aggressive wasm-opt

[chenyan-dfinity]

base: new metering, no wasm-opt

[github-actions]

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	297_254 ($\textcolor{red}{86.93\%}$)	7_842_361_244 ($\textcolor{green}{-17.42\%}$)	61_987_732	321_307 ($\textcolor{green}{-18.21\%}$)	6_155_781_309 ($\textcolor{green}{-15.69\%}$)	345_252 ($\textcolor{green}{-18.43\%}$)
triemap	333_856 ($\textcolor{red}{106.48\%}$)	11_407_728_709 ($\textcolor{green}{-34.06\%}$)	74_216_052	195_746 ($\textcolor{green}{-43.43\%}$)	533_776 ($\textcolor{green}{-36.54\%}$)	524_332 ($\textcolor{green}{-36.45\%}$)
rbtree	296_811 ($\textcolor{red}{83.20\%}$)	6_513_814_073 ($\textcolor{green}{-23.03\%}$)	57_995_940	98_058 ($\textcolor{green}{-38.18\%}$)	289_073 ($\textcolor{green}{-24.93\%}$)	312_663 ($\textcolor{green}{-26.84\%}$)
splay	286_117 ($\textcolor{red}{81.79\%}$)	11_942_316_508 ($\textcolor{green}{-31.46\%}$)	53_995_876	563_736 ($\textcolor{green}{-32.98\%}$)	593_264 ($\textcolor{green}{-32.94\%}$)	821_312 ($\textcolor{green}{-33.49\%}$)
btree	612_964 ($\textcolor{red}{186.62\%}$)	8_499_914_744 ($\textcolor{green}{-35.95\%}$)	31_103_892	287_553 ($\textcolor{green}{-37.65\%}$)	394_692 ($\textcolor{green}{-37.28\%}$)	437_636 ($\textcolor{green}{-38.07\%}$)
zhenya_hashmap	327_958 ($\textcolor{red}{94.71\%}$)	2_966_591_069 ($\textcolor{green}{-23.47\%}$)	65_987_480	77_126 ($\textcolor{green}{-27.67\%}$)	89_303 ($\textcolor{green}{-31.98\%}$)	88_334 ($\textcolor{green}{-43.41\%}$)
btreemap_rs	614_219 ($\textcolor{red}{37.63\%}$)	1_797_482_744 ($\textcolor{green}{-0.01\%}$)	13_762_560	75_522 ($\textcolor{red}{1.31\%}$)	125_349 ($\textcolor{green}{-0.62\%}$)	91_276 ($\textcolor{green}{-1.68\%}$)
imrc_hashmap_rs	613_876 ($\textcolor{red}{37.59\%}$)	2_573_781_688 ($\textcolor{red}{0.07\%}$)	122_454_016	38_896 ($\textcolor{green}{-0.15\%}$)	178_602 ($\textcolor{green}{-0.28\%}$)	115_825 ($\textcolor{red}{0.23\%}$)
hashmap_rs	604_242 ($\textcolor{red}{37.53\%}$)	430_766_072 ($\textcolor{green}{-3.77\%}$)	36_536_320	21_623 ($\textcolor{green}{-2.72\%}$)	26_760 ($\textcolor{green}{-3.27\%}$)	24_943 ($\textcolor{green}{-1.37\%}$)

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	272_461 ($\textcolor{red}{78.86\%}$)	5_176_386_504 ($\textcolor{green}{-29.05\%}$)	29_995_836	557_538 ($\textcolor{green}{-31.44\%}$)	207_483 ($\textcolor{green}{-30.27\%}$)
heap_rs	602_644 ($\textcolor{red}{37.82\%}$)	140_170_047 ($\textcolor{green}{-1.92\%}$)	9_109_504	59_253 ($\textcolor{green}{-1.00\%}$)	23_240 ($\textcolor{green}{-2.05\%}$)

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	364_175 ($\textcolor{red}{125.59\%}$)	2_313_561 ($\textcolor{green}{-29.10\%}$)	65_508	92_884 ($\textcolor{green}{-25.89\%}$)	714_218 ($\textcolor{green}{-31.50\%}$)	143_884 ($\textcolor{green}{-31.59\%}$)
vector	479_368 ($\textcolor{red}{198.16\%}$)	1_808_257 ($\textcolor{green}{-34.61\%}$)	24_764	143_018 ($\textcolor{green}{-27.60\%}$)	174_110 ($\textcolor{green}{-34.12\%}$)	167_657 ($\textcolor{green}{-37.26\%}$)
vec_rs	600_293 ($\textcolor{red}{37.73\%}$)	289_309 ($\textcolor{green}{-0.29\%}$)	655_360	17_099 ($\textcolor{green}{-2.87\%}$)	30_520 ($\textcolor{green}{-1.59\%}$)	25_655 ($\textcolor{red}{1.00\%}$)

Statistics

• binary_size: 87.90% [62.83%, 112.97%]
• max_mem: no change
• cycles: -20.43% [-23.94%, -16.91%]

SHA-2

	binary_size	SHA-256	SHA-512	account_id	neuron_id
Motoko	384_505 ($\textcolor{red}{96.00\%}$)	271_650_289 ($\textcolor{green}{-23.01\%}$)	261_095_203 ($\textcolor{green}{-23.00\%}$)	33_645 ($\textcolor{green}{-24.93\%}$)	24_286 ($\textcolor{green}{-23.91\%}$)
Rust	713_795 ($\textcolor{red}{35.13\%}$)	83_780_628 ($\textcolor{red}{1.20\%}$)	57_415_417 ($\textcolor{red}{1.09\%}$)	49_652 ($\textcolor{green}{-1.97\%}$)	52_343 ($\textcolor{green}{-2.22\%}$)

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	930_337 ($\textcolor{red}{353.66\%}$)	4_641_230_534 ($\textcolor{green}{-25.77\%}$)	3_429_924	550_465 ($\textcolor{green}{-25.86\%}$)	365_185 ($\textcolor{green}{-27.89\%}$)
Rust	744_723 ($\textcolor{red}{58.47\%}$)	6_355_506_025 ($\textcolor{green}{-0.06\%}$)	1_081_344	1_011_262 ($\textcolor{green}{-0.09\%}$)	303_453 ($\textcolor{green}{-0.55\%}$)

Statistics

• binary_size: 135.81% [-37.60%, 309.23%]
• max_mem: no change
• cycles: -12.64% [-18.71%, -6.57%]

Basic DAO

	binary_size	init	transfer_token	submit_proposal	vote_proposal
Motoko	1_030_946 ($\textcolor{red}{271.66\%}$)	44_715 ($\textcolor{green}{-12.82\%}$)	20_929 ($\textcolor{green}{-17.00\%}$)	17_672 ($\textcolor{green}{-15.69\%}$)	18_567 ($\textcolor{green}{-17.63\%}$)
Rust	1_034_916 ($\textcolor{red}{35.63\%}$)	533_132 ($\textcolor{green}{-3.43\%}$)	101_918 ($\textcolor{green}{-3.12\%}$)	125_469 ($\textcolor{green}{-2.55\%}$)	135_796 ($\textcolor{green}{-2.68\%}$)

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	430_376 ($\textcolor{red}{87.10\%}$)	16_971 ($\textcolor{green}{-11.92\%}$)	28_261 ($\textcolor{green}{-12.50\%}$)	8_614 ($\textcolor{green}{-11.44\%}$)
Rust	1_157_204 ($\textcolor{red}{39.72\%}$)	142_282 ($\textcolor{green}{-2.72\%}$)	369_309 ($\textcolor{green}{-2.88\%}$)	91_478 ($\textcolor{green}{-2.44\%}$)

Statistics

• binary_size: 108.53% [-22.36%, 239.42%]
• max_mem: no change
• cycles: -8.49% [-11.39%, -5.59%]

Heartbeat

	binary_size	heartbeat
Motoko	235_377 ($\textcolor{red}{65.63\%}$)	19_034 ($\textcolor{green}{-20.92\%}$)
Rust	27_599 ($\textcolor{red}{7.60\%}$)	1_162 ($\textcolor{red}{111.66\%}$)

Timer

	binary_size	setTimer	cancelTimer
Motoko	289_586 ($\textcolor{red}{94.03\%}$)	50_678 ($\textcolor{green}{-7.02\%}$)	4_491 ($\textcolor{green}{-9.78\%}$)
Rust	650_758 ($\textcolor{red}{38.26\%}$)	67_826 ($\textcolor{green}{-2.73\%}$)	11_037 ($\textcolor{green}{-3.23\%}$)

Statistics

• binary_size: 66.14% [-109.93%, 242.22%]
• max_mem: no change
• cycles: -5.69% [-9.62%, -1.76%]

Garbage Collection

Note
Same as main branch, skipping.

Actor class

	binary size	put new bucket	put existing bucket	get
Map	499_549 ($\textcolor{red}{67.79\%}$)	775_680 ($\textcolor{green}{-1.01\%}$)	15_832 ($\textcolor{green}{-7.12\%}$)	16_330 ($\textcolor{green}{-6.85\%}$)

Statistics

• binary_size: no change
• max_mem: no change
• cycles: 13.20% [-29.75%, 56.15%]

Publisher & Subscriber

	pub_binary_size	sub_binary_size	subscribe_caller	subscribe_callee	publish_caller	publish_callee
Motoko	305_839 ($\textcolor{red}{83.38\%}$)	265_681 ($\textcolor{red}{74.91\%}$)	28_092 ($\textcolor{green}{-6.22\%}$)	11_466 ($\textcolor{green}{-8.68\%}$)	22_385 ($\textcolor{green}{-6.96\%}$)	6_218 ($\textcolor{green}{-9.52\%}$)
Rust	692_066 ($\textcolor{red}{35.20\%}$)	767_763 ($\textcolor{red}{35.79\%}$)	69_878 ($\textcolor{green}{-2.58\%}$)	42_967 ($\textcolor{green}{-3.05\%}$)	93_307 ($\textcolor{green}{-2.57\%}$)	52_511 ($\textcolor{green}{-2.65\%}$)

Statistics

• binary_size: 57.32% [27.39%, 87.25%]
• max_mem: no change
• cycles: -5.28% [-7.23%, -3.32%]

Overall Statistics

• binary_size: 91.77% [67.19%, 116.35%]
• max_mem: no change
• cycles: -14.40% [-17.14%, -11.66%]

[github-actions]

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	297_254	7_842_361_244	61_987_732	321_307	6_155_781_309	345_252
triemap	333_856	11_407_728_709	74_216_052	195_746	533_776	524_332
rbtree	296_811	6_513_814_073	57_995_940	98_058	289_073	312_663
splay	286_117	11_942_316_508	53_995_876	563_736	593_264	821_312
btree	612_964	8_499_914_744	31_103_892	287_553	394_692	437_636
zhenya_hashmap	327_958	2_966_591_069	65_987_480	77_126	89_303	88_334
btreemap_rs	614_219	1_797_482_744	13_762_560	75_522	125_349	91_276
imrc_hashmap_rs	613_876	2_573_781_688	122_454_016	38_896	178_602	115_825
hashmap_rs	604_242	430_766_072	36_536_320	21_623	26_760	24_943

Priority queue

	binary_size	heapify 1m	max mem	pop_min 50	put 50
heap	272_461	5_176_386_504	29_995_836	557_538	207_483
heap_rs	602_644	140_170_047	9_109_504	59_253	23_240

Growable array

	binary_size	generate 5k	max mem	batch_get 500	batch_put 500	batch_remove 500
buffer	364_175	2_313_561	65_508	92_884	714_218	143_884
vector	479_368	1_808_257	24_764	143_018	174_110	167_657
vec_rs	600_293	289_309	655_360	17_099	30_520	25_655

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	384_505	271_650_289	261_095_203	33_645	24_286
Rust	713_795	83_780_628	57_415_417	49_652	52_343

Certified map

	binary_size	generate 10k	max mem	inc	witness
Motoko	930_337	4_641_230_534	3_429_924	550_465	365_185
Rust	744_723	6_355_506_025	1_081_344	1_011_262	303_453

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	1_030_946	44_715	20_929	17_672	18_567
Rust	1_034_916	533_132	101_918	125_469	135_796

DIP721 NFT

	binary_size	init	mint_token	transfer_token
Motoko	430_376	16_971	28_261	8_614
Rust	1_157_204	142_282	369_309	91_478

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	235_377	19_034
Rust	27_599	1_162

Timer

	binary_size	setTimer	cancelTimer
Motoko	289_586	50_678	4_491
Rust	650_758	67_826	11_037

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_338_231_405	59_396_776	118	118	118
copying	1_338_231_287	59_396_776	1_337_913_569	1_338_002_371	1_337_919_144
compacting	1_911_420_608	59_396_776	1_473_824_186	1_756_485_066	1_787_369_954
generational	2_891_818_643	59_405_240	1_141_865_993	1_217_376	1_117_840
incremental	33_436_719	1_136_155_048	333_734_166	336_829_512	336_860_690

Actor class

	binary size	put new bucket	put existing bucket	get
Map	499_549	775_680	15_832	16_330

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	305_839	265_681	28_092	11_466	22_385	6_218
Rust	692_066	767_763	69_878	42_967	93_307	52_511