Advanced Sequential Circuit Design
Lecturer: Tzu-Hsuan, Hung
System Integration and Silicon Implementation (Si2) Lab
Institute of Electronics
National Yang Ming Chiao Tung University, Hsinchu, Taiwan
1
Outline
Section 1- Timing
Section 2- Designware
ICLAB NCTU Institute of Electronics 2
Outline
Section 1- Timing
Setup/hold time
Pipeline
Section 2- Designware
ICLAB NCTU Institute of Electronics 3
Timing of D Flip-Flop
Term definition
Setup time (tsetup): The time that the input signal must be stabilized
before the clock edge.
Hold time (thold): The time that the input signal must be stabilized
after the clock edge.
Clk-to-Q contamination delay (tccq): The contamination time that Q
is first changed after the clock edge.
Clk-to-Q propagation delay (tpcq): The propagation time that Q
reaches steady state after the clock edge.
ICLAB NCTU Institute of Electronics 4
Timing Violation
Timing violation Q_1
D_2
Tpd: Logic Propagation Delay
Tcd: Logic Contamination Delay
Tskew: Difference between CLK_1 and CLK_2
ICLAB NCTU Institute of Electronics 5
Timing Check (1/2)
Setup time check
The $setup system task determines whether a data signal remains
stable for a minimum specified time before a transition in an
enabling, such as a clock event.
in this red region
Hold time check
The $hold system task determines whether a data signal remains
stable for a minimum specified time after a transition in an
enabling signal, such as a clock event.
in this red region
ICLAB NCTU Institute of Electronics 6
Timing Check (2/2)
Timing report: setup time
Timing report: hold time Slacks should be MET!
(non-negative)
ICLAB NCTU Institute of Electronics 7
Setup Time Criterion
Setup time margin: ( Tcycle + tskew ) > ( tpcq + tpd +
tsetup )
ICLAB NCTU Institute of Electronics 8
Hold Time Criterion
Hold time margin: ( tccq + tcd ) > ( thold + tskew )
tcc tc
q d
ICLAB NCTU Institute of Electronics 9
Adjust data path to meet the constraints
Setup violation too many works in one cycle
Apply pipelining
Hold violation insufficient delay
add delays the violated path, such as buffers/inverters/Muxes
Increase clock period for setup violation
In most practical cases, hold violations are fixed
during the backend work (after clock tree synthesis)
ICLAB NCTU Institute of Electronics 10
Outline
Section 1- Timing
Setup/hold time
Pipeline
Section 2- Designware
ICLAB NCTU Institute of Electronics 11
Trade-off between Area and Timing
Area: 1 unit
Time: 40 mins (Wash: 20 mins + Dry: 20 mins)
ICLAB NCTU Institute of Electronics 12
Trade-off between Area and Timing
0 40 80 120 160
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 13
Trade-off between Area and Timing
0 40 80 120 160
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 14
Trade-off between Area and Timing
0 40 80 120 160
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 15
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 16
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash and Dry = 40 mins
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 17
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash and Dry = 40 mins
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 18
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash and Dry = 40 mins
Area: 2 units
Time: 80 mins
Wash and Dry = 40 mins
ICLAB NCTU Institute of Electronics 19
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Area 0.7 units
Dry 20 mins
Area 0.7 units
ICLAB NCTU Institute of Electronics 20
Trade-off between Area and Timing
0 20 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Dry 20 mins
ICLAB NCTU Institute of Electronics 21
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Dry 20 mins
ICLAB NCTU Institute of Electronics 22
Trade-off between Area and Timing
0 40 60 80 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Dry 20 mins
ICLAB NCTU Institute of Electronics 23
Trade-off between Area and Timing
0 40 80 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Dry 20 mins
ICLAB NCTU Institute of Electronics 24
Trade-off between Area and Timing
0 40 80 100 120 160
Area: 1 unit
Time: 160 mins
Wash 20 mins
Area: 0.7+0.7 =1.4 units
Time: 100 mins
Dry 20 mins
ICLAB NCTU Institute of Electronics 25
Trade-off between Area and Timing
Area: 1 unit
Basic
Time: 160 mins
Area: 2 units
Parallel
Time: 80 mins
Area: 0.7+0.7 = 1.4 units
Pipeline
Time: 100 mins
ICLAB NCTU Institute of Electronics 26
Trade-off between Area and Timing
a [7:0] , b [7:0] , c [3:0] , d [3:0]
Q: (a + b + c + d) x 4 iterations ?
Basic
Parallel
Pipeline
ICLAB NCTU Institute of Electronics 27
Trade-off between Area and Timing
a [7:0] , b [7:0] , c [3:0] , d [3:0]
Q: (a + b + c + d) x 4 iterations ?
5
c 5-bit 9
Add 10
d 9-bit
Area: 24 units
er
Basic a Adder 10-bit output
Time: 24 x 4 = 96 units
b Adder
Parallel
Pipeline
ICLAB NCTU Institute of Electronics 28
Trade-off between Area and Timing
a [7:0] , b [7:0] , c [3:0] , d [3:0]
Q: (a + b + c + d) x 4 iterations ?
5
c 5-bit 9
Add 10
d 9-bit
Area: 24 units
er
Basic a Adder 10-bit output
Time: 24 x 4 = 96 units
b Adder
a 9-bit
c Adder 10-bit Area: 28 units
output
Parallel b 9-bit Adder
Adder
Time: 19 x 4 = 76 units
d
Pipeline
ICLAB NCTU Institute of Electronics 29
Trade-off between Area and Timing
a [7:0] , b [7:0] , c [3:0] , d [3:0]
Q: (a + b + c + d) x 4 iterations ?
5
c 5-bit 9
Add 10
d 9-bit
Area: 24 units
er
Basic a Adder 10-bit output
Time: 24 x 4 = 96 units
b Adder
a 9-bit
c Adder 10-bit Area: 28 units
output
Parallel b 9-bit Adder
Adder
Time: 19 x 4 = 76 units
d
c 5-bit
Add
d
Pipeline a
er 9-bit
Adder 10-bit output
b Adder
ICLAB NCTU Institute of Electronics 30
Trade-off between Area and Timing
5
9
c 5-bit 10
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
ICLAB NCTU Institute of Electronics 31
Trade-off between Area and Timing
c 5-bit
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
T=1 c2 + d2 a1 + (c1 + d1)
ICLAB NCTU Institute of Electronics 32
Trade-off between Area and Timing
c 5-bit
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
T=1 c2 + d2 a1 + (c1 + d1)
T=2 c3 + d3 a2 + (c2 + d2) b1 + (a1 + c1 + d1)
ICLAB NCTU Institute of Electronics 33
Trade-off between Area and Timing
c 5-bit
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
T=1 c2 + d2 a1 + (c1 + d1)
T=2 c3 + d3 a2 + (c2 + d2) b1 + (a1 + c1 + d1)
T=3 c4 + d4 a3 + (c3 + d3) b2 + (a2 + c2 + d2)
ICLAB NCTU Institute of Electronics 34
Trade-off between Area and Timing
c 5-bit
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
T=1 c2 + d2 a1 + (c1 + d1)
T=2 c3 + d3 a2 + (c2 + d2) b1 + (a1 + c1 + d1)
T=3 c4 + d4 a3 + (c3 + d3) b2 + (a2 + c2 + d2)
T=4 a4 + (c4 + d4) b3 + (a3 + c3 + d3)
ICLAB NCTU Institute of Electronics 35
Trade-off between Area and Timing
c 5-bit
Pipe d Adder
9-bit
a Adder out
line 10-bit
b Adder put
T=0 c1 + d1
T=1 c2 + d2 a1 + (c1 + d1)
T=2 c3 + d3 a2 + (c2 + d2) b1 + (a1 + c1 + d1)
T=3 c4 + d4 a3 + (c3 + d3) b2 + (a2 + c2 + d2)
T=4 a4 + (c4 + d4) b3 + (a3 + c3 + d3)
T=5 b4 + (a4 + c4 + d4)
ICLAB NCTU Institute of Electronics 36
Trade-off between Area and Timing
a [7:0] , b [7:0] , c [3:0] , d [3:0]
Q: (a + b + c + d) x 4 iterations ?
5
c 5-bit 9
Add 10
d 9-bit
Area: 24 units
er
Basic a Adder 10-bit output
Time: 24 x 4 = 96 units
b Adder
a 9-bit
c Adder 10-bit Area: 28 units
output
Parallel b 9-bit Adder
Adder
Time: 19 x 4 = 76 units
d
c 5-bit
Add Area: 24+ units
d
Pipeline a
er 9-bit
Adder 10-bit output Time: 60 units
b Adder
ICLAB NCTU Institute of Electronics 37
Outline
Section 1- Timing
Section 2- Designware
ICLAB NCTU Institute of Electronics 38
Overview of DesignWare
IP (Intellectual Property )
Hard IP : GDSII format, high performance but technology dependent.
Firm IP : Netlist resource, less used.
Soft IP : RTL design, requires verification.
DesignWare library
Provides synthesizable and verification IPs.
Supports the method to optimize the area or the speed and reduce the
timing.
DesignWare IP library categories
Building Block IPs (formally called Foundation Library)
CoreTools
Implementation IPs
Smart Model Library
Memory Models
AMBA OCB Family
Verification IPs
ICLAB NCTU Institute of Electronics 39
DesignWare Building Block IPs (1/2)
DesignWare building block IPs
A collection of reusable IP blocks integrated into the SYNOPSYS
synthesis environment.
Characteristics
Pre-verified for quality and better quality of results (QOR) in
synthesis, decreasing design and technology risk.
Allows high-level optimization of performance during synthesis.
Increased design reusability, productivity
Parameterized in size and also in functionality for some IP
Technology-independent
Provide synthesizable models, simulation models, datasheets, and
examples.
ICLAB NCTU Institute of Electronics 40
DesignWare Building Block IPs (2/2)
Library categories
Basic Library : A set of components bundled with HDL
Compiler that implements several common
arithmetic and logic functions.
Logic : Combinational and sequential components
Math : Arithmetic and trigonometric components
Memory : Registers, FIFOs, and FIFO controllers, sync. And
async. RAMs and stack components.
DSP Library : Digital filters for digital signal processing (DSP)
applications, ex: FIR, IIR filter
Application Specific: Data integrity, interface, and JTAG components.
GTECH Library : Genetic technology library, a technology-
Independent, gate-level library.
ICLAB NCTU Institute of Electronics 41
Usage of DesignWare Building Block IP
Usage of DesignWare Building Block IP
Operator inference
Convenient, but sometimes it is inefficient when synthesizing.
Supply default function only, can not use special function.
Instantiate IP
Use SYNOPSYS design compiler shell script.
Supply different architecture for implementation.
Applying pre-compiling sub-blocks speeds up the synthesis for large design.
ICLAB NCTU Institute of Electronics 42
Operator Inference (1/3)
Operator inference
Use the HDL operator in description, and the operator
must include in synthetic operator definition.
HDL compiler will infer synthetic operator in HDL code.
HDL compiler supply high-level synthesis.
The / operator is required for the DesignWare license.
The HDL operator defined in standard synthetic operator:
Synthetic Operators HDL Operator
adder +, +1
subtractor -, -1
comparator ==, <, <=, >, >=
multiplier *
selector If, case
ICLAB NCTU Institute of Electronics 43
Operator Inference (2/3)
High-level synthesis
Arithmetic optimization
Arithmetic level optimization, ex: a+b+c+d -> (a+b)+(c+d)
Resource sharing
Allows similar operations that do not overlap in time to be carried out
by the same physical hardware.
ICLAB NCTU Institute of Electronics 44
Operator inference (3/3)
High-level synthesis flow
ICLAB NCTU Institute of Electronics 45
Instantiate IP (1/9)
Instantiation IP
To instantiate a synthetic module manually and explicitly.
Need to include a reference to the synthetic module in HDL code.
SYNOPSYS online document
Command:
evince /RAID2/EDA/synopsys/synthesis/2020.09/dw/doc/manuals/dwbb_userguide.pdf &
remember execute Xwin and setenv DISPLAY your_IP:0
ICLAB NCTU Institute of Electronics 46
Instantiate IP (2/9)
SYNOPSYS online document
Select section1.6
ICLAB NCTU Institute of Electronics 47
Instantiate IP (3/9)
ICLAB NCTU Institute of Electronics 48
Instantiate IP (4/9)
Module
name
input & output
Argument
assignment:
DW02_ mult #(N,N)
ICLAB NCTU Institute of Electronics 49
Instantiate IP (5/9)
User implementation type
specification
Simulation model path specification
Functional parameter specification
ICLAB NCTU Institute of Electronics 50
Instantiate IP (6/9)
Instantiate module
Instantiate the synthetic module and specify parameters defined in
document.
Table1-
2
Table1-1 I/O port
ICLAB NCTU Institute of Electronics 51
Instantiate IP (7/9)
RTL behavior simulation
Specify the behavioral simulation models (Table1-4).
Absolute path
Relative path
Absolute path
`include /usr/synthesis/dw/sim_ver/<model_name>.v
Relative path
`include <model_name>.v
Command: irun <file_name>.v incdir <directory>
Ex : irun DW02_multi_inst.v incdir /usr/synthesis/dw/sim_ver/
ICLAB NCTU Institute of Electronics 52
Instantiate IP (8/9)
Synthesis
Apply //synopsys translate_off
//synopsys translate_on
//synopsys translate_off (DA synthesis off)
//synopsys translate_on (DA synthesis on)
Set the implementation type of IP
User specify the implementation type of IP manually.
//synopsys dc_script_begin
//set_implementation wall U1 (instance name of IP)
implementation type from (Table1-3)
//synopsys dc_script_end
ICLAB NCTU Institute of Electronics 53
Instantiate IP (9/9)
Example
RTL/Gate simulation description
//synopsys translate_off
(Table1-4)
//synopsys translate_on
module SignedMultiplier(a, b, product);
input [7 : 0] a;
input [7 : 0] b;
output [15: 0] product;
(cell name) (Table1-2) (Table1-1)
//synopsys dc_script_begin
//set_implementation csa U1
(Table1-3)
//synopsys dc_script_end
Note : If you use Designware, you should use clean
command after each simulation. (./09_clean_up)!
endmodule
ICLAB NCTU Institute of Electronics 54
