1. 13 Apr, 2018 1 commit
  2. 09 Apr, 2018 2 commits
  3. 03 Apr, 2018 1 commit
  4. 28 Mar, 2018 1 commit
  5. 20 Mar, 2018 1 commit
  6. 19 Mar, 2018 1 commit
  7. 16 Mar, 2018 2 commits
  8. 15 Mar, 2018 3 commits
  9. 13 Mar, 2018 3 commits
  10. 07 Mar, 2018 3 commits
  11. 06 Mar, 2018 1 commit
  12. 05 Mar, 2018 1 commit
  13. 22 Feb, 2018 1 commit
    • Samuel Pitoiset's avatar
      radv: enable lowering of fpow to fexp2 and flog2 · 7aa008d1
      Samuel Pitoiset authored
      
      
      There is no fpow in hardware, so it's always lowered somewhere,
      but it appears that lowering at NIR level is better. Figured while
      comparing compute shaders between RadeonSI and RADV.
      
      Polaris10:
      Totals from affected shaders:
      SGPRS: 18936 -> 18904 (-0.17 %)
      VGPRS: 12240 -> 12220 (-0.16 %)
      Spilled SGPRs: 2809 -> 2809 (0.00 %)
      Code Size: 718116 -> 719848 (0.24 %) bytes
      Max Waves: 1409 -> 1410 (0.07 %)
      
      Vega10:
      Totals from affected shaders:
      SGPRS: 18392 -> 18392 (0.00 %)
      VGPRS: 12008 -> 11920 (-0.73 %)
      Spilled SGPRs: 3001 -> 2981 (-0.67 %)
      Code Size: 777444 -> 778788 (0.17 %) bytes
      Max Waves: 1503 -> 1504 (0.07 %)
      Signed-off-by: default avatarSamuel Pitoiset <samuel.pitoiset@gmail.com>
      Reviewed-by: default avatarBas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
      7aa008d1
  14. 15 Feb, 2018 1 commit
  15. 06 Feb, 2018 1 commit
    • Samuel Pitoiset's avatar
      radv: run nir_opt_shrink_load · 3488a3f0
      Samuel Pitoiset authored
      
      
      LLVM can't shrink loads.
      
      Polaris10:
      Totals from affected shaders:
      SGPRS: 62528 -> 59955 (-4.11 %)
      VGPRS: 44708 -> 44616 (-0.21 %)
      Spilled SGPRs: 16 -> 8 (-50.00 %)
      Code Size: 1355504 -> 1355172 (-0.02 %) bytes
      Max Waves: 11710 -> 11670 (-0.34 %)
      
      Vega10:
      Totals from affected shaders:
      SGPRS: 51448 -> 50371 (-2.09 %)
      VGPRS: 39140 -> 39048 (-0.24 %)
      Spilled SGPRs: 16 -> 16 (0.00 %)
      Code Size: 1307188 -> 1304296 (-0.22 %) bytes
      Max Waves: 11312 -> 11292 (-0.18 %)
      
      This reduces SGPRs spilling in MadMax, and it also reduces
      number of SGPRs in DOW3 and F12017. The number of waves slightly
      decreases in F1 but I don't see any performance changes after
      benchmarking it. Talos and Serious Sam are not affected because
      they don't use any push constants.
      Signed-off-by: default avatarSamuel Pitoiset <samuel.pitoiset@gmail.com>
      Reviewed-by: default avatarBas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
      3488a3f0
  16. 29 Jan, 2018 1 commit
  17. 22 Jan, 2018 1 commit
  18. 18 Jan, 2018 1 commit
  19. 13 Jan, 2018 1 commit
  20. 10 Jan, 2018 2 commits
  21. 08 Jan, 2018 2 commits
  22. 04 Jan, 2018 1 commit
  23. 19 Dec, 2017 1 commit
  24. 18 Dec, 2017 1 commit
  25. 14 Dec, 2017 2 commits
  26. 02 Dec, 2017 1 commit
  27. 01 Dec, 2017 1 commit
  28. 13 Nov, 2017 1 commit
  29. 06 Nov, 2017 1 commit