1 / 34

ECE 551: Digital System Design & Synthesis

ECE 551: Digital System Design & Synthesis . Lecture Set 9 9.1: Constraints and Timing (In separate file) 9.2: Optimization - Part 1 9.3: Optimization - Part 2. ECE 551 - Digital System Design & Synthesis Lecture 9.2 - Optimization and Timing Analysis. Overview

betty_james
Télécharger la présentation

ECE 551: Digital System Design & Synthesis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ECE 551: Digital System Design & Synthesis Lecture Set 9 9.1: Constraints and Timing (In separate file) 9.2: Optimization - Part 1 9.3: Optimization - Part 2

  2. ECE 551 - Digital System Design & SynthesisLecture 9.2 - Optimization and Timing Analysis Overview • Optimization Phases • Compilation Types • Optimization Flow • Compilation Strategies • Controlling Specific Optimization Steps • Controlling Logic-Level and Gate-Level Optimization

  3. References • Design Compiler Reference Manual: Optimization and Timing Analysis • Design Compiler User Guide

  4. Optimization Process • Optimization modifies the initial netlist resulting from elaboration. • Uses cells from the technology library • Attempts to meet all specified constraints • The process is divided into major phases • All or some selection of the major phases may be performed during optimization • Phase selection can be controlled by the user

  5. Major Optimization Phases • Combinational optimization • Technology-independent • Technology-dependent • Sequential optimization • Initial • Final • I/O pad optimization • Local optimizations

  6. Combinational Optimization • Converts a logic-level description to a gate-level netlist • Example: See next slide • Subphases: See slide after next • Technology-independent • Operates at the Boolean level with interconnected gates represented by Boolean equations. • Applies algebraic and Boolean techniques • Flattening and Structuring (covered later!) • Technology-specific (mapping) • Operates at the gate level • Selects components from library to approach defined area and delay goals (covered later!)

  7. Combinational Optimization

  8. Combinational Optimization

  9. Initial Sequential Optimization • Maps sequential logic to cells in the library • First phase of technology-specific optimization • Cannot select optimum sequential cell due to lack of combinational delay info • Defines sequential cells producing “islands” of combinational logic • Sets timing constraints on combinational “islands” required to meet setup and hold constraints on the sequential cells

  10. Final Sequential Optimization • Optimizes timing critical sequential logic • Later phase of technology-dependent optimization • Has accurate values for delays through I/Os and through combinational logic • Effects: • Improve delay by using higher-performance sequential cells • Reduce delay and area by use of complex sequential cells • Improve area by remapping sequential elements (due to setting compile_sequential_area_recovery to true)

  11. I/O Pad Optimization • Insertion and mapping of I/O pads • Follows mapping of combinational logic and initial mapping of sequential logic • Adds input and output buffers to each port in top-level design • Sizes buffers to meet timing constraints on paths involving inputs and outputs • Uses smallest buffer that meets timing specs to reduce current and power consumption

  12. Local Optimization • Incremental modifications to locally optimize gate level logic • Final phase of gate level optimization • Example: See next slide • Effects • Improve delay • Improve area

  13. Local Optimization

  14. Compilation Types • Full - optimizes a complete RTL design • Can include all of the various phases both technology-independent and technology-dependent • Removes and rebuilds logic/gate structure across entire design • Invoked by compile

  15. Compilation Types (continued) • Test-Ready - integrates optimization and scan insertion • Accounts for timing impact of scan design throughout the synthesis process • Eliminates recompile after scan insertion • Invoked by compile -scan

  16. Compilation Types (continued) • Incremental - focuses on areas of design where constraints not met • Structures where constraints met are preserved • Incremental mapping begins with existing gates from earlier compilation • Restructuring only if area or speed improved • Invoked with compile -incremental_mapping

  17. Compilation Types (continued) • Top Level - fixes constraint violations at top level as subblocks in the design are assembled. • Assumes subblocks compiled and meeting timing constraints • Design rule violations fixed even if in subblocks • Invoked with compile -top

  18. Compilation Types (continued) • Critical Path Resynthesis • Finds critical path and attempts full compilation on logic along that path • Repeats on new critical path • Requires DC Expert version • Invoked with compile -map_effort_high

  19. Compilation Types (continued) • Logic Duplication and Mapping to Wide-Fanin Gates (DC Ultra only) • Evaluates cells along critical path • Determines if timing or area can be improved by replacing cell groups with complex, wide-fanin cells • Also improves timing of high-fanout or heavily-loaded nets by duplicating and restructuring large sections of logic • Improves timing at expense of area

  20. Compilation Types (continued) • Optimizing Once for Best- and Worst-Case Conditions • Optimizing with Multiple Libraries

  21. Optimization Flow • FSM Optimization • Subdesign Ungrouping (set_ungroup) • High-Level Optimization • Synthetic Library Implementation Selection • Sequential Inference • Flattening • Structuring • Initial Combinational Mapping • I/O Pad Insertion and Optimization

  22. Optimization Flow (continued) • Mapping Optimization • Phase 1: Delay Optimization • Phase 2: Design Rule Fixing, Part 1 (no delay cost increase) • Phase 3: Design Rule Fixing, Part 2 (delay cost increase; notset_cost_priority delay) • Phase 4: Area Optimization • Verifying Design Functionality

  23. Compilation Strategies • Strategies • Top-down hierarchical • Compile-characterize-write script-recompile

  24. Top-Down Hierarchical Compile (TDHC) • Hierarchy not collapsed • Constraints at top level • Continues to optimize circuit until all constraints met after each module compiled • Requires recompile of subdesigns on critical paths • Stops when no further improvement or goals achieved

  25. TDHC (continued) • Note that entire design is being manipulated at once • Default strategy • Intermodule dependencies handled automatically • Multiple instances of same design must be handled using: • uniquify • set_dont_touch • ungroup

  26. TDHC (continued) • Procedure: • Read in entire design • Make top level the current design • Resolve multiple references: dcug 8-18 to 8-26 • Apply constraints and attributes at the top level • Compile • Preserves hierarchy and optimizes individual levels based on constraints at top level • Example: dcrmo pp. 2-5 to 2-7

  27. Compile-Characterize- Write Script-Recompile (CCWSR) • Optimize (nonunique) designs using context information or time budgets • Mark lower level blocks that have been optimized don’t_touch • Optimize higher level blocks in succession • Useful for medium/large designs without good interblock specifications • Facilitates large designs since only single block in optimization at a time

  28. CCWSR (continued) • Requires iteration until design is stable • Requires manual revision control • Procedure: 1. Compile subblocks independently using a default script for drive and load estimates 2. Read in entire compiled design 3. Characterize one subblock 4. Use write_script to save characterization info

  29. CCWSR (continued) 5. Clear memory; read in previously characterized subblock and recompile using saved script 6. Read in the entire compiled design again without the old subblock; use recompiled subblock 7. Choose another subblock and repeat steps 3 through 7 until all subblocks are recompiled, using their actual environments.

  30. CCWSR (continued) • Comments • Compile automatically goes down hierarchy to submodules - to restrict to current design, remove or omit submodule from database or use set_dont_touch • Compile is bottom-up • Characterize is top-down • Example: dcrmo pp. 2-9 and 2-10

  31. CCWSR (Continued) • What is really being done if pure CCWSR used? • Compile all subblocks bottom up using rough estimates of drive, load and timing constraints. Note that some constraints such as clock period are known. • Read in the entire design and apply the top level constraints. Then the attributes and constraints of any subblock can be determined based on the prior compile of all subblocks.

  32. CCWSR (Continued) • The above is precisely what characterize does. • The script written captures the attributes and constraints of the subblock chosen (one with worst violations is a good choice) • The subblock can then be recompiled based on this current “state” of the optimization process. • The new subblock replaces the old and process repeated for a different subblock • See Chapter 7 of User Guide for Script!

  33. Mixing Strategies • See figure on next slide • Detailed time budget for A, B, C, and D means they can be compiled separately. • Procedure • Compile A • Use CCWSR for the ascending the hierarchy at B • Compile C • Use TDHC for the hierarchy at D

  34. Mixing Strategies

More Related