[Performance] Use Lua scripts for scattered index operations by vmoens · Pull Request #1571 · pytorch/tensordict

vmoens · 2026-02-14T01:21:37Z

Stack from ghstack (oldest at bottom):

Replace per-row GETRANGE/SETRANGE pipeline commands with server-side
Lua scripts for tensor/list/bool indices. Each key now emits exactly
one EVAL command regardless of the number of indexed positions.

Hybrid strategy:

int / slice (any step): single GETRANGE + local stride (unchanged)
tensor / list / bool: Lua GETRANGES script (new)
step>1 writes: covering-range RMW (unchanged)
scattered writes: Lua SETRANGES script (new)

This gives deterministic O(K) commands for K keys with O(N*row_size)
bandwidth, eliminating the covering-range waste for sparse indices
like td[tensor([0, 1000])].

Fancy/bool writes improved ~2.5x (5.9ms -> 2.4ms).

[ghstack-poisoned]

Replace per-row GETRANGE/SETRANGE pipeline commands with server-side Lua scripts for tensor/list/bool indices. Each key now emits exactly one EVAL command regardless of the number of indexed positions. Hybrid strategy: - int / slice (any step): single GETRANGE + local stride (unchanged) - tensor / list / bool: Lua GETRANGES script (new) - step>1 writes: covering-range RMW (unchanged) - scattered writes: Lua SETRANGES script (new) This gives deterministic O(K) commands for K keys with O(N*row_size) bandwidth, eliminating the covering-range waste for sparse indices like td[tensor([0, 1000])]. Fancy/bool writes improved ~2.5x (5.9ms -> 2.4ms). ghstack-source-id: 59fb136 Pull-Request: #1571

github-actions · 2026-02-14T01:39:40Z

$\color{#D29922}\textsf{\Large&#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of GPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}16$. Worsened: $\large\color{#d91a1a}7$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	35.9310μs	14.5620μs	68.6718 KOps/s	69.3438 KOps/s	$\color{#d91a1a}-0.97\%$
test_plain_set_stack_nested	36.5010μs	14.7715μs	67.6979 KOps/s	69.0744 KOps/s	$\color{#d91a1a}-1.99\%$
test_plain_set_nested_inplace	40.6010μs	16.1382μs	61.9648 KOps/s	61.8277 KOps/s	$\color{#35bf28}+0.22\%$
test_plain_set_stack_nested_inplace	46.5110μs	15.9019μs	62.8857 KOps/s	62.5348 KOps/s	$\color{#35bf28}+0.56\%$
test_items	34.0610μs	5.5538μs	180.0572 KOps/s	181.3411 KOps/s	$\color{#d91a1a}-0.71\%$
test_items_nested	0.5763ms	0.5225ms	1.9140 KOps/s	1.9554 KOps/s	$\color{#d91a1a}-2.11\%$
test_items_nested_locked	0.5717ms	0.5213ms	1.9184 KOps/s	1.9419 KOps/s	$\color{#d91a1a}-1.21\%$
test_items_nested_leaf	0.1180ms	91.9692μs	10.8732 KOps/s	10.8536 KOps/s	$\color{#35bf28}+0.18\%$
test_items_stack_nested	0.6002ms	0.5216ms	1.9173 KOps/s	1.9492 KOps/s	$\color{#d91a1a}-1.63\%$
test_items_stack_nested_leaf	0.1223ms	91.7531μs	10.8988 KOps/s	10.8081 KOps/s	$\color{#35bf28}+0.84\%$
test_items_stack_nested_locked	0.5772ms	0.5225ms	1.9138 KOps/s	1.9367 KOps/s	$\color{#d91a1a}-1.19\%$
test_keys	30.2800μs	4.1159μs	242.9602 KOps/s	241.8597 KOps/s	$\color{#35bf28}+0.46\%$
test_keys_nested	0.1470ms	0.1179ms	8.4794 KOps/s	8.5366 KOps/s	$\color{#d91a1a}-0.67\%$
test_keys_nested_locked	88.8760ms	0.1398ms	7.1518 KOps/s	7.9307 KOps/s	$\textbf{\color{#d91a1a}-9.82\%}$
test_keys_nested_leaf	0.1384ms	0.1090ms	9.1707 KOps/s	9.2673 KOps/s	$\color{#d91a1a}-1.04\%$
test_keys_stack_nested	0.1493ms	0.1192ms	8.3893 KOps/s	8.5260 KOps/s	$\color{#d91a1a}-1.60\%$
test_keys_stack_nested_leaf	0.1353ms	0.1085ms	9.2190 KOps/s	9.2959 KOps/s	$\color{#d91a1a}-0.83\%$
test_keys_stack_nested_locked	0.1610ms	0.1285ms	7.7795 KOps/s	7.9669 KOps/s	$\color{#d91a1a}-2.35\%$
test_values	6.1380μs	0.9961μs	1.0039 MOps/s	1.0025 MOps/s	$\color{#35bf28}+0.14\%$
test_values_nested	73.8410μs	46.7094μs	21.4090 KOps/s	21.3443 KOps/s	$\color{#35bf28}+0.30\%$
test_values_nested_locked	83.6020μs	49.5439μs	20.1841 KOps/s	20.2728 KOps/s	$\color{#d91a1a}-0.44\%$
test_values_nested_leaf	82.5720μs	52.8102μs	18.9357 KOps/s	18.9705 KOps/s	$\color{#d91a1a}-0.18\%$
test_values_stack_nested	0.1224ms	46.1585μs	21.6645 KOps/s	21.4608 KOps/s	$\color{#35bf28}+0.95\%$
test_values_stack_nested_leaf	80.6520μs	52.2008μs	19.1568 KOps/s	18.9401 KOps/s	$\color{#35bf28}+1.14\%$
test_values_stack_nested_locked	85.5520μs	49.4662μs	20.2158 KOps/s	20.1835 KOps/s	$\color{#35bf28}+0.16\%$
test_membership	4.1852μs	0.8129μs	1.2301 MOps/s	1.2377 MOps/s	$\color{#d91a1a}-0.61\%$
test_membership_nested	24.2410μs	2.9749μs	336.1479 KOps/s	335.5553 KOps/s	$\color{#35bf28}+0.18\%$
test_membership_nested_leaf	26.1900μs	3.0172μs	331.4340 KOps/s	335.7182 KOps/s	$\color{#d91a1a}-1.28\%$
test_membership_stacked_nested	33.8410μs	3.0119μs	332.0129 KOps/s	334.0291 KOps/s	$\color{#d91a1a}-0.60\%$
test_membership_stacked_nested_leaf	27.6210μs	3.0132μs	331.8706 KOps/s	336.2902 KOps/s	$\color{#d91a1a}-1.31\%$
test_membership_nested_last	33.7900μs	4.3708μs	228.7928 KOps/s	230.4465 KOps/s	$\color{#d91a1a}-0.72\%$
test_membership_nested_leaf_last	29.6510μs	4.4110μs	226.7035 KOps/s	229.5998 KOps/s	$\color{#d91a1a}-1.26\%$
test_membership_stacked_nested_last	34.0710μs	4.3407μs	230.3781 KOps/s	228.7252 KOps/s	$\color{#35bf28}+0.72\%$
test_membership_stacked_nested_leaf_last	25.1300μs	4.3407μs	230.3787 KOps/s	230.0922 KOps/s	$\color{#35bf28}+0.12\%$
test_nested_getleaf	48.1110μs	20.6752μs	48.3672 KOps/s	46.7511 KOps/s	$\color{#35bf28}+3.46\%$
test_nested_get	58.1810μs	19.1893μs	52.1123 KOps/s	49.8121 KOps/s	$\color{#35bf28}+4.62\%$
test_stacked_getleaf	57.6110μs	20.2224μs	49.4500 KOps/s	46.7939 KOps/s	$\textbf{\color{#35bf28}+5.68\%}$
test_stacked_get	45.6710μs	19.4251μs	51.4798 KOps/s	49.2546 KOps/s	$\color{#35bf28}+4.52\%$
test_nested_getitemleaf	46.8310μs	21.2409μs	47.0791 KOps/s	46.6567 KOps/s	$\color{#35bf28}+0.91\%$
test_nested_getitem	45.9710μs	20.1906μs	49.5280 KOps/s	49.2473 KOps/s	$\color{#35bf28}+0.57\%$
test_stacked_getitemleaf	45.4710μs	21.3594μs	46.8178 KOps/s	46.6820 KOps/s	$\color{#35bf28}+0.29\%$
test_stacked_getitem	52.5710μs	20.3599μs	49.1161 KOps/s	48.8905 KOps/s	$\color{#35bf28}+0.46\%$
test_lock_nested	7.8776ms	0.4591ms	2.1780 KOps/s	2.1835 KOps/s	$\color{#d91a1a}-0.25\%$
test_lock_stack_nested	0.5002ms	0.4586ms	2.1807 KOps/s	2.1493 KOps/s	$\color{#35bf28}+1.46\%$
test_unlock_nested	0.4685ms	0.3659ms	2.7328 KOps/s	2.7084 KOps/s	$\color{#35bf28}+0.90\%$
test_unlock_stack_nested	0.4215ms	0.3682ms	2.7156 KOps/s	2.6606 KOps/s	$\color{#35bf28}+2.07\%$
test_flatten_speed	0.1655ms	0.1178ms	8.4881 KOps/s	8.5872 KOps/s	$\color{#d91a1a}-1.15\%$
test_unflatten_speed	0.6387ms	0.5727ms	1.7460 KOps/s	1.7513 KOps/s	$\color{#d91a1a}-0.30\%$
test_common_ops	0.8418ms	0.6848ms	1.4603 KOps/s	1.4680 KOps/s	$\color{#d91a1a}-0.53\%$
test_creation	0.1226ms	2.7499μs	363.6447 KOps/s	366.3530 KOps/s	$\color{#d91a1a}-0.74\%$
test_creation_empty	29.5500μs	5.7996μs	172.4252 KOps/s	175.2394 KOps/s	$\color{#d91a1a}-1.61\%$
test_creation_nested_1	31.8500μs	10.2736μs	97.3371 KOps/s	97.4462 KOps/s	$\color{#d91a1a}-0.11\%$
test_creation_nested_2	39.7210μs	11.3006μs	88.4907 KOps/s	90.4903 KOps/s	$\color{#d91a1a}-2.21\%$
test_creation_many_keys[10]	51.6210μs	17.1622μs	58.2675 KOps/s	59.5464 KOps/s	$\color{#d91a1a}-2.15\%$
test_creation_many_keys[50]	95.8520μs	72.0606μs	13.8772 KOps/s	13.9863 KOps/s	$\color{#d91a1a}-0.78\%$
test_creation_many_keys[100]	0.2025ms	0.1429ms	6.9960 KOps/s	7.0912 KOps/s	$\color{#d91a1a}-1.34\%$
test_creation_nested_many_keys[10]	63.3110μs	36.9392μs	27.0715 KOps/s	27.2077 KOps/s	$\color{#d91a1a}-0.50\%$
test_creation_nested_many_keys[50]	0.1865ms	0.1494ms	6.6947 KOps/s	6.7845 KOps/s	$\color{#d91a1a}-1.32\%$
test_clone	44.7710μs	12.9296μs	77.3420 KOps/s	75.3930 KOps/s	$\color{#35bf28}+2.59\%$
test_getitem[int]	1.6189ms	14.0510μs	71.1693 KOps/s	57.6310 KOps/s	$\textbf{\color{#35bf28}+23.49\%}$
test_getitem[slice_int]	0.1346ms	24.0831μs	41.5229 KOps/s	40.9517 KOps/s	$\color{#35bf28}+1.39\%$
test_getitem[range]	0.1627ms	60.2635μs	16.5938 KOps/s	16.4883 KOps/s	$\color{#35bf28}+0.64\%$
test_getitem[tuple]	0.1426ms	23.4613μs	42.6234 KOps/s	42.3250 KOps/s	$\color{#35bf28}+0.70\%$
test_getitem[list]	0.1778ms	57.9105μs	17.2680 KOps/s	17.7068 KOps/s	$\color{#d91a1a}-2.48\%$
test_setitem_dim[int]	49.2310μs	25.5992μs	39.0638 KOps/s	39.1756 KOps/s	$\color{#d91a1a}-0.29\%$
test_setitem_dim[slice_int]	87.9920μs	43.3572μs	23.0642 KOps/s	23.1140 KOps/s	$\color{#d91a1a}-0.22\%$
test_setitem_dim[range]	0.1286ms	91.4503μs	10.9349 KOps/s	10.9478 KOps/s	$\color{#d91a1a}-0.12\%$
test_setitem_dim[tuple]	77.9910μs	39.8109μs	25.1188 KOps/s	24.7635 KOps/s	$\color{#35bf28}+1.43\%$
test_setitem	53.5810μs	17.4864μs	57.1872 KOps/s	55.5233 KOps/s	$\color{#35bf28}+3.00\%$
test_set	53.3910μs	16.6399μs	60.0964 KOps/s	58.8948 KOps/s	$\color{#35bf28}+2.04\%$
test_set_shared	0.5056ms	0.2029ms	4.9294 KOps/s	4.8883 KOps/s	$\color{#35bf28}+0.84\%$
test_update	0.3462ms	21.3863μs	46.7589 KOps/s	45.6155 KOps/s	$\color{#35bf28}+2.51\%$
test_update_nested	78.6410μs	33.6984μs	29.6750 KOps/s	29.4573 KOps/s	$\color{#35bf28}+0.74\%$
test_update__nested	0.4699ms	33.0917μs	30.2191 KOps/s	29.6685 KOps/s	$\color{#35bf28}+1.86\%$
test_set_nested	50.3810μs	18.6610μs	53.5878 KOps/s	52.0623 KOps/s	$\color{#35bf28}+2.93\%$
test_set_nested_new	61.8910μs	23.4307μs	42.6790 KOps/s	41.5905 KOps/s	$\color{#35bf28}+2.62\%$
test_select	80.4820μs	40.7945μs	24.5131 KOps/s	24.0638 KOps/s	$\color{#35bf28}+1.87\%$
test_select_nested	97.9020μs	71.1593μs	14.0530 KOps/s	14.2497 KOps/s	$\color{#d91a1a}-1.38\%$
test_exclude_nested	0.1237ms	93.0099μs	10.7515 KOps/s	11.0332 KOps/s	$\color{#d91a1a}-2.55\%$
test_empty[True]	0.4792ms	0.4233ms	2.3626 KOps/s	2.4148 KOps/s	$\color{#d91a1a}-2.16\%$
test_empty[False]	8.4250μs	1.2655μs	790.2289 KOps/s	799.3109 KOps/s	$\color{#d91a1a}-1.14\%$
test_to	0.1007ms	73.0300μs	13.6930 KOps/s	13.9880 KOps/s	$\color{#d91a1a}-2.11\%$
test_to_nonblocking	0.1172ms	63.6546μs	15.7098 KOps/s	15.7518 KOps/s	$\color{#d91a1a}-0.27\%$
test_unbind_speed	0.3490ms	0.3139ms	3.1861 KOps/s	3.1830 KOps/s	$\color{#35bf28}+0.10\%$
test_unbind_speed_stack0	0.3772ms	0.3131ms	3.1938 KOps/s	3.2299 KOps/s	$\color{#d91a1a}-1.12\%$
test_unbind_speed_stack1	0.1032s	0.8844ms	1.1307 KOps/s	1.1191 KOps/s	$\color{#35bf28}+1.04\%$
test_split	1.1557ms	1.0930ms	914.9265 Ops/s	915.3271 Ops/s	$\color{#d91a1a}-0.04\%$
test_chunk	0.1029s	1.1638ms	859.2684 Ops/s	958.8233 Ops/s	$\textbf{\color{#d91a1a}-10.38\%}$
test_to_cpu_blocking	19.1052ms	19.0012ms	52.6283 Ops/s	40.1713 Ops/s	$\textbf{\color{#35bf28}+31.01\%}$
test_to_cpu_global_sync	11.0030ms	10.9206ms	91.5701 Ops/s	89.1188 Ops/s	$\color{#35bf28}+2.75\%$
test_to_cpu_event_sync	0.1151s	13.1330ms	76.1440 Ops/s	81.9111 Ops/s	$\textbf{\color{#d91a1a}-7.04\%}$
test_to_cpu_default	12.2201ms	11.9364ms	83.7771 Ops/s	82.0550 Ops/s	$\color{#35bf28}+2.10\%$
test_consolidate[False-None]	4.0531ms	3.9457ms	253.4393 Ops/s	223.5667 Ops/s	$\textbf{\color{#35bf28}+13.36\%}$
test_consolidate[default-None]	2.0082ms	1.9179ms	521.3978 Ops/s	494.8778 Ops/s	$\textbf{\color{#35bf28}+5.36\%}$
test_consolidate[reduce-overhead-None]	1.9546ms	1.8544ms	539.2481 Ops/s	515.8135 Ops/s	$\color{#35bf28}+4.54\%$
test_consolidate_njt[False-None]	8.3073ms	8.1317ms	122.9748 Ops/s	120.7718 Ops/s	$\color{#35bf28}+1.82\%$
test_to[False-False-None]	2.0990ms	1.9854ms	503.6722 Ops/s	490.1707 Ops/s	$\color{#35bf28}+2.75\%$
test_to[True-False-None]	2.1327ms	1.8684ms	535.2163 Ops/s	535.3325 Ops/s	$\color{#d91a1a}-0.02\%$
test_to[within-False-None]	6.1437ms	5.8949ms	169.6383 Ops/s	167.4598 Ops/s	$\color{#35bf28}+1.30\%$
test_to[True-default-None]	7.4509ms	7.3428ms	136.1878 Ops/s	125.6146 Ops/s	$\textbf{\color{#35bf28}+8.42\%}$
test_to_njt[False-False-None]	8.3378ms	8.2460ms	121.2715 Ops/s	116.8855 Ops/s	$\color{#35bf28}+3.75\%$
test_to_njt[True-False-None]	6.8284ms	6.7145ms	148.9306 Ops/s	142.0950 Ops/s	$\color{#35bf28}+4.81\%$
test_to_njt[within-False-None]	15.1388ms	15.0029ms	66.6539 Ops/s	64.3186 Ops/s	$\color{#35bf28}+3.63\%$
test_creation[device0]	0.4683ms	0.1135ms	8.8113 KOps/s	8.3675 KOps/s	$\textbf{\color{#35bf28}+5.30\%}$
test_creation_from_tensor	0.4633ms	0.1117ms	8.9510 KOps/s	8.6310 KOps/s	$\color{#35bf28}+3.71\%$
test_add_one[memmap_tensor0]	0.2264ms	6.2435μs	160.1659 KOps/s	152.0520 KOps/s	$\textbf{\color{#35bf28}+5.34\%}$
test_contiguous[memmap_tensor0]	28.2000μs	0.6149μs	1.6264 MOps/s	2.2730 MOps/s	$\textbf{\color{#d91a1a}-28.45\%}$
test_stack[memmap_tensor0]	24.7300μs	4.3388μs	230.4759 KOps/s	216.5658 KOps/s	$\textbf{\color{#35bf28}+6.42\%}$
test_memmaptd_index	1.0700ms	0.2560ms	3.9066 KOps/s	3.9381 KOps/s	$\color{#d91a1a}-0.80\%$
test_memmaptd_index_astensor	0.5001ms	0.3429ms	2.9166 KOps/s	2.8997 KOps/s	$\color{#35bf28}+0.58\%$
test_memmaptd_index_op	0.8295ms	0.5830ms	1.7154 KOps/s	1.6490 KOps/s	$\color{#35bf28}+4.03\%$
test_serialize_model	0.1379s	0.1365s	7.3252 Ops/s	7.2578 Ops/s	$\color{#35bf28}+0.93\%$
test_serialize_model_pickle	1.6670s	1.2574s	0.7953 Ops/s	0.8259 Ops/s	$\color{#d91a1a}-3.71\%$
test_serialize_weights	0.1366s	0.1340s	7.4610 Ops/s	7.3250 Ops/s	$\color{#35bf28}+1.86\%$
test_serialize_weights_returnearly	0.4074s	88.2321ms	11.3337 Ops/s	11.3279 Ops/s	$\color{#35bf28}+0.05\%$
test_serialize_weights_pickle	1.3759s	1.2185s	0.8207 Ops/s	0.8170 Ops/s	$\color{#35bf28}+0.45\%$
test_reshape_pytree	0.2008ms	31.8796μs	31.3680 KOps/s	30.4521 KOps/s	$\color{#35bf28}+3.01\%$
test_reshape_td	74.7620μs	42.3039μs	23.6385 KOps/s	23.1217 KOps/s	$\color{#35bf28}+2.24\%$
test_view_pytree	0.2006ms	31.4213μs	31.8255 KOps/s	30.7457 KOps/s	$\color{#35bf28}+3.51\%$
test_view_td	89.6810μs	49.5621μs	20.1767 KOps/s	19.5401 KOps/s	$\color{#35bf28}+3.26\%$
test_unbind_pytree	0.2357ms	35.3877μs	28.2584 KOps/s	27.9350 KOps/s	$\color{#35bf28}+1.16\%$
test_unbind_td	0.1068ms	46.5485μs	21.4830 KOps/s	21.4871 KOps/s	$\color{#d91a1a}-0.02\%$
test_split_pytree	0.2457ms	40.5302μs	24.6730 KOps/s	24.1129 KOps/s	$\color{#35bf28}+2.32\%$
test_split_td	0.1147ms	62.1331μs	16.0945 KOps/s	15.6781 KOps/s	$\color{#35bf28}+2.66\%$
test_add_pytree	0.1898ms	41.2913μs	24.2182 KOps/s	24.2468 KOps/s	$\color{#d91a1a}-0.12\%$
test_add_td	0.1056ms	51.3955μs	19.4570 KOps/s	19.0094 KOps/s	$\color{#35bf28}+2.35\%$
test_compile_add_one_nested[tensordict-compile]	0.2032ms	0.1358ms	7.3653 KOps/s	6.9480 KOps/s	$\textbf{\color{#35bf28}+6.01\%}$
test_compile_add_one_nested[tensordict-eager]	0.4104ms	0.1884ms	5.3082 KOps/s	5.3728 KOps/s	$\color{#d91a1a}-1.20\%$
test_compile_add_one_nested[pytree-compile]	0.1739ms	0.1110ms	9.0107 KOps/s	9.2231 KOps/s	$\color{#d91a1a}-2.30\%$
test_compile_add_one_nested[pytree-eager]	0.4291ms	0.1759ms	5.6854 KOps/s	5.6691 KOps/s	$\color{#35bf28}+0.29\%$
test_compile_copy_nested[tensordict-compile]	60.3220μs	28.8608μs	34.6491 KOps/s	30.5084 KOps/s	$\textbf{\color{#35bf28}+13.57\%}$
test_compile_copy_nested[tensordict-eager]	79.0320μs	49.7057μs	20.1184 KOps/s	20.1272 KOps/s	$\color{#d91a1a}-0.04\%$
test_compile_copy_nested[pytree-compile]	0.1482ms	9.5114μs	105.1369 KOps/s	105.7495 KOps/s	$\color{#d91a1a}-0.58\%$
test_compile_copy_nested[pytree-eager]	0.4493ms	66.6374μs	15.0066 KOps/s	15.0087 KOps/s	$\color{#d91a1a}-0.01\%$
test_compile_add_one_flat[tensordict-compile]	0.2220ms	0.1717ms	5.8226 KOps/s	5.5139 KOps/s	$\textbf{\color{#35bf28}+5.60\%}$
test_compile_add_one_flat[tensordict-eager]	0.3087ms	0.2477ms	4.0376 KOps/s	4.0288 KOps/s	$\color{#35bf28}+0.22\%$
test_compile_add_one_flat[tensorclass-compile]	0.1553ms	0.1127ms	8.8732 KOps/s	8.5601 KOps/s	$\color{#35bf28}+3.66\%$
test_compile_add_one_flat[tensorclass-eager]	0.1128ms	66.7636μs	14.9782 KOps/s	14.9741 KOps/s	$\color{#35bf28}+0.03\%$
test_compile_add_one_flat[pytree-compile]	0.2240ms	0.1549ms	6.4553 KOps/s	6.2238 KOps/s	$\color{#35bf28}+3.72\%$
test_compile_add_one_flat[pytree-eager]	0.8910ms	0.5187ms	1.9278 KOps/s	1.9040 KOps/s	$\color{#35bf28}+1.25\%$
test_compile_add_self_flat[tensordict-eager]	0.3368ms	0.3013ms	3.3186 KOps/s	3.2952 KOps/s	$\color{#35bf28}+0.71\%$
test_compile_add_self_flat[tensordict-compile]	0.2309ms	0.1818ms	5.5015 KOps/s	5.5024 KOps/s	$\color{#d91a1a}-0.02\%$
test_compile_add_self_flat[tensorclass-eager]	0.1433ms	84.1232μs	11.8873 KOps/s	12.1298 KOps/s	$\color{#d91a1a}-2.00\%$
test_compile_add_self_flat[tensorclass-compile]	0.1939ms	0.1157ms	8.6428 KOps/s	8.4293 KOps/s	$\color{#35bf28}+2.53\%$
test_compile_add_self_flat[pytree-eager]	0.6498ms	0.4284ms	2.3343 KOps/s	2.3152 KOps/s	$\color{#35bf28}+0.82\%$
test_compile_add_self_flat[pytree-compile]	0.2459ms	0.1628ms	6.1426 KOps/s	6.3447 KOps/s	$\color{#d91a1a}-3.18\%$
test_compile_copy_flat[tensordict-compile]	56.2410μs	25.3189μs	39.4961 KOps/s	40.8239 KOps/s	$\color{#d91a1a}-3.25\%$
test_compile_copy_flat[tensordict-eager]	71.9610μs	39.8086μs	25.1202 KOps/s	25.0221 KOps/s	$\color{#35bf28}+0.39\%$
test_compile_copy_flat[pytree-compile]	0.1272ms	10.3785μs	96.3527 KOps/s	95.3937 KOps/s	$\color{#35bf28}+1.01\%$
test_compile_copy_flat[pytree-eager]	0.4170ms	51.1698μs	19.5428 KOps/s	19.7130 KOps/s	$\color{#d91a1a}-0.86\%$
test_compile_assign_and_add[tensordict-compile]	1.9882ms	0.1698ms	5.8883 KOps/s	5.6537 KOps/s	$\color{#35bf28}+4.15\%$
test_compile_assign_and_add[tensordict-eager]	3.5138ms	3.2575ms	306.9813 Ops/s	303.3602 Ops/s	$\color{#35bf28}+1.19\%$
test_compile_assign_and_add[pytree-compile]	1.8803ms	0.1577ms	6.3424 KOps/s	6.2618 KOps/s	$\color{#35bf28}+1.29\%$
test_compile_assign_and_add[pytree-eager]	2.8960ms	2.7696ms	361.0595 Ops/s	360.9127 Ops/s	$\color{#35bf28}+0.04\%$
test_compile_indexing[tensor-tensordict-compile]	0.2256ms	0.1054ms	9.4840 KOps/s	9.3731 KOps/s	$\color{#35bf28}+1.18\%$
test_compile_indexing[tensor-tensordict-eager]	0.3504ms	70.5159μs	14.1812 KOps/s	14.0349 KOps/s	$\color{#35bf28}+1.04\%$
test_compile_indexing[tensor-tensorclass-compile]	0.1385ms	93.5312μs	10.6916 KOps/s	10.5406 KOps/s	$\color{#35bf28}+1.43\%$
test_compile_indexing[tensor-tensorclass-eager]	0.2459ms	45.2819μs	22.0839 KOps/s	22.2790 KOps/s	$\color{#d91a1a}-0.88\%$
test_compile_indexing[tensor-pytree-compile]	0.1456ms	95.7553μs	10.4433 KOps/s	10.4425 KOps/s	$+0.01\%$
test_compile_indexing[tensor-pytree-eager]	0.2679ms	47.7385μs	20.9475 KOps/s	22.2248 KOps/s	$\textbf{\color{#d91a1a}-5.75\%}$
test_compile_indexing[slice-tensordict-compile]	0.2165ms	57.8222μs	17.2944 KOps/s	17.9740 KOps/s	$\color{#d91a1a}-3.78\%$
test_compile_indexing[slice-tensordict-eager]	0.2290ms	28.9019μs	34.5999 KOps/s	36.9335 KOps/s	$\textbf{\color{#d91a1a}-6.32\%}$
test_compile_indexing[slice-tensorclass-compile]	0.1495ms	46.9594μs	21.2950 KOps/s	22.2643 KOps/s	$\color{#d91a1a}-4.35\%$
test_compile_indexing[slice-tensorclass-eager]	0.2711ms	23.1290μs	43.2357 KOps/s	45.2093 KOps/s	$\color{#d91a1a}-4.37\%$
test_compile_indexing[slice-pytree-compile]	0.1416ms	44.6762μs	22.3833 KOps/s	21.7462 KOps/s	$\color{#35bf28}+2.93\%$
test_compile_indexing[slice-pytree-eager]	0.2643ms	21.6736μs	46.1391 KOps/s	45.0502 KOps/s	$\color{#35bf28}+2.42\%$
test_compile_indexing[int-tensordict-compile]	0.2340ms	55.2173μs	18.1103 KOps/s	17.4530 KOps/s	$\color{#35bf28}+3.77\%$
test_compile_indexing[int-tensordict-eager]	0.2277ms	26.9743μs	37.0724 KOps/s	37.0184 KOps/s	$\color{#35bf28}+0.15\%$
test_compile_indexing[int-tensorclass-compile]	80.8420μs	44.2830μs	22.5820 KOps/s	21.6281 KOps/s	$\color{#35bf28}+4.41\%$
test_compile_indexing[int-tensorclass-eager]	0.2671ms	22.0402μs	45.3715 KOps/s	44.9964 KOps/s	$\color{#35bf28}+0.83\%$
test_compile_indexing[int-pytree-compile]	80.1920μs	44.6408μs	22.4010 KOps/s	21.7069 KOps/s	$\color{#35bf28}+3.20\%$
test_compile_indexing[int-pytree-eager]	0.2611ms	21.4755μs	46.5646 KOps/s	45.1241 KOps/s	$\color{#35bf28}+3.19\%$
test_mod_add[eager]	0.1045ms	49.0265μs	20.3971 KOps/s	19.8954 KOps/s	$\color{#35bf28}+2.52\%$
test_mod_add[compile]	0.4700ms	0.1067ms	9.3703 KOps/s	9.2756 KOps/s	$\color{#35bf28}+1.02\%$
test_mod_add[compile-overhead]	0.3824ms	0.1440ms	6.9452 KOps/s	6.8098 KOps/s	$\color{#35bf28}+1.99\%$
test_mod_wrap[eager]	0.3629ms	0.2986ms	3.3491 KOps/s	3.4976 KOps/s	$\color{#d91a1a}-4.25\%$
test_mod_wrap[compile]	0.4414ms	0.3546ms	2.8204 KOps/s	2.9048 KOps/s	$\color{#d91a1a}-2.91\%$
test_mod_wrap[compile-overhead]	7.3767ms	4.0461ms	247.1489 Ops/s	246.2275 Ops/s	$\color{#35bf28}+0.37\%$
test_mod_wrap_and_backward[eager]	1.6020ms	1.4859ms	672.9949 Ops/s	668.6457 Ops/s	$\color{#35bf28}+0.65\%$
test_mod_wrap_and_backward[compile]	1.5071ms	1.4112ms	708.6105 Ops/s	699.7653 Ops/s	$\color{#35bf28}+1.26\%$
test_mod_wrap_and_backward[compile-overhead]	1.2261ms	0.8627ms	1.1591 KOps/s	1.1317 KOps/s	$\color{#35bf28}+2.42\%$
test_seq_add[eager]	0.2049ms	0.1526ms	6.5542 KOps/s	6.5149 KOps/s	$\color{#35bf28}+0.60\%$
test_seq_add[compile]	0.2053ms	0.1117ms	8.9490 KOps/s	8.5914 KOps/s	$\color{#35bf28}+4.16\%$
test_seq_add[compile-overhead]	0.1838ms	0.1494ms	6.6952 KOps/s	6.4165 KOps/s	$\color{#35bf28}+4.34\%$
test_seq_wrap[eager]	0.8940ms	0.5137ms	1.9468 KOps/s	1.9382 KOps/s	$\color{#35bf28}+0.45\%$
test_seq_wrap[compile]	0.4032ms	0.3568ms	2.8026 KOps/s	2.7371 KOps/s	$\color{#35bf28}+2.40\%$
test_seq_wrap[compile-overhead]	0.3205ms	0.2555ms	3.9138 KOps/s	3.7810 KOps/s	$\color{#35bf28}+3.51\%$
test_func_call_runtime[False-eager]	0.9288ms	0.8245ms	1.2129 KOps/s	1.2048 KOps/s	$\color{#35bf28}+0.67\%$
test_func_call_runtime[False-compile]	0.9466ms	0.8850ms	1.1299 KOps/s	1.1147 KOps/s	$\color{#35bf28}+1.36\%$
test_func_call_runtime[False-compile-overhead]	0.4914ms	0.4435ms	2.2550 KOps/s	2.2185 KOps/s	$\color{#35bf28}+1.64\%$
test_func_call_runtime[True-eager]	1.1454ms	1.0619ms	941.7006 Ops/s	939.6734 Ops/s	$\color{#35bf28}+0.22\%$
test_func_call_runtime[True-compile]	1.0000ms	0.8926ms	1.1203 KOps/s	1.1010 KOps/s	$\color{#35bf28}+1.76\%$
test_func_call_runtime[True-compile-overhead]	0.5024ms	0.4525ms	2.2102 KOps/s	2.1575 KOps/s	$\color{#35bf28}+2.44\%$
test_func_call_cm_runtime[False-eager]	0.8838ms	0.8199ms	1.2196 KOps/s	1.2076 KOps/s	$\color{#35bf28}+1.00\%$
test_func_call_cm_runtime[False-compile]	0.9637ms	0.8868ms	1.1277 KOps/s	1.0879 KOps/s	$\color{#35bf28}+3.66\%$
test_func_call_cm_runtime[False-compile-overhead]	0.4911ms	0.4462ms	2.2409 KOps/s	2.2156 KOps/s	$\color{#35bf28}+1.14\%$
test_func_call_cm_runtime[True-eager]	1.2903ms	1.1912ms	839.4815 Ops/s	824.0648 Ops/s	$\color{#35bf28}+1.87\%$
test_func_call_cm_runtime[True-compile]	0.9786ms	0.9288ms	1.0767 KOps/s	1.0613 KOps/s	$\color{#35bf28}+1.45\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5429ms	0.4865ms	2.0555 KOps/s	2.0206 KOps/s	$\color{#35bf28}+1.73\%$
test_vmap_func_call_cm_runtime[eager]	2.8163ms	2.3217ms	430.7264 Ops/s	424.7075 Ops/s	$\color{#35bf28}+1.42\%$
test_vmap_func_call_cm_runtime[compile]	1.0654ms	0.9471ms	1.0558 KOps/s	1.0034 KOps/s	$\textbf{\color{#35bf28}+5.23\%}$
test_vmap_func_call_cm_runtime[compile-overhead]	0.5430ms	0.4929ms	2.0288 KOps/s	1.9930 KOps/s	$\color{#35bf28}+1.80\%$
test_distributed	2.6297ms	0.1632ms	6.1260 KOps/s	6.5214 KOps/s	$\textbf{\color{#d91a1a}-6.06\%}$
test_tdmodule	75.5420μs	27.8906μs	35.8544 KOps/s	34.5936 KOps/s	$\color{#35bf28}+3.64\%$
test_tdmodule_dispatch	76.1120μs	46.1871μs	21.6510 KOps/s	21.7286 KOps/s	$\color{#d91a1a}-0.36\%$
test_tdseq	45.8010μs	26.9731μs	37.0739 KOps/s	36.7498 KOps/s	$\color{#35bf28}+0.88\%$
test_tdseq_dispatch	69.2720μs	47.5839μs	21.0155 KOps/s	20.7978 KOps/s	$\color{#35bf28}+1.05\%$
test_instantiation_functorch	2.1670ms	1.9647ms	508.9750 Ops/s	502.5926 Ops/s	$\color{#35bf28}+1.27\%$
test_exec_functorch	0.2140ms	0.1741ms	5.7443 KOps/s	5.6974 KOps/s	$\color{#35bf28}+0.82\%$
test_exec_functional_call	0.1988ms	0.1585ms	6.3103 KOps/s	6.2686 KOps/s	$\color{#35bf28}+0.67\%$
test_exec_td_decorator	0.4466ms	0.2301ms	4.3461 KOps/s	4.3200 KOps/s	$\color{#35bf28}+0.60\%$
test_vmap_mlp_speed_decorator[True-True]	1.0177ms	0.8107ms	1.2336 KOps/s	1.2273 KOps/s	$\color{#35bf28}+0.51\%$
test_vmap_mlp_speed_decorator[True-False]	0.9836ms	0.8111ms	1.2329 KOps/s	1.2267 KOps/s	$\color{#35bf28}+0.51\%$
test_vmap_mlp_speed_decorator[False-True]	0.8491ms	0.7027ms	1.4231 KOps/s	1.4076 KOps/s	$\color{#35bf28}+1.10\%$
test_vmap_mlp_speed_decorator[False-False]	0.8906ms	0.7016ms	1.4252 KOps/s	1.4119 KOps/s	$\color{#35bf28}+0.94\%$
test_vmap_transformer_speed_decorator[True-True]	20.3016ms	20.1978ms	49.5103 Ops/s	48.4598 Ops/s	$\color{#35bf28}+2.17\%$
test_vmap_transformer_speed_decorator[True-False]	20.3492ms	20.2443ms	49.3966 Ops/s	48.5519 Ops/s	$\color{#35bf28}+1.74\%$
test_vmap_transformer_speed_decorator[False-True]	20.1639ms	20.0555ms	49.8616 Ops/s	48.9241 Ops/s	$\color{#35bf28}+1.92\%$
test_vmap_transformer_speed_decorator[False-False]	20.2047ms	20.0841ms	49.7907 Ops/s	48.9483 Ops/s	$\color{#35bf28}+1.72\%$
test_to_module_speed[True]	1.5025ms	1.4206ms	703.9347 Ops/s	704.6668 Ops/s	$\color{#d91a1a}-0.10\%$
test_to_module_speed[False]	1.4868ms	1.4027ms	712.9128 Ops/s	710.5752 Ops/s	$\color{#35bf28}+0.33\%$
test_tc_init	65.7920μs	44.8455μs	22.2988 KOps/s	22.2967 KOps/s	$+0.01\%$
test_tc_init_tensor_only	32.7300μs	9.4073μs	106.3004 KOps/s	107.4280 KOps/s	$\color{#d91a1a}-1.05\%$
test_tc_init_nested	0.1187ms	89.9750μs	11.1142 KOps/s	11.0648 KOps/s	$\color{#35bf28}+0.45\%$
test_tc_init_many_fields	37.7510μs	15.7732μs	63.3988 KOps/s	63.9272 KOps/s	$\color{#d91a1a}-0.83\%$
test_tc_first_layer_tensor	19.2500μs	1.7441μs	573.3553 KOps/s	586.8186 KOps/s	$\color{#d91a1a}-2.29\%$
test_tc_first_layer_tensor_only	5.5687μs	0.7110μs	1.4065 MOps/s	1.4271 MOps/s	$\color{#d91a1a}-1.44\%$
test_tc_first_layer_tensor_set	31.5300μs	4.0038μs	249.7650 KOps/s	253.9699 KOps/s	$\color{#d91a1a}-1.66\%$
test_tc_first_layer_tensor_only_set	28.3610μs	3.0020μs	333.1152 KOps/s	332.5503 KOps/s	$\color{#35bf28}+0.17\%$
test_tc_first_layer_nontensor	2.3987ms	5.8798μs	170.0735 KOps/s	172.7744 KOps/s	$\color{#d91a1a}-1.56\%$
test_tc_second_layer_tensor	36.8600μs	4.1609μs	240.3347 KOps/s	244.2335 KOps/s	$\color{#d91a1a}-1.60\%$
test_tc_second_layer_nontensor	30.6800μs	8.2575μs	121.1016 KOps/s	126.8796 KOps/s	$\color{#d91a1a}-4.55\%$
test_unbind	0.2411s	13.2413ms	75.5214 Ops/s	69.3823 Ops/s	$\textbf{\color{#35bf28}+8.85\%}$
test_full_like	5.4671ms	4.3308ms	230.9068 Ops/s	227.3441 Ops/s	$\color{#35bf28}+1.57\%$
test_zeros_like	5.0319ms	4.3547ms	229.6389 Ops/s	228.6198 Ops/s	$\color{#35bf28}+0.45\%$
test_ones_like	4.5191ms	4.3574ms	229.4938 Ops/s	228.2051 Ops/s	$\color{#35bf28}+0.56\%$
test_clone	6.7262ms	6.4512ms	155.0088 Ops/s	153.6532 Ops/s	$\color{#35bf28}+0.88\%$
test_squeeze	0.1786ms	13.5712μs	73.6852 KOps/s	66.4650 KOps/s	$\textbf{\color{#35bf28}+10.86\%}$
test_unsqueeze	0.1536ms	0.1102ms	9.0741 KOps/s	8.8629 KOps/s	$\color{#35bf28}+2.38\%$
test_split	0.3475ms	0.1760ms	5.6814 KOps/s	5.3556 KOps/s	$\textbf{\color{#35bf28}+6.08\%}$
test_permute	0.2670ms	0.2015ms	4.9639 KOps/s	4.7897 KOps/s	$\color{#35bf28}+3.64\%$
test_stack	52.2511ms	51.4144ms	19.4498 Ops/s	19.4800 Ops/s	$\color{#d91a1a}-0.16\%$
test_cat	51.6251ms	51.3769ms	19.4640 Ops/s	19.4537 Ops/s	$\color{#35bf28}+0.05\%$

github-actions · 2026-02-14T01:39:49Z

$\color{#D29922}\textsf{\Large&#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of CPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}14$. Worsened: $\large\color{#d91a1a}12$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	57.7130μs	14.8731μs	67.2353 KOps/s	67.0241 KOps/s	$\color{#35bf28}+0.32\%$
test_plain_set_stack_nested	40.2320μs	15.2653μs	65.5080 KOps/s	66.3270 KOps/s	$\color{#d91a1a}-1.23\%$
test_plain_set_nested_inplace	43.4020μs	16.5050μs	60.5877 KOps/s	59.9730 KOps/s	$\color{#35bf28}+1.02\%$
test_plain_set_stack_nested_inplace	59.0140μs	16.4465μs	60.8033 KOps/s	61.0029 KOps/s	$\color{#d91a1a}-0.33\%$
test_items	35.4630μs	5.8655μs	170.4884 KOps/s	172.7090 KOps/s	$\color{#d91a1a}-1.29\%$
test_items_nested	0.5975ms	0.5322ms	1.8790 KOps/s	1.8670 KOps/s	$\color{#35bf28}+0.64\%$
test_items_nested_locked	0.6395ms	0.5261ms	1.9008 KOps/s	1.8457 KOps/s	$\color{#35bf28}+2.99\%$
test_items_nested_leaf	0.1346ms	95.5464μs	10.4661 KOps/s	10.4033 KOps/s	$\color{#35bf28}+0.60\%$
test_items_stack_nested	0.6039ms	0.5420ms	1.8449 KOps/s	1.8717 KOps/s	$\color{#d91a1a}-1.43\%$
test_items_stack_nested_leaf	0.1372ms	97.7904μs	10.2260 KOps/s	10.4305 KOps/s	$\color{#d91a1a}-1.96\%$
test_items_stack_nested_locked	0.6230ms	0.5373ms	1.8612 KOps/s	1.8475 KOps/s	$\color{#35bf28}+0.74\%$
test_keys	30.0520μs	4.1966μs	238.2897 KOps/s	223.5863 KOps/s	$\textbf{\color{#35bf28}+6.58\%}$
test_keys_nested	0.1629ms	0.1195ms	8.3695 KOps/s	8.2741 KOps/s	$\color{#35bf28}+1.15\%$
test_keys_nested_locked	88.4651ms	0.1406ms	7.1117 KOps/s	7.7009 KOps/s	$\textbf{\color{#d91a1a}-7.65\%}$
test_keys_nested_leaf	0.1612ms	0.1108ms	9.0223 KOps/s	9.0666 KOps/s	$\color{#d91a1a}-0.49\%$
test_keys_stack_nested	0.1647ms	0.1207ms	8.2849 KOps/s	8.3159 KOps/s	$\color{#d91a1a}-0.37\%$
test_keys_stack_nested_leaf	0.1478ms	0.1105ms	9.0487 KOps/s	9.0245 KOps/s	$\color{#35bf28}+0.27\%$
test_keys_stack_nested_locked	0.1672ms	0.1282ms	7.7990 KOps/s	7.7077 KOps/s	$\color{#35bf28}+1.18\%$
test_values	6.0184μs	1.0173μs	983.0414 KOps/s	976.3358 KOps/s	$\color{#35bf28}+0.69\%$
test_values_nested	92.0660μs	47.7094μs	20.9602 KOps/s	20.6621 KOps/s	$\color{#35bf28}+1.44\%$
test_values_nested_locked	85.7150μs	50.7432μs	19.7071 KOps/s	19.6926 KOps/s	$\color{#35bf28}+0.07\%$
test_values_nested_leaf	85.6250μs	54.2007μs	18.4499 KOps/s	18.2171 KOps/s	$\color{#35bf28}+1.28\%$
test_values_stack_nested	84.9150μs	47.8319μs	20.9065 KOps/s	20.6023 KOps/s	$\color{#35bf28}+1.48\%$
test_values_stack_nested_leaf	93.0550μs	53.8305μs	18.5768 KOps/s	18.3135 KOps/s	$\color{#35bf28}+1.44\%$
test_values_stack_nested_locked	0.7847ms	50.4946μs	19.8041 KOps/s	19.4765 KOps/s	$\color{#35bf28}+1.68\%$
test_membership	4.1587μs	0.8595μs	1.1635 MOps/s	1.1748 MOps/s	$\color{#d91a1a}-0.96\%$
test_membership_nested	26.5420μs	3.2050μs	312.0151 KOps/s	312.2958 KOps/s	$\color{#d91a1a}-0.09\%$
test_membership_nested_leaf	33.9320μs	3.1655μs	315.9061 KOps/s	310.6203 KOps/s	$\color{#35bf28}+1.70\%$
test_membership_stacked_nested	27.1020μs	3.2246μs	310.1147 KOps/s	312.4235 KOps/s	$\color{#d91a1a}-0.74\%$
test_membership_stacked_nested_leaf	36.9020μs	3.1729μs	315.1667 KOps/s	313.0597 KOps/s	$\color{#35bf28}+0.67\%$
test_membership_nested_last	34.4620μs	4.6629μs	214.4610 KOps/s	217.1786 KOps/s	$\color{#d91a1a}-1.25\%$
test_membership_nested_leaf_last	32.3120μs	4.6911μs	213.1694 KOps/s	215.3491 KOps/s	$\color{#d91a1a}-1.01\%$
test_membership_stacked_nested_last	25.5720μs	4.6472μs	215.1837 KOps/s	213.6375 KOps/s	$\color{#35bf28}+0.72\%$
test_membership_stacked_nested_leaf_last	34.8520μs	4.6454μs	215.2661 KOps/s	215.1544 KOps/s	$\color{#35bf28}+0.05\%$
test_nested_getleaf	46.7630μs	22.0064μs	45.4413 KOps/s	45.6060 KOps/s	$\color{#d91a1a}-0.36\%$
test_nested_get	51.8030μs	20.5686μs	48.6177 KOps/s	49.1273 KOps/s	$\color{#d91a1a}-1.04\%$
test_stacked_getleaf	47.8220μs	22.1499μs	45.1470 KOps/s	46.2246 KOps/s	$\color{#d91a1a}-2.33\%$
test_stacked_get	49.8630μs	20.8434μs	47.9767 KOps/s	48.1779 KOps/s	$\color{#d91a1a}-0.42\%$
test_nested_getitemleaf	51.4730μs	22.3251μs	44.7925 KOps/s	44.2700 KOps/s	$\color{#35bf28}+1.18\%$
test_nested_getitem	47.1030μs	21.1587μs	47.2620 KOps/s	47.0771 KOps/s	$\color{#35bf28}+0.39\%$
test_stacked_getitemleaf	52.8030μs	22.6638μs	44.1232 KOps/s	44.6083 KOps/s	$\color{#d91a1a}-1.09\%$
test_stacked_getitem	47.0930μs	21.3417μs	46.8567 KOps/s	46.3611 KOps/s	$\color{#35bf28}+1.07\%$
test_lock_nested	7.5813ms	0.4749ms	2.1059 KOps/s	2.0760 KOps/s	$\color{#35bf28}+1.44\%$
test_lock_stack_nested	0.5365ms	0.4751ms	2.1049 KOps/s	2.0483 KOps/s	$\color{#35bf28}+2.76\%$
test_unlock_nested	0.5114ms	0.3784ms	2.6428 KOps/s	2.5966 KOps/s	$\color{#35bf28}+1.78\%$
test_unlock_stack_nested	0.4307ms	0.3815ms	2.6210 KOps/s	2.5684 KOps/s	$\color{#35bf28}+2.05\%$
test_flatten_speed	0.1557ms	0.1212ms	8.2510 KOps/s	8.0311 KOps/s	$\color{#35bf28}+2.74\%$
test_unflatten_speed	0.6597ms	0.5912ms	1.6916 KOps/s	1.6663 KOps/s	$\color{#35bf28}+1.52\%$
test_common_ops	0.8116ms	0.6858ms	1.4582 KOps/s	1.4275 KOps/s	$\color{#35bf28}+2.15\%$
test_creation	69.7640μs	2.8121μs	355.6042 KOps/s	340.2768 KOps/s	$\color{#35bf28}+4.50\%$
test_creation_empty	31.0220μs	6.1402μs	162.8616 KOps/s	161.4765 KOps/s	$\color{#35bf28}+0.86\%$
test_creation_nested_1	31.5010μs	10.9426μs	91.3863 KOps/s	91.1254 KOps/s	$\color{#35bf28}+0.29\%$
test_creation_nested_2	34.1630μs	12.0034μs	83.3097 KOps/s	82.4825 KOps/s	$\color{#35bf28}+1.00\%$
test_creation_many_keys[10]	55.2740μs	18.2064μs	54.9258 KOps/s	53.8795 KOps/s	$\color{#35bf28}+1.94\%$
test_creation_many_keys[50]	0.1062ms	77.9965μs	12.8211 KOps/s	12.6029 KOps/s	$\color{#35bf28}+1.73\%$
test_creation_many_keys[100]	0.1907ms	0.1521ms	6.5744 KOps/s	6.5242 KOps/s	$\color{#35bf28}+0.77\%$
test_creation_nested_many_keys[10]	63.5340μs	39.3007μs	25.4449 KOps/s	25.0940 KOps/s	$\color{#35bf28}+1.40\%$
test_creation_nested_many_keys[50]	0.1907ms	0.1585ms	6.3106 KOps/s	6.2140 KOps/s	$\color{#35bf28}+1.55\%$
test_clone	43.4130μs	13.4554μs	74.3196 KOps/s	73.8232 KOps/s	$\color{#35bf28}+0.67\%$
test_getitem[int]	1.6938ms	14.6725μs	68.1549 KOps/s	56.5448 KOps/s	$\textbf{\color{#35bf28}+20.53\%}$
test_getitem[slice_int]	0.1403ms	25.2186μs	39.6533 KOps/s	39.5271 KOps/s	$\color{#35bf28}+0.32\%$
test_getitem[range]	0.1789ms	61.9277μs	16.1479 KOps/s	15.9466 KOps/s	$\color{#35bf28}+1.26\%$
test_getitem[tuple]	0.1525ms	24.4048μs	40.9755 KOps/s	41.0721 KOps/s	$\color{#d91a1a}-0.24\%$
test_getitem[list]	0.1868ms	56.7096μs	17.6337 KOps/s	17.2111 KOps/s	$\color{#35bf28}+2.46\%$
test_setitem_dim[int]	47.5530μs	26.1626μs	38.2225 KOps/s	38.1474 KOps/s	$\color{#35bf28}+0.20\%$
test_setitem_dim[slice_int]	67.2650μs	44.9331μs	22.2553 KOps/s	22.3278 KOps/s	$\color{#d91a1a}-0.32\%$
test_setitem_dim[range]	0.1185ms	92.9428μs	10.7593 KOps/s	10.6128 KOps/s	$\color{#35bf28}+1.38\%$
test_setitem_dim[tuple]	64.0340μs	40.9599μs	24.4141 KOps/s	24.0456 KOps/s	$\color{#35bf28}+1.53\%$
test_setitem	56.4730μs	18.6125μs	53.7272 KOps/s	54.2373 KOps/s	$\color{#d91a1a}-0.94\%$
test_set	44.8130μs	17.5337μs	57.0330 KOps/s	56.9990 KOps/s	$\color{#35bf28}+0.06\%$
test_set_shared	0.5886ms	0.2072ms	4.8263 KOps/s	4.8242 KOps/s	$\color{#35bf28}+0.04\%$
test_update	0.1620ms	22.6801μs	44.0915 KOps/s	44.4513 KOps/s	$\color{#d91a1a}-0.81\%$
test_update_nested	75.1550μs	35.0669μs	28.5169 KOps/s	28.7394 KOps/s	$\color{#d91a1a}-0.77\%$
test_update__nested	0.4492ms	34.4823μs	29.0003 KOps/s	28.7670 KOps/s	$\color{#35bf28}+0.81\%$
test_set_nested	63.9340μs	19.5732μs	51.0904 KOps/s	51.4158 KOps/s	$\color{#d91a1a}-0.63\%$
test_set_nested_new	62.4040μs	24.4314μs	40.9310 KOps/s	40.1596 KOps/s	$\color{#35bf28}+1.92\%$
test_select	80.7650μs	42.6247μs	23.4606 KOps/s	23.2035 KOps/s	$\color{#35bf28}+1.11\%$
test_select_nested	0.1071ms	75.3083μs	13.2788 KOps/s	13.2925 KOps/s	$\color{#d91a1a}-0.10\%$
test_exclude_nested	0.1345ms	98.0142μs	10.2026 KOps/s	10.2168 KOps/s	$\color{#d91a1a}-0.14\%$
test_empty[True]	0.5031ms	0.4415ms	2.2648 KOps/s	2.2652 KOps/s	$\color{#d91a1a}-0.02\%$
test_empty[False]	7.4430μs	1.3264μs	753.8969 KOps/s	759.1225 KOps/s	$\color{#d91a1a}-0.69\%$
test_to	0.1031ms	73.6931μs	13.5698 KOps/s	13.6123 KOps/s	$\color{#d91a1a}-0.31\%$
test_to_nonblocking	0.1004ms	65.6954μs	15.2218 KOps/s	15.1455 KOps/s	$\color{#35bf28}+0.50\%$
test_unbind_speed	0.3704ms	0.3261ms	3.0666 KOps/s	3.0538 KOps/s	$\color{#35bf28}+0.42\%$
test_unbind_speed_stack0	0.3758ms	0.3236ms	3.0899 KOps/s	3.0464 KOps/s	$\color{#35bf28}+1.43\%$
test_unbind_speed_stack1	0.1026s	0.9122ms	1.0963 KOps/s	1.1870 KOps/s	$\textbf{\color{#d91a1a}-7.64\%}$
test_split	1.2214ms	1.1458ms	872.7158 Ops/s	784.4757 Ops/s	$\textbf{\color{#35bf28}+11.25\%}$
test_chunk	0.1018s	1.2085ms	827.4524 Ops/s	916.4067 Ops/s	$\textbf{\color{#d91a1a}-9.71\%}$
test_to_cpu_blocking	28.6797ms	28.5949ms	34.9713 Ops/s	46.8310 Ops/s	$\textbf{\color{#d91a1a}-25.32\%}$
test_to_cpu_global_sync	11.3268ms	11.2254ms	89.0834 Ops/s	89.2523 Ops/s	$\color{#d91a1a}-0.19\%$
test_to_cpu_event_sync	12.3964ms	12.2001ms	81.9663 Ops/s	81.8864 Ops/s	$\color{#35bf28}+0.10\%$
test_to_cpu_default	0.1141s	13.4967ms	74.0924 Ops/s	81.9372 Ops/s	$\textbf{\color{#d91a1a}-9.57\%}$
test_consolidate[False-None]	4.1992ms	4.1425ms	241.4012 Ops/s	219.7046 Ops/s	$\textbf{\color{#35bf28}+9.88\%}$
test_consolidate[default-None]	2.4503ms	2.0288ms	492.8990 Ops/s	489.2368 Ops/s	$\color{#35bf28}+0.75\%$
test_consolidate[reduce-overhead-None]	2.3563ms	1.9410ms	515.2033 Ops/s	513.1994 Ops/s	$\color{#35bf28}+0.39\%$
test_consolidate_njt[False-None]	8.9696ms	8.5466ms	117.0054 Ops/s	117.0215 Ops/s	$\color{#d91a1a}-0.01\%$
test_to[False-False-None]	2.4840ms	2.0686ms	483.4101 Ops/s	478.3350 Ops/s	$\color{#35bf28}+1.06\%$
test_to[True-False-None]	2.3379ms	1.9444ms	514.3039 Ops/s	511.3351 Ops/s	$\color{#35bf28}+0.58\%$
test_to[within-False-None]	6.6408ms	6.2160ms	160.8741 Ops/s	161.4492 Ops/s	$\color{#d91a1a}-0.36\%$
test_to[True-default-None]	8.0069ms	7.6223ms	131.1943 Ops/s	129.0495 Ops/s	$\color{#35bf28}+1.66\%$
test_to_njt[False-False-None]	8.9258ms	8.5347ms	117.1685 Ops/s	115.4941 Ops/s	$\color{#35bf28}+1.45\%$
test_to_njt[True-False-None]	7.4462ms	7.0399ms	142.0474 Ops/s	137.3797 Ops/s	$\color{#35bf28}+3.40\%$
test_to_njt[within-False-None]	16.9708ms	16.0244ms	62.4048 Ops/s	63.6655 Ops/s	$\color{#d91a1a}-1.98\%$
test_creation[device0]	0.5119ms	0.1215ms	8.2306 KOps/s	8.4433 KOps/s	$\color{#d91a1a}-2.52\%$
test_creation_from_tensor	0.5345ms	0.1156ms	8.6474 KOps/s	8.6200 KOps/s	$\color{#35bf28}+0.32\%$
test_add_one[memmap_tensor0]	0.4496ms	6.4104μs	155.9960 KOps/s	151.0374 KOps/s	$\color{#35bf28}+3.28\%$
test_contiguous[memmap_tensor0]	14.4910μs	0.7123μs	1.4038 MOps/s	1.9360 MOps/s	$\textbf{\color{#d91a1a}-27.49\%}$
test_stack[memmap_tensor0]	27.5920μs	4.6022μs	217.2870 KOps/s	222.4626 KOps/s	$\color{#d91a1a}-2.33\%$
test_memmaptd_index	1.1114ms	0.2660ms	3.7599 KOps/s	3.7686 KOps/s	$\color{#d91a1a}-0.23\%$
test_memmaptd_index_astensor	0.7878ms	0.3628ms	2.7566 KOps/s	2.7754 KOps/s	$\color{#d91a1a}-0.68\%$
test_memmaptd_index_op	1.0371ms	0.6061ms	1.6500 KOps/s	1.6322 KOps/s	$\color{#35bf28}+1.09\%$
test_serialize_model	0.1383s	0.1363s	7.3374 Ops/s	7.3509 Ops/s	$\color{#d91a1a}-0.18\%$
test_serialize_model_pickle	1.3654s	1.2167s	0.8219 Ops/s	0.8264 Ops/s	$\color{#d91a1a}-0.55\%$
test_serialize_weights	0.1391s	0.1351s	7.4044 Ops/s	7.3257 Ops/s	$\color{#35bf28}+1.07\%$
test_serialize_weights_returnearly	0.4290s	92.8567ms	10.7693 Ops/s	5.9994 Ops/s	$\textbf{\color{#35bf28}+79.51\%}$
test_serialize_weights_pickle	1.3659s	1.1890s	0.8410 Ops/s	0.8218 Ops/s	$\color{#35bf28}+2.34\%$
test_reshape_pytree	0.1996ms	33.6639μs	29.7054 KOps/s	29.8638 KOps/s	$\color{#d91a1a}-0.53\%$
test_reshape_td	72.1040μs	44.1659μs	22.6419 KOps/s	21.4426 KOps/s	$\textbf{\color{#35bf28}+5.59\%}$
test_view_pytree	0.2195ms	33.4321μs	29.9114 KOps/s	30.0761 KOps/s	$\color{#d91a1a}-0.55\%$
test_view_td	85.4450μs	51.5663μs	19.3925 KOps/s	18.9994 KOps/s	$\color{#35bf28}+2.07\%$
test_unbind_pytree	0.2375ms	36.9467μs	27.0660 KOps/s	26.5668 KOps/s	$\color{#35bf28}+1.88\%$
test_unbind_td	0.1993ms	48.7119μs	20.5289 KOps/s	20.2239 KOps/s	$\color{#35bf28}+1.51\%$
test_split_pytree	0.2305ms	43.5573μs	22.9583 KOps/s	23.4280 KOps/s	$\color{#d91a1a}-2.01\%$
test_split_td	0.1852ms	65.6773μs	15.2260 KOps/s	15.3263 KOps/s	$\color{#d91a1a}-0.65\%$
test_add_pytree	0.2371ms	42.6719μs	23.4346 KOps/s	23.7134 KOps/s	$\color{#d91a1a}-1.18\%$
test_add_td	85.3950μs	53.7773μs	18.5952 KOps/s	18.9010 KOps/s	$\color{#d91a1a}-1.62\%$
test_compile_add_one_nested[tensordict-compile]	0.1912ms	0.1390ms	7.1933 KOps/s	6.7170 KOps/s	$\textbf{\color{#35bf28}+7.09\%}$
test_compile_add_one_nested[tensordict-eager]	0.4088ms	0.1925ms	5.1955 KOps/s	5.1568 KOps/s	$\color{#35bf28}+0.75\%$
test_compile_add_one_nested[pytree-compile]	0.1550ms	0.1083ms	9.2362 KOps/s	9.1199 KOps/s	$\color{#35bf28}+1.28\%$
test_compile_add_one_nested[pytree-eager]	0.4393ms	0.1825ms	5.4793 KOps/s	5.5181 KOps/s	$\color{#d91a1a}-0.70\%$
test_compile_copy_nested[tensordict-compile]	0.1505ms	29.9605μs	33.3773 KOps/s	30.2496 KOps/s	$\textbf{\color{#35bf28}+10.34\%}$
test_compile_copy_nested[tensordict-eager]	90.9460μs	52.6651μs	18.9879 KOps/s	18.4666 KOps/s	$\color{#35bf28}+2.82\%$
test_compile_copy_nested[pytree-compile]	45.0430μs	10.0935μs	99.0736 KOps/s	102.4606 KOps/s	$\color{#d91a1a}-3.31\%$
test_compile_copy_nested[pytree-eager]	0.4365ms	70.7620μs	14.1319 KOps/s	14.2688 KOps/s	$\color{#d91a1a}-0.96\%$
test_compile_add_one_flat[tensordict-compile]	0.3040ms	0.1790ms	5.5852 KOps/s	5.3138 KOps/s	$\textbf{\color{#35bf28}+5.11\%}$
test_compile_add_one_flat[tensordict-eager]	0.3052ms	0.2530ms	3.9520 KOps/s	3.8473 KOps/s	$\color{#35bf28}+2.72\%$
test_compile_add_one_flat[tensorclass-compile]	0.1777ms	0.1183ms	8.4561 KOps/s	8.2495 KOps/s	$\color{#35bf28}+2.50\%$
test_compile_add_one_flat[tensorclass-eager]	0.1038ms	70.2497μs	14.2349 KOps/s	14.2666 KOps/s	$\color{#d91a1a}-0.22\%$
test_compile_add_one_flat[pytree-compile]	0.2189ms	0.1594ms	6.2747 KOps/s	6.1735 KOps/s	$\color{#35bf28}+1.64\%$
test_compile_add_one_flat[pytree-eager]	0.8351ms	0.5285ms	1.8920 KOps/s	1.8701 KOps/s	$\color{#35bf28}+1.17\%$
test_compile_add_self_flat[tensordict-eager]	0.4601ms	0.3082ms	3.2442 KOps/s	3.1722 KOps/s	$\color{#35bf28}+2.27\%$
test_compile_add_self_flat[tensordict-compile]	0.2312ms	0.1795ms	5.5706 KOps/s	5.2660 KOps/s	$\textbf{\color{#35bf28}+5.78\%}$
test_compile_add_self_flat[tensorclass-eager]	0.1263ms	86.1686μs	11.6052 KOps/s	11.6273 KOps/s	$\color{#d91a1a}-0.19\%$
test_compile_add_self_flat[tensorclass-compile]	0.2020ms	0.1202ms	8.3224 KOps/s	8.0717 KOps/s	$\color{#35bf28}+3.11\%$
test_compile_add_self_flat[pytree-eager]	0.6595ms	0.4360ms	2.2938 KOps/s	2.2500 KOps/s	$\color{#35bf28}+1.94\%$
test_compile_add_self_flat[pytree-compile]	0.1948ms	0.1595ms	6.2700 KOps/s	6.1377 KOps/s	$\color{#35bf28}+2.16\%$
test_compile_copy_flat[tensordict-compile]	75.4440μs	24.2088μs	41.3073 KOps/s	39.6893 KOps/s	$\color{#35bf28}+4.08\%$
test_compile_copy_flat[tensordict-eager]	97.9260μs	41.5771μs	24.0517 KOps/s	23.8854 KOps/s	$\color{#35bf28}+0.70\%$
test_compile_copy_flat[pytree-compile]	36.7020μs	11.0939μs	90.1396 KOps/s	91.2262 KOps/s	$\color{#d91a1a}-1.19\%$
test_compile_copy_flat[pytree-eager]	0.1777s	62.5607μs	15.9845 KOps/s	18.9695 KOps/s	$\textbf{\color{#d91a1a}-15.74\%}$
test_compile_assign_and_add[tensordict-compile]	2.0217ms	0.1742ms	5.7416 KOps/s	5.3944 KOps/s	$\textbf{\color{#35bf28}+6.44\%}$
test_compile_assign_and_add[tensordict-eager]	3.4254ms	3.3108ms	302.0375 Ops/s	298.5079 Ops/s	$\color{#35bf28}+1.18\%$
test_compile_assign_and_add[pytree-compile]	1.9667ms	0.1620ms	6.1711 KOps/s	6.0638 KOps/s	$\color{#35bf28}+1.77\%$
test_compile_assign_and_add[pytree-eager]	2.9084ms	2.7747ms	360.3953 Ops/s	357.4238 Ops/s	$\color{#35bf28}+0.83\%$
test_compile_indexing[tensor-tensordict-compile]	0.1443ms	0.1091ms	9.1664 KOps/s	8.8923 KOps/s	$\color{#35bf28}+3.08\%$
test_compile_indexing[tensor-tensordict-eager]	0.3139ms	73.1973μs	13.6617 KOps/s	13.5795 KOps/s	$\color{#35bf28}+0.61\%$
test_compile_indexing[tensor-tensorclass-compile]	0.1441ms	96.7933μs	10.3313 KOps/s	10.1571 KOps/s	$\color{#35bf28}+1.71\%$
test_compile_indexing[tensor-tensorclass-eager]	0.2568ms	44.8076μs	22.3176 KOps/s	21.8954 KOps/s	$\color{#35bf28}+1.93\%$
test_compile_indexing[tensor-pytree-compile]	0.1607ms	97.9076μs	10.2137 KOps/s	10.1779 KOps/s	$\color{#35bf28}+0.35\%$
test_compile_indexing[tensor-pytree-eager]	0.2588ms	44.4854μs	22.4793 KOps/s	21.9225 KOps/s	$\color{#35bf28}+2.54\%$
test_compile_indexing[slice-tensordict-compile]	0.1013ms	56.0597μs	17.8381 KOps/s	17.4101 KOps/s	$\color{#35bf28}+2.46\%$
test_compile_indexing[slice-tensordict-eager]	0.2277ms	27.7994μs	35.9720 KOps/s	34.2563 KOps/s	$\textbf{\color{#35bf28}+5.01\%}$
test_compile_indexing[slice-tensorclass-compile]	0.1050ms	45.9639μs	21.7562 KOps/s	21.7703 KOps/s	$\color{#d91a1a}-0.06\%$
test_compile_indexing[slice-tensorclass-eager]	0.2682ms	22.5207μs	44.4035 KOps/s	42.8107 KOps/s	$\color{#35bf28}+3.72\%$
test_compile_indexing[slice-pytree-compile]	90.5760μs	46.4560μs	21.5258 KOps/s	21.5461 KOps/s	$\color{#d91a1a}-0.09\%$
test_compile_indexing[slice-pytree-eager]	0.2694ms	22.6455μs	44.1589 KOps/s	43.3897 KOps/s	$\color{#35bf28}+1.77\%$
test_compile_indexing[int-tensordict-compile]	95.1760μs	56.5719μs	17.6766 KOps/s	17.2876 KOps/s	$\color{#35bf28}+2.25\%$
test_compile_indexing[int-tensordict-eager]	0.2197ms	27.6473μs	36.1699 KOps/s	34.6056 KOps/s	$\color{#35bf28}+4.52\%$
test_compile_indexing[int-tensorclass-compile]	86.4950μs	46.0830μs	21.7000 KOps/s	21.5035 KOps/s	$\color{#35bf28}+0.91\%$
test_compile_indexing[int-tensorclass-eager]	0.2594ms	22.5338μs	44.3777 KOps/s	43.3487 KOps/s	$\color{#35bf28}+2.37\%$
test_compile_indexing[int-pytree-compile]	92.5550μs	46.3123μs	21.5926 KOps/s	21.5913 KOps/s	$+0.01\%$
test_compile_indexing[int-pytree-eager]	0.2784ms	22.5845μs	44.2782 KOps/s	43.4472 KOps/s	$\color{#35bf28}+1.91\%$
test_mod_add[eager]	0.1031ms	51.9254μs	19.2584 KOps/s	19.5087 KOps/s	$\color{#d91a1a}-1.28\%$
test_mod_add[compile]	0.1818ms	0.1056ms	9.4732 KOps/s	9.3938 KOps/s	$\color{#35bf28}+0.85\%$
test_mod_add[compile-overhead]	0.2339ms	0.1493ms	6.6990 KOps/s	6.5770 KOps/s	$\color{#35bf28}+1.86\%$
test_mod_wrap[eager]	0.3867ms	0.2942ms	3.3986 KOps/s	3.4054 KOps/s	$\color{#d91a1a}-0.20\%$
test_mod_wrap[compile]	0.4236ms	0.3497ms	2.8600 KOps/s	2.8368 KOps/s	$\color{#35bf28}+0.82\%$
test_mod_wrap[compile-overhead]	7.4151ms	4.0725ms	245.5487 Ops/s	249.4051 Ops/s	$\color{#d91a1a}-1.55\%$
test_mod_wrap_and_backward[eager]	1.6404ms	1.4990ms	667.1143 Ops/s	641.6356 Ops/s	$\color{#35bf28}+3.97\%$
test_mod_wrap_and_backward[compile]	1.9471ms	1.4896ms	671.3062 Ops/s	680.1041 Ops/s	$\color{#d91a1a}-1.29\%$
test_mod_wrap_and_backward[compile-overhead]	1.2648ms	0.8990ms	1.1123 KOps/s	1.1067 KOps/s	$\color{#35bf28}+0.50\%$
test_seq_add[eager]	0.5942ms	0.1681ms	5.9474 KOps/s	6.3232 KOps/s	$\textbf{\color{#d91a1a}-5.94\%}$
test_seq_add[compile]	0.5637ms	0.1217ms	8.2194 KOps/s	8.4139 KOps/s	$\color{#d91a1a}-2.31\%$
test_seq_add[compile-overhead]	0.6055ms	0.1641ms	6.0952 KOps/s	6.2743 KOps/s	$\color{#d91a1a}-2.85\%$
test_seq_wrap[eager]	1.0336ms	0.5575ms	1.7937 KOps/s	1.8452 KOps/s	$\color{#d91a1a}-2.79\%$
test_seq_wrap[compile]	0.4700ms	0.3682ms	2.7162 KOps/s	2.5907 KOps/s	$\color{#35bf28}+4.84\%$
test_seq_wrap[compile-overhead]	0.3730ms	0.2645ms	3.7810 KOps/s	3.7111 KOps/s	$\color{#35bf28}+1.88\%$
test_func_call_runtime[False-eager]	0.9143ms	0.8432ms	1.1860 KOps/s	1.1039 KOps/s	$\textbf{\color{#35bf28}+7.44\%}$
test_func_call_runtime[False-compile]	0.9995ms	0.9144ms	1.0936 KOps/s	1.0425 KOps/s	$\color{#35bf28}+4.90\%$
test_func_call_runtime[False-compile-overhead]	0.5232ms	0.4628ms	2.1607 KOps/s	2.1482 KOps/s	$\color{#35bf28}+0.58\%$
test_func_call_runtime[True-eager]	1.1588ms	1.0958ms	912.5944 Ops/s	913.0356 Ops/s	$\color{#d91a1a}-0.05\%$
test_func_call_runtime[True-compile]	0.9918ms	0.9228ms	1.0836 KOps/s	1.0758 KOps/s	$\color{#35bf28}+0.73\%$
test_func_call_runtime[True-compile-overhead]	0.5375ms	0.4726ms	2.1159 KOps/s	2.0724 KOps/s	$\color{#35bf28}+2.10\%$
test_func_call_cm_runtime[False-eager]	0.9085ms	0.8414ms	1.1884 KOps/s	1.1740 KOps/s	$\color{#35bf28}+1.23\%$
test_func_call_cm_runtime[False-compile]	1.1157ms	0.9185ms	1.0887 KOps/s	1.0839 KOps/s	$\color{#35bf28}+0.45\%$
test_func_call_cm_runtime[False-compile-overhead]	0.5525ms	0.4633ms	2.1584 KOps/s	2.1344 KOps/s	$\color{#35bf28}+1.13\%$
test_func_call_cm_runtime[True-eager]	1.3616ms	1.2323ms	811.5107 Ops/s	797.2967 Ops/s	$\color{#35bf28}+1.78\%$
test_func_call_cm_runtime[True-compile]	1.1708ms	0.9541ms	1.0481 KOps/s	1.0329 KOps/s	$\color{#35bf28}+1.47\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5887ms	0.5088ms	1.9652 KOps/s	1.9475 KOps/s	$\color{#35bf28}+0.91\%$
test_vmap_func_call_cm_runtime[eager]	2.8421ms	2.3551ms	424.6035 Ops/s	419.1372 Ops/s	$\color{#35bf28}+1.30\%$
test_vmap_func_call_cm_runtime[compile]	1.0543ms	0.9801ms	1.0203 KOps/s	1.0132 KOps/s	$\color{#35bf28}+0.71\%$
test_vmap_func_call_cm_runtime[compile-overhead]	0.5594ms	0.5125ms	1.9511 KOps/s	1.9274 KOps/s	$\color{#35bf28}+1.23\%$
test_distributed	0.7950ms	0.1533ms	6.5217 KOps/s	6.4605 KOps/s	$\color{#35bf28}+0.95\%$
test_tdmodule	0.3039ms	29.3107μs	34.1172 KOps/s	33.3891 KOps/s	$\color{#35bf28}+2.18\%$
test_tdmodule_dispatch	78.2150μs	47.5921μs	21.0119 KOps/s	20.6943 KOps/s	$\color{#35bf28}+1.53\%$
test_tdseq	48.6130μs	27.9095μs	35.8301 KOps/s	35.3509 KOps/s	$\color{#35bf28}+1.36\%$
test_tdseq_dispatch	71.4640μs	49.1078μs	20.3634 KOps/s	19.6114 KOps/s	$\color{#35bf28}+3.83\%$
test_instantiation_functorch	2.1494ms	2.0778ms	481.2786 Ops/s	480.6254 Ops/s	$\color{#35bf28}+0.14\%$
test_exec_functorch	0.2228ms	0.1812ms	5.5196 KOps/s	5.5931 KOps/s	$\color{#d91a1a}-1.31\%$
test_exec_functional_call	0.2319ms	0.1618ms	6.1802 KOps/s	6.2185 KOps/s	$\color{#d91a1a}-0.62\%$
test_exec_td_decorator	0.4500ms	0.2373ms	4.2142 KOps/s	4.1639 KOps/s	$\color{#35bf28}+1.21\%$
test_vmap_mlp_speed_decorator[True-True]	1.0394ms	0.8224ms	1.2160 KOps/s	1.1970 KOps/s	$\color{#35bf28}+1.59\%$
test_vmap_mlp_speed_decorator[True-False]	0.9949ms	0.8218ms	1.2169 KOps/s	1.2004 KOps/s	$\color{#35bf28}+1.37\%$
test_vmap_mlp_speed_decorator[False-True]	0.8870ms	0.7108ms	1.4069 KOps/s	1.3834 KOps/s	$\color{#35bf28}+1.70\%$
test_vmap_mlp_speed_decorator[False-False]	0.8879ms	0.7111ms	1.4063 KOps/s	1.3873 KOps/s	$\color{#35bf28}+1.37\%$
test_vmap_transformer_speed_decorator[True-True]	21.3341ms	20.5888ms	48.5702 Ops/s	48.4520 Ops/s	$\color{#35bf28}+0.24\%$
test_vmap_transformer_speed_decorator[True-False]	20.9925ms	20.5484ms	48.6656 Ops/s	48.4422 Ops/s	$\color{#35bf28}+0.46\%$
test_vmap_transformer_speed_decorator[False-True]	21.2605ms	20.3721ms	49.0868 Ops/s	48.8741 Ops/s	$\color{#35bf28}+0.44\%$
test_vmap_transformer_speed_decorator[False-False]	20.5323ms	20.3666ms	49.1000 Ops/s	48.8326 Ops/s	$\color{#35bf28}+0.55\%$
test_to_module_speed[True]	1.5587ms	1.4730ms	678.8871 Ops/s	669.3019 Ops/s	$\color{#35bf28}+1.43\%$
test_to_module_speed[False]	1.5779ms	1.4379ms	695.4353 Ops/s	684.2293 Ops/s	$\color{#35bf28}+1.64\%$
test_tc_init	78.5750μs	46.2440μs	21.6244 KOps/s	20.9390 KOps/s	$\color{#35bf28}+3.27\%$
test_tc_init_tensor_only	29.0510μs	10.0325μs	99.6762 KOps/s	99.9967 KOps/s	$\color{#d91a1a}-0.32\%$
test_tc_init_nested	0.1283ms	93.3694μs	10.7101 KOps/s	10.4802 KOps/s	$\color{#35bf28}+2.19\%$
test_tc_init_many_fields	46.2130μs	16.6994μs	59.8823 KOps/s	59.8179 KOps/s	$\color{#35bf28}+0.11\%$
test_tc_first_layer_tensor	19.9310μs	1.8698μs	534.8207 KOps/s	542.5069 KOps/s	$\color{#d91a1a}-1.42\%$
test_tc_first_layer_tensor_only	5.1631μs	0.7544μs	1.3256 MOps/s	1.3106 MOps/s	$\color{#35bf28}+1.14\%$
test_tc_first_layer_tensor_set	33.6220μs	4.2429μs	235.6856 KOps/s	235.6754 KOps/s	$+0.00\%$
test_tc_first_layer_tensor_only_set	20.7510μs	3.1855μs	313.9203 KOps/s	309.2807 KOps/s	$\color{#35bf28}+1.50\%$
test_tc_first_layer_nontensor	55.0040μs	6.2494μs	160.0162 KOps/s	160.7073 KOps/s	$\color{#d91a1a}-0.43\%$
test_tc_second_layer_tensor	35.7630μs	4.4877μs	222.8313 KOps/s	225.1473 KOps/s	$\color{#d91a1a}-1.03\%$
test_tc_second_layer_nontensor	67.3940μs	8.8011μs	113.6223 KOps/s	112.6837 KOps/s	$\color{#35bf28}+0.83\%$
test_unbind	0.2429s	14.1877ms	70.4837 Ops/s	57.3203 Ops/s	$\textbf{\color{#35bf28}+22.96\%}$
test_full_like	4.6989ms	4.3685ms	228.9135 Ops/s	227.8356 Ops/s	$\color{#35bf28}+0.47\%$
test_zeros_like	11.0574ms	10.5606ms	94.6920 Ops/s	228.9871 Ops/s	$\textbf{\color{#d91a1a}-58.65\%}$
test_ones_like	10.6587ms	10.5408ms	94.8692 Ops/s	229.1657 Ops/s	$\textbf{\color{#d91a1a}-58.60\%}$
test_clone	15.1738ms	15.0345ms	66.5135 Ops/s	154.1283 Ops/s	$\textbf{\color{#d91a1a}-56.85\%}$
test_squeeze	0.1521ms	14.2765μs	70.0453 KOps/s	69.3117 KOps/s	$\color{#35bf28}+1.06\%$
test_unsqueeze	0.2615ms	0.1111ms	9.0018 KOps/s	9.0721 KOps/s	$\color{#d91a1a}-0.78\%$
test_split	0.2464ms	0.1834ms	5.4523 KOps/s	5.3162 KOps/s	$\color{#35bf28}+2.56\%$
test_permute	0.2802ms	0.2046ms	4.8867 KOps/s	4.8910 KOps/s	$\color{#d91a1a}-0.09\%$
test_stack	53.1967ms	51.6606ms	19.3571 Ops/s	19.4468 Ops/s	$\color{#d91a1a}-0.46\%$
test_cat	51.7298ms	51.5539ms	19.3972 Ops/s	23.2310 Ops/s	$\textbf{\color{#d91a1a}-16.50\%}$

Update

1e29e0a

[ghstack-poisoned]

meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Feb 14, 2026

vmoens added the Performance label Feb 16, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Performance] Use Lua scripts for scattered index operations#1571

[Performance] Use Lua scripts for scattered index operations#1571
vmoens wants to merge 1 commit intogh/vmoens/57/basefrom
gh/vmoens/57/head

vmoens commented Feb 14, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Feb 14, 2026

Uh oh!

github-actions bot commented Feb 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

vmoens commented Feb 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Feb 14, 2026

$\color{#D29922}\textsf{\Large&amp;#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of GPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}16$. Worsened: $\large\color{#d91a1a}7$.

Uh oh!

github-actions bot commented Feb 14, 2026

$\color{#D29922}\textsf{\Large&amp;#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of CPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}14$. Worsened: $\large\color{#d91a1a}12$.

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

vmoens commented Feb 14, 2026 •

edited

Loading

$\color{#D29922}\textsf{\Large⚠\kern{0.2cm}\normalsize Warning}$ Result of GPU Benchmark Tests

$\color{#D29922}\textsf{\Large⚠\kern{0.2cm}\normalsize Warning}$ Result of CPU Benchmark Tests