architectural and system synthesis
DESCRIPTION
Architectural and System Synthesis. Camposano, J. Hofstede, Knapp, MacMillen Lin. SOURCES- DeMicheli Mark Manwaring Kia Bazargan Giovanni De Micheli Gupta Youn-Long Lin. Outline. Motivation. Compiling language models into abstract models . - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/1.jpg)
Architectural and System Architectural and System SynthesisSynthesis
SOURCES- DeMicheliMark ManwaringKia BazarganGiovanni De Micheli GuptaYoun-Long Lin
Camposano, Camposano, J. Hofstede, J. Hofstede, Knapp,Knapp,MacMillenMacMillenLinLin
![Page 2: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/2.jpg)
OutlineOutline
• Motivation.
• Compiling language models into abstract models.
• Behavioral-level optimization and program-level transformations.
• Architectural synthesis: an overview.
![Page 3: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/3.jpg)
Architectural SynthesisArchitectural Synthesis
![Page 4: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/4.jpg)
Architectural Synthesis ProblemArchitectural Synthesis Problem• Specification
• A sequencing graph• A set of functional resources• characterized by area and execution delay
• Constraints• Tasks
• Place operations in time and space• Determine detailed interconnection and control
• This is what we need to do in behavioral synthesis! :)
• Constraints include: area, cycle time, latency, and throughput.
• Area: number of modules/resources available or size of your silicon die.
• Cycle time: how fast your clock runs
• Latency: number of cycles for input data to result in a solution or result.
• Throughput: Amount of data that can be processed in a given amount of time (usually involves pipelining)
![Page 5: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/5.jpg)
Behavioral OptimizationBehavioral Optimization
![Page 6: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/6.jpg)
Resource BindingResource Binding• May have a pool of resources larger than required for
problem• Map a constrained set of resources to given
operations• Dedicated resources: each operation is bound to a
single resource.
1. Resource pool: may include various kinds of multipliers (booth, array, etc) adders (tree, carry-lookahead, etc.) multipurpose units (ALUs, multiplier/divider, etc.)
2. Mapping a given set of resources to a set of known operations is one type of problem to solve.
3. Dedicated resource allocation is a one-to-one mapping.
![Page 7: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/7.jpg)
Overview of Hardware SynthesisOverview of Hardware Synthesis
assign times to operations under given constraints
reduce the amount of hardware, optimize the design in general. May be done with the consideration of additional constraints.
assign operations to physical resources under given constraints
![Page 8: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/8.jpg)
Architectural versus Logic SynthesisArchitectural versus Logic Synthesis
• Transform behavioral into structural view.
• Architectural-level synthesis:• Architectural abstraction level.• Determine macroscopic structure.• Example of synthesis: major building blocks.
• Logic-level Logic-level synthesis:• Logic abstraction level.• Determine microscopic structure.• Example of synthesis: logic gate interconnection.
![Page 9: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/9.jpg)
Synthesis and optimizationSynthesis and optimization
![Page 10: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/10.jpg)
Example of HDL description of Example of HDL description of architecturearchitecture
diffeq { read (x, y, u, dx, a); repeat { xl = x + dx; ul = u - (3 * x * u * dx) - (3 * y * dx); yl = y + u * dx; c = x < a; x = xl; u = ul; y = yl; } until ( c ) ; write (y);}
![Page 11: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/11.jpg)
Example of structures to implement this architectureExample of structures to implement this architecture
Processes Processes control and control and
datadata
![Page 12: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/12.jpg)
Principle of scheduling and Principle of scheduling and allocationallocation
+
+
<
*
unittime
1
4
3
2+
+
*
<< *+
+
Scheduling and allocation Control assignment
CDFG
control
Control execution
![Page 13: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/13.jpg)
Scheduling and AllocationScheduling and Allocation
a dcb e f g h
1234
+1
*1
+3
a b c d
e
gf
h
+2
controlstep
1
2
3
A1
A1A2
M1
![Page 14: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/14.jpg)
Internal representationsInternal representations• Internal
representation is design back-bone of synthesis
• Representations• Parse tree• Control-flow
graph (CFG)• Data-flow
graph (DFG, SFG)
• Control/data-flow graph (CDFG)
CDFG( control data flow
graph )
+1+2
+3
*1
a b c d
eg
f
h
e=a+b;g=c+d;f=e+b;h=f*g;
![Page 15: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/15.jpg)
![Page 16: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/16.jpg)
Example of trade-off in architectural designExample of trade-off in architectural design
![Page 17: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/17.jpg)
Architectural-level synthesis Architectural-level synthesis motivationmotivation
• Raise input abstraction level.1. Reduce specification of details.
2. Extend designer base.
3. Self-documenting design specifications.
4. Ease modifications and extensions.
• Reduce design time.
• Explore and optimize macroscopic structure:• Series/parallel execution of operations.
![Page 18: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/18.jpg)
Design Space ExplorationDesign Space Exploration
Arch I
Arch II
Arch III
Dela
y
Area
We consider here totally
different architectures
![Page 19: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/19.jpg)
Stages of architectural-level synthesisStages of architectural-level synthesis
1. Translate HDL models into sequencing graphs.
2. Behavioral-levelBehavioral-level optimization:1. Optimize abstract models independently from
the implementation parameters.
3. ArchitecturalArchitectural synthesis and optimization:1. Create macroscopic structure:
• data-path and control-unit.
2. Consider area and delay information of the implementation. (on the globalglobal level)
![Page 20: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/20.jpg)
Hardware and software compilation.Hardware and software compilation.
software compilation.software compilation.
hardware compilation.hardware compilation.
![Page 21: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/21.jpg)
High Level Synthesis Compilation Flow High Level Synthesis Compilation Flow
![Page 22: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/22.jpg)
Compilation and behavioral Compilation and behavioral optimizationoptimization
• Software compilation:• Compile program into intermediate form.• Optimize intermediate form.• Generate target code for an architecture.
• Hardware compilationHardware compilation:• Compile HDL model into sequencing graph.• Optimize sequencing graph.• Generate gate-level interconnection for a cell library.
![Page 23: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/23.jpg)
CompilationCompilation• Front-end:
1. Lexical and syntax analysis.
2. Parse-tree generation.
3. Macro-expansion.
4. Expansion of meta-variables.
• Semantic analysis:1. Data-flow and control-flow analysis.
2. Type checking.
3. Resolve arithmetic and relational operators.
![Page 24: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/24.jpg)
Parse tree exampleParse tree examplea = p +q a = p +q rr
![Page 25: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/25.jpg)
Behavioral-level optimizationBehavioral-level optimization
• Semantic-preserving transformations aiming at simplifying the model.
• Applied to parse-trees or during their generation.
• Taxonomy:1. Data-flow based transformations.2. Control-flow based transformations.
![Page 26: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/26.jpg)
1. Tree-height reduction.
2. Constant and variable propagation.
3. Common sub-expression elimination.
4. Dead-code elimination.
5. Operator-strength reduction.
6. Code motion.
Data-Flow Based Data-Flow Based Transformations (Transformations (reviewreview))
We will illustrate eachWe will illustrate each
These are transformations
These are transformations
done during compilation.
done during compilation.
There are similar transformations
There are similar transformations
done during optimization to be discussed
done during optimization to be discussed
![Page 27: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/27.jpg)
Tree-height reductionTree-height reduction
• Applied to arithmetic expressions.
• GoalGoal:• Split into two-operand expressions to exploit
hardware parallelism at best.
• Techniques:Techniques:• Balance the expression tree.• Exploit commutativity, associativity commutativity, associativity and and
distributivitdistributivity.y.
![Page 28: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/28.jpg)
Example of tree-height reductionExample of tree-height reductionusing using commutativitycommutativity and and associativityassociativity
x = ( a + (b * c ) ) + d x = (a +d) + (b * c)
![Page 29: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/29.jpg)
Example of tree-height reductionExample of tree-height reductionusing distributivityusing distributivity
x = a * (b c d +e) x = (a b) (c d) + (a e);
![Page 30: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/30.jpg)
Examples of propagationExamples of propagation
• First Transformation type: Constant propagation:• a = 0, b = a +1, c = 2 * b,• a = 0, b = 1, c = 2,
• Second Transformation type: Variable propagation:• a = x, b = a +1, c = 2 * a,• a = x, b = x +1, c = 2 * x,
![Page 31: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/31.jpg)
Sub-expression eliminationSub-expression elimination
• Logic expressions:• Performed by logic optimization.• Kernel-based methods.
• We discussed with factorization
• Arithmetic expressions:• Search isomorphic patterns in the parse trees.
• Example:• a = x +y, b = a +1, c = x +y,• a = x +y, b = a +1, c = a.
![Page 32: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/32.jpg)
Examples of other transformationsExamples of other transformations
• Dead-code elimination:• a = x; b = x +1; c = 2 * x;• a = x; can be removed if not referenced.
• Operator-strength reduction:• a = x 2 ; b = 3 * x;• a = x * x; t = x << 1; b = x + t;
• Code motion:• for (i = 1; i a * b) { }• t = a * b; for (i = 1; i t) { }
• Multiplication only once.
![Page 33: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/33.jpg)
Control- flow based transformationsControl- flow based transformations
1. Model expansion.
2. Conditional expansion.
3. Loop expansion.
4. Block-level transformations. • (will be discussed in more detail separately, presented
on Friday)
Next slides
![Page 34: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/34.jpg)
Model expansionModel expansion• Expand subroutine and flatten hierarchy as the
result.
• Useful to expand scope of other optimization techniques.
• Problematic when routine is called more than once.
• Example of model expansion:Example of model expansion:• x = a +b; y = a * b; z = foo(x; y);• foo(p; q) {t = q - p; return(t); }• By expanding foo:• x = a +b; y = a * b; z = y - x
foo does subtraction
![Page 35: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/35.jpg)
Conditional expansionConditional expansion• Transform conditional into parallel execution with test at the
end.
• Useful when test depends on late signals.
• May preclude hardware sharing.
• Always useful for logic expressions.
• Example:• y = ab; ifif (a) {x = b + d; } else else {x = bd;}
• can be expanded to: x = a(b +d) +a’a’ bd• and simplified as: y = ab; x = y +d(a +b)
Moves conditionals from Moves conditionals from control unit to data pathcontrol unit to data path
![Page 36: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/36.jpg)
Loop expansionLoop expansion
• Applicable to loops with data-independent exit conditions.
• Useful to expand scope of other optimization techniques.
• Problematic when loop has many iterations.• Example of loop expansion:
• x = 0;
for (i = 1; i 3; i ++) {x = x +1; }• Expanded to: x = 0; x = x +1; x = x +2; x = x +3
Can use various variable semantics
![Page 37: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/37.jpg)
What is architectural synthesis and What is architectural synthesis and optimizationoptimization
• Synthesize macroscopic structuremacroscopic structure in terms of building-blocks.
• Explore area/performance trade-offs:1. maximum performance implementations subject to area
constraints.2. minimum area implementations subject to performance
constraints.
• Determine an optimal optimal implementation.
• Create logic modellogic model for data-path and control.
![Page 38: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/38.jpg)
Design space and objectives in Design space and objectives in architectural synthesisarchitectural synthesis
• Design space:• Set of all feasible implementations.
• Implementation parameters:• Area.• Performance:
• Cycle-time.• Latency.• Throughput (for pipelined implementations).
• Power consumption
![Page 39: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/39.jpg)
Three dimensional Design evaluation spaceThree dimensional Design evaluation space
![Page 40: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/40.jpg)
Hardware modelingHardware modeling1. Circuit behavior:
• Sequencing graphs.
2. Building blocks:• Resources.Resources.
3. Constraints:• Timing and resource usage.
Our methods and data structures have to model
them for architectural design
![Page 41: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/41.jpg)
What are What are ResourcesResources??
1. Functional resources:• Perform operations on data.• Example: arithmetic and logic blocks.
2. Memory resources:• Store data.• Example: memory and registers.
3. Interface resources:• Example: busses and ports.
![Page 42: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/42.jpg)
1. Standard resources:• Existing macro-cells.• Well characterized (area/delay).• Example: adders, multipliers, ...
2. Application-specific resources:• Circuits for specific tasks.• Yet to be synthesized.• Example: instruction decoder.
Functional resourcesFunctional resources
![Page 43: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/43.jpg)
Resources and circuit familiesResources and circuit families
• Resource-dominated circuits.• Area and performance depend on few, well-characterized
blocks.• Example: DSP circuits.
• Non resource-dominated circuits.• Area and performance are strongly influenced by sparse logic,
control and wiring.• Example: some ASIC circuits.
![Page 44: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/44.jpg)
Implementation constraintsImplementation constraints
• Timing constraints:• Cycle-time.• Latency of a set of operations.• Time spacing between operation pairs.
• Resource constraints:• Resource usage (or allocation).• Partial binding.
![Page 45: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/45.jpg)
Synthesis in the Synthesis in the temporal temporal domaindomain
• Scheduling:• Associate a start-time with each operation.• Determine latency and parallelism of the
implementation.
• Scheduled sequencing graph:• Sequencing graph with start-time annotation.
Result of scheduling
![Page 46: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/46.jpg)
Example of Synthesis in the temporal domainExample of Synthesis in the temporal domain
ASAP
Here we use
sequencing graph
![Page 47: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/47.jpg)
Synthesis in the spatial domainSynthesis in the spatial domain
1. Binding:• Associate a resource with each operation with the same type.• Determine area of the implementation.
2. Sharing:• Bind a resource to more than one operation.• Operations must not execute concurrently.
3. Bound sequencing graph:• Sequencing graph with resource annotationwith resource annotation.
![Page 48: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/48.jpg)
Example of Synthesis in the spatial domainExample of Synthesis in the spatial domain• First
multiplier
• Second multiplier • Third
multiplier • Fourth multiplier
• First ALU
• Second ALU
• Solution• Four
Multipliers• Two ALUs• Four Cycles
![Page 49: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/49.jpg)
Binding specificationBinding specification• Mapping from the vertex set to the set of resource
instances, for each given type.
1. Partial binding:• Partial mapping,
• given as design constraintdesign constraint.
2. Compatible binding:• Binding which is satisfying the constraints of the partial constraints of the partial
binding.binding.
cont
![Page 50: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/50.jpg)
Example of Binding specificationExample of Binding specification• Binding to the
same multiplier
![Page 51: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/51.jpg)
EstimationEstimation: area, latency, : area, latency, cycle timecycle time
• Resource-dominated circuits.• Area = sum of the area of the resources bound to the
operations.• Determined by binding.
• Latency = start time of the sink operation (minus start time of the source operation).
• Determined by scheduling
• Non resource-dominated circuits.• Area also affected by:
• registers, steering logic, wiring and control.
• Cycle-time also affected by: • steering logic, wiring and (possibly) control.
![Page 52: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/52.jpg)
What are the approaches to What are the approaches to architectural optimization?architectural optimization?
• Architectural Optimization is the Multiple-criteria optimization problem:• area, latency, cycle-time.
• Determine Pareto optimal points:• Implementations such that no other has all parameters
with inferior values.
• Draw trade-off curves:• discontinuous and highly nonlinear.
![Page 53: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/53.jpg)
Approaches to architectural Approaches to architectural optimizationoptimization
1. Area/latency trade-off,• for some values of the cycle-time.
2. Cycle-time/latency trade-off,• for some binding (area).
3. Area/cycle-time trade-off,• for some schedules (latency).
![Page 54: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/54.jpg)
Area/latency trade-off for various cycle timesArea/latency trade-off for various cycle times• Area/Latency
for cycle time=30
• Area/Latency for cycle time=40
Pareto points Pareto points in three in three
dimensionsdimensions
![Page 55: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/55.jpg)
Area-latency trade-offArea-latency trade-off• Rationale:
• Cycle-time dictated by system constraints.
• Resource-dominated circuits:• Area is determined by resource usage.
• Approaches:1. Schedule for minimum latency under resource
constraints2. Schedule for minimum resource usage under latency
constraints • for varying constraints.
![Page 56: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/56.jpg)
Summary on behavioral and Summary on behavioral and architectural synthesis and optimizationarchitectural synthesis and optimization
• Behavioral optimization:• Create abstract models from HDL models.• Optimize models without considering implementation
parameters.
• Architectural synthesis and optimization.• Consider resource parameters.• Multiple-criteria optimization problem:
• area, latency, cycle-time.
![Page 57: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/57.jpg)
High-LevelHigh-Level SynthesisSynthesis
Some authors treat
architectural synthesis as
part of high level synthesis
In some systems there is no architectural
synthesis but there are elements of specialized
high-level synthesis
specialized high-level synthesis:
1. For low power
2. For high testability
3. For high manufacturability
![Page 58: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/58.jpg)
High Level Synthesis for Low Low PowerPower
![Page 59: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/59.jpg)
High Level Synthesis for low High Level Synthesis for low power power
Instructions
Operations
Variables
Arrays
signals
specification
Control
Datapath Memory
Operators, Registers,
Memory, Multiplexor
Control
Scheduling
Hardware allocation
Memory inferencing
Register sharing
Control interencing
for(I=0;I<=2;I=I+1begin
@(posedge clk);
if(fgb[I]%8; begin
p=rgb[I]%8;
g=filter(x,y)*8;
end
............
constraints
high level synthesis
RTL(register transfer level) architecture
![Page 60: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/60.jpg)
Low Power designLow Power design
MUX
Register
DeMux
Control Control
ijk
i*j*k*
C out,Mux C in,Register C out,RegisterC in,DeMux
yx
Power(Register) =
switching(x)(Cout,Mux+Cin,Register)+switching(y) x (Cout,Register+Cin,DeMux)
switching(x)=switching(y) …. Power(Register)=switching(y) x Ctotal
![Page 61: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/61.jpg)
comparison of benchmarks for low comparison of benchmarks for low power synthesis methodspower synthesis methods
0
5
10
15
20
25
Example Cascade F ir11 IIR Wavel e t
Power Reduction%
benchmark 회로
0
10
20
30
40
50
60
70
80
90
100
Example Cascade Fir11 IIR Wavel e t
Non low PowerLow Power
mW
![Page 62: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/62.jpg)
Role of CDFG in High Level Role of CDFG in High Level SynthesisSynthesis
CDFG
Parsing
Transformation
Synthesis
StructuralRTL
BehavioralDescription
This exists in any kind of high level synthesis
![Page 63: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/63.jpg)
![Page 64: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/64.jpg)
Design Flow of specialized high level Design Flow of specialized high level synthesis systemssynthesis systems
• Synthesizable (and executable) specification
• High level verification and design space exploration
• Synthesis / estimation / resynthesis
• Low level validation• formal• simulation
Time-to-market often more Time-to-market often more important than chip areaimportant than chip area
![Page 65: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/65.jpg)
Objective function 1Objective function 1
Main goals in classical approach1. Minimum area
• Functional units, registers, memory, interconnect
2. Maximum speed• Number of clock cycles
Generally one parameter is set as a constrained
and the other one is optimized
![Page 66: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/66.jpg)
More sophisticated Objective functions More sophisticated Objective functions for high-level and system designfor high-level and system design
Additional goals in modern approaches
• More accurate estimation, such as• Size of operands• Sharing of hardware for similar operations (e.g. + and -)
• Testability
• Low power• Power down, clock disabling
• Reliability• Fault tolerance, self-test
• Controller
![Page 67: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/67.jpg)
Other steps in HLSOther steps in HLS• Chaining / multi-cycle operations
• Loop pipelining
• Retiming
• Memory design
• Reset, clock
• Interface design
• Estimation, integration with Logic Synthesis
• Real libraries (Higher level components)
![Page 68: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/68.jpg)
Specification issuesSpecification issues• Timing I/O operations
• Cycle-fixed• Superstate-fixed (pipelined)• Free-floating (order only)
• Clocks
• Resets
• Registered outputs
• Loop pipelining
![Page 69: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/69.jpg)
Behavioral Specification Behavioral Specification LanguagesLanguages
• Add hardware-specific constructs to existing languages• HardwareC
• Popular HDL• Verilog, VHDL
• Synthesis-oriented HDL• UDL/I
![Page 70: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/70.jpg)
VHDL synthesis toolsVHDL synthesis toolsRTL-synthesis• FU allocation• Limited register
allocation• Interconnect
allocation• Binding• Logic and physical
synthesis
Behavioral synthesis• HL Optimizations• Scheduling• RTL-synthesis
![Page 71: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/71.jpg)
ChipChip Synthesis Synthesis
Now every stage ofsynthesis must take
space into consideration
Many issues do not exist in FPGA or architectural synthesis that use ready blocks but they exist in VLSI chip design.
System on a chip
![Page 72: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/72.jpg)
Layout, pins, power, temperature,Reliability, manufacturability,
testability, test generation
Chip synthesis
![Page 73: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/73.jpg)
Layout and partitioning must be considered, must be
iterated
![Page 74: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/74.jpg)
Various models are used in the same synthesis process
![Page 75: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/75.jpg)
![Page 76: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/76.jpg)
Structure to Structure to layoutlayout
![Page 77: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/77.jpg)
![Page 78: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/78.jpg)
Software engineering
![Page 79: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/79.jpg)
System SynthesisSystem SynthesisSYSTEM
specification for a robot
![Page 80: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/80.jpg)
SYSTEM specification for
a robot
Several ASIC chips are part of the entire
system automatically
designed
![Page 81: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/81.jpg)
SYSTEM SYSTEM specification specification for a robotfor a robot
![Page 82: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/82.jpg)
Modern Modern Experimental Experimental
High-Level High-Level Synthesis Synthesis
SystemSystem
![Page 83: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/83.jpg)
System selects System selects interactively or interactively or
automatically the automatically the realization realization
technology or technology or mixture of themmixture of them
![Page 84: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/84.jpg)
To allow communication and integration, user’s feedback
![Page 85: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/85.jpg)
System SynthesisSystem SynthesisSystem in
chip versus system using a chip-set
Variants of the robot system
![Page 86: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/86.jpg)
Decomposition is Decomposition is not the same as not the same as
partitioningpartitioning
System “knows” System “knows” typical blocks typical blocks
and libraries of and libraries of commercial commercial componentscomponents
![Page 87: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/87.jpg)
Example of a System-on-a-ChipExample of a System-on-a-Chip
Processor Memory
ExternalMemoryInterface
IPBus Master UART
WirelessB
ridg
e US
B
Everything in one chip – floorplanning and communicationEverything in one chip – floorplanning and communication
![Page 88: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/88.jpg)
SOC with SOC with PLDsPLDs
Processor Memory
ExternalMemoryInterface
FPGAFPGABus Master FPGAFPGA
WirelessB
ridg
e US
B
Everything in one chip – two FPGAs are inside, reconfigurable dynamicallyEverything in one chip – two FPGAs are inside, reconfigurable dynamically
![Page 89: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/89.jpg)
WaferFoundry
SystemHouses/
IC Vendors(Fabless)
Integrators
Library/IP
Vendors(Chipless)
EDAVendors Paradigm ShiftParadigm Shift
Move of EDA vendors to production
![Page 90: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/90.jpg)
![Page 91: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/91.jpg)
![Page 92: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/92.jpg)
Essential Current and Open Essential Current and Open Issues in Design AutomationIssues in Design Automation• Behavioral Specification Languages
• From Matlab to chip, from Prolog to chip, etc.
• Target Architectures• Network on a chip, sensors and motion control integrated.
• Intermediate Representation• For users to exchange, to understand the design better
• Operation Scheduling• On the level of complex operations such as transforms or filters.
• Allocation/Binding• On many levels of operations and processors
• Control Generation• State machine optimization for large controllers• New technologies , integrate FSM-logic-layout
Still Still areas of areas of active active
researchresearch
![Page 93: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/93.jpg)
Future researchFuture research areas in High areas in High Level SyntesisLevel Syntesis
• System level design • Software-hardware system co-design
• Reuse• Intellectual Property (IP) or Virtual Components (VC)
• More emphasis on verification• currently often > 60% of design effort• correctness by construction
![Page 94: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/94.jpg)
Future Research: IP and Future Research: IP and SynthesisSynthesis
• Authoring IP for Synthesis
• Synthesis utilizing IP
• Synthesizing IPs
Executable Data Sheets
![Page 95: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/95.jpg)
Executable Data SheetsExecutable Data Sheets
IP
IP WrapperMore thanjust thePort Interface
![Page 96: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/96.jpg)
Future Directions for system Future Directions for system designdesign
• Realistic Methodology• Evolutional Transition from Current Practice• Domain Specific
• IP-Centric• As both Authoring Aid and Integrator
• Software• Co-design and Code Generation
Needs better collaboration of
research universities and
companies
![Page 97: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/97.jpg)
![Page 98: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/98.jpg)
[1] D. Gajski and N. Dutt, High-level Synthesis : Introduction to Chip and System Design. Kluwer Academic Publishers, 1992.
[2] G. D. Micheli, Synthesis and Optimization of Digital Circuits. New York : McGraw Hill. Inc, 1994.
[3] A. P. Chandrakasan, S. Sheng, and R. W. Brodersen, "Low-Power CMOS digital design", IEEE J. of Solid-State Circuits, pp. 473-484, 1992.
[4] A. P. Chandrakasan, M. Potkonjak, R. Mehra, J. Rabaey, and R. W. Brodersen, "Optimizing power using transformation," IEEE Tr. on CAD/ICAS, pp. 12-31, Jan. 1995.
[5] E. Musool and J. Cortadella, "Scheduling and resource binding for low power", Int'l Symp on Synstem Syntheiss, pp. 104-109, Apr. 1995.
[6] Y. Fang and A. Albicki, "Joint scheduling and allocation for low power," in Proc. of Int'l Symp. on Circuits & Systems, pp. 556-559, May. 1996.
[7] J. Monteiro and Pranav Ashar, "Scheduling techniques to enable power management", 33rd Design Automation Conference, 1996.
[8] R. S. Martin, J. P. Knight, "Optimizing Power in ASIC Behavioral Synthesis", IEEE Design & Test of Computers, pp. 58-70, 1995.
[9] R. Mehra, J. Rabaey, "Exploting Regularity for Low Power Design", IEEE Custom Integrated Circuits Conference, pp.177-182. 1996.
[10] A. Chandrakasan, T. Sheng, and R. W. Brodersen, "Low Power CMOS Digital Design", Journal of Solid State Circuits, pp. 473-484, 1992.
[11] R. Mehra and J. Rabaey, "Behavioral level power estimation and exploration," in Proc. of Int'l Symp. on Low Power Design, pp. 197-202, Apr. 1994.
[12] A. Raghunathan and N. K. Jha, "An iterative improvement algorithm for low power data path synthesis," in Proc. of Int'l Conf. on Computer-Aided Design, pp. 597-602, Nov. 1995.
[13] R. Mehra, J. Rabaey, "Low power architectural synthesis and the impact of exploiting locality," Journal of VLSI Signal Processing, 1996.
LiteratureLiterature
![Page 99: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/99.jpg)
[14] M. B. Srivastava, A. P. Chandrakasan, and R. W. Brodersen, "Predictive system shutdown and other architectural techniques for energy efficient programmable computation," IEEE Tr. on VLSI Systems, pp. 42-55, Mar. 1996.
[15] A. Abnous and J. M. Rabaey, "Ultra low power domain specific multimedia processors," in Proc. of IEEE VLSI Signal Processing Workshop, Oct. 1996.
[16] M. C. Mcfarland, A. C. Parker, R. Camposano, "The high level synthesis of digital systems," Proceedings of the IEEE. Vol 78. No 2 , February, 1990.
[17] A. Chandrakasan, S. Sheng, R. Brodersen, "Low power CMOS digital design,", IEEE Solid State Circuit, April, 1992.
[18] A. Chandrakasan, R. Brodersen, "Low power digital CMOS design, Kluwer Academic Publishers, 1995.
[19] M. Alidina, J. Moteiro, S. Devadas, A. Ghosh, M. Papaefthymiou, "Precomputation based sequential logic optimization for low power," IEEE International Conference on Computer Aided Design, 1994.
[20] J. Monterio, S. Devadas and A. Ghosh, "Retiming sequential circuits for low power," In Proceeding of the IEEE International Conference on Computer Aided Design, November, 1993.
[21] F. J. Kurdahi, A. C. Parker, REAL: A Program for Register Allocation,: in Proc. of the 24th Design Automation Conference, ACM/IEEE, June. pp. 210-215, 1987.
[22] A. Wolfe. A case study in low-power system level design. In Proc.of the IEEE International Conference on Computer Design, Oct., 1995.
[23] T.D. Burd and R.W. Brothersen. Energy ecient CMOS micropro-cessor design. In Proc. 28th Annual Hawaii International Conf. On System Sciences, January 1995.
[24] A. Dasgupta and R. Karri. Simultaneous scheduling and binding for power minimization during microarchitectural synthesis. In Int. Symposium on Low Power Design, pages 69-74, April 1995.
[25] R.S. Martin. Optimizing power consumption, area and delay in behavioral synthesis. PhD thesis, Department of Electronics, Faculty of Enginering, Carleton University, March 1995.
[26] A. Matsuzawa. Low-power portable design. In Proc. International Symposium on Advanced Research in Asynchronous Circuits and Systems, March 1996. Invited lecture.
[27] J.D. Meindl. Low-power microelectronics: retrospect and prospect. Proceedings of the IEEE 83(4):619-635, April 1995.
![Page 100: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/100.jpg)
Exam Exam Problem 1Problem 1
• Write set of equations for solving some type of equations by an iterative method
• Find Data Flow Graph for this set of equations
• Schedule• Allocate• Bind and share• Design final data path• Find control unit• Optimize partitioning and
communication
• Too long for one exam.• Can be a take-home exam
![Page 101: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/101.jpg)
Exam Exam ProbleProblem 2m 2
1. Allocate to time
2. Allocate to logic blocks
3. Design a complete controller
4. Design a controller for pipelined design
![Page 102: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/102.jpg)
Exam Problem 3: SchedulingExam Problem 3: Scheduling• Set area constraint
• 2 multipliers• 2 general-purpose ALUs
• Set the cycle time = latency of a multiplier• Goal: minimize latency of circuit
![Page 103: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/103.jpg)
Exam Exam Problem Problem 44
1. Give the set of functional resources: two multipliers, two ALUs.2. Scheduling example with the constraints (two set constraints,
then optimize the third)3. We need to maintain the data dependencies. (e.g. vertex 6
must be scheduled at least one cycle after vertex 1.)4. This is the same differential equation dataflow graph from a
previous slide.5. Edges that are not necessary to !show dependencies between
vertices have been removed.6. Complete this problem
![Page 104: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/104.jpg)
Exam Problem 5: BindingExam Problem 5: Binding
![Page 105: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/105.jpg)
Exam Problem 5Exam Problem 5
![Page 106: Architectural and System Synthesis](https://reader035.vdocuments.mx/reader035/viewer/2022070405/56813e2e550346895da81043/html5/thumbnails/106.jpg)
Second Exam Problem 5: Second Exam Problem 5: Competition for StudentsCompetition for Students
1. The student with the smallest area gets a prize, the student with the smallest latency gets a prize.
2. Bring exam submission to the next lecture to be eligible for competition.
3. You are not required to give an optimal solution, since that may prove to be more difficult than can be done in a reasonable amount of time.
Not for this year