Skip to content
GitLab

freedreno: logging improvements

Merged Rob Clark requested to merge robclark/mesa:wip/android-profiling into master

In particular:

  1. android support
  2. better handling for multiple log-chunks per batch (such as is the case with higher tile count)
  3. parser improvements to show breakdown of difference phases of the GMEM passes within a frame:
FRAME[109]: 11 blits (440,180 ns), 0 SYSMEM (0 ns), 14 GMEM (16,390,348 ns), 0 COMPUTE (0 ns)
  GMEM[0]: 1920x1080 (4x2 tiles), 4 draws, lrz clear: 0 ns, binning: 22,568 ns, restore/clear: 1,768 ns, draw: 1,872 ns, resolve: 3,120 ns, total: 69,940 ns, rt/zs: PIPE_FORMAT_Z24X8_UNORM
  GMEM[1]: 1024x1024 (1x2 tiles), 0 draws, lrz clear: 13,780 ns, binning: 0 ns, restore/clear: 1,404 ns, draw: 0 ns, resolve: 129,896 ns, total: 147,836 ns, rt/zs: PIPE_FORMAT_Z16_UNORM
  GMEM[2]: 1920x1080 (9x6 tiles), 79 draws, lrz clear: 23,036 ns, binning: 987,740 ns, restore/clear: 160,108 ns, draw: 3,293,368 ns, resolve: 1,212,224 ns, total: 5,860,712 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_Z24X8_UNORM
  GMEM[3]: 1920x1080 (5x4 tiles), 46 draws, lrz clear: 0 ns, binning: 64,116 ns, restore/clear: 246,948 ns, draw: 4,948,944 ns, resolve: 140,296 ns, total: 5,490,056 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_Z24X8_UNORM
  GMEM[4]: 1920x1080 (5x4 tiles), 45 draws, lrz clear: 0 ns, binning: 100,880 ns, restore/clear: 253,396 ns, draw: 1,752,140 ns, resolve: 225,680 ns, total: 2,421,432 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_Z24X8_UNORM
  GMEM[5]: 960x540 (2x1 tiles), 1 draws, lrz clear: 0 ns, binning: 2,964 ns, restore/clear: 8,268 ns, draw: 111,748 ns, resolve: 38,740 ns, total: 192,452 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[6]: 960x540 (2x1 tiles), 1 draws, lrz clear: 0 ns, binning: 2,392 ns, restore/clear: 7,696 ns, draw: 156,520 ns, resolve: 36,296 ns, total: 229,008 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[7]: 480x270 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 2,756 ns, draw: 58,448 ns, resolve: 10,556 ns, total: 75,712 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[8]: 480x270 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 988 ns, draw: 41,444 ns, resolve: 9,620 ns, total: 55,640 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[9]: 240x135 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 1,300 ns, draw: 20,748 ns, resolve: 3,016 ns, total: 28,600 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[10]: 240x135 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 676 ns, draw: 16,536 ns, resolve: 2,964 ns, total: 23,816 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[11]: 120x67 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 1,248 ns, draw: 11,180 ns, resolve: 780 ns, total: 16,952 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[12]: 120x67 (1x1 tiles), 1 draws, lrz clear: 0 ns, binning: 0 ns, restore/clear: 1,196 ns, draw: 9,828 ns, resolve: 884 ns, total: 15,496 ns, rt/zs: PIPE_FORMAT_R8G8B8A8_UNORM
  GMEM[13]: 1920x1080 (5x4 tiles), 1 draws, lrz clear: 18,876 ns, binning: 1,768 ns, restore/clear: 61,828 ns, draw: 1,081,756 ns, resolve: 522,184 ns, total: 1,762,696 ns, rt/zs: PIPE_FORMAT_B8G8R8A8_UNORM, PIPE_FORMAT_Z24_UNORM_S8_UINT
  TOTAL: lrz clear: 55,692 ns (0.33089870977309804%), binning: 1,182,428 ns (7.025495575658708%), restore/clear: 749,580 ns (4.453692718374611%), draw: 11,504,532 ns (68.35514607741361%), resolve: 2,336,256 ns (13.881061841910128%), blit: 440,180 ns (2.6153665529685104%), sysmem: 0 ns (0.0%), total: 16,830,528 ns

Merge request reports