Skip to content

README.md

Latest commit

 

History

History
83 lines (59 loc) · 2.46 KB

README.md

File metadata and controls

83 lines (59 loc) · 2.46 KB

light-program-profiler

Profiler macros with custom syscalls for measuring CU and heap consumption in Solana programs.

Usage

Solana program

Annotate functions with #[profile]:

use light_program_profiler::profile;

#[profile]
pub fn my_function() -> u64 {
    let a: u64 = 1000;
    let b: u64 = 2000;
    a.wrapping_add(b)
}
[dependencies]
light-program-profiler = { git = "https://github.com/Lightprotocol/light-program-profiler", features = ["profile-program", "inline"] }

Test harness

Option 1: Mollusk (preferred)

[dev-dependencies]
light-program-profiler = { git = "https://github.com/Lightprotocol/light-program-profiler", features = ["mollusk"] }
mollusk-svm = "0.3.0"
use light_program_profiler::mollusk::{
    register_profiling_syscalls, take_profiling_results,
    extract_category_and_file, write_categorized_readme, ReadmeConfig,
};
use mollusk_svm::Mollusk;

let mut mollusk = Mollusk::default();
register_profiling_syscalls(&mut mollusk);
mollusk.add_program(&program_id, "my_program", &mollusk_svm::program::loader_keys::LOADER_V3);

mollusk.process_instruction(&instruction, &accounts);

for (func_name, cu_consumed, file_location) in take_profiling_results() {
    println!("{}: {} CU at {}", func_name, cu_consumed, file_location);
}

ReadmeConfig and write_categorized_readme generate benchmark README files from collected results. See the doc example on ReadmeConfig for the full workflow.

Option 2: LiteSVM with agave fork

Patch solana-program-runtime and solana-bpf-loader-program to use the Lightprotocol agave fork, which has the profiling syscalls built into the runtime:

[dev-dependencies]
litesvm = "0.7"

[patch.crates-io]
solana-program-runtime = { git = "https://github.com/Lightprotocol/agave", rev = "be34dc76559e14921a2c8610f2fa2402bf0684bb" }
solana-bpf-loader-program = { git = "https://github.com/Lightprotocol/agave", rev = "be34dc76559e14921a2c8610f2fa2402bf0684bb" }

Profiling results are written to transaction logs and must be parsed by the caller.

Features

  • profile-program -- enables #[profile] macro instrumentation
  • profile-heap -- includes heap usage tracking
  • inline -- adds #[inline(always)] to profiled functions
  • mollusk -- Mollusk test harness integration (syscall registration, README generation)

Solana versions

  • Solana program: 2.2+
  • Mollusk feature: solana-program-runtime 2.3+, mollusk-svm 0.3.0
  • Agave fork: 2.3