pushing the branch predictability limits with the multi
play

Pushing the branch predictability limits with the multi-poTAGE+SC - PowerPoint PPT Presentation

Pushing the branch predictability limits with the multi-poTAGE+SC predictor Pierre Michaud + Andr Seznec june 2014 Competition track: Unlimited size 2 We did not modify the predictor algorithm after the submission We just corrected a bug


  1. Pushing the branch predictability limits with the multi-poTAGE+SC predictor Pierre Michaud + André Seznec june 2014

  2. Competition track: Unlimited size 2

  3. We did not modify the predictor algorithm after the submission We just corrected a bug (out of bound array write) that had almost no impact on prediction accuracy 3

  4. What we did 4

  5. What we did André TAGE statistical corrector (SC) 5

  6. What we did André Pierre multi-poTAGE TAGE (MP) statistical corrector (SC) 6

  7. What we did André Pierre multi-poTAGE TAGE (MP) statistical corrector (SC) Approximately same prediction accuracy on average, but significant differences on individual traces 7

  8. What we did multi-poTAGE TAGE (MP) statistical corrector (SC) 8

  9. What we did TAGE multi-poTAGE (MP) statistical corrector (SC) 9

  10. What we did multi-poTAGE (MP) statistical corrector (SC) 10

  11. What we did multi-poTAGE (MP) statistical corrector (SC) 11

  12. What we did multi-poTAGE (MP) statistical corrector (SC) change a few parameters because of the memory size constraint 12

  13. Multi-poTAGE + Statistical Corrector MP SC 13

  14. Multi-poTAGE + Statistical Corrector MP -5% MPKI SC 14

  15. Conclusion • Performance gain of MP+SC over TAGE-SC comes mainly from the non-global components of multi-poTAGE • With the Statistical Corrector, the post-predictor in poTAGE is almost superfluous - on isolated poTAGE, removing the post-predictor  +10% MPKI - with SC, removing post-predictor  +1% MPKI • The Statistical Corrector solves the cold-counter problem more effectively than the post-predictor 15

  16. Questions ? 16

Recommend


More recommend