Skip to content

[MISC] Add lazy velocity update to kinematic entity. #2558

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
duburcqa wants to merge 1 commit into
base: main
Choose a base branch
from

Add lazy velocity update to kinematic entity.

5ae1cd3
Select commit
Loading
Failed to load commit list.
Sign in for the full log view
Draft

[MISC] Add lazy velocity update to kinematic entity. #2558

Add lazy velocity update to kinematic entity.
5ae1cd3
Select commit
Loading
Failed to load commit list.
GitHub Actions / Benchmark Comparison failed Mar 16, 2026 in 0s

Benchmark Comparison

🔴 Regressions detected. See tables below.

Details

Thresholds: runtime ± 8%, compile ± 30%, mem ± 4%

Runtime FPS

status env batch_size constraint_solver use_contact_island dtype backend gjk_collision current FPS baseline FPS [last (mean ± std)] (*1) Δ FPS (*2)
✅ anymal_random 30000 - False field cuda - 8,634,979 8,657,922 (8,762,218 ± 189,593) -0.3%
✅ anymal_random 30000 - False ndarray cuda - 6,925,657 6,902,501 (7,004,952 ± 123,170) +0.3%
✅ anymal_uniform 30000 - False field cuda - 11,622,755 11,619,630 (11,887,984 ± 318,220) +0.0%
✅ anymal_uniform 30000 - False ndarray cuda - 9,446,928 9,603,654 (9,712,555 ± 286,016) -1.6%
✅ anymal_uniform_kinematic 0 - False field cpu - 1,800 1,777 (1,778 ± 12) +1.3%
✅ anymal_uniform_kinematic 0 - False ndarray cpu - 959 952 (949 ± 7) +0.7%
✅ anymal_uniform_kinematic 30000 - False field cuda - 9,754,988 9,901,823 (9,954,472 ± 289,637) -1.5%
✅ anymal_uniform_kinematic 30000 - False ndarray cuda - 7,486,510 7,349,393 (7,469,073 ± 112,218) +1.9%
✅ anymal_zero 0 - False field cpu - 6,116 6,125 (6,164 ± 58) -0.1%
✅ anymal_zero 0 - False ndarray cpu - 3,498 3,389 (3,426 ± 35) +3.2%
✅ anymal_zero 30000 - False field cuda - 18,212,782 17,845,687 (18,408,778 ± 586,613) +2.1%
✅ anymal_zero 30000 - False ndarray cuda - 12,990,761 12,921,731 (13,254,442 ± 392,161) +0.5%
✅ box_pyramid_3 4096 - False field cuda - 636,434 637,942 (643,671 ± 10,812) -0.2%
✅ box_pyramid_3 4096 - False ndarray cuda - 464,759 469,215 (474,076 ± 10,877) -0.9%
✅ box_pyramid_4 4096 - False field cuda - 204,755 205,922 (205,153 ± 881) -0.6%
✅ box_pyramid_4 4096 - False ndarray cuda - 153,214 153,613 (155,519 ± 2,032) -0.3%
✅ box_pyramid_5 4096 - False field cuda - 78,938 79,017 (78,872 ± 1,499) -0.1%
✅ box_pyramid_5 4096 - False ndarray cuda - 59,573 59,770 (59,374 ± 327) -0.3%
✅ box_pyramid_6 4096 - False field cuda False 23,785 23,598 (22,867 ± 368) +0.8%
⚠️ box_pyramid_6 4096 - False field cuda True 25,351 23,274 (23,971 ± 1,013) +8.9%
✅ box_pyramid_6 4096 - False ndarray cuda False 17,741 18,614 (18,428 ± 124) -4.7%
✅ box_pyramid_6 4096 - False ndarray cuda True 19,855 20,002 (18,673 ± 990) -0.7%
✅ dex_hand 4096 - False field cuda - 7,804 7,728 (6,724 ± 1,220) +1.0%
✅ dex_hand 4096 - False ndarray cuda - 5,553 5,570 (4,926 ± 752) -0.3%
✅ duck_in_box_easy 30000 - False field cuda False 28,263,138 28,532,887 (28,146,348 ± 273,587) -0.9%
✅ duck_in_box_easy 30000 - False field cuda True 11,052,836 11,033,576 (11,165,187 ± 232,563) +0.2%
✅ duck_in_box_easy 30000 - False ndarray cuda False 19,457,944 19,415,642 (19,694,520 ± 315,941) +0.2%
✅ duck_in_box_easy 30000 - False ndarray cuda True 9,229,443 9,218,344 (9,363,804 ± 156,377) +0.1%
✅ duck_in_box_hard 0 - False field cpu - 4,560 4,704 (4,686 ± 77) -3.1%
✅ duck_in_box_hard 0 - False ndarray cpu - 2,947 2,932 (2,874 ± 78) +0.5%
✅ duck_in_box_hard 30000 - False field cuda False 10,124,955 9,880,631 (8,796,823 ± 1,295,191) +2.5%
✅ duck_in_box_hard 30000 - False field cuda True 3,687,312 3,671,496 (3,316,446 ± 450,021) +0.4%
✅ duck_in_box_hard 30000 - False ndarray cuda False 6,717,825 6,774,406 (6,396,540 ± 417,717) -0.8%
✅ duck_in_box_hard 30000 - False ndarray cuda True 2,875,075 2,893,207 (2,612,209 ± 330,723) -0.6%
✅ franka 30000 - False field cuda - 19,996,477 20,636,558 (21,394,249 ± 1,285,686) -3.1%
✅ franka 30000 - False ndarray cuda - 14,255,497 14,194,344 (14,623,146 ± 438,051) +0.4%
✅ franka_accessors 0 - False field cpu - 1,780 1,738 (1,776 ± 23) +2.4%
✅ franka_accessors 0 - False ndarray cpu - 1,277 1,297 (1,302 ± 22) -1.5%
✅ franka_accessors 30000 - False field cuda - 20,347,338 20,289,660 (20,698,618 ± 568,915) +0.3%
✅ franka_accessors 30000 - False ndarray cuda - 14,267,128 14,184,486 (14,474,045 ± 242,889) +0.6%
✅ franka_free 30000 - False field cuda - 37,323,648 37,235,611 (38,396,870 ± 1,756,689) +0.2%
✅ franka_free 30000 - False ndarray cuda - 22,155,572 22,139,364 (22,925,211 ± 754,556) +0.1%
✅ franka_random 0 - False field cpu - 5,402 5,376 (5,389 ± 20) +0.5%
✅ franka_random 0 - False ndarray cpu - 3,164 3,041 (3,181 ± 80) +4.0%
✅ franka_random 30000 CG False field cuda - 15,274,989 15,547,790 (15,074,872 ± 562,833) -1.8%
✅ franka_random 30000 CG False ndarray cuda - 10,837,460 10,699,694 (10,561,811 ± 272,971) +1.3%
✅ franka_random 30000 Newton False field cuda - 14,911,137 14,967,178 (14,663,269 ± 349,363) -0.4%
✅ franka_random 30000 Newton False ndarray cuda - 10,683,733 10,715,627 (10,644,163 ± 202,845) -0.3%
✅ franka_random 30000 - False field cuda False 14,941,126 14,856,509 (14,525,942 ± 479,737) +0.6%
✅ franka_random 30000 - False field cuda True 11,876,747 11,653,721 (10,113,926 ± 1,849,154) +1.9%
✅ franka_random 30000 - False ndarray cuda False 10,694,844 10,712,923 (10,641,374 ± 171,193) -0.2%
✅ franka_random 30000 - False ndarray cuda True 7,967,240 8,067,276 (7,211,390 ± 960,572) -1.2%
✅ g1_fall 4096 Newton False field cuda - 609,316 612,811 (608,152 ± 6,151) -0.6%
✅ g1_fall 4096 Newton False ndarray cuda - 432,913 434,437 (429,295 ± 5,826) -0.4%
✅ go2 4096 CG False field cuda False 2,890,813 2,914,705 (2,917,999 ± 15,492) -0.8%
🔴 go2 4096 CG False ndarray cuda False 1,682,497 2,022,575 (2,028,017 ± 14,195) -16.8%
✅ go2 4096 Newton False field cuda False 3,289,220 3,280,059 (3,326,611 ± 60,398) +0.3%
✅ go2 4096 Newton False ndarray cuda False 2,218,367 2,208,550 (2,251,243 ± 28,593) +0.4%
✅ go2 4096 - False field cuda True 2,647,889 2,616,846 (2,506,075 ± 121,650) +1.2%
✅ go2 4096 - False ndarray cuda True 1,785,415 1,792,370 (1,704,091 ± 93,183) -0.4%

Baselines considered: 5 commits

Compile Time

status env batch_size constraint_solver use_contact_island dtype backend gjk_collision current compile baseline compile [last (mean ± std)] (*1) Δ compile (*2)
✅ anymal_random 30000 - False field cuda - 93.80 92.20 (78.18 ± 13.52) +1.7%
✅ anymal_random 30000 - False ndarray cuda - 72.60 69.00 (59.66 ± 12.13) +5.2%
✅ anymal_uniform 30000 - False field cuda - 93.20 93.00 (80.24 ± 13.32) +0.2%
✅ anymal_uniform 30000 - False ndarray cuda - 69.50 67.20 (58.58 ± 11.16) +3.4%
✅ anymal_uniform_kinematic 0 - False field cpu - 55.50 56.20 (56.86 ± 2.33) -1.2%
✅ anymal_uniform_kinematic 0 - False ndarray cpu - 38 38 (38 ± 0) +0.0%
✅ anymal_uniform_kinematic 30000 - False field cuda - 95.50 92.80 (81.54 ± 12.04) +2.9%
✅ anymal_uniform_kinematic 30000 - False ndarray cuda - 72 67 (58 ± 12) +6.5%
✅ anymal_zero 0 - False field cpu - 52.80 51.00 (51.26 ± 1.28) +3.5%
✅ anymal_zero 0 - False ndarray cpu - 35 34 (35 ± 0) +1.2%
✅ anymal_zero 30000 - False field cuda - 93.40 92.20 (78.84 ± 13.35) +1.3%
✅ anymal_zero 30000 - False ndarray cuda - 71.60 68.30 (59.56 ± 12.57) +4.8%
✅ box_pyramid_3 4096 - False field cuda - 87.70 85.20 (74.46 ± 13.92) +2.9%
✅ box_pyramid_3 4096 - False ndarray cuda - 69.70 66.50 (57.00 ± 11.54) +4.8%
✅ box_pyramid_4 4096 - False field cuda - 87.70 90.70 (76.42 ± 14.89) -3.3%
✅ box_pyramid_4 4096 - False ndarray cuda - 65.70 63.70 (54.78 ± 11.32) +3.1%
✅ box_pyramid_5 4096 - False field cuda - 85.50 91.40 (75.54 ± 15.25) -6.5%
✅ box_pyramid_5 4096 - False ndarray cuda - 65.60 64.10 (55.10 ± 11.25) +2.3%
✅ box_pyramid_6 4096 - False field cuda False 85.80 91.20 (76.06 ± 15.25) -5.9%
✅ box_pyramid_6 4096 - False field cuda True 54.90 55.20 (48.10 ± 8.54) -0.5%
✅ box_pyramid_6 4096 - False ndarray cuda False 68.40 67.00 (57.24 ± 11.09) +2.1%
✅ box_pyramid_6 4096 - False ndarray cuda True 42.10 40.90 (35.92 ± 6.31) +2.9%
✅ dex_hand 4096 - False field cuda - 98.20 94.90 (83.86 ± 13.87) +3.5%
✅ dex_hand 4096 - False ndarray cuda - 80 75 (66 ± 11) +5.8%
✅ duck_in_box_easy 30000 - False field cuda False 91.30 88.50 (80.54 ± 11.48) +3.2%
✅ duck_in_box_easy 30000 - False field cuda True 62.20 61.50 (55.22 ± 7.43) +1.1%
✅ duck_in_box_easy 30000 - False ndarray cuda False 71.70 68.60 (60.48 ± 12.42) +4.5%
✅ duck_in_box_easy 30000 - False ndarray cuda True 48.80 49.50 (43.66 ± 6.66) -1.4%
✅ duck_in_box_hard 0 - False field cpu - 56.70 54.60 (53.60 ± 2.09) +3.8%
✅ duck_in_box_hard 0 - False ndarray cpu - 39.70 38.90 (37.72 ± 1.68) +2.1%
✅ duck_in_box_hard 30000 - False field cuda False 90.30 84.50 (76.10 ± 12.93) +6.9%
✅ duck_in_box_hard 30000 - False field cuda True 56.10 55.00 (48.32 ± 7.55) +2.0%
✅ duck_in_box_hard 30000 - False ndarray cuda False 68.60 66.50 (56.64 ± 12.37) +3.2%
✅ duck_in_box_hard 30000 - False ndarray cuda True 44.20 39.50 (36.34 ± 4.40) +11.9%
✅ franka 30000 - False field cuda - 88.20 84.60 (72.46 ± 13.28) +4.3%
✅ franka 30000 - False ndarray cuda - 66.70 66.10 (55.18 ± 12.43) +0.9%
✅ franka_accessors 0 - False field cpu - 50.40 52.20 (51.48 ± 1.16) -3.4%
✅ franka_accessors 0 - False ndarray cpu - 34.50 34.60 (34.84 ± 0.98) -0.3%
✅ franka_accessors 30000 - False field cuda - 86.60 87.10 (74.86 ± 13.42) -0.6%
✅ franka_accessors 30000 - False ndarray cuda - 66.90 65.80 (56.44 ± 11.29) +1.7%
✅ franka_free 30000 - False field cuda - 88.20 91.70 (76.28 ± 14.76) -3.8%
✅ franka_free 30000 - False ndarray cuda - 66.50 64.60 (55.74 ± 11.12) +2.9%
✅ franka_random 0 - False field cpu - 49.50 53.40 (51.68 ± 2.09) -7.3%
✅ franka_random 0 - False ndarray cpu - 34.60 34.50 (35.00 ± 0.38) +0.3%
✅ franka_random 30000 CG False field cuda - 83.20 81.10 (71.84 ± 12.39) +2.6%
✅ franka_random 30000 CG False ndarray cuda - 64.70 65.40 (54.58 ± 11.77) -1.1%
✅ franka_random 30000 Newton False field cuda - 86.50 84.90 (73.16 ± 12.51) +1.9%
✅ franka_random 30000 Newton False ndarray cuda - 68 64 (55 ± 11) +4.9%
✅ franka_random 30000 - False field cuda False 84.40 85.50 (73.00 ± 11.63) -1.3%
✅ franka_random 30000 - False field cuda True 53.20 55.50 (47.36 ± 8.25) -4.1%
✅ franka_random 30000 - False ndarray cuda False 65.70 62.80 (54.48 ± 11.91) +4.6%
✅ franka_random 30000 - False ndarray cuda True 41.40 40.70 (35.82 ± 6.22) +1.7%
✅ g1_fall 4096 Newton False field cuda - 86.50 86.70 (73.88 ± 13.50) -0.2%
✅ g1_fall 4096 Newton False ndarray cuda - 65.80 65.40 (55.00 ± 11.60) +0.6%
✅ go2 4096 CG False field cuda False 86.40 84.30 (73.16 ± 13.94) +2.5%
✅ go2 4096 CG False ndarray cuda False 66.80 67.80 (56.78 ± 11.12) -1.5%
✅ go2 4096 Newton False field cuda False 87.40 86.50 (74.60 ± 13.11) +1.0%
✅ go2 4096 Newton False ndarray cuda False 67.40 65.10 (56.00 ± 11.70) +3.5%
✅ go2 4096 - False field cuda True 54.70 54.20 (47.66 ± 7.62) +0.9%
✅ go2 4096 - False ndarray cuda True 41.80 41.10 (36.30 ± 6.26) +1.7%

Baselines considered: 5 commits

Memory usage

status env batch_size backend dtype gjk_collision constraint_solver current memory baseline memory [last (mean ± std)] (*1) Δ memory (*2)
✅ anymal_random 30000 gpu field - - 12,528 12,592 (11,814 ± 863) -0.5%
✅ anymal_random 30000 gpu ndarray - - 12,496 12,472 (11,769 ± 879) +0.2%
✅ anymal_uniform 30000 gpu field - - 12,592 12,592 (11,832 ± 877) +0.0%
✅ anymal_uniform 30000 gpu ndarray - - 12,560 12,560 (11,817 ± 891) +0.0%
✅ anymal_uniform_kinematic 0 cpu field - - 806 806 (806 ± 0) +0.0%
✅ anymal_uniform_kinematic 0 cpu ndarray - - 806 806 (806 ± 0) +0.0%
✅ anymal_uniform_kinematic 30000 gpu field - - 12,960 12,960 (12,217 ± 891) +0.0%
✅ anymal_uniform_kinematic 30000 gpu ndarray - - 12,928 12,928 (12,185 ± 891) +0.0%
✅ anymal_zero 0 cpu field - - 806 806 (806 ± 0) +0.0%
✅ anymal_zero 0 cpu ndarray - - 806 806 (806 ± 0) +0.0%
✅ anymal_zero 30000 gpu field - - 12,592 12,536 (11,777 ± 894) +0.4%
✅ anymal_zero 30000 gpu ndarray - - 12,560 12,472 (11,800 ± 877) +0.7%
✅ box_pyramid_3 4096 gpu field - - 2,174 2,174 (2,097 ± 92) +0.0%
✅ box_pyramid_3 4096 gpu ndarray - - 2,174 2,174 (2,097 ± 92) +0.0%
✅ box_pyramid_4 4096 gpu field - - 3,056 3,056 (2,940 ± 138) +0.0%
✅ box_pyramid_4 4096 gpu ndarray - - 3,056 3,056 (2,928 ± 153) +0.0%
✅ box_pyramid_5 4096 gpu field - - 6,684 6,684 (6,492 ± 229) +0.0%
✅ box_pyramid_5 4096 gpu ndarray - - 6,684 6,684 (6,492 ± 230) +0.0%
✅ box_pyramid_6 4096 gpu field False - 9,722 9,722 (9,490 ± 277) +0.0%
✅ box_pyramid_6 4096 gpu field True - 10,192 10,192 (9,744 ± 537) +0.0%
✅ box_pyramid_6 4096 gpu ndarray False - 9,722 9,722 (9,490 ± 277) +0.0%
✅ box_pyramid_6 4096 gpu ndarray True - 10,192 10,192 (9,744 ± 537) +0.0%
✅ dex_hand 4096 gpu field - - 9,140 9,140 (8,840 ± 348) +0.0%
✅ dex_hand 4096 gpu ndarray - - 9,112 9,140 (8,847 ± 351) -0.3%
✅ duck_in_box_easy 30000 gpu field False - 3,384 3,384 (3,222 ± 195) +0.0%
✅ duck_in_box_easy 30000 gpu field True - 3,992 3,992 (3,603 ± 460) +0.0%
✅ duck_in_box_easy 30000 gpu ndarray False - 3,440 3,352 (3,228 ± 189) +2.6%
✅ duck_in_box_easy 30000 gpu ndarray True - 3,952 3,952 (3,568 ± 460) +0.0%
✅ duck_in_box_hard 0 cpu field - - 8 8 (8 ± 0) +0.0%
✅ duck_in_box_hard 0 cpu ndarray - - 8 8 (8 ± 0) +0.0%
✅ duck_in_box_hard 30000 gpu field False - 13,464 13,464 (12,220 ± 1,527) +0.0%
✅ duck_in_box_hard 30000 gpu field True - 14,104 14,008 (12,547 ± 1,789) +0.7%
✅ duck_in_box_hard 30000 gpu ndarray False - 13,432 13,432 (12,171 ± 1,552) +0.0%
✅ duck_in_box_hard 30000 gpu ndarray True - 14,064 13,976 (12,534 ± 1,769) +0.6%
✅ franka 30000 gpu field - - 16,418 16,354 (15,010 ± 1,638) +0.4%
✅ franka 30000 gpu ndarray - - 16,322 16,322 (14,997 ± 1,641) +0.0%
✅ franka_accessors 0 cpu field - - 1,222 1,222 (1,222 ± 0) +0.0%
✅ franka_accessors 0 cpu ndarray - - 1,222 1,222 (1,222 ± 0) +0.0%
✅ franka_accessors 30000 gpu field - - 16,346 16,346 (14,989 ± 1,628) +0.0%
✅ franka_accessors 30000 gpu ndarray - - 16,314 16,314 (14,963 ± 1,620) +0.0%
✅ franka_free 30000 gpu field - - 15,950 15,950 (14,612 ± 1,605) +0.0%
✅ franka_free 30000 gpu ndarray - - 15,918 15,918 (14,561 ± 1,628) +0.0%
✅ franka_random 0 cpu field - - 1,222 1,222 (1,222 ± 0) +0.0%
✅ franka_random 0 cpu ndarray - - 1,222 1,222 (1,222 ± 0) +0.0%
✅ franka_random 30000 gpu field False - 15,950 16,078 (14,625 ± 1,641) -0.8%
✅ franka_random 30000 gpu field True - 16,568 16,600 (15,006 ± 1,899) -0.2%
✅ franka_random 30000 gpu field - CG 16,000 16,000 (14,662 ± 1,605) +0.0%
✅ franka_random 30000 gpu field - Newton 16,352 16,352 (14,995 ± 1,628) +0.0%
✅ franka_random 30000 gpu ndarray False - 15,918 15,918 (14,612 ± 1,670) +0.0%
✅ franka_random 30000 gpu ndarray True - 16,536 16,568 (14,968 ± 1,894) -0.2%
✅ franka_random 30000 gpu ndarray - CG 15,968 16,032 (14,630 ± 1,631) -0.4%
✅ franka_random 30000 gpu ndarray - Newton 16,384 16,384 (14,976 ± 1,638) +0.0%
✅ g1_fall 4096 gpu field - Newton 4,830 4,830 (4,599 ± 276) +0.0%
✅ g1_fall 4096 gpu ndarray - Newton 4,830 4,830 (4,644 ± 233) +0.0%
✅ go2 4096 gpu field False CG 3,710 3,710 (3,479 ± 276) +0.0%
✅ go2 4096 gpu field False Newton 4,062 4,062 (3,831 ± 276) +0.0%
✅ go2 4096 gpu field True - 4,246 4,246 (3,810 ± 522) +0.0%
✅ go2 4096 gpu ndarray False CG 3,710 3,710 (3,466 ± 291) +0.0%
✅ go2 4096 gpu ndarray False Newton 4,062 4,062 (3,818 ± 291) +0.0%
✅ go2 4096 gpu ndarray True - 4,246 4,246 (3,798 ± 537) +0.0%

Baselines considered: 5 commits

  • Commit 1: 9238414

  • Commit 2: cfef9de

  • Commit 3: 5fd3857

  • Commit 4: 64429b6

  • Commit 5: 986a116

  • (*1) last: last commit on main, mean/std: stats over commit hashes 5 commits if available.

  • (*2) Δ: relative difference between PR and last commit on main, i.e. (PR - main) / main * 100%.

Artifact: Download raw data

View more details on GitHub Actions