Skip to content

Commit 6c4c8a6

Browse files
committed
add config_fast_builds.toml file + cranelift
1 parent 82bc17f commit 6c4c8a6

File tree

2 files changed

+180
-23
lines changed

2 files changed

+180
-23
lines changed

.cargo/config.toml

Lines changed: 0 additions & 23 deletions
This file was deleted.

.cargo/config_fast_builds.toml

Lines changed: 180 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,180 @@
1+
# Copy this file to `config.toml` to speed up your builds.
2+
#
3+
# # Faster linker
4+
#
5+
# One of the slowest aspects of compiling large Rust programs is the linking time. This file configures an
6+
# alternate linker that may improve build times. When choosing a new linker, you have two options:
7+
#
8+
# ## LLD
9+
#
10+
# LLD is a linker from the LLVM project that supports Linux, Windows, macOS, and Wasm. It has the greatest
11+
# platform support and the easiest installation process. It is enabled by default in this file for Linux
12+
# and Windows. On macOS, the default linker yields higher performance than LLD and is used instead.
13+
#
14+
# To install, please scroll to the corresponding table for your target (eg. `[target.x86_64-pc-windows-msvc]`
15+
# for Windows) and follow the steps under `LLD linker`.
16+
#
17+
# For more information, please see LLD's website at <https://lld.llvm.org>.
18+
#
19+
# ## Mold
20+
#
21+
# Mold is a newer linker written by one of the authors of LLD. It boasts even greater performance, specifically
22+
# through its high parallelism, though it only supports Linux.
23+
#
24+
# Mold is disabled by default in this file. If you wish to enable it, follow the installation instructions for
25+
# your corresponding target, disable LLD by commenting out its `-Clink-arg=...` line, and enable Mold by
26+
# *uncommenting* its `-Clink-arg=...` line.
27+
#
28+
# There is a fork of Mold named Sold that supports macOS, but it is unmaintained and is about the same speed as
29+
# the default ld64 linker. For this reason, it is not included in this file.
30+
#
31+
# For more information, please see Mold's repository at <https://github.com/rui314/mold>.
32+
#
33+
# # Nightly configuration
34+
#
35+
# Be warned that the following features require nightly Rust, which is experimental and may contain bugs. If you
36+
# are having issues, skip this section and use stable Rust instead.
37+
#
38+
# There are a few unstable features that can improve performance. To use them, first install nightly Rust
39+
# through Rustup:
40+
#
41+
# ```
42+
# rustup toolchain install nightly
43+
# ```
44+
#
45+
# Finally, uncomment the lines under the `Nightly` heading for your corresponding target table (eg.
46+
# `[target.x86_64-unknown-linux-gnu]` for Linux) to enable the following features:
47+
#
48+
# ## `share-generics`
49+
#
50+
# Usually rustc builds each crate separately, then combines them all together at the end. `share-generics` forces
51+
# crates to share monomorphized generic code, so they do not duplicate work.
52+
#
53+
# In other words, instead of crate 1 generating `Foo<String>` and crate 2 generating `Foo<String>` separately,
54+
# only one crate generates `Foo<String>` and the other adds on to the pre-existing work.
55+
#
56+
# Note that you may have some issues with this flag on Windows. If compiling fails due to the 65k symbol limit,
57+
# you may have to disable this setting. For more information and possible solutions to this error, see
58+
# <https://github.com/bevyengine/bevy/issues/1110>.
59+
#
60+
# ## `threads`
61+
#
62+
# This option enables rustc's parallel frontend, which improves performance when parsing, type checking, borrow
63+
# checking, and more. We currently set `threads=0`, which defaults to the amount of cores in your CPU.
64+
#
65+
# For more information, see the blog post at <https://blog.rust-lang.org/2023/11/09/parallel-rustc.html>.
66+
67+
[target.x86_64-unknown-linux-gnu]
68+
linker = "clang"
69+
rustflags = [
70+
# LLD linker
71+
#
72+
# You may need to install it:
73+
#
74+
# - Ubuntu: `sudo apt-get install lld clang`
75+
# - Fedora: `sudo dnf install lld clang`
76+
# - Arch: `sudo pacman -S lld clang`
77+
# "-Clink-arg=-fuse-ld=lld",
78+
79+
# Mold linker
80+
#
81+
# You may need to install it:
82+
#
83+
# - Ubuntu: `sudo apt-get install mold clang`
84+
# - Fedora: `sudo dnf install mold clang`
85+
# - Arch: `sudo pacman -S mold clang`
86+
"-Clink-arg=-fuse-ld=mold",
87+
88+
# Nightly
89+
"-Zshare-generics=y",
90+
"-Zthreads=0",
91+
]
92+
# Some systems may experience linker performance issues when running doc tests.
93+
# See https://github.com/bevyengine/bevy/issues/12207 for details.
94+
rustdocflags = [
95+
# LLD linker
96+
# "-Clink-arg=-fuse-ld=lld",
97+
98+
# Mold linker
99+
"-Clink-arg=-fuse-ld=mold",
100+
]
101+
102+
[target.x86_64-apple-darwin]
103+
rustflags = [
104+
# LLD linker
105+
#
106+
# The default ld64 linker is faster, you should continue using it instead.
107+
#
108+
# You may need to install it:
109+
#
110+
# Brew: `brew install llvm`
111+
# Manually: <https://lld.llvm.org/MachO/index.html>
112+
# "-Clink-arg=-fuse-ld=/usr/local/opt/llvm/bin/ld64.lld",
113+
114+
# Nightly
115+
"-Zshare-generics=y",
116+
"-Zthreads=0",
117+
]
118+
119+
[target.aarch64-apple-darwin]
120+
rustflags = [
121+
# LLD linker
122+
#
123+
# The default ld64 linker is faster, you should continue using it instead.
124+
#
125+
# You may need to install it:
126+
#
127+
# Brew: `brew install llvm`
128+
# Manually: <https://lld.llvm.org/MachO/index.html>
129+
# "-Clink-arg=-fuse-ld=/opt/homebrew/opt/llvm/bin/ld64.lld",
130+
131+
# Nightly
132+
"-Zshare-generics=y",
133+
"-Zthreads=0",
134+
]
135+
136+
[target.x86_64-pc-windows-msvc]
137+
# LLD linker
138+
#
139+
# You may need to install it:
140+
#
141+
# ```
142+
# cargo install -f cargo-binutils
143+
# rustup component add llvm-tools
144+
# ```
145+
linker = "rust-lld.exe"
146+
rustdocflags = ["-Clinker=rust-lld.exe"]
147+
rustflags = [
148+
# Nightly
149+
"-Zshare-generics=n", # This needs to be off if you use dynamic linking on Windows.
150+
"-Zthreads=0",
151+
]
152+
153+
# Optional: Uncommenting the following improves compile times, but reduces the amount of debug info to 'line number tables only'.
154+
# In most cases the gains are negligible, but if you are on macOS and have slow compile times you should see significant gains.
155+
# [profile.dev]
156+
# debug = 1
157+
158+
# Cranelift
159+
[unstable]
160+
codegen-backend = true
161+
162+
[profile]
163+
incremental = true
164+
165+
[profile.dev]
166+
codegen-backend = "cranelift"
167+
# If you want to attach a debugger, set this to true
168+
debug = "line-tables-only"
169+
170+
# Consider compiling deps with cranelift if you want cold-compilation to be faster
171+
[profile.dev.package."*"]
172+
codegen-backend = "cranelift"
173+
174+
# cranelift is `panic = abort`, so you need to compile with llvm to get `#[should_panic]` working
175+
[profile.test.package."*"]
176+
codegen-backend = "llvm"
177+
178+
# Disable cranelift for release profile
179+
[profile.release]
180+
codegen-backend = "llvm"

0 commit comments

Comments
 (0)