1 / 15

Graph Matching and Clock Gating

Graph Matching and Clock Gating. Prepared by Shmuel Wimer Bar-Ilan Univ., Eng. Faculty. Outline. How data-driven clock gating works Flip-flops activity and correlation The optimal fan-out of a gater Delay implications The complexity of flip-flop grouping. Motivation.

malcolmt
Download Presentation

Graph Matching and Clock Gating

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. Graph Matching and Clock Gating Prepared by Shmuel Wimer Bar-Ilan Univ., Eng. Faculty Graph Matching and Clock Gating

  2. Outline How data-driven clock gating works Flip-flops activity and correlation The optimal fan-out of a gater Delay implications The complexity of flip-flop grouping Graph Matching and Clock Gating

  3. Motivation • Clocking consumes 30% to 70% of dynamic power • Typically, only 3% of the clock pulses are useful, namely, only 3% of the clock switching occurs with data toggling! • Clock enabling is easier at high design levels but harder in logic and gate level • Clock enabling is easier in register files and data path, but harder in control • Designers are conservative, leaving on table a lot of “hidden disabling” Graph Matching and Clock Gating

  4. k O X CL CL D1 D2 clk_g FF FF Q2 Q1 clk_g A clk D3 Latch Q3 Data-Driven CLK Gating There is timing overhead! Graph Matching and Clock Gating

  5. TC TC clk Q1 clk_g D2 D3 Timing Implications  Graph Matching and Clock Gating

  6. The Optimal Clock Gater Fan-out There is a tradeoff between hardware overhead and amount of saved clock pulses (power savings). FFs’ activities and their correlations is a key. Worst case assumption: All FF are toggling independently of each other. Graph Matching and Clock Gating

  7. Latch overhead amortized over k FFs Net saving at a leaf flip-flop Switching probability of FF enabling Gater’s disabling probability Derivate by k: k: # flip-flops, q: FF probability for D=Q Graph Matching and Clock Gating

  8. Graph Matching and Clock Gating

  9. Optimal k-size Flip-flop Grouping Graph Matching and Clock Gating

  10. FF Pairwise Activity Model Graph Matching and Clock Gating

  11. Assume that FF grouping in pairs (k=2). Total power, normalized to number of clock switching: Essential + Waste Graph Matching and Clock Gating

  12. Graph Matching and Clock Gating

  13. What happens when k>2 ? Graph Matching and Clock Gating

  14. Is repeated perfect matching optimal ? Graph Matching and Clock Gating

  15. No! Here is the optimal 4-size grouping Finding optimal k-grouping is NP-hard Graph Matching and Clock Gating

More Related