experiment: perf run with tagging disabled

[crusso]

Using code from https://github.com/dfinity/motoko/pull/4400 which incorporates but disables tagging, wideing scalars to their original width.

To see ramifications of checking the tagging feature in, but gated off, until future use.

[github-actions[bot]]

Note Diffing the performance result against the published result from main branch. Unchanged benchmarks are omitted.

Map

	binary_size	generate 1m	max mem	batch_get 50	batch_put 50	batch_remove 50	upgrade
hashmap	189_916 ($\textcolor{red}{0.16\%}$)	8_184_618_025 ($\textcolor{green}{-2.22\%}$)	56_000_256 ($\textcolor{green}{-9.66\%}$)	342_784 ($\textcolor{green}{-0.62\%}$)	6_462_528_122 ($\textcolor{green}{-1.98\%}$)	368_420 ($\textcolor{green}{-0.76\%}$)	10_728_193_099 ($\textcolor{green}{-2.71\%}$)
triemap	195_458 ($\textcolor{green}{-0.03\%}$)	13_661_315_924 ($\textcolor{green}{-1.40\%}$)	68_228_576 ($\textcolor{green}{-8.07\%}$)	252_649 ($\textcolor{green}{-0.76\%}$)	657_794 ($\textcolor{green}{-0.56\%}$)	648_084 ($\textcolor{green}{-0.42\%}$)	15_499_470_884 ($\textcolor{green}{-2.01\%}$)
rbtree	185_901 ($\textcolor{green}{-0.35\%}$)	7_009_043_570 ($\textcolor{green}{-1.66\%}$)	52_000_464 ($\textcolor{green}{-10.34\%}$)	116_348 ($\textcolor{red}{1.79\%}$)	318_320 ($\textcolor{green}{-0.01\%}$)	330_226 ($\textcolor{red}{0.59\%}$)	6_870_900_152 ($\textcolor{green}{-4.16\%}$)
splay	190_447 ($\textcolor{green}{-0.04\%}$)	13_157_617_583 ($\textcolor{green}{-0.68\%}$)	48_000_400 ($\textcolor{green}{-11.10\%}$)	631_329 ($\textcolor{red}{0.42\%}$)	662_998 ($\textcolor{red}{0.21\%}$)	928_144 ($\textcolor{red}{0.67\%}$)	4_308_925_798 ($\textcolor{green}{-5.67\%}$)
btree	230_305 ($\textcolor{red}{0.19\%}$)	10_223_929_607 ($\textcolor{green}{-0.41\%}$)	25_108_416 ($\textcolor{green}{-19.28\%}$)	357_912 ($\textcolor{red}{1.21\%}$)	485_794 ($\textcolor{red}{0.76\%}$)	539_490 ($\textcolor{red}{1.04\%}$)	2_861_974_825 ($\textcolor{green}{-8.68\%}$)
zhenya_hashmap	188_888 ($\textcolor{green}{-0.21\%}$)	2_360_638_679 ($\textcolor{green}{-8.17\%}$)	16_777_504 ($\textcolor{green}{-26.33\%}$)	58_204 ($\textcolor{green}{-3.31\%}$)	66_552 ($\textcolor{green}{-5.11\%}$)	79_675 ($\textcolor{green}{-3.37\%}$)	3_018_208_083 ($\textcolor{green}{-8.69\%}$)
btreemap_rs	537_393	1_793_333_047	27_590_656	75_328	125_166	86_260	2_937_041_107
imrc_hashmap_rs	542_882	2_584_501_850	244_973_568	37_762	178_926	115_385	5_796_587_958
hashmap_rs	529_458	439_248_112	73_138_176	21_501	26_711	25_024	1_298_646_667

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50	pop_min 50.1	upgrade
heap	166_898 ($\textcolor{green}{-0.36\%}$)	5_554_617_018 ($\textcolor{green}{-2.51\%}$)	24_000_360 ($\textcolor{green}{-19.99\%}$)	621_690 ($\textcolor{red}{0.06\%}$)	227_224 ($\textcolor{green}{-0.63\%}$)	592_588 ($\textcolor{red}{0.07\%}$)	3_189_831_485 ($\textcolor{green}{-3.62\%}$)
heap_rs	525_853	139_669_830	18_284_544	57_419	23_051	57_545	510_960_192

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500	upgrade
buffer	173_895 ($\textcolor{green}{-0.01\%}$)	2_601_059 ($\textcolor{red}{1.17\%}$)	65_644	95_506 ($\textcolor{red}{0.02\%}$)	803_474 ($\textcolor{red}{0.38\%}$)	173_506 ($\textcolor{red}{1.77\%}$)	3_091_310 ($\textcolor{red}{0.98\%}$)
vector	171_927 ($\textcolor{green}{-0.19\%}$)	1_952_689 ($\textcolor{red}{1.65\%}$)	24_580	126_130 ($\textcolor{red}{0.01\%}$)	186_485 ($\textcolor{red}{1.69\%}$)	176_123 ($\textcolor{red}{0.08\%}$)	4_675_192 ($\textcolor{green}{-0.43\%}$)
vec_rs	520_881	289_040	1_376_256	17_251	30_571	23_331	3_161_017

Stable structures

Note Same as main branch, skipping.

Statistics

• binary_size: -0.09% [-0.22%, 0.03%]
• max_mem: -14.97% [-20.03%, -9.90%]
• cycles: -1.24% [-1.91%, -0.58%]

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	193_678 ($\textcolor{green}{-1.43\%}$)	267_743_355 ($\textcolor{green}{-1.94\%}$)	247_834_501 ($\textcolor{green}{-4.62\%}$)	33_636 ($\textcolor{green}{-2.14\%}$)	24_532 ($\textcolor{green}{-1.47\%}$)
Rust	537_397	82_787_911	56_792_991	47_914	50_388

Certified map

	binary_size	generate 10k	max mem	inc	witness	upgrade
Motoko	243_625 ($\textcolor{green}{-0.64\%}$)	4_666_119_661 ($\textcolor{green}{-2.04\%}$)	3_430_044	553_629 ($\textcolor{green}{-2.03\%}$)	407_936 ($\textcolor{red}{1.46\%}$)	274_434_719 ($\textcolor{red}{0.06\%}$)
Rust	565_792	6_409_147_805	2_228_224	1_019_959	303_897	6_019_483_730

Statistics

• binary_size: -1.03% [-3.55%, 1.48%]
• max_mem: no change
• cycles: -1.59% [-2.78%, -0.40%]

Basic DAO

	binary_size	init	transfer_token	submit_proposal	vote_proposal	upgrade
Motoko	273_924 ($\textcolor{green}{-0.56\%}$)	510_825 ($\textcolor{red}{0.03\%}$)	22_316 ($\textcolor{red}{0.24\%}$)	18_599 ($\textcolor{green}{-0.07\%}$)	19_665 ($\textcolor{red}{0.15\%}$)	157_946 ($\textcolor{red}{0.22\%}$)
Rust	849_921	599_916 ($\textcolor{green}{-0.01\%}$)	99_156	123_702	136_655	1_799_828 ($\textcolor{green}{-0.00\%}$)

DIP721 NFT

	binary_size	init	mint_token	transfer_token	upgrade
Motoko	220_392 ($\textcolor{green}{-1.00\%}$)	481_158	30_204 ($\textcolor{red}{1.32\%}$)	8_764 ($\textcolor{green}{-0.14\%}$)	89_833 ($\textcolor{red}{0.42\%}$)
Rust	869_104	236_542	368_044	91_941	1_999_207 ($\textcolor{green}{-0.00\%}$)

Statistics

• binary_size: -0.78% [-2.16%, 0.61%]
• max_mem: no change
• cycles: 0.20% [-0.03%, 0.42%]

Heartbeat

	binary_size	heartbeat
Motoko	137_178 ($\textcolor{green}{-0.52\%}$)	19_507 ($\textcolor{green}{-0.02\%}$)
Rust	23_637	1_112

Timer

	binary_size	setTimer	cancelTimer
Motoko	145_613 ($\textcolor{green}{-0.46\%}$)	51_778 ($\textcolor{red}{0.24\%}$)	4_626 ($\textcolor{red}{0.35\%}$)
Rust	487_585	68_173	11_184

Statistics

• binary_size: -0.46%
• max_mem: no change
• cycles: 0.29% [-0.05%, 0.64%]

Garbage Collection

	generate 700k	max mem	batch_get 50	batch_put 50	batch_remove 50
default	1_068_192_695 ($\textcolor{green}{-8.81\%}$)	47_793_792 ($\textcolor{green}{-8.07\%}$)	119	119	119
copying	1_068_192_577 ($\textcolor{green}{-8.81\%}$)	47_793_792 ($\textcolor{green}{-8.07\%}$)	1_067_924_316 ($\textcolor{green}{-8.81\%}$)	1_068_004_203 ($\textcolor{green}{-8.81\%}$)	1_067_925_853 ($\textcolor{green}{-8.81\%}$)
compacting	1_545_586_176 ($\textcolor{green}{-7.57\%}$)	47_793_792 ($\textcolor{green}{-8.07\%}$)	1_192_139_528 ($\textcolor{green}{-7.59\%}$)	1_415_425_189 ($\textcolor{green}{-7.69\%}$)	1_439_317_325 ($\textcolor{green}{-8.00\%}$)
generational	2_304_140_531 ($\textcolor{green}{-8.89\%}$)	47_802_256 ($\textcolor{green}{-8.07\%}$)	882_208_645 ($\textcolor{green}{-11.74\%}$)	1_211_144 ($\textcolor{green}{-1.75\%}$)	1_103_549 ($\textcolor{green}{-0.03\%}$)
incremental	29_503_170	976_097_188 ($\textcolor{green}{-0.99\%}$)	471_911_803 ($\textcolor{green}{-1.27\%}$)	497_465_467 ($\textcolor{red}{0.78\%}$)	1_221_308_722 ($\textcolor{red}{11.81\%}$)

Actor class

	binary size	put new bucket	put existing bucket	get
Map	299_202 ($\textcolor{green}{-0.33\%}$)	813_521 ($\textcolor{green}{-0.31\%}$)	16_115 ($\textcolor{red}{0.10\%}$)	16_660 ($\textcolor{red}{0.10\%}$)

Statistics

• binary_size: no change
• max_mem: -6.66% [-9.68%, -3.64%]
• cycles: -4.32% [-6.52%, -2.12%]

Publisher & Subscriber

	pub_binary_size	sub_binary_size	subscribe_caller	subscribe_callee	publish_caller	publish_callee
Motoko	161_274 ($\textcolor{green}{-0.41\%}$)	145_736 ($\textcolor{green}{-0.33\%}$)	28_593	11_963	22_854 ($\textcolor{green}{-0.04\%}$)	6_446 ($\textcolor{red}{0.25\%}$)
Rust	519_866	570_028	68_903	42_634	92_131	51_818

Statistics

• binary_size: -0.37% [-0.60%, -0.14%]

• max_mem: no change

• cycles: 0.10% [-0.82%, 1.03%]

  Overall Statistics

• binary_size: -0.35% [-0.54%, -0.17%]

• max_mem: -11.50% [-15.09%, -7.92%]

• cycles: -1.73% [-2.37%, -1.08%]

[github-actions[bot]]

Note The flamegraph link only works after you merge. Unchanged benchmarks are omitted.

Collection libraries

Measure different collection libraries written in both Motoko and Rust. The library names with _rs suffix are written in Rust; the rest are written in Motoko. The _stable and _stable_rs suffix represents that the library directly writes the state to stable memory using Region in Motoko and ic-stable-stuctures in Rust.

We use the same random number generator with fixed seed to ensure that all collections contain the same elements, and the queries are exactly the same. Below we explain the measurements of each column in the table:

• generate 1m. Insert 1m Nat64 integers into the collection. For Motoko collections, it usually triggers the GC; the rest of the column are not likely to trigger GC.
• max mem. For Motoko, it reports rts_max_heap_size after generate call; For Rust, it reports the Wasm's memory page * 32Kb.
• batch_get 50. Find 50 elements from the collection.
• batch_put 50. Insert 50 elements to the collection.
• batch_remove 50. Remove 50 elements from the collection.
• upgrade. Upgrade the canister with the same Wasm module. For non-stable benchmarks, the map state is persisted by serializing and deserializing states into stable memory. For stable benchmarks, the upgrade takes no cycles, as the state is already in the stable memory.

💎 Takeaways

• The platform only charges for instruction count. Data structures which make use of caching and locality have no impact on the cost.
• We have a limit on the maximal cycles per round. This means asymptotic behavior doesn't matter much. We care more about the performance up to a fixed N. In the extreme cases, you may see an $O(10000 n\log n)$ algorithm hitting the limit, while an $O(n^2)$ algorithm runs just fine.
• Amortized algorithms/GC may need to be more eager to avoid hitting the cycle limit on a particular round.
• Rust costs more cycles to process complicated Candid data, but it is more efficient in performing core computations.

Note

• The Candid interface of the benchmark is minimal, therefore the serialization cost is negligible in this measurement.
• Due to the instrumentation overhead and cycle limit, we cannot profile computations with very large collections.
• The upgrade column uses Candid for serializing stable data. In Rust, you may get better cycle cost by using a different serialization format. Another slowdown in Rust is that ic-stable-structures tends to be slower than the region memory in Motoko.
• Different library has different ways for persisting data during upgrades, there are mainly three categories:
  • Use stable variable directly in Motoko: zhenya_hashmap, btree, vector
  • Expose and serialize external state (share/unshare in Motoko, candid::Encode in Rust): rbtree, heap, btreemap_rs, hashmap_rs, heap_rs, vector_rs
  • Use pre/post-upgrade hooks to convert data into an array: hashmap, splay, triemap, buffer, imrc_hashmap_rs
• The stable benchmarks are much more expensive than their non-stable counterpart, because the stable memory API is much more expensive. The benefit is that they get fast upgrade. The upgrade still needs to parse the metadata when initializing the upgraded Wasm module.
• hashmap uses amortized data structure. When the initial capacity is reached, it has to copy the whole array, thus the cost of batch_put 50 is much higher than other data structures.
• btree comes from mops.one/stableheapbtreemap.
• zhenya_hashmap comes from mops.one/map.
• vector comes from mops.one/vector. Compare with buffer, put has better worst case time and space complexity ($O(\sqrt{n})$ vs $O(n)$); get has a slightly larger constant overhead.
• hashmap_rs uses the fxhash crate, which is the same as std::collections::HashMap, but with a deterministic hasher. This ensures reproducible result.
• imrc_hashmap_rs uses the im-rc crate, which is the immutable version hashmap in Rust.

Map

	binary_size	generate 1m	max mem	batch_get 50	batch_put 50	batch_remove 50	upgrade
hashmap	189_916	8_184_618_025	56_000_256	342_784	6_462_528_122	368_420	10_728_193_099
triemap	195_458	13_661_315_924	68_228_576	252_649	657_794	648_084	15_499_470_884
rbtree	185_901	7_009_043_570	52_000_464	116_348	318_320	330_226	6_870_900_152
splay	190_447	13_157_617_583	48_000_400	631_329	662_998	928_144	4_308_925_798
btree	230_305	10_223_929_607	25_108_416	357_912	485_794	539_490	2_861_974_825
zhenya_hashmap	188_888	2_360_638_679	16_777_504	58_204	66_552	79_675	3_018_208_083
btreemap_rs	537_393	1_793_333_047	27_590_656	75_328	125_166	86_260	2_937_041_107
imrc_hashmap_rs	542_882	2_584_501_850	244_973_568	37_762	178_926	115_385	5_796_587_958
hashmap_rs	529_458	439_248_112	73_138_176	21_501	26_711	25_024	1_298_646_667

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50	pop_min 50	upgrade
heap	166_898	5_554_617_018	24_000_360	621_690	227_224	592_588	3_189_831_485
heap_rs	525_853	139_669_830	18_284_544	57_419	23_051	57_545	510_960_192

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500	upgrade
buffer	173_895	2_601_059	65_644	95_506	803_474	173_506	3_091_310
vector	171_927	1_952_689	24_580	126_130	186_485	176_123	4_675_192
vec_rs	520_881	289_040	1_376_256	17_251	30_571	23_331	3_161_017

Stable structures

	binary_size	generate 50k	max mem	batch_get 50	batch_put 50	batch_remove 50	upgrade
btreemap_rs	537_393	76_200_333	2_555_904	64_886	97_044	85_272	126_265_270
btreemap_stable_rs	543_609	4_561_985_735	2_031_616	2_707_064	5_026_642	8_594_683	729_311
heap_rs	525_853	7_051_730	2_293_760	49_928	23_299	49_894	26_768_703
heap_stable_rs	506_559	271_553_517	458_752	2_294_851	238_596	2_277_771	729_317
vec_rs	520_881	3_079_382	2_293_760	17_251	18_421	17_719	24_671_551
vec_stable_rs	503_829	63_394_912	458_752	62_491	79_685	81_633	729_320

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0

  Cryptographic libraries

Measure different cryptographic libraries written in both Motoko and Rust.

• SHA-2 benchmarks
  • SHA-256/SHA-512. Compute the hash of a 1M Wasm binary.
  • account_id. Compute the ledger account id from principal, based on SHA-224.
  • neuron_id. Compute the NNS neuron id from principal, based on SHA-256.
• Certified map. Merkle Tree for storing key-value pairs and generate witness according to the IC Interface Specification.
  • generate 10k. Insert 10k 7-character word as both key and value into the certified map.
  • max mem. For Motoko, it reports rts_max_heap_size after generate call; For Rust, it reports the Wasm's memory page * 32Kb.
  • inc. Increment a counter and insert the counter value into the map.
  • witness. Generate the root hash and a witness for the counter.
  • upgrade. Upgrade the canister with the same Wasm. In Motoko, we use stable variable. In Rust, we convert the tree to a vector before serialization.

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	193_678	267_743_355	247_834_501	33_636	24_532
Rust	537_397	82_787_911	56_792_991	47_914	50_388

Certified map

	binary_size	generate 10k	max mem	inc	witness	upgrade
Motoko	243_625	4_666_119_661	3_430_044	553_629	407_936	274_434_719
Rust	565_792	6_409_147_805	2_228_224	1_019_959	303_897	6_019_483_730

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0

  Sample Dapps

Measure the performance of some typical dapps:

• Basic DAO, with heartbeat disabled to make profiling easier. We have a separate benchmark to measure heartbeat performance.
• DIP721 NFT

Note

• The cost difference is mainly due to the Candid serialization cost.
• Motoko statically compiles/specializes the serialization code for each method, whereas in Rust, we use serde to dynamically deserialize data based on data on the wire.
• We could improve the performance on the Rust side by using parser combinators. But it is a challenge to maintain the ergonomics provided by serde.
• For real-world applications, we tend to send small data for each endpoint, which makes the Candid overhead in Rust tolerable.

Basic DAO

	binary_size	init	transfer_token	submit_proposal	vote_proposal	upgrade
Motoko	273_924	510_825	22_316	18_599	19_665	157_946
Rust	849_921	599_916	99_156	123_702	136_655	1_799_828

DIP721 NFT

	binary_size	init	mint_token	transfer_token	upgrade
Motoko	220_392	481_158	30_204	8_764	89_833
Rust	869_104	236_542	368_044	91_941	1_999_207

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0

  Heartbeat / Timer

Measure the cost of empty heartbeat and timer job.

• setTimer measures both the setTimer(0) method and the execution of empty job.
• It is not easy to reliably capture the above events in one flamegraph, as the implementation detail of the replica can affect how we measure this. Typically, a correct flamegraph contains both setTimer and canister_global_timer function. If it's not there, we may need to adjust the script.

Heartbeat

	binary_size	heartbeat
Motoko	137_178	19_507
Rust	23_637	1_112

Timer

	binary_size	setTimer	cancelTimer
Motoko	145_613	51_778	4_626
Rust	487_585	68_173	11_184

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0

  Motoko Specific Benchmarks

Measure various features only available in Motoko.

• Garbage Collection. Measure Motoko garbage collection cost using the Triemap benchmark. The max mem column reports rts_max_heap_size after generate call. The cycle cost numbers reported here are garbage collection cost only. Some flamegraphs are truncated due to the 2M log size limit. The dfx/ic-wasm optimizer is disabled for the garbage collection test cases due to how the optimizer affects function names, making profiling trickier.

  • default. Compile with the default GC option. With the current GC scheduler, generate will trigger the copying GC. The rest of the methods will not trigger GC.
  • copying. Compile with --force-gc --copying-gc.
  • compacting. Compile with --force-gc --compacting-gc.
  • generational. Compile with --force-gc --generational-gc.
  • incremental. Compile with --force-gc --incremental-gc.

• Actor class. Measure the cost of spawning actor class, using the Actor classes example.

Garbage Collection

	generate 700k	max mem	batch_get 50	batch_put 50	batch_remove 50
default	1_068_192_695	47_793_792	119	119	119
copying	1_068_192_577	47_793_792	1_067_924_316	1_068_004_203	1_067_925_853
compacting	1_545_586_176	47_793_792	1_192_139_528	1_415_425_189	1_439_317_325
generational	2_304_140_531	47_802_256	882_208_645	1_211_144	1_103_549
incremental	29_503_170	976_097_188	471_911_803	497_465_467	1_221_308_722

Actor class

	binary size	put new bucket	put existing bucket	get
Map	299_202	813_521	16_115	16_660

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0

  Publisher & Subscriber

Measure the cost of inter-canister calls from the Publisher & Subscriber example.

	pub_binary_size	sub_binary_size	subscribe_caller	subscribe_callee	publish_caller	publish_callee
Motoko	161_274	145_736	28_593	11_963	22_854	6_446
Rust	519_866	570_028	68_903	42_634	92_131	51_818

Environment

• dfx 0.16.1
• Motoko compiler 0.10.4 (source kfk8w3jy-5b63clja-3y6iac6j-vqw16jlw)
• rustc 1.75.0 (82e1608df 2023-12-21)
• ic-repl 0.6.2
• ic-wasm 0.7.0