rust-release-profile

name	rust-release-profile
description	Configure release builds for maximum performance with LTO, optimizations, and binary stripping. Use for production deployments.

Release Profile

Optimize Rust builds for production performance and binary size.

Production Cargo.toml Profile

[profile.release]
opt-level = 3           # Maximum optimization
lto = true              # Link-time optimization (slower build, faster binary)
codegen-units = 1       # Single codegen unit (better optimization)
strip = true            # Strip debug symbols
panic = 'abort'         # Smaller binary, no unwinding

[profile.release-with-debug]
inherits = "release"
debug = true            # Keep debug info for profiling
strip = false

Optimization Levels

# opt-level options:
# 0 - No optimization (fastest compile)
# 1 - Basic optimization
# 2 - Default release optimization
# 3 - Maximum optimization (default for release)
# "s" - Optimize for size
# "z" - Optimize for size, disable loop vectorization

opt-level = 3   # Best performance
opt-level = "s" # Smallest binary with good performance
opt-level = "z" # Smallest binary

Link-Time Optimization (LTO)

# LTO options:
# false - No LTO (fastest build)
# true/"fat" - Full LTO (slowest build, best performance)
# "thin" - Incremental LTO (good balance)

lto = true       # Best performance, slow build
lto = "thin"     # Good performance, faster build
lto = false      # Fastest build, no cross-crate optimization

Codegen Units

# Fewer units = better optimization, slower parallel compilation
codegen-units = 1    # Best optimization
codegen-units = 16   # Default, faster parallel builds

Debug Symbols

# For release builds
strip = true          # Remove all symbols
strip = "symbols"     # Remove symbols, keep sections
strip = "debuginfo"   # Remove debug info only
strip = "none"        # Keep everything

# Keep debug info for profiling
[profile.release-perf]
inherits = "release"
debug = true
strip = false

Panic Behavior

# 'unwind' - Default, allows catch_unwind, larger binary
# 'abort' - Smaller binary, faster, no unwinding

panic = 'abort'  # Production: smaller, faster
panic = 'unwind' # Development: catch panics, better debugging

Target-Specific Optimization

# Enable CPU-specific features
[build]
rustflags = ["-C", "target-cpu=native"]

# Or in .cargo/config.toml
[target.x86_64-unknown-linux-gnu]
rustflags = ["-C", "target-cpu=native"]

Profile for Different Use Cases

# Fast iteration during development
[profile.dev]
opt-level = 0
debug = true

# Quick release for testing
[profile.release-quick]
inherits = "release"
lto = false
codegen-units = 16

# Maximum performance
[profile.release]
opt-level = 3
lto = true
codegen-units = 1
strip = true
panic = 'abort'

# Profiling build
[profile.release-perf]
inherits = "release"
debug = true
strip = false

# Smallest binary
[profile.release-small]
inherits = "release"
opt-level = "z"
lto = true
codegen-units = 1
strip = true
panic = 'abort'

Workspace Settings

# In workspace Cargo.toml
[profile.release]
opt-level = 3
lto = true

# Override for specific packages
[profile.release.package.expensive-crate]
opt-level = 2  # Faster build for this crate

Build Commands

# Standard release build
cargo build --release

# With specific profile
cargo build --profile release-perf

# Check binary size
ls -la target/release/my-binary

# Strip binary manually (if not using strip = true)
strip target/release/my-binary

# Check for debug info
file target/release/my-binary

Measuring Impact

# Build time comparison
time cargo build --release
time cargo build --release --config 'profile.release.lto=false'

# Binary size comparison
cargo build --release
ls -la target/release/my-binary

cargo build --profile release-small
ls -la target/release-small/my-binary

# Performance comparison
hyperfine './target/release/my-binary' './target/release-quick/my-binary'

Guidelines

• Use lto = true and codegen-units = 1 for production
• Use strip = true and panic = 'abort' for smallest binaries
• Use --profile release-perf with debug symbols for profiling
• Use lto = "thin" for faster builds with good optimization
• Test different opt-level values for your workload
• Use target-cpu=native for deployment on known hardware

Examples

See hercules-local-algo/Cargo.toml for production profile.