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
hashmap	133_772 ($\textcolor{green}{-0.04\%}$)	9_120_083_788 ($\textcolor{red}{31.03\%}$)	61_987_732	371_774 ($\textcolor{red}{29.33\%}$)	7_046_555_134 ($\textcolor{red}{27.75\%}$)	402_302 ($\textcolor{red}{30.21\%}$)
triemap	135_185 ($\textcolor{green}{-0.10\%}$)	17_215_517_319 ($\textcolor{red}{50.60\%}$)	74_216_052	345_840 ($\textcolor{red}{55.25\%}$)	836_239 ($\textcolor{red}{52.70\%}$)	819_816 ($\textcolor{red}{52.10\%}$)
rbtree	135_861 ($\textcolor{green}{-0.19\%}$)	8_149_335_945 ($\textcolor{red}{36.29\%}$)	57_995_940	116_820 ($\textcolor{red}{31.41\%}$)	367_123 ($\textcolor{red}{36.70\%}$)	396_460 ($\textcolor{red}{42.44\%}$)
splay	131_698 ($\textcolor{green}{-0.13\%}$)	17_095_892_890 ($\textcolor{red}{47.78\%}$)	53_995_876	821_820 ($\textcolor{red}{48.90\%}$)	865_787 ($\textcolor{red}{48.85\%}$)	1_195_419 ($\textcolor{red}{47.54\%}$)
btree	176_246 ($\textcolor{green}{-0.12\%}$)	12_563_923_452 ($\textcolor{red}{52.77\%}$)	31_103_892	424_386 ($\textcolor{red}{52.91\%}$)	592_838 ($\textcolor{red}{54.32\%}$)	659_333 ($\textcolor{red}{53.68\%}$)
zhenya_hashmap	141_702 ($\textcolor{green}{-0.00\%}$)	3_623_441_658 ($\textcolor{red}{37.61\%}$)	65_987_480	92_003 ($\textcolor{red}{40.81\%}$)	114_956 ($\textcolor{red}{43.42\%}$)	137_180 ($\textcolor{red}{44.77\%}$)
btreemap_rs	419_620 ($\textcolor{red}{1.49\%}$)	1_818_061_882 ($\textcolor{red}{10.20\%}$)	13_762_560	76_267 ($\textcolor{red}{14.15\%}$)	127_354 ($\textcolor{red}{13.44\%}$)	92_934 ($\textcolor{red}{14.36\%}$)
imrc_hashmap_rs	419_303 ($\textcolor{red}{1.38\%}$)	2_582_553_741 ($\textcolor{red}{8.25\%}$)	122_454_016	39_485 ($\textcolor{red}{20.21\%}$)	179_841 ($\textcolor{red}{10.53\%}$)	122_297 ($\textcolor{red}{24.17\%}$)
hashmap_rs	413_086 ($\textcolor{red}{1.72\%}$)	437_767_476 ($\textcolor{red}{11.51\%}$)	36_536_320	22_388 ($\textcolor{red}{35.70\%}$)	27_813 ($\textcolor{red}{33.31\%}$)	25_646 ($\textcolor{red}{28.40\%}$)

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	127_702 ($\textcolor{green}{-0.04\%}$)	7_251_557_847 ($\textcolor{red}{54.80\%}$)	29_995_836	797_941 ($\textcolor{red}{56.00\%}$)	294_639 ($\textcolor{red}{58.02\%}$)
heap_rs	410_722 ($\textcolor{red}{1.68\%}$)	140_171_516 ($\textcolor{red}{13.87\%}$)	9_109_504	60_210 ($\textcolor{red}{12.92\%}$)	23_974 ($\textcolor{red}{32.18\%}$)

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	135_417 ($\textcolor{green}{-0.03\%}$)	3_217_931 ($\textcolor{red}{54.51\%}$)	65_508	126_168 ($\textcolor{red}{72.63\%}$)	1_012_069 ($\textcolor{red}{50.71\%}$)	199_668 ($\textcolor{red}{56.50\%}$)
vector	133_898 ($\textcolor{green}{-0.00\%}$)	2_920_678 ($\textcolor{red}{68.97\%}$)	24_764	198_390 ($\textcolor{red}{63.67\%}$)	278_928 ($\textcolor{red}{70.14\%}$)	264_366 ($\textcolor{red}{63.59\%}$)
vec_rs	409_517 ($\textcolor{red}{1.70\%}$)	290_479 ($\textcolor{red}{9.24\%}$)	655_360	17_874 ($\textcolor{red}{39.38\%}$)	31_274 ($\textcolor{red}{23.84\%}$)	26_410 ($\textcolor{red}{25.67\%}$)

Statistics

binary_size: 0.52% [0.13%, 0.92%]
max_mem: no change
cycles: 38.89% [34.83%, 42.95%]

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	170_122 ($\textcolor{red}{0.01\%}$)	341_183_942 ($\textcolor{red}{29.16\%}$)	309_559_018 ($\textcolor{red}{31.67\%}$)	47_134 ($\textcolor{red}{34.12\%}$)	32_225 ($\textcolor{red}{38.60\%}$)
Rust	497_788 ($\textcolor{red}{1.41\%}$)	82_789_862 ($\textcolor{red}{0.34\%}$)	56_794_773 ($\textcolor{red}{0.48\%}$)	51_161 ($\textcolor{red}{20.67\%}$)	54_143 ($\textcolor{red}{30.16\%}$)

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	162_071 ($\textcolor{green}{-0.21\%}$)	27_613_961_279 ($\textcolor{red}{48.62\%}$)	3_429_924	3_286_019 ($\textcolor{red}{48.74\%}$)	491_221 ($\textcolor{red}{49.87\%}$)
Rust	441_351 ($\textcolor{red}{1.73\%}$)	6_341_165_388 ($\textcolor{red}{2.16\%}$)	1_081_344	1_009_424 ($\textcolor{red}{2.50\%}$)	304_952 ($\textcolor{red}{5.58\%}$)

Statistics

binary_size: 0.73% [-0.42%, 1.88%]
max_mem: no change
cycles: 24.48% [15.45%, 33.50%]

Basic DAO

	binary_size	init	transfer_token	submit_proposal	vote_proposal
Motoko	225_908 ($\textcolor{red}{0.05\%}$)	50_463 ($\textcolor{red}{34.59\%}$)	24_244 ($\textcolor{red}{48.59\%}$)	20_204 ($\textcolor{red}{59.66\%}$)	21_570 ($\textcolor{red}{52.94\%}$)
Rust	717_720 ($\textcolor{red}{1.82\%}$)	555_159 ($\textcolor{red}{17.64\%}$)	106_187 ($\textcolor{red}{22.72\%}$)	131_091 ($\textcolor{red}{25.31\%}$)	141_461 ($\textcolor{red}{22.20\%}$)

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	183_999 ($\textcolor{red}{0.06\%}$)	18_403 ($\textcolor{red}{51.08\%}$)	31_180 ($\textcolor{red}{39.70\%}$)	9_225 ($\textcolor{red}{95.86\%}$)
Rust	777_672 ($\textcolor{red}{1.43\%}$)	146_963 ($\textcolor{red}{17.54\%}$)	384_008 ($\textcolor{red}{18.34\%}$)	95_147 ($\textcolor{red}{23.38\%}$)

Statistics

binary_size: 0.84% [-0.24%, 1.92%]
max_mem: no change
cycles: 37.83% [27.31%, 48.34%]

Heartbeat

	binary_size	heartbeat
Motoko	118_910 ($\textcolor{red}{0.00\%}$)	23_725 ($\textcolor{red}{220.96\%}$)
Rust	23_573 ($\textcolor{green}{-0.53\%}$)	1_195 ($\textcolor{red}{51.07\%}$)

Timer

	binary_size	setTimer	cancelTimer
Motoko	125_155 ($\textcolor{green}{-0.01\%}$)	53_493 ($\textcolor{red}{251.74\%}$)	4_819 ($\textcolor{red}{187.02\%}$)
Rust	442_874 ($\textcolor{red}{1.85\%}$)	70_359 ($\textcolor{red}{61.60\%}$)	11_528 ($\textcolor{red}{50.05\%}$)

Statistics

binary_size: 0.92% [-4.94%, 6.78%]
max_mem: no change
cycles: 137.60% [22.08%, 253.12%]

Garbage Collection

	generate 800k	max mem	batch_get 50	batch_put 50	batch_remove 50
default	1_491_263_035 ($\textcolor{red}{47.32\%}$)	59_396_776	123 ($\textcolor{red}{146.00\%}$)	123 ($\textcolor{red}{146.00\%}$)	123 ($\textcolor{red}{146.00\%}$)
copying	1_491_262_912 ($\textcolor{red}{47.32\%}$)	59_396_776	1_490_940_704 ($\textcolor{red}{47.29\%}$)	1_491_039_746 ($\textcolor{red}{47.29\%}$)	1_490_946_844 ($\textcolor{red}{47.29\%}$)
compacting	2_007_811_833 ($\textcolor{red}{19.87\%}$)	59_396_776	1_551_336_811 ($\textcolor{red}{19.98\%}$)	1_851_015_361 ($\textcolor{red}{20.80\%}$)	1_883_761_399 ($\textcolor{red}{20.87\%}$)
generational	3_073_273_610 ($\textcolor{red}{22.10\%}$)	59_405_240	1_200_499_985 ($\textcolor{red}{22.80\%}$)	1_296_067 ($\textcolor{red}{23.11\%}$)	1_189_105 ($\textcolor{red}{22.92\%}$)
incremental	33_437_135 ($\textcolor{red}{3.45\%}$)	1_136_153_832	355_581_986 ($\textcolor{red}{22.51\%}$)	358_878_332 ($\textcolor{red}{22.50\%}$)	358_911_360 ($\textcolor{red}{22.50\%}$)

Actor class

	binary size	put new bucket	put existing bucket	get
Map	254_149 ($\textcolor{red}{0.03\%}$)	702_838 ($\textcolor{red}{10.06\%}$)	16_754 ($\textcolor{red}{276.58\%}$)	17_269 ($\textcolor{red}{251.78\%}$)

Statistics

binary_size: no change
max_mem: no change
cycles: 60.68% [34.18%, 87.19%]

Publisher & Subscriber

	pub_binary_size	sub_binary_size	subscribe_caller	subscribe_callee	publish_caller	publish_callee
Motoko	139_908 ($\textcolor{red}{0.02\%}$)	126_847 ($\textcolor{red}{0.02\%}$)	29_430 ($\textcolor{red}{101.01\%}$)	12_477 ($\textcolor{red}{47.64\%}$)	23_710 ($\textcolor{red}{125.17\%}$)	6_660 ($\textcolor{red}{81.87\%}$)
Rust	479_139 ($\textcolor{red}{1.48\%}$)	529_168 ($\textcolor{red}{1.78\%}$)	72_411 ($\textcolor{red}{40.36\%}$)	44_728 ($\textcolor{red}{29.04\%}$)	96_875 ($\textcolor{red}{30.61\%}$)	54_605 ($\textcolor{red}{31.21\%}$)

Statistics

binary_size: 0.82% [-0.28%, 1.93%]
max_mem: no change
cycles: 60.86% [36.07%, 85.66%]

Overall Statistics
binary_size: 0.67% [0.40%, 0.94%]
max_mem: no change
cycles: 46.32% [39.28%, 53.36%]

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.

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.

💎 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 large collections. Hopefully, when deterministic time slicing is ready, we can measure the performance on larger memory footprint.

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
hashmap	133_772	9_120_083_788	61_987_732	371_774	7_046_555_134	402_302
triemap	135_185	17_215_517_319	74_216_052	345_840	836_239	819_816
rbtree	135_861	8_149_335_945	57_995_940	116_820	367_123	396_460
splay	131_698	17_095_892_890	53_995_876	821_820	865_787	1_195_419
btree	176_246	12_563_923_452	31_103_892	424_386	592_838	659_333
zhenya_hashmap	141_702	3_623_441_658	65_987_480	92_003	114_956	137_180
btreemap_rs	419_620	1_818_061_882	13_762_560	76_267	127_354	92_934
imrc_hashmap_rs	419_303	2_582_553_741	122_454_016	39_485	179_841	122_297
hashmap_rs	413_086	437_767_476	36_536_320	22_388	27_813	25_646

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	127_702	7_251_557_847	29_995_836	797_941	294_639	760_200
heap_rs	410_722	140_171_516	9_109_504	60_210	23_974	60_423

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	135_417	3_217_931	65_508	126_168	1_012_069	199_668
vector	133_898	2_920_678	24_764	198_390	278_928	264_366
vec_rs	409_517	290_479	655_360	17_874	31_274	26_410

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.

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	170_122	341_183_942	309_559_018	47_134	32_225
Rust	497_788	82_789_862	56_794_773	51_161	54_143

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	162_071	27_613_961_279	3_429_924	3_286_019	491_221
Rust	441_351	6_341_165_388	1_081_344	1_009_424	304_952

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
Motoko	225_908	50_463	24_244	20_204	21_570
Rust	717_720	555_159	106_187	131_091	141_461

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	183_999	18_403	31_180	9_225
Rust	777_672	146_963	384_008	95_147

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	118_910	23_725
Rust	23_573	1_195

Timer

	binary_size	setTimer	cancelTimer
Motoko	125_155	53_493	4_819
Rust	442_874	70_359	11_528

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 800k	max mem	batch_get 50	batch_put 50	batch_remove 50
default	1_491_263_035	59_396_776	123	123	123
copying	1_491_262_912	59_396_776	1_490_940_704	1_491_039_746	1_490_946_844
compacting	2_007_811_833	59_396_776	1_551_336_811	1_851_015_361	1_883_761_399
generational	3_073_273_610	59_405_240	1_200_499_985	1_296_067	1_189_105
incremental	33_437_135	1_136_153_832	355_581_986	358_878_332	358_911_360

Actor class

	binary size	put new bucket	put existing bucket	get
Map	254_149	702_838	16_754	17_269

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	139_908	126_847	29_430	12_477	23_710	6_660
Rust	479_139	529_168	72_411	44_728	96_875	54_605

dfinity / canister-profiling

Compare cost for the new metering #74

Map

Priority queue

Growable array

Statistics

SHA-2

Certified map

Statistics

Basic DAO

DIP721 NFT

Statistics

Heartbeat

Timer

Statistics

Garbage Collection

Actor class

Statistics

Publisher & Subscriber

Statistics

Overall Statistics

Collection libraries

💎 Takeaways

Map

Priority queue

Growable array

Cryptographic libraries

SHA-2

Certified map

Sample Dapps

Basic DAO

DIP721 NFT

Heartbeat / Timer

Heartbeat

Timer

Motoko Specific Benchmarks

Garbage Collection

Actor class

Publisher & Subscriber