL


Altera EPLDs contain 10s-100s of independently programmed macrocells

Personalizedby EPROMbits: Flipflop controlled

by global clock signal

local signal computesoutput enable

Flipflop controlledby locally generatedclock signal

+ Seq Logic: could be D, T positive or negative edge triggered+ product term to implement clear function

Synchronous Mode

Asynchronous Mode

Global CLK

OE/Local CLK

EPROM Cell

1

Global CLK

OE/Local CLK

EPROM Cell

1

Clk MUX

Clk MUX

Q

Q

Altera EPLD: Synchronous vs. Asynchronous Mode


LAB A LAB H

LAB B LAB G

LAB C LAB F

LAB D LAB E

P I A

AND-OR structures are relatively limited Cannot share signals/product terms among macrocells

LogicArrayBlocks

(similar tomacrocells)

Global Routing:ProgrammableInterconnect

Array

8 Fixed Inputs52 I/O Pins8 LABs16 Macrocells/LAB32 Expanders/LAB

EPM5128:

Altera Multiple Array Matrix (MAX)


LAB Architecture

Expander Terms shared among allmacrocells within the LAB• Efficient way to use AND plane resources

Macrocell ARRAY

I/O Block

Expander Product

Term ARRAY

I NPUTS

P I

A

I/O Pad

I/O Pad

Macrocell P-Terms

Expander P-Terms


0ASYNCHRONOUS RESET (TO ALL REGISTERS)

23AR

88132176220264308352396

44

22

2

OUTPUT LOGIC

MACROCELL

P - 5810 R - 5811

528572616660704748792836

484

880

440

21

3

OUTPUT LOGIC

MACROCELL

P - 5812 R - 5813

10561100114411881232127613201364

1012

1408

924

968

1452

20

4

OUTPUT LOGIC

MACROCELL

P - 5814 R - 5815

16721716176018041848189219361980

1628

2024

1496

1584

2068

1540

2112

19

5

OUTPUT LOGIC

MACROCELL

P - 5816 R - 5817

23762420246425082552259626402684

2332

2728

2156

2288

2772

22442200

28162860

1

1

0

0

1

0

0

1

D Q

QSP

10

5808

P

R

5809

10 4 8 12 16 20 24 28 32 36 40

INCREMENT

FIRST FUSE NUMBERS

15

9

OUTPUT LOGIC

MACROCELL

P - 5824 R - 5825

49725016506051045148519252365280

4928

5324

4884

17

7

OUTPUT LOGIC

MACROCELL

P - 5820 R - 5821

38283872391639604004404840924136

3784

4180

3652

3740

4224

3696

4268

16

8

OUTPUT LOGIC

MACROCELL

P - 5822 R - 5823

44444488453245764620466447084752

4400

4796

4312

4356

4840

18

6

OUTPUT LOGIC

MACROCELL

P - 5818 R - 5819

31243168321232563300334433883432

3080

3476

2904

3036

3520

29922948

35643608

14

10

OUTPUT LOGIC

MACROCELL

P - 5826 R - 5827

54125456550055445588563256765720

5368

11

5764

13

SYNCHRONOUS PRESET (TO ALL REGISTERS)

0 4 8 12 16 20 24 28 32 36 40INCREMENT

Supports large number of product terms per outputLatches and muxes associated with output pins

P22V10 PAL


Rows of programmablelogic building blocks

+

rows of interconnect

Anti-fuse Technology:Program Once

8 input, single output combinational logic blocksFFs constructed from discrete cross coupled gates

Use Anti-fuses to buildup long wiring runs from

short segments

Actel Programmable Gate Arrays


Basic Module is aModified 4:1 Multiplexer

Example: Implementation of S-R Latch

2:1 MUXD0

D1

SOA

2:1 MUXD2

D3

SOB

2:1 MUX

S0

Y

S1

2:1 MUX"0"

R

2:1 MUX"1"

S

2:1 MUX Q

"0"

Actel Logic Module


Interconnection Fabric

Logic Module

Horizontal Track

Vertical Track

Anti-fuse

Actel Interconnect


Jogs cross an anti-fuse

minimize the # of jogs for speed critical circuits

2 - 3 hops for most interconnections

Logic Module

Logic ModuleLogic Module Output

Input

Input

Actel Routing Example


Actel’s Next Generation: Axcelerator

C-Cell Basic multiplexer logic plus

more inputs and support for fast carry calculation

Carry connections are “direct” and do not require propagation through the programmable interconnect


Actel’s Next Generation: Accelerator R-Cell

Core is D flip-flop Muxes for altering the clock

and selecting an input Feed back path for current

value of the flip-flop for simple hold

Direct connection from one C-cell output of logic module to an R-cell input; Eliminates need to use the programmable interconnect

Interconnection Fabric Partitioned wires Special long wires


IOB IOB IOB IOB

CLB CLB

CLB CLB

IOB

IOB

IOB

IOB

Wiring Channels

Xilinx Programmable Gate Arrays CLB - Configurable Logic Block

5-input, 1 output function or 2 4-input, 1 output functions optional register on outputs

Built-in fast carry logic Can be used as memory Three types of routing

direct general-purpose long lines of various lengths

RAM-programmable can be reconfigured

CLB

CLB

CLB

CLB

SwitchMatrix

ProgrammableInterconnect

I/O Blocks (IOBs)

ConfigurableLogic Blocks (CLBs)

D Q

SlewRate

Control

PassivePull-Up,

Pull-Down

Delay

Vcc

OutputBuffer

InputBuffer

Q D

Pad

D QSD

RD

EC

S/RControl

D QSD

RD

EC

S/RControl

1

1

F'

G'

H'

DIN

F'

G'

H'

DIN

F'

G'

H'

H'

HFunc.Gen.

GFunc.Gen.

FFunc.Gen.

G4G3G2G1

F4F3F2F1

C4C1 C2 C3

K

Y

X

H1 DIN S/R EC


The Xilinx 4000 CLB


Two 4-input functions, registered output


5-input function, combinational output


CLB Used as RAM


Fast Carry Logic


Xilinx 4000 Interconnect


Switch Matrix


Xilinx 4000 Interconnect Details


Global Signals - Clock, Reset, Control


Xilinx 4000 IOB


Xilinx FPGA Combinational Logic Examples Key: General functions are limited to 5 inputs

(4 even better - 1/2 CLB) No limitation on function complexity

Example 2-bit comparator:

A B = C D and A B > C D implemented with 1 CLB(GT) F = A C' + A B D' + B C' D'(EQ) G = A'B'C'D'+ A'B C'D + A B'C D'+ A

B C D

Can implement some functions of > 5 input


CLB

5-input Majority Circuit

CLB

CLB

CLB

7-input Majority Circuit

Xilinx FPGA Combinational Logic Examples

N-input majority function: 1 whenever n/2 or more inputs are 1

N-input parity functions: 5 input/1 CLB; 2 levels yield 25 inputs!

CLB

CLB

9 Input Parity Logic


Xilinx FPGA Adder Example Example

2-bit binary adder - inputs: A1, A0, B1, B0, CIN outputs: S0, S1, Cout

CLB

A0 B0 Cin

S0

CLB

A1 B1

S1

CLB

A2 B2

C1S2

CLB

A3 B3

C2S3 C0Cout

S0

S1

C2

A1 B1 CinA0 B0

CLBS2

S3

Cout

A3 B3 A2 B2

CLB

Full Adder, 4 CLB delays tofinal carry out

2 x Two-bit Adders (3 CLBseach) yields 2 CLBs to finalcarry out


Xilinx Vertex-II Family 88-1000+ pins 64-10000+ CLBs

Combinational and sequential logic using lookup tables and flip-flops

Random-access memory Shift registers for use as buffer storage

Multipliers regularly placed throughout the CLB array to accelerate digital signal processing applications

E.g., the XC2V8000: 11,648 CLBs, 1108 IOBs, 90,000+ FFs, 3Mbits RAM (168 x 18Kbit blocks), 168 multipliers Equivalent to eight million two-input gates!


Xilinx Vertex-II Family IOB Tri-state/bidirectional driver Registers for each of three

signals involved: input, output, tri-state enable.

Two registers to latch values with separate clocks.

For large pinouts, separate clocks stagger signals changes to avoid large current spikes

FFs used for synchronization as well as latching


Xilinx Vertex-II Family CLB

Four basic slices in two groups

Each has a fast carry-chain

Local interconnect to wire logic of each slice and connect to the CLB array: switch matrix is large collection of programmable switches


Xilinx Vertex-II Family CLB Internals

Just ½ of one slice!

4-input LUT + FF

Fast carry logic

Many programmable interconnections for sync vs. async operation


Xilinx Vertex-II Family Fast Carry Logic

LUT

LUT

AB

AB

Mux0 1

Mux0 1

Ci

Co

0

(AB)Ci

(AB)

(ABCi)

(AB)Ci+AB

AB

1

A

C

B

111 1


Xilinx Vertex-II Family CLB

Sequential Portion Two positive edge-

triggered flip-flops Transparent latches or flip-

flops Asynchronous or

synchronous sets and resets

Initialize to different values at power-up

Clocks and load enables complemented or not


Xilinx Vertex-II Family Slice Personality

4-input function generator OR 16 bits of dual-ported

random-access memory (with separate address inputs for read - G1 to G4 - and write - WG1 to WG4)

OR a 16-bit variable-tap shift register

With muxes, CLB can implement any function of 8 inputs and some functions of 9 inputs

Registered and unregistered versions of function block outputs


Xilinx Vertex-II Family Interconnections

Methods of interconnecting CLBs and IOBs: (1) direct fast connections

within a CLB(2) direct-connections between

adjacent CLBs(3) double-lines to fanout

signals to CLBs one or two away

(4) hex lines to connect to CLBs three or six away

(5) long lines that span the entire chip

Fast access to neighbors vertically and horizontally with direct connections

Double and hex lines provide a slightly larger range

Long lines saved for time-critical signals w/ min signal skew


Programmable Logic Summary

Discrete Gates

Packaged Logic

PLAs

Ever more general architectures of programmable combinational + sequential logic and interconnect Altera Actel Xilinx—4000 series to Vertex

CLBs implementing logic function generators, RAMs, Shift registers, fast carry logic

Local, inter-CLB, and long line interconnections

Documents

CS 150 – Fall 2005 - Lec #27: FPGA Evolution – 1 trend toward higher levels of integration Evolution of Implementation Technologies zDiscrete devices: