Open drahnr opened 2 years ago
Sure - here's a rough breakdown of how we optimize performance:
Across All OS's:
Ramdisk (if the user is using a hard disk) OR always if the user is using WSL (3x speed improvement from our benchmarks)
Sccache (cache pre-compiled crates, especially effective with speeding up incremental builds at multi-crate cargo workspaces)
Share-generics (might be made optional soon, as our issues indicate we're trying to move to stable
): -Zshare-generics=y
Increasing codegen-units
, turning opt-level
to 0
for debug builds.
Linux: LLD (We're looking to change this to mold soon) Linker Clang
Windows:
rust-lld.exe
MacOS: ZLD Linker -Csplit-debuginfo=unpacked
On a different note, cachepot
looks super cool 😄
+
Hey there, I am a contributor to sccache and initiator of cachepot and just stumbled upon this project.
Could you clarify how exactly the speedup gains are achieved? The README.md doesn't explain that and I see that under the hood it builds upon the
sccache
binary artifact.CC @Xanewok