Oliver Tale-Yazdi
9543d31474
This MR contains two major changes and some maintenance cleanup. ## 1. Free Standing Pallet Benchmark Runner Closes https://github.com/paritytech/polkadot-sdk/issues/3045, depends on your runtime exposing the `GenesisBuilderApi` (like https://github.com/paritytech/polkadot-sdk/pull/1492). Introduces a new binary crate: `frame-omni-bencher`. It allows to directly benchmark a WASM blob - without needing a node or chain spec. This makes it much easier to generate pallet weights and should allow us to remove bloaty code from the node. It should work for all FRAME runtimes that dont use 3rd party host calls or non `BlakeTwo256` block hashing (basically all polkadot parachains should work). It is 100% backwards compatible with the old CLI args, when the `v1` compatibility command is used. This is done to allow for forwards compatible addition of new commands. ### Example (full example in the Rust docs) Installing the CLI: ```sh cargo install --locked --path substrate/utils/frame/omni-bencher frame-omni-bencher --help ``` Building the Westend runtime: ```sh cargo build -p westend-runtime --release --features runtime-benchmarks ``` Benchmarking the runtime: ```sh frame-omni-bencher v1 benchmark pallet --runtime target/release/wbuild/westend-runtime/westend_runtime.compact.compressed.wasm --all ``` ## 2. Building the Benchmark Genesis State in the Runtime Closes https://github.com/paritytech/polkadot-sdk/issues/2664 This adds `--runtime` and `--genesis-builder=none|runtime|spec` arguments to the `benchmark pallet` command to make it possible to generate the genesis storage by the runtime. This can be used with both the node and the freestanding benchmark runners. It utilizes the new `GenesisBuilder` RA and depends on having https://github.com/paritytech/polkadot-sdk/pull/3412 deployed. ## 3. Simpler args for `PalletCmd::run` You can do three things here to integrate the changes into your node: - nothing: old code keeps working as before but emits a deprecated warning - delete: remove the pallet benchmarking code from your node and use the omni-bencher instead - patch: apply the patch below and keep using as currently. This emits a deprecated warning at runtime, since it uses the old way to generate a genesis state, but is the smallest change. ```patch runner.sync_run(|config| cmd - .run::<HashingFor<Block>, ReclaimHostFunctions>(config) + .run_with_spec::<HashingFor<Block>, ReclaimHostFunctions>(Some(config.chain_spec)) ) ``` ## 4. Maintenance Change - `pallet-nis` get a `BenchmarkSetup` config item to prepare its counterparty asset. - Add percent progress print when running benchmarks. - Dont immediately exit on benchmark error but try to run as many as possible and print errors last. --------- Signed-off-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io> Co-authored-by: Liam Aharon <liam.aharon@hotmail.com>
3.5 KiB
The FRAME Benchmarking CLI
This crate contains commands to benchmark various aspects of Substrate and the hardware.
The goal is to have a comprehensive suite of benchmarks that cover all aspects of Substrate and the hardware that its
running on.
There exist fundamentally two ways to use this crate. A node-integrated CLI version, and a freestanding CLI. If you are
only interested in pallet benchmarking, then skip ahead to the Freestanding CLI.
Node Integrated CLI
Mostly all Substrate nodes will expose some commands for benchmarking. You can refer to the staging-node-cli crate as
an example on how to integrate those. Note that for solely benchmarking pallets, the freestanding CLI is more suitable.
Usage
Here we invoke the root command on the staging-node-cli. Most Substrate nodes should have a similar output, depending
on their integration of these commands.
$ cargo run -p staging-node-cli --profile=production --features=runtime-benchmarks -- benchmark
Sub-commands concerned with benchmarking.
USAGE:
substrate benchmark <SUBCOMMAND>
OPTIONS:
-h, --help Print help information
-V, --version Print version information
SUBCOMMANDS:
block Benchmark the execution time of historic blocks
machine Command to benchmark the hardware.
overhead Benchmark the execution overhead per-block and per-extrinsic
pallet Benchmark the extrinsic weight of FRAME Pallets
storage Benchmark the storage speed of a chain snapshot
All examples use the production profile for correctness which makes the compilation very slow; for testing you can
use --release.
For the final results the production profile and reference hardware should be used, otherwise the results are not
comparable.
Freestanding CLI
The freestanding is a standalone CLI that does not rely on any node integration. It can be used to benchmark pallets of
any FRAME runtime that does not utilize 3rd party host functions.
It currently only supports pallet benchmarking, since the other commands still rely on a node.
Installation
Installing from local source repository:
cargo install --locked --path substrate/utils/frame/omni-bencher --profile=production
Usage
The exposed pallet sub-command is identical as the node-integrated CLI. The only difference is that it needs to be prefixed
with a v1 to ensure drop-in compatibility.
First we need to ensure that there is a runtime available. As example we will build the Westend runtime:
cargo build -p westend-runtime --profile production --features runtime-benchmarks
Now the benchmarking can be started with:
frame-omni-bencher v1 \
benchmark pallet \
--runtime target/release/wbuild/westend-runtime/westend-runtime.compact.compressed.wasm \
--pallet "pallet_balances" --extrinsic ""
For the exact arguments of the pallet command, please refer to the pallet sub-module.
Commands
The sub-commands of both CLIs have the same semantics and are documented in their respective sub-modules:
- block Compare the weight of a historic block to its actual resource usage
- machine Gauges the speed of the hardware
- overhead Creates weight files for the Block- and Extrinsic-base weights
- pallet Creates weight files for a Pallet
- storage Creates weight files for Read and Write storage operations
License: Apache-2.0