How to do performance profiling?

[XuJiandong]

Currently there are 3 choices:

1. Smart Contracts without syscalls
   It doesn't require any syscall. For example, a pure algorithm which doesn't call any syscall. You can choose
   ckb-vm-pprof directly.
2. Using ckb-standalone-debugger with option (--pprof)
3. Use ckb-vm-pprof as lib
   see example

// Steps to use pprof in Rust test cases:
// 1. add ckb-vm-pprof in Cargo.tom, e.g
// ckb-vm-pprof = { path = "../../../ckb-vm-pprof" }
// clone from https://github.com/nervosnetwork/ckb-vm-pprof
// and enable feature:
// #[features]
// #pprof = []
// 2. remove [ignore] and run the test case
// 3. after "build/rc_lock.pprof-result" is generated, use
// inferno-flamegraph build/rc_lock.pprof-result > build/rc_lock.svg
// to generate to final plot
#[test]
#[ignore]
#[cfg(feature = "pprof")]
fn test_pubkey_hash_on_wl_pprof() {
    let mut data_loader = DummyDataLoader::new();

    let mut config = TestConfig::new(IDENTITY_FLAGS_PUBKEY_HASH, true);
    config.scheme = TestScheme::OnWhiteList;

    let tx = gen_tx(&mut data_loader, &mut config);
    let tx = sign_tx(&mut data_loader, tx, &mut config);
    let resolved_tx = build_resolved_tx(&data_loader, &tx);

    let consensus = gen_consensus();
    let tx_env = gen_tx_env();
    let verifier =
        TransactionScriptsVerifier::new(&resolved_tx, &consensus, &data_loader, &tx_env);

    let mut pass = false;
    for (_, _, script_group) in verifier.groups() {
        let syscalls = verifier.generate_syscalls(ScriptVersion::V1, script_group);
        let result = ckb_vm_pprof::quick_start(syscalls, "../../build/rc_lock.debug", Default::default(), "../../build/rc_lock.pprof-result");
        result.expect("pass verification");
        pass = true;
        break;
    }
    assert!(pass);
    // verifier.set_debug_printer(debug_printer);
    // let verify_result = verifier.verify(MAX_CYCLES);
    // verify_result.expect("pass verification");
}

[XuJiandong]

Update: now ckb-debugger can support it by "--pprof"
https://github.com/nervosnetwork/ckb-standalone-debugger