JILP Computer Architecture Workshop: Branch Prediction Championship Results

4th jilp workshop on computer architecture n.w
1 / 9
Embed
Share

The 4th JILP Workshop on Computer Architecture focused on Branch Prediction Championship, showcasing winners, MPKI results, and detailed category outcomes for 4KB and 32KB predictors. Find out who emerged victorious and how MPKI performance impacted results.

  • JILP Workshop
  • Computer Architecture
  • Branch Prediction
  • Championship
  • MPKI
rayaanacharya
bega_yah Follow

Uploaded on | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.

Download Presentation

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.

E N D

Presentation Transcript

  1. 4th JILP Workshop on Computer Architecture Competitions Championship Branch Prediction (CBP-4) -Moinuddin Qureshi (GT)

  2. Drum Rolls The Reward 1. Trophy for winner in each category 2. Certificates for winner and runners up in each category For reference, we also have MPKI for the winning predictor from CBP 2011 (thanks to Andre for providing MPKI for CBP4 traces)

  3. For the 4KB category, the winner is The winning predictor from 2011 has MPKI: 3.882 Rank 1 2 3 4 5 Authors Andre Seznec Yasuo Ishii Samir Otiv et al. Daniel Jimenez Jorge Albericio et al. MPKI 3.315 3.629 3.692 3.953 3.995 Winning predictor of CBP-4 reduced MPKI by 14.6%

  4. Detailed Results for 4KB Category DanielJimenez JorgeAlbericio SamirOtiv Seznec YasuoIshii MIN-ALL LONG-SPEC2K6-00 LONG-SPEC2K6-01 LONG-SPEC2K6-02 LONG-SPEC2K6-03 LONG-SPEC2K6-04 LONG-SPEC2K6-05 LONG-SPEC2K6-06 LONG-SPEC2K6-07 LONG-SPEC2K6-08 LONG-SPEC2K6-09 LONG-SPEC2K6-10 LONG-SPEC2K6-11 LONG-SPEC2K6-12 LONG-SPEC2K6-13 LONG-SPEC2K6-14 LONG-SPEC2K6-15 LONG-SPEC2K6-16 LONG-SPEC2K6-17 LONG-SPEC2K6-18 LONG-SPEC2K6-19 SHORT-FP-1 SHORT-FP-2 SHORT-FP-3 SHORT-FP-4 SHORT-FP-5 SHORT-INT-1 SHORT-INT-2 SHORT-INT-3 SHORT-INT-4 SHORT-INT-5 SHORT-MM-1 SHORT-MM-2 SHORT-MM-3 SHORT-MM-4 SHORT-MM-5 SHORT-SERV-1 SHORT-SERV-2 SHORT-SERV-3 SHORT-SERV-4 SHORT-SERV-5 AMEAN 2.666 7.045 2.345 1.145 9.094 4.819 0.538 15.567 0.967 4.496 1.578 0.632 11.262 9.145 0.004 1.231 3.355 3.467 0.778 1.318 1.235 0.965 0.423 0.095 0.029 2.373 7.072 8.155 1.22 0.298 7.153 9.086 0.15 1.456 5.487 6.134 6.892 5.978 6.527 5.925 3.953 2.491 7.728 1.856 1.267 9.268 5.067 0.629 16.89 0.723 4.193 1.734 0.761 9.538 11.041 0.001 0.97 3.217 3.262 0.004 1.354 1.55 0.851 0.015 0.015 0.008 6.427 8.312 10.407 1.316 0.145 7.564 10.133 0.064 1.019 5.369 4.991 5.299 4.646 4.858 4.825 3.995 2.36 7.514 1.505 1.137 8.816 4.882 0.739 15.663 0.778 4.016 1.29 0.74 11.338 8.484 0.001 0.877 3.151 2.93 0.161 1.31 1.349 0.834 0.016 0.024 0.043 1.328 6.807 8.926 1.048 0.152 7.338 9.525 0.071 1.073 5.104 5.319 5.812 4.837 5.462 2.201 7.413 1.919 1.023 8.866 5.034 0.701 13.561 0.734 4.136 1.481 0.712 11.318 7.394 0.001 0.849 3.046 3.096 0.054 1.246 1.474 0.745 0.016 0.018 0.009 1.355 6.121 8.693 0.992 0.152 7.095 9.313 0.146 1.02 4.784 2.382 2.557 4.13 3.812 3.012 3.315 2.369 7.373 1.978 1.086 9.338 5.154 0.795 15.193 0.816 4.056 1.417 0.61 11.515 8.023 0.001 0.873 3.118 3.414 0.063 1.406 1.43 0.747 0.015 0.091 0.027 1.739 6.366 9.034 1.038 0.263 7.385 9.561 0.087 1.242 4.892 4.327 4.611 4.676 4.823 4.223 3.629 2.201 7.045 1.505 1.023 8.816 4.819 0.538 13.561 0.723 4.016 1.290 0.610 9.538 7.394 0.001 0.849 3.046 2.930 0.004 1.246 1.235 0.745 0.015 0.015 0.008 1.328 6.121 8.155 0.992 0.145 7.095 9.086 0.064 1.019 4.784 2.382 2.557 4.130 3.812 3.012 3.196 MPKI with Min-ALL is 3.196 (3.6% lower) 4.9 3.69

  5. For the 32KB category, the winner is The winning predictor from 2011 has MPKI: 2.491 Rank 1 2 3 4 5 6 Authors Andre Seznec Yasuo Ishii Samir Otiv et al. Daniel Jimenez Dibakar Gope et al. J. Albericio et al. MPKI 2.365 2.401 2.542 2.566 2.733 2.794 Winning predictor of CBP-4 reduced MPKI by 5.1%

  6. Detailed Results for 32KB Category DanielJimenez DibakarGope JorgeAlbericio SamirOtiv Seznec YasuoIshii MIN LONG-SPEC2K6-00 LONG-SPEC2K6-01 LONG-SPEC2K6-02 LONG-SPEC2K6-03 LONG-SPEC2K6-04 LONG-SPEC2K6-05 LONG-SPEC2K6-06 LONG-SPEC2K6-07 LONG-SPEC2K6-08 LONG-SPEC2K6-09 LONG-SPEC2K6-10 LONG-SPEC2K6-11 LONG-SPEC2K6-12 LONG-SPEC2K6-13 LONG-SPEC2K6-14 LONG-SPEC2K6-15 LONG-SPEC2K6-16 LONG-SPEC2K6-17 LONG-SPEC2K6-18 LONG-SPEC2K6-19 SHORT-FP-1 SHORT-FP-2 SHORT-FP-3 SHORT-FP-4 SHORT-FP-5 SHORT-INT-1 SHORT-INT-2 SHORT-INT-3 SHORT-INT-4 SHORT-INT-5 SHORT-MM-1 SHORT-MM-2 SHORT-MM-3 SHORT-MM-4 SHORT-MM-5 SHORT-SERV-1 SHORT-SERV-2 SHORT-SERV-3 SHORT-SERV-4 SHORT-SERV-5 AMEAN 1.84 6.668 0.696 0.861 7.862 4.412 0.528 8.708 0.709 4.063 0.698 0.462 10.343 5.34 0.001 0.462 2.984 2.523 0.183 1.115 0.984 0.766 0.019 0.041 0.027 0.186 4.683 5.924 0.647 0.274 6.546 8.494 0.065 1.267 3.447 0.932 0.898 3.002 2.167 1.825 2.566 2.069 6.841 1.215 0.795 8.975 4.708 0.697 8.691 0.708 3.865 0.771 0.479 10.879 5.301 0.002 0.399 2.84 7.07 1.239 0.618 8.742 4.617 0.596 12.626 0.675 5.323 0.967 0.522 8.893 7.494 0.001 0.355 3.097 2.863 0.003 0.991 1.151 0.468 0.014 0.014 0.007 0.144 5.298 7.208 0.556 0.073 6.923 8.835 0.058 0.532 3.383 0.787 0.755 2.641 1.86 1.535 2.794 1.438 6.997 0.313 0.638 8.801 4.547 0.601 8.373 0.594 3.392 0.632 0.507 10.975 5.314 0.001 0.284 2.928 2.369 0.003 1.01 1.073 0.462 0.014 0.014 0.008 0.133 4.548 7.245 0.551 0.061 6.832 8.785 0.06 0.913 3.48 0.789 0.761 2.759 1.924 1.538 2.542 1.375 6.524 0.407 0.611 8.03 4.361 0.543 7.321 0.578 3.349 0.564 0.412 10.86 4.596 0.001 0.283 2.601 2.423 0.013 0.947 0.942 0.442 0.014 0.014 0.009 0.144 3.997 6.197 0.483 0.063 6.479 8.498 0.046 0.876 3.198 0.783 0.754 2.562 1.785 1.52 2.365 1.441 6.636 0.37 0.637 7.933 4.335 0.592 7.422 0.567 3.277 0.588 0.394 10.773 4.912 0.001 0.301 2.715 2.441 0.003 0.963 1.034 0.449 0.015 0.015 0.008 0.121 4.206 6.453 0.51 0.068 6.467 8.632 0.055 0.872 3.281 0.782 0.759 2.694 1.789 1.535 2.401 1.375 6.524 0.313 0.611 7.862 4.335 0.528 7.321 0.567 3.277 0.564 0.394 8.893 4.596 0.001 0.283 2.601 2.369 0.003 0.947 0.942 0.442 0.014 0.014 0.007 0.121 3.997 5.924 0.483 0.061 6.467 8.494 0.046 0.532 3.198 0.782 0.754 2.562 1.785 1.52 2.288 3 2.539 0.018 1.082 1.043 0.798 0.017 0.02 0.009 0.749 4.701 7.634 0.672 0.084 6.945 8.726 0.073 0.975 3.762 1.097 1.077 3.379 2.474 2.065 2.733 MPKI with Min-ALL is 2.288 (3.3% lower)

  7. For the Unlimited Category, the winner is 64KB version of 2011 predictor has MPKI: 2.358 Rank 1 2 3 4 5 6 7 8 9 10 Authors MPKI 1.691 1.781 1.782 1.919 2.014 2.055 2.105 2.129 2.682 3.487 Pierre Michaud & Andre Seznec Andre Seznec Pierre Michaud Ibrahim Burak Karsli et al. J. Albericio et al. Samir Otiv et al. Dibakar Gope et al. Daniel Jimenez Sharath Babu Yongmin Shen & Michael Ferdman

  8. Detailed Results for Unlimited Category BurakKarsli DanielJimenez JorgeAlbericio MichaudSeznec PierreMichaud SamirOtiv Seznec SharathBabu YongmingShen MIN-ALL LONG-SPEC2K6-00 LONG-SPEC2K6-01 LONG-SPEC2K6-02 LONG-SPEC2K6-03 LONG-SPEC2K6-04 LONG-SPEC2K6-05 LONG-SPEC2K6-06 LONG-SPEC2K6-07 LONG-SPEC2K6-08 LONG-SPEC2K6-09 LONG-SPEC2K6-10 LONG-SPEC2K6-11 LONG-SPEC2K6-12 LONG-SPEC2K6-13 LONG-SPEC2K6-14 LONG-SPEC2K6-15 LONG-SPEC2K6-16 LONG-SPEC2K6-17 LONG-SPEC2K6-18 LONG-SPEC2K6-19 SHORT-FP-1 SHORT-FP-2 SHORT-FP-3 SHORT-FP-4 SHORT-FP-5 SHORT-INT-1 SHORT-INT-2 SHORT-INT-3 SHORT-INT-4 SHORT-INT-5 SHORT-MM-1 SHORT-MM-2 SHORT-MM-3 SHORT-MM-4 SHORT-MM-5 SHORT-SERV-1 SHORT-SERV-2 SHORT-SERV-3 SHORT-SERV-4 SHORT-SERV-5 AMEAN 0.79 6.201 0.209 0.109 6.084 3.894 0.515 3.183 0.533 2.478 0.386 0.304 10.689 3.584 0.001 0.162 2.427 1.557 0.003 0.77 0.875 0.374 0.014 0.014 0.008 1.26 6.397 0.417 0.635 6.871 4.145 0.491 3.94 0.659 3.652 0.482 0.395 9.998 3.801 0.001 0.274 2.55 2.128 0.144 0.965 0.988 0.666 0.025 0.039 0.009 0.148 3.342 5.22 0.499 0.282 6.327 8.23 0.043 1.272 2.49 0.695 0.68 2.154 1.482 1.36 2.129 1.064 6.499 0.272 0.132 6.336 4.366 0.567 3.7 0.587 2.672 0.494 0.389 8.721 4.026 0.001 0.204 2.796 1.691 0.003 0.842 1.017 0.433 0.013 0.014 0.007 0.128 3.695 5.863 0.439 0.053 6.602 8.297 0.036 0.521 2.332 0.624 0.609 1.879 1.379 1.265 2.014 0.756 5.75 0.175 0.086 5.485 2.804 0.464 2.536 0.493 2.277 0.304 0.276 9.748 2.652 0.001 0.125 2.148 1.462 0.002 0.656 0.763 0.273 0.013 0.013 0.007 0.076 2.636 4.33 0.202 0.045 5.992 7.599 0.034 0.757 1.781 0.531 0.521 1.541 1.22 1.09 1.691 0.773 5.973 0.178 0.089 5.861 2.87 0.5 2.737 0.55 2.322 0.324 0.292 10.421 3.043 0.001 0.132 2.27 1.486 0.003 0.687 0.798 0.284 0.013 0.013 0.007 0.087 2.898 4.648 0.233 0.044 6.178 7.93 0.034 0.768 1.83 0.548 0.525 1.571 1.253 1.118 1.782 1.052 6.428 0.268 0.133 6.394 4.288 0.605 3.606 0.595 2.667 0.469 0.381 10.826 4.028 0.001 0.196 2.83 1.673 0.003 0.834 0.976 0.423 0.013 0.014 0.007 0.131 3.578 5.63 0.423 0.06 6.546 8.261 0.06 0.859 2.257 0.611 0.602 1.861 1.365 1.256 2.055 0.873 5.913 0.212 0.119 5.566 3.787 0.471 2.776 0.497 2.483 0.341 0.288 9.708 2.906 0.001 0.147 2.213 1.546 0.003 0.739 0.826 0.34 0.013 0.013 0.007 0.085 2.749 4.476 0.332 0.048 6.019 7.778 0.035 0.804 1.928 0.567 0.558 1.659 1.262 1.157 1.781 1.866 6.933 0.772 0.959 8.077 4.853 0.881 6.87 0.831 3.696 0.862 0.582 11.271 5.578 0.003 0.56 3.033 2.415 0.746 1.116 1.173 0.815 0.308 0.112 0.038 0.391 4.347 7.132 0.633 0.334 7.233 2.416 8.793 0.644 0.224 8.729 5.737 1.531 8.613 1.131 3.562 1.292 0.54 12.723 7.241 0.001 0.526 4.005 3.576 0.006 1.651 2.38 0.773 0.042 0.029 0.013 0.438 7.818 9.838 0.954 0.154 9.702 10.591 0.163 1.167 4.42 2.302 2.379 4.584 4.567 4.215 3.487 0.756 5.75 0.175 0.086 5.485 2.804 0.464 2.536 0.493 2.277 0.304 0.276 8.721 2.652 0.001 0.125 2.148 1.462 0.002 0.656 0.763 0.273 0.013 0.013 0.007 0.076 2.636 4.33 0.202 0.044 5.992 7.599 0.034 0.521 1.781 0.531 0.521 1.541 1.22 1.09 1.659 MPKI with Min-ALL is 1.659 (1.9% lower) 0.1 3.198 5.087 0.322 0.05 6.265 8.102 0.038 0.937 2.122 0.577 0.569 1.718 1.31 1.204 1.919 8.7 0.132 1.291 3.237 1.197 1.158 3.056 2.167 1.933 2.682

  9. Endnotes Code, paper, and presentation will be uploaded on the CBP4 website shortly Branch prediction remains a very important problem. We need the next big insight ! The below (from CBP3, Hongliang) still holds:

Related


No related presentations.

More Related Content