SLIDE 1
Competition track:
2
Pushing the branch predictability limits with the multi-poTAGE+SC - - PowerPoint PPT Presentation
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
SLIDE 2
SLIDE 3
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
SLIDE 4
4
What we did
SLIDE 5
5
What we did
TAGE statistical corrector (SC) André
SLIDE 6
6
What we did
TAGE statistical corrector (SC) André multi-poTAGE (MP) Pierre
SLIDE 7
7
What we did
TAGE statistical corrector (SC) André multi-poTAGE (MP) Pierre Approximately same prediction accuracy
- n average, but significant differences on
individual traces
SLIDE 8
8
What we did
TAGE statistical corrector (SC) multi-poTAGE (MP)
SLIDE 9
9
What we did
TAGE statistical corrector (SC) multi-poTAGE (MP)
SLIDE 10
10
What we did
statistical corrector (SC) multi-poTAGE (MP)
SLIDE 11
11
What we did
statistical corrector (SC) multi-poTAGE (MP)
SLIDE 12
12
What we did
statistical corrector (SC) multi-poTAGE (MP) change a few parameters because
- f the memory size constraint
SLIDE 13
13
Multi-poTAGE + Statistical Corrector
SC MP
SLIDE 14
14
Multi-poTAGE + Statistical Corrector
SC MP
- 5% MPKI
SLIDE 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
SLIDE 16
16