Profiles
Profiles provide a way to alter the compiler settings, influencing things like optimizations and debugging symbols.
Cargo has 4 built-in profiles: dev
, release
, test
, and bench
. The
profile is automatically chosen based on which command is being run if a
profile is not specified on the command-line. In addition to the built-in
profiles, custom user-defined profiles can also be specified.
Profile settings can be changed in Cargo.toml
with the
[profile]
table. Within each named profile, individual settings can be changed
with key/value pairs like this:
[profile.dev]
opt-level = 1 # Use slightly better optimizations.
overflow-checks = false # Disable integer overflow checks.
Cargo only looks at the profile settings in the Cargo.toml
manifest at the
root of the workspace. Profile settings defined in dependencies will be
ignored.
Additionally, profiles can be overridden from a config definition.
Specifying a profile in a config file or environment variable will override
the settings from Cargo.toml
.
Profile settings
The following is a list of settings that can be controlled in a profile.
opt-level
The opt-level
setting controls the -C opt-level
flag which controls the level
of optimization. Higher optimization levels may produce faster runtime code at
the expense of longer compiler times. Higher levels may also change and
rearrange the compiled code which may make it harder to use with a debugger.
The valid options are:
0
: no optimizations1
: basic optimizations2
: some optimizations3
: all optimizations"s"
: optimize for binary size"z"
: optimize for binary size, but also turn off loop vectorization.
It is recommended to experiment with different levels to find the right
balance for your project. There may be surprising results, such as level 3
being slower than 2
, or the "s"
and "z"
levels not being necessarily
smaller. You may also want to reevaluate your settings over time as newer
versions of rustc
changes optimization behavior.
See also Profile Guided Optimization for more advanced optimization techniques.
debug
The debug
setting controls the -C debuginfo
flag which controls the
amount of debug information included in the compiled binary.
The valid options are:
0
,false
, or"none"
: no debug info at all, default forrelease
"line-directives-only"
: line info directives only. For the nvptx* targets this enables profiling. For other use cases,line-tables-only
is the better, more compatible choice."line-tables-only"
: line tables only. Generates the minimal amount of debug info for backtraces with filename/line number info, but not anything else, i.e. no variable or function parameter info.1
or"limited"
: debug info without type or variable-level information. Generates more detailed module-level info thanline-tables-only
.2
,true
, or"full"
: full debug info, default fordev
For more information on what each option does see rustc
’s docs on debuginfo.
You may wish to also configure the split-debuginfo
option
depending on your needs as well.
split-debuginfo
The split-debuginfo
setting controls the -C split-debuginfo
flag which
controls whether debug information, if generated, is either placed in the
executable itself or adjacent to it.
This option is a string and acceptable values are the same as those the
compiler accepts. The default value for this option
is unpacked
on macOS for profiles that have debug information otherwise
enabled. Otherwise the default for this option is documented with rustc and is platform-specific. Some options are only
available on the nightly channel. The Cargo default may change in the future
once more testing has been performed, and support for DWARF is stabilized.
Be aware that Cargo and rustc have different defaults for this option. This option exists to allow Cargo to experiment on different combinations of flags thus providing better debugging and developer experience.
strip
The strip
option controls the -C strip
flag, which directs rustc to
strip either symbols or debuginfo from a binary. This can be enabled like so:
[package]
# ...
[profile.release]
strip = "debuginfo"
Possible string values of strip
are "none"
, "debuginfo"
, and "symbols"
.
The default is "none"
.
You can also configure this option with the boolean values true
or false
.
strip = true
is equivalent to strip = "symbols"
. strip = false
is
equivalent to strip = "none"
and disables strip
completely.
debug-assertions
The debug-assertions
setting controls the -C debug-assertions
flag which
turns cfg(debug_assertions)
conditional compilation on or off. Debug
assertions are intended to include runtime validation which is only available
in debug/development builds. These may be things that are too expensive or
otherwise undesirable in a release build. Debug assertions enables the
debug_assert!
macro in the standard library.
The valid options are:
true
: enabledfalse
: disabled
overflow-checks
The overflow-checks
setting controls the -C overflow-checks
flag which
controls the behavior of runtime integer overflow. When overflow-checks are
enabled, a panic will occur on overflow.
The valid options are:
true
: enabledfalse
: disabled
lto
The lto
setting controls rustc
’s -C lto
, -C linker-plugin-lto
, and
-C embed-bitcode
options, which control LLVM’s link time optimizations.
LTO can produce better optimized code, using whole-program analysis, at the cost
of longer linking time.
The valid options are:
false
: Performs “thin local LTO” which performs “thin” LTO on the local crate only across its codegen units. No LTO is performed if codegen units is 1 or opt-level is 0.true
or"fat"
: Performs “fat” LTO which attempts to perform optimizations across all crates within the dependency graph."thin"
: Performs “thin” LTO. This is similar to “fat”, but takes substantially less time to run while still achieving performance gains similar to “fat”."off"
: Disables LTO.
See the linker-plugin-lto chapter if you are interested in cross-language LTO.
This is not yet supported natively in Cargo, but can be performed via
RUSTFLAGS
.
panic
The panic
setting controls the -C panic
flag which controls which panic
strategy to use.
The valid options are:
"unwind"
: Unwind the stack upon panic."abort"
: Terminate the process upon panic.
When set to "unwind"
, the actual value depends on the default of the target
platform. For example, the NVPTX platform does not support unwinding, so it
always uses "abort"
.
Tests, benchmarks, build scripts, and proc macros ignore the panic
setting.
The rustc
test harness currently requires unwind
behavior. See the
panic-abort-tests
unstable flag which enables abort
behavior.
Additionally, when using the abort
strategy and building a test, all of the
dependencies will also be forced to build with the unwind
strategy.
incremental
The incremental
setting controls the -C incremental
flag which controls
whether or not incremental compilation is enabled. Incremental compilation
causes rustc
to save additional information to disk which will be reused
when recompiling the crate, improving re-compile times. The additional
information is stored in the target
directory.
The valid options are:
true
: enabledfalse
: disabled
Incremental compilation is only used for workspace members and “path” dependencies.
The incremental value can be overridden globally with the CARGO_INCREMENTAL
environment variable or the build.incremental
config variable.
codegen-units
The codegen-units
setting controls the -C codegen-units
flag which
controls how many “code generation units” a crate will be split into. More
code generation units allows more of a crate to be processed in parallel
possibly reducing compile time, but may produce slower code.
This option takes an integer greater than 0.
The default is 256 for incremental builds, and 16 for non-incremental builds.
rpath
The rpath
setting controls the -C rpath
flag which controls
whether or not rpath
is enabled.
Default profiles
dev
The dev
profile is used for normal development and debugging. It is the
default for build commands like cargo build
, and is used for cargo install --debug
.
The default settings for the dev
profile are:
[profile.dev]
opt-level = 0
debug = true
split-debuginfo = '...' # Platform-specific.
debug-assertions = true
overflow-checks = true
lto = false
panic = 'unwind'
incremental = true
codegen-units = 256
rpath = false
release
The release
profile is intended for optimized artifacts used for releases
and in production. This profile is used when the --release
flag is used, and
is the default for cargo install
.
The default settings for the release
profile are:
[profile.release]
opt-level = 3
debug = false
split-debuginfo = '...' # Platform-specific.
debug-assertions = false
overflow-checks = false
lto = false
panic = 'unwind'
incremental = false
codegen-units = 16
rpath = false
test
The test
profile is the default profile used by cargo test
.
The test
profile inherits the settings from the dev
profile.
bench
The bench
profile is the default profile used by cargo bench
.
The bench
profile inherits the settings from the release
profile.
Build Dependencies
To compile quickly, all profiles, by default, do not optimize build dependencies (build scripts, proc macros, and their dependencies), and avoid computing debug info when a build dependency is not used as a runtime dependency. The default settings for build overrides are:
[profile.dev.build-override]
opt-level = 0
codegen-units = 256
debug = false # when possible
[profile.release.build-override]
opt-level = 0
codegen-units = 256
However, if errors occur while running build dependencies, turning full debug info on will improve backtraces and debuggability when needed:
debug = true
Build dependencies otherwise inherit settings from the active profile in use, as described in Profile selection.
Custom profiles
In addition to the built-in profiles, additional custom profiles can be
defined. These may be useful for setting up multiple workflows and build
modes. When defining a custom profile, you must specify the inherits
key to
specify which profile the custom profile inherits settings from when the
setting is not specified.
For example, let’s say you want to compare a normal release build with a
release build with LTO optimizations, you can specify something like
the following in Cargo.toml
:
[profile.release-lto]
inherits = "release"
lto = true
The --profile
flag can then be used to choose this custom profile:
cargo build --profile release-lto
The output for each profile will be placed in a directory of the same name
as the profile in the target
directory. As in the example above, the
output would go into the target/release-lto
directory.
Profile selection
The profile used depends on the command, the command-line flags like
--release
or --profile
, and the package (in the case of
overrides). The default profile if none is specified is:
Command | Default Profile |
---|---|
cargo run , cargo build ,cargo check , cargo rustc | dev profile |
cargo test | test profile |
cargo bench | bench profile |
cargo install | release profile |
You can switch to a different profile using the --profile=NAME
option which will used the given profile.
The --release
flag is equivalent to --profile=release
.
The selected profile applies to all Cargo targets, including library, binary, example, test, and benchmark.
The profile for specific packages can be specified with overrides, described below.
Overrides
Profile settings can be overridden for specific packages and build-time
crates. To override the settings for a specific package, use the package
table to change the settings for the named package:
# The `foo` package will use the -Copt-level=3 flag.
[profile.dev.package.foo]
opt-level = 3
The package name is actually a Package ID Spec, so you can
target individual versions of a package with syntax such as
[profile.dev.package."foo:2.1.0"]
.
To override the settings for all dependencies (but not any workspace member),
use the "*"
package name:
# Set the default for dependencies.
[profile.dev.package."*"]
opt-level = 2
To override the settings for build scripts, proc macros, and their
dependencies, use the build-override
table:
# Set the settings for build scripts and proc-macros.
[profile.dev.build-override]
opt-level = 3
Note: When a dependency is both a normal dependency and a build dependency, Cargo will try to only build it once when
--target
is not specified. When usingbuild-override
, the dependency may need to be built twice, once as a normal dependency and once with the overridden build settings. This may increase initial build times.
The precedence for which value is used is done in the following order (first match wins):
[profile.dev.package.name]
— A named package.[profile.dev.package."*"]
— For any non-workspace member.[profile.dev.build-override]
— Only for build scripts, proc macros, and their dependencies.[profile.dev]
— Settings inCargo.toml
.- Default values built-in to Cargo.
Overrides cannot specify the panic
, lto
, or rpath
settings.
Overrides and generics
The location where generic code is instantiated will influence the optimization settings used for that generic code. This can cause subtle interactions when using profile overrides to change the optimization level of a specific crate. If you attempt to raise the optimization level of a dependency which defines generic functions, those generic functions may not be optimized when used in your local crate. This is because the code may be generated in the crate where it is instantiated, and thus may use the optimization settings of that crate.
For example, nalgebra is a library which defines vectors and matrices making
heavy use of generic parameters. If your local code defines concrete nalgebra
types like Vector4<f64>
and uses their methods, the corresponding nalgebra
code will be instantiated and built within your crate. Thus, if you attempt to
increase the optimization level of nalgebra
using a profile override, it may
not result in faster performance.
Further complicating the issue, rustc
has some optimizations where it will
attempt to share monomorphized generics between crates. If the opt-level is 2
or 3, then a crate will not use monomorphized generics from other crates, nor
will it export locally defined monomorphized items to be shared with other
crates. When experimenting with optimizing dependencies for development,
consider trying opt-level 1, which will apply some optimizations while still
allowing monomorphized items to be shared.