Download presentation

Presentation is loading. Please wait.

Published byCaleb Crymes Modified over 3 years ago

1
© 2000 M. CiesielskiDatapath Synthesis1 ANALYTICAL APPROACH TO CUSTOM DATAPATH SYNTHESIS Serkan Askar, Maciej Ciesielski Department of Electrical & Computer Engineering University of Massachusetts, Amherst {saskar, ciesiel}@ecs.umass.edu

2
© 2000 M. CiesielskiDatapath Synthesis2 Outline Introduction Problem definition Approach, design phases Analytical formulation (MILP) Results Conclusions and future work

3
© 2000 M. CiesielskiDatapath Synthesis3 Motivation Datapath: highly regular structure – array of bit/row slices – rows: identical cells – hand-crafted layout Datapath cell power rails Limitations of custom design – ability of the designer – slow development time

4
© 2000 M. CiesielskiDatapath Synthesis4 Typical Datapath Cell - schematic General topoloy generated by floorplanning –power rails –bristles Fixed width (pitch)

5
© 2000 M. CiesielskiDatapath Synthesis5 Datapath Cell - layout

6
© 2000 M. CiesielskiDatapath Synthesis6 Problem Definition Automation of transistor level placement –diffusion-limited, routing ignored (to certain degree) Goal: – minimize layout area, size – improve manufacturability, routability Constraints – bit-slice constraints – component constraints – custom design issues fixed

7
© 2000 M. CiesielskiDatapath Synthesis7 Previous Work Very little work on custom quality datapath synthesis Stochastic methods – simulated annealing – genetic algorithms Deterministic methods – analytical (QP, ILP, etc) – constructive methods

8
© 2000 M. CiesielskiDatapath Synthesis8 Bit-slice Constraints Fixed width (pitch) – power rails Vertical bristles – data lines Horizontal bristles – control lines control lines pitch VDDGND data lines

9
© 2000 M. CiesielskiDatapath Synthesis9 Component Types Transistors, logic gates Transistor chains, folded transistors

10
© 2000 M. CiesielskiDatapath Synthesis10 Component Geometries RectangularL-shape

11
© 2000 M. CiesielskiDatapath Synthesis11 Custom Design Issues Multiple instances – chaining, folding Merging ( same diffusion type, sharing common net) –diffusion sharing (same cell)

12
© 2000 M. CiesielskiDatapath Synthesis12 Custom Design Issues (cont’d) Component sharing –combine components from adjacent cells with common independent net VDD GND cell icell i+1

13
© 2000 M. CiesielskiDatapath Synthesis13 Design Objective Secondary objectives: – performance – routability – well minimization Main objective: – minimize height of datapath cell fixed min

14
© 2000 M. CiesielskiDatapath Synthesis14 Post-processing Connectivity - no geometry info Geometry - no connectivity info Managing complexity Design Flow Initial relative placement Component grouping Geometric placement Final layout Well minimization Net length minimization Pre-processing Input files

15
© 2000 M. CiesielskiDatapath Synthesis15 Pre-processing Connectivity analysis Initial component merging (to minimize number of objects to place)

16
© 2000 M. CiesielskiDatapath Synthesis16 Initial Relative Placement Derive a relative initial placement – connectivity to components and bristles – geometry is ignored Standard force-directed technique – component locations: centers of net gravity

17
© 2000 M. CiesielskiDatapath Synthesis17 Initial Relative Placement Generates relative placement of components based on electrical connectivity

18
© 2000 M. CiesielskiDatapath Synthesis18 Component Grouping Limit the number of relations between components by grouping them together for placement before grouping – 6 relations after grouping – 3 relations

19
© 2000 M. CiesielskiDatapath Synthesis19 Geometric Placement Generate non-overlapping placement –component geometry –design rules –fixed connectivity (initial relative placement) Model: Mixed Integer Linear Program (MILP)

20
© 2000 M. CiesielskiDatapath Synthesis20 Component Modeling Component modeled as a pair of squares R i = orientation parameter X i2, Y i2 X i1, Y i1 R i = 1 R i = 0 X i1 - X i2 = (D i - d i ) * (1- R i ) Y i1 - Y i2 = (D i - d i ) * R i d i = min (W i, H i ) D i = max (W i, H i ) where

21
© 2000 M. CiesielskiDatapath Synthesis21 Geometric Orientation R i = orientation parameter X, Y R i = 1 Xdim Ydim Xdim – perpendicular to poly Ydim – along poly gate WiWi HiHi X, Y R i = 0 Xdim Ydim W i = Xdim i R i + Ydim (1- R i ) H i = Ydim i R i + Xdim (1- R i )

22
© 2000 M. CiesielskiDatapath Synthesis22 Modeling of L-shape Components Modeled as a pair of abutting rectangles H2H2 W1W1 W2W2 H1H1 P i = P j = 1 W1W1 H1H1 H2H2 W2W2 P i = P j = 0 X 2 – X 1 = W 2 P - (H 1 – H 2 ) (1- P) Y 2 – Y 1 = W 2 (1 - P) - (H 1 - H 2 ) P P = P 1 = P 2

23
© 2000 M. CiesielskiDatapath Synthesis23 Selection of Component Instances Selection parameter, S i (k = 2 instances) Easy extension to arbitrary k. S i = 0S i = 1

24
© 2000 M. CiesielskiDatapath Synthesis24 Boundary Constraints bi < 0 for sharable components bi > 0 for others bi Margin: X i – W i > bi X i < X0 – bi Y i – H i > bi Y i < Y0 – bi For each component i Recall: W i, H i = f(R i )

25
© 2000 M. CiesielskiDatapath Synthesis25 Non-overlapping Constraints i Q ij = 0 j Q ij = 1 j Binary variable Q ij = 0 (vert), = 1 (horiz) X j - X i >= DimX j - L * (1 - Q ij ) + sep Y j - Y i >= DimY j - L * Q ij + sep

26
© 2000 M. CiesielskiDatapath Synthesis26 Optimization Problem (MILP) Minimize –the cell height, Y0 Subject to – boundary constraints – non-overlapping constraints – component shape constraints – component selection constraints fixed Y0

27
© 2000 M. CiesielskiDatapath Synthesis27 Sample Flow

28
© 2000 M. CiesielskiDatapath Synthesis28 Final layout

29
© 2000 M. CiesielskiDatapath Synthesis29 MILP: Complexity Issues Reduce number of placeable components and variables – merging (pre-processing) Reduce number of integer variables – component grouping Q AB Group AGroup B Q xy = Q AB for x A and y B

30
© 2000 M. CiesielskiDatapath Synthesis30 Results - routed layouts PG9PB7CS15MU9 Cell height

31
© 2000 M. CiesielskiDatapath Synthesis31 Conclusion & Future Work Automation of datapath layout – fast – within acceptable area overhead (5 - 10%) Post-processing Iterative improvement Hierarchical placement Grouping algorithm

32
© 2000 M. CiesielskiDatapath Synthesis32 Post-processing Improve the result of geometric placement Well minimization: improve manufacturability Mirroring along both axes Efficient MILP model and formulation Net length minimization: simplify routing Mirroring along both axes Constrained optimization problem Simple but effective MILP model and formulation

33
© 2000 M. CiesielskiDatapath Synthesis33 Post-processing –Net length minimization Mirroring Swapping Moving –Well minimization Mirroring –Layout compaction

34
© 2000 M. CiesielskiDatapath Synthesis34 Well minimization – Model

35
© 2000 M. CiesielskiDatapath Synthesis35 Well minimization – overlap weights Mirroring affects the overlap weights

36
© 2000 M. CiesielskiDatapath Synthesis36 Well minimization – Formulation Objective: Max Z ij subject to Z ij = (1 - m i - m j + m ij ) v ij 0 + (m i - m ij ) v ij 1 + (m j - m ij ) v ij 2 + m ij v ij 3 where m i = 1 if i is mirrored 0 otherwise (i,j) m ij = m i AND m j m ij <= m i m ij <= m j m ij m i + m j - 1

37
© 2000 M. CiesielskiDatapath Synthesis37 Well minimization – Example

38
© 2000 M. CiesielskiDatapath Synthesis38 Net length minimization - Approach Net length model –Half perimeter of the bounding box of the net scope (pins) Minimize the weighted sum of net length

39
© 2000 M. CiesielskiDatapath Synthesis39 Net length minimization - Modeling Mirroring effects the pin location

40
© 2000 M. CiesielskiDatapath Synthesis40 Net length minimization - Formulation Objective: Min L i subject to Xp j = Xc q - (1 - Mx q - My q + Mxy q ) d x + (Mx q – Mxy q ) d x - (My q – Mxy q ) d x + Mxy q d x where : j Pins, q Comps,Mxy = Mx AND My Xc q = center coordinate of component q i Nets

41
© 2000 M. CiesielskiDatapath Synthesis41 Net length minimization – Example

42
© 2000 M. CiesielskiDatapath Synthesis42 Results - interval overlaps Interval overlap PG9PB7CS15MU9CMU82X

43
© 2000 M. CiesielskiDatapath Synthesis43 Results - net length PG9PB7CS15MU9 Net length CMU82X

44
© 2000 M. CiesielskiDatapath Synthesis44 Conclusions & Future Work Automation of datapath layout – fast within acceptable area overhead Improvement –routability, manufacturability –delay, area Component swapping Global routing phase after post-processing Iterative group relaxation Efficient algorithms for grouping

Similar presentations

OK

1 Exploring Custom Instruction Synthesis for Application-Specific Instruction Set Processors with Multiple Design Objectives Lin, Hai Fei, Yunsi ACM/IEEE.

1 Exploring Custom Instruction Synthesis for Application-Specific Instruction Set Processors with Multiple Design Objectives Lin, Hai Fei, Yunsi ACM/IEEE.

© 2018 SlidePlayer.com Inc.

All rights reserved.

To ensure the functioning of the site, we use **cookies**. We share information about your activities on the site with our partners and Google partners: social networks and companies engaged in advertising and web analytics. For more information, see the Privacy Policy and Google Privacy & Terms.
Your consent to our cookies if you continue to use this website.

Ads by Google

Ppt on history of pie in math Ppt on human nutrition and digestion of flat Pptx to ppt online free Ppt on food trucks Ppt on art of war sun Ppt on waste management in india Ppt on obesity prevention source Free ppt on team building and leadership Ppt on endangered animals of india Ppt on current account deficit in australia