Skip to content

Commit

Permalink
Add Proof Size to Weight Output (paritytech#11637)
Browse files Browse the repository at this point in the history
* initial impl

* add template test

* linear fit proof size

* always record proof when tracking storage

* calculate worst case pov

* remove duplicate worst case

* cargo run --quiet --profile=production  --features=runtime-benchmarks --manifest-path=bin/node/cli/Cargo.toml -- benchmark pallet --chain=dev --steps=50 --repeat=20 --pallet=pallet_assets --extrinsic=* --execution=wasm --wasm-execution=compiled --heap-pages=4096 --output=./frame/assets/src/weights.rs --template=./.maintain/frame-weight-template.hbs

* more comment output

* add cli for worst case map size

* update name

* clap does not support underscores

* rename

* expose worst case map values

* improve some comments

* cargo run --quiet --profile=production  --features=runtime-benchmarks --manifest-path=bin/node/cli/Cargo.toml -- benchmark pallet --chain=dev --steps=50 --repeat=20 --pallet=pallet_assets --extrinsic=* --execution=wasm --wasm-execution=compiled --heap-pages=4096 --output=./frame/assets/src/weights.rs --template=./.maintain/frame-weight-template.hbs

* update template

* cargo run --quiet --profile=production  --features=runtime-benchmarks --manifest-path=bin/node/cli/Cargo.toml -- benchmark pallet --chain=dev --steps=50 --repeat=20 --pallet=pallet_assets --extrinsic=* --execution=wasm --wasm-execution=compiled --heap-pages=4096 --output=./frame/assets/src/weights.rs --template=./.maintain/frame-weight-template.hbs

* fix fmt

* more fmt

* more fmt

* Dont panic when there is no proof

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Fix test features

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Whitelist :extrinsic_index

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Use whitelist when recording proof

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add logs

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add PoV testing pallet

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Deploy PoV testing pallet

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Storage benches reside in the PoV pallet

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Linear regress PoV per component

Splits the PoV calculation into "measured" and "estimated".
The measured part is reported by the Proof recorder and linear
regressed over all components at once.
The estimated part is calculated as worst-case by using the max
PoV size per storage access and calculating one linear regress per
component. This gives each component a (possibly) independent PoV.
For now the measured size will always be lower than the PoV on
Polkadot since it is measured on an empty snapshot. The measured
part is therefor only used as diagnostic for debugging.

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Put PoV into the weight templates

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* fmt

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Extra alanysis choise for PoV

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add+Fix tests

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Make benches faster

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Cleanup

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Use same template comments

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* ".git/.scripts/bench-bot.sh" pallet dev pallet_balances

* ".git/.scripts/bench-bot.sh" pallet dev pallet_democracy

* Update referenda mock BlockWeights

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Take measured value size into account

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* clippy

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* ".git/.scripts/bench-bot.sh" pallet dev pallet_scheduler

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* proof_size: None

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* ugly, but works

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* wup

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* WIP

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add pov_mode attribute to the benchmarks! macro

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Use pov_mode attribute in PoV benchmarking

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Update tests

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Scheduler, Whitelist: Add pov_mode attr

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Update PoV weights

* Add CLI arg: default-pov-mode

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Fix tests

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* fmt

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* fix

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Revert "Update PoV weights"

This reverts commit 2f3ac23.

* Revert "WIP"

This reverts commit c34b538.

* Revert first approach

This reverts commit range 8ddaa2f^..4c84f87

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Clippy

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add extra benchmarks

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* ".git/.scripts/commands/bench/bench.sh" pallet dev pallet_alliance

* ".git/.scripts/commands/bench/bench.sh" pallet dev pallet_whitelist

* ".git/.scripts/commands/bench/bench.sh" pallet dev pallet_scheduler

* fmt

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Clippy

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Clippy 🤦

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add reference benchmarks

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Fix doc comments

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Undo logging

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Add 'Ignored' pov_mode

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Allow multiple attributes per benchmark

Turns out that the current benchmarking syntax does not support
multiple attributes per bench 🤦. Changing it to support that
since otherwise the `pov_mode` would conflict with the others.

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Validate pov_mode syntax

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Ignore PoV for all contract benchmarks

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Test

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* test

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Bump macro recursion limit

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* fmt

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Update contract weights

They dont have a PoV component anymore.

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* fix test ffs

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* pov_mode is unsupported in V2 syntax

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Fix pallet ui tests

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* update pallet ui

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Fix pallet ui tests

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

* Update weights

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>

Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
Co-authored-by: Parity Bot <admin@parity.io>
Co-authored-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
Co-authored-by: command-bot <>
Co-authored-by: Your Name <you@example.com>
  • Loading branch information
4 people authored and ltfschoen committed Feb 22, 2023
1 parent 57747a2 commit e243fd6
Show file tree
Hide file tree
Showing 74 changed files with 19,128 additions and 9,571 deletions.
35 changes: 28 additions & 7 deletions .maintain/frame-weight-template.hbs
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,8 @@
//! Autogenerated weights for {{pallet}}
//!
//! THIS FILE WAS AUTO-GENERATED USING THE SUBSTRATE BENCHMARK CLI VERSION {{version}}
//! DATE: {{date}}, STEPS: `{{cmd.steps}}`, REPEAT: {{cmd.repeat}}, LOW RANGE: `{{cmd.lowest_range_values}}`, HIGH RANGE: `{{cmd.highest_range_values}}`
//! DATE: {{date}}, STEPS: `{{cmd.steps}}`, REPEAT: `{{cmd.repeat}}`, LOW RANGE: `{{cmd.lowest_range_values}}`, HIGH RANGE: `{{cmd.highest_range_values}}`
//! WORST CASE MAP SIZE: `{{cmd.worst_case_map_values}}`
//! HOSTNAME: `{{hostname}}`, CPU: `{{cpuname}}`
//! EXECUTION: {{cmd.execution}}, WASM-EXECUTION: {{cmd.wasm_execution}}, CHAIN: {{cmd.chain}}, DB CACHE: {{cmd.db_cache}}

Expand Down Expand Up @@ -38,7 +39,7 @@ impl<T: frame_system::Config> WeightInfo for SubstrateWeight<T> {
{{/if}}
{{#each benchmarks as |benchmark|}}
{{#each benchmark.comments as |comment|}}
// {{comment}}
/// {{comment}}
{{/each}}
{{#each benchmark.component_ranges as |range|}}
/// The range of component `{{range.name}}` is `[{{range.min}}, {{range.max}}]`.
Expand All @@ -48,24 +49,34 @@ impl<T: frame_system::Config> WeightInfo for SubstrateWeight<T> {
{{~#each benchmark.components as |c| ~}}
{{~#if (not c.is_used)}}_{{/if}}{{c.name}}: u32, {{/each~}}
) -> Weight {
// Proof Size summary in bytes:
// Measured: `{{benchmark.base_recorded_proof_size}}{{#each benchmark.component_recorded_proof_size as |cp|}} + {{cp.name}} * ({{cp.slope}} ±{{underscore cp.error}}){{/each}}`
// Estimated: `{{benchmark.base_calculated_proof_size}}{{#each benchmark.component_calculated_proof_size as |cp|}} + {{cp.name}} * ({{cp.slope}} ±{{underscore cp.error}}){{/each}}`
// Minimum execution time: {{underscore benchmark.min_execution_time}} nanoseconds.
{{#if (ne benchmark.base_calculated_proof_size "0")}}
Weight::from_parts({{underscore benchmark.base_weight}}, {{benchmark.base_calculated_proof_size}})
{{else}}
Weight::from_ref_time({{underscore benchmark.base_weight}})
{{/if}}
{{#each benchmark.component_weight as |cw|}}
// Standard Error: {{underscore cw.error}}
.saturating_add(Weight::from_ref_time({{underscore cw.slope}}).saturating_mul({{cw.name}}.into()))
{{/each}}
{{#if (ne benchmark.base_reads "0")}}
.saturating_add(T::DbWeight::get().reads({{benchmark.base_reads}}))
.saturating_add(T::DbWeight::get().reads({{benchmark.base_reads}}_u64))
{{/if}}
{{#each benchmark.component_reads as |cr|}}
.saturating_add(T::DbWeight::get().reads(({{cr.slope}}_u64).saturating_mul({{cr.name}}.into())))
{{/each}}
{{#if (ne benchmark.base_writes "0")}}
.saturating_add(T::DbWeight::get().writes({{benchmark.base_writes}}))
.saturating_add(T::DbWeight::get().writes({{benchmark.base_writes}}_u64))
{{/if}}
{{#each benchmark.component_writes as |cw|}}
.saturating_add(T::DbWeight::get().writes(({{cw.slope}}_u64).saturating_mul({{cw.name}}.into())))
{{/each}}
{{#each benchmark.component_calculated_proof_size as |cp|}}
.saturating_add(Weight::from_proof_size({{cp.slope}}).saturating_mul({{cp.name}}.into()))
{{/each}}
}
{{/each}}
}
Expand All @@ -74,7 +85,7 @@ impl<T: frame_system::Config> WeightInfo for SubstrateWeight<T> {
impl WeightInfo for () {
{{#each benchmarks as |benchmark|}}
{{#each benchmark.comments as |comment|}}
// {{comment}}
/// {{comment}}
{{/each}}
{{#each benchmark.component_ranges as |range|}}
/// The range of component `{{range.name}}` is `[{{range.min}}, {{range.max}}]`.
Expand All @@ -84,24 +95,34 @@ impl WeightInfo for () {
{{~#each benchmark.components as |c| ~}}
{{~#if (not c.is_used)}}_{{/if}}{{c.name}}: u32, {{/each~}}
) -> Weight {
// Proof Size summary in bytes:
// Measured: `{{benchmark.base_recorded_proof_size}}{{#each benchmark.component_recorded_proof_size as |cp|}} + {{cp.name}} * ({{cp.slope}} ±{{underscore cp.error}}){{/each}}`
// Estimated: `{{benchmark.base_calculated_proof_size}}{{#each benchmark.component_calculated_proof_size as |cp|}} + {{cp.name}} * ({{cp.slope}} ±{{underscore cp.error}}){{/each}}`
// Minimum execution time: {{underscore benchmark.min_execution_time}} nanoseconds.
{{#if (ne benchmark.base_calculated_proof_size "0")}}
Weight::from_parts({{underscore benchmark.base_weight}}, {{benchmark.base_calculated_proof_size}})
{{else}}
Weight::from_ref_time({{underscore benchmark.base_weight}})
{{/if}}
{{#each benchmark.component_weight as |cw|}}
// Standard Error: {{underscore cw.error}}
.saturating_add(Weight::from_ref_time({{underscore cw.slope}}).saturating_mul({{cw.name}}.into()))
{{/each}}
{{#if (ne benchmark.base_reads "0")}}
.saturating_add(RocksDbWeight::get().reads({{benchmark.base_reads}}))
.saturating_add(RocksDbWeight::get().reads({{benchmark.base_reads}}_u64))
{{/if}}
{{#each benchmark.component_reads as |cr|}}
.saturating_add(RocksDbWeight::get().reads(({{cr.slope}}_u64).saturating_mul({{cr.name}}.into())))
{{/each}}
{{#if (ne benchmark.base_writes "0")}}
.saturating_add(RocksDbWeight::get().writes({{benchmark.base_writes}}))
.saturating_add(RocksDbWeight::get().writes({{benchmark.base_writes}}_u64))
{{/if}}
{{#each benchmark.component_writes as |cw|}}
.saturating_add(RocksDbWeight::get().writes(({{cw.slope}}_u64).saturating_mul({{cw.name}}.into())))
{{/each}}
{{#each benchmark.component_calculated_proof_size as |cp|}}
.saturating_add(Weight::from_proof_size({{cp.slope}}).saturating_mul({{cp.name}}.into()))
{{/each}}
}
{{/each}}
}
15 changes: 15 additions & 0 deletions Cargo.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,7 @@ members = [
"frame/beefy-mmr",
"frame/beefy-mmr/primitives",
"frame/benchmarking",
"frame/benchmarking/pov",
"frame/bounties",
"frame/child-bounties",
"frame/collective",
Expand Down
8 changes: 6 additions & 2 deletions bin/node/runtime/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -43,7 +43,8 @@ sp-io = { version = "7.0.0", default-features = false, path = "../../../primitiv

# frame dependencies
frame-executive = { version = "4.0.0-dev", default-features = false, path = "../../../frame/executive" }
frame-benchmarking = { version = "4.0.0-dev", default-features = false, path = "../../../frame/benchmarking", optional = true }
frame-benchmarking = { version = "4.0.0-dev", default-features = false, path = "../../../frame/benchmarking" }
frame-benchmarking-pallet-pov = { version = "4.0.0-dev", default-features = false, path = "../../../frame/benchmarking/pov" }
frame-support = { version = "4.0.0-dev", default-features = false, path = "../../../frame/support" }
frame-system = { version = "4.0.0-dev", default-features = false, path = "../../../frame/system" }
frame-system-benchmarking = { version = "4.0.0-dev", default-features = false, path = "../../../frame/system/benchmarking", optional = true }
Expand Down Expand Up @@ -178,7 +179,8 @@ std = [
"sp-session/std",
"pallet-sudo/std",
"frame-support/std",
"frame-benchmarking?/std",
"frame-benchmarking/std",
"frame-benchmarking-pallet-pov/std",
"frame-system-rpc-runtime-api/std",
"frame-system/std",
"pallet-election-provider-multi-phase/std",
Expand Down Expand Up @@ -209,6 +211,7 @@ std = [
]
runtime-benchmarks = [
"frame-benchmarking/runtime-benchmarks",
"frame-benchmarking-pallet-pov/runtime-benchmarks",
"frame-support/runtime-benchmarks",
"frame-system/runtime-benchmarks",
"sp-runtime/runtime-benchmarks",
Expand Down Expand Up @@ -263,6 +266,7 @@ runtime-benchmarks = [
]
try-runtime = [
"frame-try-runtime/try-runtime",
"frame-benchmarking-pallet-pov/try-runtime",
"frame-executive/try-runtime",
"frame-system/try-runtime",
"frame-support/try-runtime",
Expand Down
6 changes: 6 additions & 0 deletions bin/node/runtime/src/lib.rs
Original file line number Diff line number Diff line change
Expand Up @@ -1696,6 +1696,10 @@ impl pallet_alliance::Config for Runtime {
type RetirementPeriod = RetirementPeriod;
}

impl frame_benchmarking_pallet_pov::Config for Runtime {
type RuntimeEvent = RuntimeEvent;
}

construct_runtime!(
pub enum Runtime where
Block = Block,
Expand Down Expand Up @@ -1762,6 +1766,7 @@ construct_runtime!(
RankedCollective: pallet_ranked_collective,
FastUnstake: pallet_fast_unstake,
MessageQueue: pallet_message_queue,
Pov: frame_benchmarking_pallet_pov,
}
);

Expand Down Expand Up @@ -1830,6 +1835,7 @@ mod mmr {
mod benches {
frame_benchmarking::define_benchmarks!(
[frame_benchmarking, BaselineBench::<Runtime>]
[frame_benchmarking_pallet_pov, Pov]
[pallet_alliance, Alliance]
[pallet_assets, Assets]
[pallet_babe, Babe]
Expand Down
4 changes: 4 additions & 0 deletions client/db/src/bench.rs
Original file line number Diff line number Diff line change
Expand Up @@ -606,10 +606,14 @@ impl<B: BlockT> StateBackend<HashFor<B>> for BenchmarkingState<B> {
let proof_recorder_root = self.proof_recorder_root.get();
if proof_recorder_root == Default::default() || proof_size == 1 {
// empty trie
log::debug!(target: "benchmark", "Some proof size: {}", &proof_size);
proof_size
} else {
if let Some(size) = proof.encoded_compact_size::<HashFor<B>>(proof_recorder_root) {
size as u32
} else if proof_recorder_root == self.root.get() {
log::debug!(target: "benchmark", "No changes - no proof");
0
} else {
panic!(
"proof rec root {:?}, root {:?}, genesis {:?}, rec_len {:?}",
Expand Down
2 changes: 1 addition & 1 deletion client/db/src/upgrade.rs
Original file line number Diff line number Diff line change
Expand Up @@ -190,7 +190,7 @@ fn version_file_path(path: &Path) -> PathBuf {
file_path
}

#[cfg(test)]
#[cfg(all(test, feature = "rocksdb"))]
mod tests {
use super::*;
use crate::{tests::Block, DatabaseSource};
Expand Down
7 changes: 5 additions & 2 deletions client/db/src/utils.rs
Original file line number Diff line number Diff line change
Expand Up @@ -583,12 +583,12 @@ mod tests {
use super::*;
use codec::Input;
use sp_runtime::testing::{Block as RawBlock, ExtrinsicWrapper};
use std::path::PathBuf;
type Block = RawBlock<ExtrinsicWrapper<u32>>;

#[cfg(any(feature = "rocksdb", test))]
#[cfg(feature = "rocksdb")]
#[test]
fn database_type_subdir_migration() {
use std::path::PathBuf;
type Block = RawBlock<ExtrinsicWrapper<u64>>;

fn check_dir_for_db_type(
Expand Down Expand Up @@ -685,6 +685,7 @@ mod tests {
assert_eq!(joined.remaining_len().unwrap(), Some(0));
}

#[cfg(feature = "rocksdb")]
#[test]
fn test_open_database_auto_new() {
let db_dir = tempfile::TempDir::new().unwrap();
Expand Down Expand Up @@ -730,6 +731,7 @@ mod tests {
}
}

#[cfg(feature = "rocksdb")]
#[test]
fn test_open_database_rocksdb_new() {
let db_dir = tempfile::TempDir::new().unwrap();
Expand Down Expand Up @@ -780,6 +782,7 @@ mod tests {
}
}

#[cfg(feature = "rocksdb")]
#[test]
fn test_open_database_paritydb_new() {
let db_dir = tempfile::TempDir::new().unwrap();
Expand Down
Loading

0 comments on commit e243fd6

Please sign in to comment.