Skip to content

Commit

Permalink
Update hardcoded weights in coretime chain runtimes (#8)
Browse files Browse the repository at this point in the history 
* Update hardcoded weights in coretime chain runtimes

* Increase buffers
  • Loading branch information
@seadanda
seadanda authored Aug 16, 2024
1 parent 93e8f05 commit f2116a9
Show file tree
Hide file tree
Showing 2 changed files with 35 additions and 32 deletions.
31 changes: 16 additions & 15 deletions system-parachains/coretime/coretime-kusama/src/coretime.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ use frame_support::{
tokens::{Fortitude, Preservation},
DefensiveResult, OnUnbalanced,
},
weights::constants::{WEIGHT_PROOF_SIZE_PER_KB, WEIGHT_REF_TIME_PER_MICROS},
};
use frame_system::Pallet as System;
use kusama_runtime_constants::{system_parachain::coretime, time::DAYS as RELAY_DAYS};
Expand Down Expand Up @@ -137,10 +138,11 @@ impl CoretimeInterface for CoretimeAllocator {
let request_core_count_call = RelayRuntimePallets::Coretime(RequestCoreCount(count));

// Weight for `request_core_count` from Kusama runtime benchmarks:
// `ref_time` = 7889000 + (3 * 25000000) + (1 * 100000000) = 182889000
// `proof_size` = 1636
// Add 5% to each component and round to 2 significant figures.
let call_weight = Weight::from_parts(190_000_000, 1700);
// `ref_time`, `proof_size`, reads, writes
// 9_670_000, 1640, 3, 1
// Add 30% to each component with a healthy round up.
let call_weight =
Weight::from_parts(250 * WEIGHT_REF_TIME_PER_MICROS, 3 * WEIGHT_PROOF_SIZE_PER_KB);

let message = Xcm(vec![
Instruction::UnpaidExecution {
Expand Down Expand Up @@ -173,13 +175,11 @@ impl CoretimeInterface for CoretimeAllocator {
RelayRuntimePallets::Coretime(RequestRevenueInfoAt(when));

// Weight for `request_revenue_at` from Kusama runtime benchmarks:
// `ref_time` = 37_637_000 + (3 * 25000000) + (6 * 100000000) = 712637000
// `proof_size` = 6428
// Add 5% to each component and round to 2 significant figures.
//
// This benchmark has been transplanted from a testnet and not rerun, so adding a healthy
// buffer. TODO refine when benchmarks are run: https://github.com/polkadot-fellows/runtimes/issues/404
let call_weight = Weight::from_parts(1_000_000_000, 20_000);
// `ref_time`, `proof_size`, reads, writes
// 94_091_000, 6384, 7, 5
// Add 30% to each component with a healthy round up.
let call_weight =
Weight::from_parts(1000 * WEIGHT_REF_TIME_PER_MICROS, 9 * WEIGHT_PROOF_SIZE_PER_KB);

let message = Xcm(vec![
Instruction::UnpaidExecution {
Expand Down Expand Up @@ -225,10 +225,11 @@ impl CoretimeInterface for CoretimeAllocator {
use crate::coretime::CoretimeProviderCalls::AssignCore;

// Weight for `assign_core` from Kusama runtime benchmarks:
// `ref_time` = 10177115 + (1 * 25000000) + (2 * 100000000) + (80 * 13932) = 236291675
// `proof_size` = 3612
// Add 5% to each component and round to 2 significant figures.
let call_weight = Weight::from_parts(248_000_000, 3800);
// `ref_time`, `proof_size`, reads, writes
// 12_042_907 + 80 * 13_919, 3545, 1, 2
// Add 30% to each component with a healthy round up.
let call_weight =
Weight::from_parts(350 * WEIGHT_REF_TIME_PER_MICROS, 5 * WEIGHT_PROOF_SIZE_PER_KB);

// The relay chain currently only allows `assign_core` to be called with a complete mask
// and only ever with increasing `begin`. The assignments must be truncated to avoid
Expand Down
36 changes: 19 additions & 17 deletions system-parachains/coretime/coretime-polkadot/src/coretime.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ use frame_support::{
tokens::{Fortitude, Preservation},
DefensiveResult, OnUnbalanced,
},
weights::constants::{WEIGHT_PROOF_SIZE_PER_KB, WEIGHT_REF_TIME_PER_MICROS},
};
use frame_system::Pallet as System;
use pallet_broker::{CoreAssignment, CoreIndex, CoretimeInterface, PartsOf57600, RCBlockNumberOf};
Expand Down Expand Up @@ -137,11 +138,12 @@ impl CoretimeInterface for CoretimeAllocator {
let request_core_count_call = RelayRuntimePallets::Coretime(RequestCoreCount(count));

// Weight for `request_core_count` from Polkadot runtime benchmarks:
// `ref_time` = 7889000 + (3 * 25000000) + (1 * 100000000) = 182889000
// `proof_size` = 1636
// Add 5% to each component and round to 2 significant figures.
// TODO check when benchmarks are rerun
let call_weight = Weight::from_parts(190_000_000, 1700);
// `ref_time`, `proof_size`, reads, writes
// 9_660_000, 1640, 3, 1
// Use overestimates for reads and writes, add 30% to each component with a healthy round
// up.
let call_weight =
Weight::from_parts(250 * WEIGHT_REF_TIME_PER_MICROS, 3 * WEIGHT_PROOF_SIZE_PER_KB);

let message = Xcm(vec![
Instruction::UnpaidExecution {
Expand Down Expand Up @@ -174,13 +176,12 @@ impl CoretimeInterface for CoretimeAllocator {
RelayRuntimePallets::Coretime(RequestRevenueInfoAt(when));

// Weight for `request_revenue_at` from Polkadot runtime benchmarks:
// `ref_time` = 37_637_000 + (3 * 25000000) + (6 * 100000000) = 712637000
// `proof_size` = 6428
// Add 5% to each component and round to 2 significant figures.
//
// These weights have been transplanted from another network and not rerun, so a healthy
// buffer is included. TODO refine when benchmarks are run.
let call_weight = Weight::from_parts(1_000_000_000, 20_000);
// `ref_time`, `proof_size`, reads, writes
// 93_731_000, 6313, 7, 5
// Use overestimates for reads and writes, add 30% to each component with a healthy round
// up.
let call_weight =
Weight::from_parts(1000 * WEIGHT_REF_TIME_PER_MICROS, 9 * WEIGHT_PROOF_SIZE_PER_KB);

let message = Xcm(vec![
Instruction::UnpaidExecution {
Expand Down Expand Up @@ -226,11 +227,12 @@ impl CoretimeInterface for CoretimeAllocator {
use crate::coretime::CoretimeProviderCalls::AssignCore;

// Weight for `assign_core` from Polkadot runtime benchmarks:
// `ref_time` = 10177115 + (1 * 25000000) + (2 * 100000000) + (80 * 13932) = 236291675
// `proof_size` = 3612
// Add 5% to each component and round to 2 significant figures.
// TODO check when benchmarks are rerun
let call_weight = Weight::from_parts(248_000_000, 3800);
// `ref_time`, `proof_size`, reads, writes
// 12_201_135 + 80 * 13_556, 3579, 1, 2
// Use overestimates for reads and writes, add 30% to each component with a healthy round
// up.
let call_weight =
Weight::from_parts(350 * WEIGHT_REF_TIME_PER_MICROS, 5 * WEIGHT_PROOF_SIZE_PER_KB);

// The relay chain currently only allows `assign_core` to be called with a complete mask
// and only ever with increasing `begin`. The assignments must be truncated to avoid
Expand Down

0 comments on commit f2116a9

Please sign in to comment.