Transcript No Slide Title

Introduction to FPGAs
Field Programmable Gate Arrays
CSET 4650
Field Programmable Logic Devices
Dan Solarek
Hierarchy of Logic Implementations
The diagram below is a modified version of the one we first
used to discuss the role of FPLDs in logic implementation
This version more closely reflects the details as we have
come to know them
Logic
Standard
Logic
TTL
ASIC
CMOS
Programmable
Logic Devices
SemiCustom
ICs
(FPLDs)
SPLDs
(e.g., PALs)
CPLDs
FPGAs
Gate
Arrays
Full Custom
ICs
Cell-Based
ICs
2
FPGA Development
FPGAs evolved from Gate Arrays
Parallel with development of CPLDs
ASIC
Programmable
Logic Devices
SemiCustom
ICs
(FPLDs)
SPLDs
(e.g., PALs)
CPLDs
FPGAs
Gate
Arrays
Full Custom
ICs
Cell-Based
ICs
3
Gate Array Technology (1970s)
Mask-Programmable Logic Devices
MPLDs as compared to FPLDs
Programmed as part of fabrication process
Mask-Programmable Gate Arrays
A specific type of MPLD
Build standard layout of transistors on chip
Customer specifies wiring to connect transistors into gates
and gates into systems
Only has to go through last few mask steps of fabrication
process
Faster than full-custom chip fabrication
4
Gate Array Technology (1970s)
Simple logic gates
Use transistors to
implement combinational
and sequential logic
Interconnect
Wires to connect inputs and
outputs to logic blocks
I/O blocks
Special blocks at periphery
for external connections
Add wires for connections
Done when chip is fabricated
“mask-programmable logic device”
Construct any circuit
5
Evolution of the FPGA
Early FPGAs
Used mainly for “glue logic” between other components
(interfacing)
Simple Combinational Logic Blocks (CLBs)
Small number of inputs and outputs
Focus was on implementing “random” logic efficiently
As capacities grew, other applications emerged
FPGAs used as an alternative to custom IC’s for entire
applications
Computing with FPGAs
6
Evolution of the FPGA
FPGAs have changed to meet new application
demands
Carry chains, better support for multi-bit operations
Integrated memories, such as the block RAMs
Specialized units, such as multipliers, to implement
functions that are slow/inefficient in CLBs
Newer devices incorporate entire CPUs:
Xilinx Virtex II Pro has 1-4 Power PC CPUs
Devices that don’t have CPU hardware generally support
synthesized CPUs
7
Current FPGAs: Major Elements
Current commercial FPGAs have the same general
structure but differ among major components:
Programmability
Technology used to program device
Internal logic cell structure
Combinational and sequential
Complexity
Routing mechanisms
Interconnecting wires and their layout
8
The Plan for Today
We will look at a generalized overview of FPGAs
and their structure
More of our examples than not will be from Xilinx
devices
Since that is what we use in the lab
Since they are recognized as a leading vendor
Over the next few meetings, we will look at greater
detail about the major FPGA elements and families
9
General FPGA Architecture
Routing mechanism
Logic cell, often called a CLB –
“configurable logic block”
10
Field-Programmable Gate Arrays
Based on Configurable Logic Blocks (CLB) as the
logic cells …
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
11
Current FPGAs: Logic Cells
Current commercial FPGAs use logic cells that are
based one one or more of the following:
Transistor pairs
Basic small gates
e.g., two-input NAND or XOR
Multiplexers
Look-up tables (LUTs)
Wide-fan-in AND-OR structures
Microprocessor-like
12
Field-Programmable Gate Arrays
Requires some form of programmable interconnect
at crossovers …
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
CLB
13
Current FPGAs: Programming
Static RAM
Switch is a pass transistor controlled by the state of the
SRAM bit
EEPROM
Switch is a floating-gate transistor that can be turned off
by injecting charge onto its floating gate
Antifuse
Switch is a device that, when electrically programmed,
forms a low resistance path
14
Current FPGAs versus MPLDs
Programmable switches
occupy larger chip areas
exhibit higher parasitic resistance and capacitance (power
dissipation and propagation delay result)
Additional chip area required for switch
programming circuitry
The more switches, the more flexible
Flexibility requires higher “overhead”
FPGAs are slower than MPLDs
15
Current FPGAs: Programming
FPGA
Antifuseprogrammed
SRAMprogrammed
Island
Actel ACT1 & 2
Quicklogic’s pASIC
Crosspoint’s CP20K
Xilinx LCA
AT&T Orca
Altera Flex
EPROMprogrammed
Cellular
Toshiba
Plesser’s ERA
Atmel’s CLi
Altera’s MAX
AMD’s Mach
Xilinx’s EPLD
16
FPGA Architectures
FPGAs are commercially available in many different
architectures and organizations.
Although each company’s offerings have unique
characteristics, FPGA architectures can be generically
classified into one of four categories:
Symmetrical Array
Row Based
Hierarchical PLD
Sea of Gates
17
FPGA Architectures
The Configurable Logic Blocks
(CLBs) are organized in a two
dimensional array separated by
horizontal and vertical wiring
channels.
Each CLB contains flip-flop(s),
multiplexers, and a combinatorial
function block which operates as
an SRAM based table look-up.
Connections between CLBs are
customized by turning on pass
transistors which selectively
connect the CLBs to the
interconnection resources
CLB
18
FPGA Architectures
Pass transistors selectively
connect the interconnect lines
between the horizontal and
vertical wiring channels.
SRAM cells which are
distributed around the chip
hold the state of the
interconnect switches.
Surrounding the CLB array
and interconnect channels are
the programmable I/O blocks
which connect to the package
pins.
CLB
19
FPGA Architectures
Xilinx XC4000 FPGA
Greater logic capacity per
CLB is achieved using a
two-level look-up table
Compared to earlier
families, the routing
resources have been more
than doubled.
number of globally
distributed signals has
increased
20
FPGA Architectures
This organization is similar to that
found in the traditional style of
Mask Programmed Gate Arrays
(MPGAs).
Vertical interconnect segments of
varying lengths are available.
Vertical segments in input tracks
are permanently connected to logic
module inputs, and vertical
segments in output tracks are
permanently connected to logic
module outputs.
21
FPGA Architectures
Long vertical segments are
available which are uncommitted
and can be assigned during
routing.
The horizontal wiring channel
resources are also segmented into
varying lengths.
The minimum horizontal segment
length is the width of a single logic
module, and the maximum
horizontal segment length spans
the full channel.
22
FPGA Architectures
Any segment that spans
more than one-third of the
row length is considered a
“long horizontal segment”.
Dedicated routing tracks are
used for global clock
distribution and for power
and ground tie-off
connections.
23
FPGA Architectures
The Actel ACT family
FPGAs a logic module
matrix is arranged as
rows of cells separated by
horizontal wiring
channels
This organization is
similar to that found in
the traditional style of
Mask Programmed Gate
Arrays (MPGAs)
24
FPGA Architectures
This architecture represents
a hierarchical arrangement
of CLBs using a twodimensional array structure.
Interconnections are via a
centralized programmable
interconnect structure
CLBs can be cascaded
I/O structures not shown
25
FPGA Architectures
The Altera Multiple Array
MatriX (MAX) architecture
represents a hierarchical
arrangement of Erasable
Programmable Logic Devices
(EPLDs) using a twodimensional array structure.
The design provides multiple
level logic, uses a programmable
routing structure, and is user
reprogrammable based on
EPROM or EEPROM
technology.
LAB A
LAB H
LAB B
LAB C
LAB D
LAB G
P
I
A
LAB F
LAB E
26
FPGA Architectures
this design has a twodimensional mesh array
structure which resembles
the gate array “sea of gates”
Static RAM programming
technology is used to
specify the function
performed by each logic
cell and to control the
switching of connections
between cells.
27
FPGA Architectures
The CAL1024 design contains
1024 identical logic cells
arranged in a 32 X 32 matrix.
The design is considered to be a
mesh-connected architecture
since each cell is directly
connected to its nearest north,
south, east, and west neighbors.
In addition to these direct
connects, two global interconnect
signals are routed to each cell to
distribute clock and other “low
skew requirement” control
signals.
28
Field-Programmable Gate Arrays
Xilinx Spartan-3 die image; note the regularity…
29
Field Programmable Gate Arrays
Xilinx FPGAs are based on Look-up Tables (LUTs) as the CLB.
A LUT is simply a representation of a truth table:
three-input truth table
C
0
0
0
0
1
1
1
1
B
0
0
1
1
0
0
1
1
A
0
1
0
1
0
1
0
1
F
1
0
1
0
0
0
1
1
a
b
c
1
0 f
1
0
0
0
1
1 LUT
The function is
programmable – any LUT
can be programmed to be
any function
three-input Look-Up Table
FPGAs are just a whole lot of LUTs with lots of interconnect
30
Synthesizing Functions to CLBs
Flexibility of CLBs is a big win -- much harder to
map to technology with less-flexible blocks
Basically, can divide logic into n-input functions,
map each onto a CLB.
Tools may have special-purpose routines for
common blocks (like adders)
Harder problem: Placing blocks to minimize
communication, particularly when using carry
chains
31
FPGA Organization
a
b
c
0x
0x f
1x
1x
0x
1x
1x
1x LUT
a
b
c
x
x f
x
x
x
x
x
x LUT
I/O1
I/O2
a
b
c
1x
1x f
0x
1x
1x
0x
1x
0x LUT
a
b
c
x
x f
x
x
x
x
x
x LUT
a
b
c
x
x f
x
x
x
x
x
x LUT
a
b
c
x
x f
x
x
x
x
x
x LUT
I/O3
I/O4
32
FPGAs
Xilinx FPGAs are based on SRAM
Lose programming when power is turned off
Can be programmed by a computer or by a special
EPROM
Capacity
May have up to 10,000,000 gate equivalent
Up to 1,200 I/O pins
33
FPGAs
FPGAs must add some kind of switch to the
equation to be user programmable.
The size and performance of the switch essentially
determines the architecture
ULM (Universal Logic Module) must be as small
as possible to maximize versatility and utilization
34
What’s in a CLB?
Inputs
Look-Up
Table
(LUT)
Out
State
Clock
Enable
35
CLB Variables
Number of inputs to LUT
Trade off number of CLBs required vs. size of CLB and
routing area
How is logic implemented
LUT vs. programmable and-or-invert vs. other
Technology used to hold configuration (program) of CLB
Flip-flop in CLB?
Additional Functionality
Carry chains
36
Switch Detail
Programmable
Switch Matrix
Connections
are controlled
by RAM bits
More later
37
Programmable Switch Matrix
programmable switch element
turning the corner, etc.
38
The Fitter’s Job
Partition logic functions into CLBs
Arrange the CLBs
Interconnect the CLBs
Minimize the number of CLBs used
Minimize the size and delay of interconnect used
Work with constraints
“Locked” I/O pins
Critical-path delays
Setup and hold times of storage elements
39
Input-Output Blocks
One IOB per FPGA pin
Allows pin to be used as input, output, or bidirectional
(tri-state)
Inputs
Direct
Registered
Drive dedicated decoder logic for address recognition
IOB may also include logic for boundary scan
(JTAG)
40
I/O blocks
Looks like
a CPLD
macrocell
41
Xilinx 4000-series FPGAs
42
FPGAs: Summary
Historically, FPGA architectures and companies
began around the same time as CPLDs
FPGAs are closer to “programmable ASICs” - large
emphasis on interconnection routing
Timing is difficult to predict - multiple hops vs. the fixed
delay of a CPLD’s switch matrix.
But more “scalable” to large sizes.
FPGA configurable logic blocks have a few inputs
and 1-2 flip-flops, but there are many more of them
compared to the number of macrocells in a CPLD.
43
Common CPLD & FPGA Problems
Pin locking
Small changes, and certainly large ones, can cause the fitter
to pick a different allocation of I/O blocks and pinout.
Locking too early may make the resulting circuit slower or
not fit at all.
Running out of resources
Design may “blow up” if it doesn’t all fit on a single device.
On-chip interconnect resources are much richer than offchip.
Larger devices are exponentially more expensive.
44
FPGAs: Pros
Reasonably Cheap
Good for low-volume parts, more expensive than IC for high-volume
parts
Short Design Cycle (~1sec programming time)
Reprogrammable
Can download bug fix into units you’ve already shipped
Large capacity (4 million gates or so, though we won’t use
any that big)
FPGAs in the lab are “rated” at 300K gates
More flexible than PLDs -- can have internal state
More compact than MSI/SSI
45
FPGAs: Cons
Lower capacity, speed and higher power
consumption than building an integrated circuit
Sub-optimal mapping of logic into CLB’s
Less dense layout and placement due to programmability
Overhead of configurable interconnect and logic blocks
PLDs may be faster than FPGA for designs they can
handle
Need sophisticated tools to map design to FPGA
46