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	159_194 ($\textcolor{red}{15.13\%}$)	7_588_838_102 ($\textcolor{red}{8.82\%}$)	61_987_732	318_951 ($\textcolor{red}{10.67\%}$)	6_008_157_311 ($\textcolor{red}{8.69\%}$)	341_403 ($\textcolor{red}{10.23\%}$)
triemap	162_066 ($\textcolor{red}{15.96\%}$)	12_638_993_478 ($\textcolor{red}{10.56\%}$)	74_216_052	250_401 ($\textcolor{red}{12.38\%}$)	611_116 ($\textcolor{red}{11.58\%}$)	600_960 ($\textcolor{red}{11.48\%}$)
rbtree	162_408 ($\textcolor{red}{15.54\%}$)	6_161_146_247 ($\textcolor{red}{3.04\%}$)	57_995_940	116_202 ($\textcolor{red}{30.70\%}$)	277_258 ($\textcolor{red}{3.23\%}$)	302_170 ($\textcolor{red}{8.56\%}$)
splay	157_707 ($\textcolor{red}{15.67\%}$)	12_285_354_629 ($\textcolor{red}{6.20\%}$)	53_995_876	589_148 ($\textcolor{red}{6.74\%}$)	619_660 ($\textcolor{red}{6.53\%}$)	870_643 ($\textcolor{red}{7.46\%}$)
btree	214_489 ($\textcolor{red}{18.21\%}$)	9_139_742_724 ($\textcolor{red}{11.13\%}$)	31_103_892	314_208 ($\textcolor{red}{13.21\%}$)	429_672 ($\textcolor{red}{11.84\%}$)	480_823 ($\textcolor{red}{12.07\%}$)
zhenya_hashmap	168_445 ($\textcolor{red}{14.99\%}$)	2_822_042_495 ($\textcolor{red}{7.13\%}$)	65_987_480	73_056 ($\textcolor{red}{11.71\%}$)	89_710 ($\textcolor{red}{11.85\%}$)	108_471 ($\textcolor{red}{14.39\%}$)
btreemap_rs	446_200 ($\textcolor{red}{6.22\%}$)	1_637_381_286 ($\textcolor{green}{-1.01\%}$)	13_762_560	65_726 ($\textcolor{green}{-1.74\%}$)	112_100 ($\textcolor{green}{-0.41\%}$)	81_944 ($\textcolor{red}{0.78\%}$)
imrc_hashmap_rs	446_100 ($\textcolor{red}{6.28\%}$)	2_388_802_442 ($\textcolor{red}{0.10\%}$)	122_454_016	33_200 ($\textcolor{red}{0.90\%}$)	163_343 ($\textcolor{red}{0.32\%}$)	97_441 ($\textcolor{green}{-1.10\%}$)
hashmap_rs	439_281 ($\textcolor{red}{6.23\%}$)	414_097_976 ($\textcolor{red}{2.93\%}$)	36_536_320	17_125 ($\textcolor{red}{2.56\%}$)	22_229 ($\textcolor{red}{2.91\%}$)	20_323 ($\textcolor{red}{1.35\%}$)

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	152_465 ($\textcolor{red}{15.31\%}$)	5_200_975_509 ($\textcolor{red}{11.02\%}$)	29_995_836	574_107 ($\textcolor{red}{12.24\%}$)	208_978 ($\textcolor{red}{12.07\%}$)
heap_rs	437_212 ($\textcolor{red}{6.33\%}$)	125_346_655 ($\textcolor{red}{1.82\%}$)	9_109_504	53_855 ($\textcolor{red}{0.89\%}$)	18_525 ($\textcolor{red}{1.77\%}$)

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	161_556 ($\textcolor{red}{15.47\%}$)	2_296_568 ($\textcolor{red}{10.27\%}$)	65_508	84_247 ($\textcolor{red}{15.26\%}$)	740_528 ($\textcolor{red}{10.28\%}$)	142_247 ($\textcolor{red}{11.49\%}$)
vector	160_836 ($\textcolor{red}{16.26\%}$)	1_890_122 ($\textcolor{red}{9.35\%}$)	24_764	137_383 ($\textcolor{red}{13.33\%}$)	179_360 ($\textcolor{red}{9.40\%}$)	181_415 ($\textcolor{red}{12.26\%}$)
vec_rs	435_769 ($\textcolor{red}{6.30\%}$)	266_276 ($\textcolor{red}{0.16\%}$)	655_360	13_221 ($\textcolor{red}{2.47\%}$)	25_648 ($\textcolor{red}{1.25\%}$)	20_782 ($\textcolor{green}{-2.04\%}$)

Statistics

binary_size: 12.42% [10.14%, 14.70%]
max_mem: no change
cycles: 7.24% [5.86%, 8.63%]

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	196_200 ($\textcolor{red}{12.37\%}$)	285_638_950 ($\textcolor{red}{8.13\%}$)	255_658_597 ($\textcolor{red}{8.74\%}$)	34_744 ($\textcolor{green}{-1.16\%}$)	23_689 ($\textcolor{red}{1.84\%}$)
Rust	528_163 ($\textcolor{red}{6.00\%}$)	82_512_868 ($\textcolor{red}{0.00\%}$)	56_526_741 ($\textcolor{red}{0.00\%}$)	43_318 ($\textcolor{red}{1.80\%}$)	45_477 ($\textcolor{red}{2.03\%}$)

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	205_806 ($\textcolor{red}{15.44\%}$)	5_050_097_115 ($\textcolor{red}{0.73\%}$)	3_429_924	598_393 ($\textcolor{red}{0.71\%}$)	361_240 ($\textcolor{red}{10.21\%}$)
Rust	469_882 ($\textcolor{red}{6.36\%}$)	6_232_657_596 ($\textcolor{red}{0.49\%}$)	1_081_344	988_880 ($\textcolor{red}{0.51\%}$)	290_927 ($\textcolor{red}{0.85\%}$)

Statistics

binary_size: 10.04% [4.59%, 15.50%]
max_mem: no change
cycles: 2.49% [0.76%, 4.22%]

Basic DAO

	binary_size	init	transfer_token	submit_proposal	vote_proposal
Motoko	277_584 ($\textcolor{red}{20.56\%}$)	39_684 ($\textcolor{red}{5.57\%}$)	17_787 ($\textcolor{red}{8.86\%}$)	13_881 ($\textcolor{red}{9.54\%}$)	15_641 ($\textcolor{red}{10.82\%}$)
Rust	762_892 ($\textcolor{red}{6.20\%}$)	484_210 ($\textcolor{red}{2.50\%}$)	89_052 ($\textcolor{red}{2.61\%}$)	107_757 ($\textcolor{red}{2.37\%}$)	118_705 ($\textcolor{red}{2.13\%}$)

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	229_941 ($\textcolor{red}{22.10\%}$)	13_217 ($\textcolor{red}{7.74\%}$)	23_778 ($\textcolor{red}{6.36\%}$)	5_267 ($\textcolor{red}{11.38\%}$)
Rust	828_098 ($\textcolor{red}{6.40\%}$)	128_336 ($\textcolor{red}{2.43\%}$)	332_435 ($\textcolor{red}{2.28\%}$)	79_221 ($\textcolor{red}{2.22\%}$)

Statistics

binary_size: 13.81% [3.58%, 24.05%]
max_mem: no change
cycles: 5.49% [3.80%, 7.17%]

Heartbeat

	binary_size	heartbeat
Motoko	142_110 ($\textcolor{red}{15.20\%}$)	7_777 ($\textcolor{red}{106.95\%}$)
Rust	25_646 ($\textcolor{red}{8.55\%}$)	789 ($\textcolor{red}{68.23\%}$)

Timer

	binary_size	setTimer	cancelTimer
Motoko	149_306 ($\textcolor{red}{15.17\%}$)	16_471 ($\textcolor{red}{8.17\%}$)	1_847 ($\textcolor{red}{9.68\%}$)
Rust	470_589 ($\textcolor{red}{6.14\%}$)	44_509 ($\textcolor{red}{2.52\%}$)	7_713 ($\textcolor{red}{2.88\%}$)

Statistics

binary_size: 10.66% [-17.86%, 39.17%]
max_mem: no change
cycles: 5.81% [1.52%, 10.11%]

Garbage Collection

Note Same as main branch, skipping.

Actor class

	binary size	put new bucket	put existing bucket	get
Map	297_686 ($\textcolor{red}{13.79\%}$)	719_907 ($\textcolor{red}{9.73\%}$)	4_854 ($\textcolor{red}{8.86\%}$)	5_288 ($\textcolor{red}{7.50\%}$)

Statistics

binary_size: no change
max_mem: no change
cycles: 9.97% [6.79%, 13.16%]

Publisher & Subscriber

	pub_binary_size	sub_binary_size	subscribe_caller	subscribe_callee	publish_caller	publish_callee
Motoko	166_785 ($\textcolor{red}{15.47\%}$)	151_855 ($\textcolor{red}{15.66\%}$)	15_317 ($\textcolor{red}{4.55\%}$)	8_758 ($\textcolor{red}{3.57\%}$)	11_042 ($\textcolor{red}{4.77\%}$)	3_872 ($\textcolor{red}{5.53\%}$)
Rust	511_787 ($\textcolor{red}{6.71\%}$)	565_325 ($\textcolor{red}{6.73\%}$)	52_965 ($\textcolor{red}{2.39\%}$)	35_499 ($\textcolor{red}{2.62\%}$)	76_563 ($\textcolor{red}{2.30\%}$)	45_397 ($\textcolor{red}{2.34\%}$)

Statistics

binary_size: 11.14% [5.14%, 17.15%]
max_mem: no change
cycles: 3.51% [2.64%, 4.37%]

Overall Statistics
binary_size: 11.97% [10.28%, 13.67%]
max_mem: no change
cycles: 6.06% [5.17%, 6.95%]

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	159_194	7_588_838_102	61_987_732	318_951	6_008_157_311	341_403
triemap	162_066	12_638_993_478	74_216_052	250_401	611_116	600_960
rbtree	162_408	6_161_146_247	57_995_940	116_202	277_258	302_170
splay	157_707	12_285_354_629	53_995_876	589_148	619_660	870_643
btree	214_489	9_139_742_724	31_103_892	314_208	429_672	480_823
zhenya_hashmap	168_445	2_822_042_495	65_987_480	73_056	89_710	108_471
btreemap_rs	446_200	1_637_381_286	13_762_560	65_726	112_100	81_944
imrc_hashmap_rs	446_100	2_388_802_442	122_454_016	33_200	163_343	97_441
hashmap_rs	439_281	414_097_976	36_536_320	17_125	22_229	20_323

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	152_465	5_200_975_509	29_995_836	574_107	208_978	546_783
heap_rs	437_212	125_346_655	9_109_504	53_855	18_525	54_078

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	161_556	2_296_568	65_508	84_247	740_528	142_247
vector	160_836	1_890_122	24_764	137_383	179_360	181_415
vec_rs	435_769	266_276	655_360	13_221	25_648	20_782

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	196_200	285_638_950	255_658_597	34_744	23_689
Rust	528_163	82_512_868	56_526_741	43_318	45_477

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	205_806	5_050_097_115	3_429_924	598_393	361_240
Rust	469_882	6_232_657_596	1_081_344	988_880	290_927

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	277_584	39_684	17_787	13_881	15_641
Rust	762_892	484_210	89_052	107_757	118_705

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	229_941	13_217	23_778	5_267
Rust	828_098	128_336	332_435	79_221

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	142_110	7_777
Rust	25_646	789

Timer

	binary_size	setTimer	cancelTimer
Motoko	149_306	16_471	1_847
Rust	470_589	44_509	7_713

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_012_258_537	59_396_776	50	50	50
copying	1_012_258_487	59_396_776	1_012_236_046	1_012_303_056	1_012_240_283
compacting	1_675_009_925	59_396_776	1_292_955_500	1_532_273_641	1_558_502_986
generational	2_498_146_508	59_405_240	977_578_983	1_044_991	960_405
incremental	32_320_754	1_136_155_048	290_257_785	292_951_006	292_977_552

Actor class

	binary size	put new bucket	put existing bucket	get
Map	297_686	719_907	4_854	5_288

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	166_785	151_855	15_317	8_758	11_042	3_872
Rust	511_787	565_325	52_965	35_499	76_563	45_397

dfinity / canister-profiling

Remove wasm-opt #79

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