Weights per runtime

[gonzamontiel]

What does it do?

This PR replaces #2939.

It modifies pallet's benchmarks and removes magic numbers to be runtime-dependant.

What important points reviewers should know?

Pallet parachain staking had some fixed numbers that I replaced by values we get from the runtime. This might go against the testing purpose, because now a change on those parameters might go unnoticed in the test. Anyway I think it is an acceptable trade off.

Is there something left for follow-up PRs?

What alternative implementations were considered?

Are there relevant PRs or issues in other repositories (Substrate, Polkadot, Frontier, Cumulus)?

What value does it bring to the blockchain users?

[github-actions]

WASM runtime size check:

Compared to target branch

Moonbase runtime: 2260 KB (no changes) ✅

Moonbeam runtime: 2224 KB (no changes) ✅

Moonriver runtime: 2224 KB (no changes) ✅

Compared to latest release (runtime-3300)

Moonbase runtime: 2260 KB (+232 KB compared to latest release) ⚠️

Moonbeam runtime: 2224 KB (+228 KB compared to latest release) ⚠️

Moonriver runtime: 2224 KB (+232 KB compared to latest release) ⚠️

[github-actions]

Coverage Report

@@                      Coverage Diff                      @@
##           master   gonza/weights-per-runtime      +/-   ##
=============================================================
- Coverage   79.34%                      74.81%   -4.53%     
+ Files         301                         369      +68     
+ Lines       87834                       94127    +6293     
=============================================================
+ Hits        69685                       70421     +736     
+ Misses      18149                       23706    +5557     

Files Changed	Coverage
/precompiles/relay-data-verifier/src/mock.rs	85.53% (-2.16%)	🔽
/runtime/moonbeam/tests/common/mod.rs	94.74% (+0.06%)	🔼
/runtime/moonriver/tests/common/mod.rs	94.85% (+0.05%)	🔼

Coverage generated Thu Nov 21 00:04:46 UTC 2024

[RomarQ]

Changes in specs should be reverted.

[gonzamontiel]

Changes in specs should be reverted.

good catch, I don't know, maybe it was a bad merge.

[crystalin]

We should flag weights that are being changed by more than 10% so we can discuss their impacts

[gonzamontiel]

We should flag weights that are being changed by more than 10% so we can discuss their impacts

Here the list of the changes greater than 5% . The negatives are improvements.
pallet_moonbeam_foreign_assets has huge changes.

File	Extrinsic	Moonbeam %	Moonriver %
pallet_treasury.rs	payout	-31.44	-31.44
pallet_xcm_transactor.rs	transact_through_sovereign	-27.23	-27.23
pallet_xcm_transactor.rs	transact_through_derivative	-21.72	-21.72
pallet_moonbeam_orbiters.rs	on_new_round	-20.18	0.20
pallet_referenda.rs	kill	-18.76	-18.76
pallet_parachain_staking.rs	notify_inactive_collator	-11.52	-0.82
pallet_whitelist.rs	dispatch_whitelisted_call	-6.55	-5.11
pallet_xcm_transactor.rs	transact_through_signed	-5.80	-5.80
pallet_xcm_transactor.rs	hrmp_manage	-5.68	-5.68
pallet_message_queue.rs	execute_overweight_page_updated	5.05	2.66
pallet_scheduler.rs	service_task_fetched	9.68	9.88
pallet_moonbeam_foreign_assets.rs	create_foreign_asset	35.60	35.60
pallet_moonbeam_foreign_assets.rs	freeze_foreign_asset	55.36	55.36
pallet_moonbeam_foreign_assets.rs	unfreeze_foreign_asset	68.09	68.09

[gonzamontiel]

We should flag weights that are being changed by more than 10% so we can discuss their impacts

Here the list of the changes greater than 5% . The negatives are improvements. pallet_moonbeam_foreign_assets has huge changes.
...

Btw, the comparison is against weights in moonbase (master). But the weights difference between moonbase (master) and moonbase (this branch) is negligible: in average -0.22 %, which is consistent becase we use the same machine.

[RomarQ]

We should flag weights that are being changed by more than 10% so we can discuss their impacts

Here the list of the changes greater than 5% . The negatives are improvements. pallet_moonbeam_foreign_assets has huge changes.
...

Btw, the comparison is against weights in moonbase (master). But the weights difference between moonbase (master) and moonbase (this branch) is negligible: in average -0.22 %, which is consistent becase we use the same machine.

The weight difference is now reasonable. Do you know why it was different before?