mirror
/
rustc-dev-guide
mirror of https://github.com/rust-lang/rustc-dev-guide.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Did more measurements on what exactly affects llvm-lines: 

optimize, codegen-units and mir-opt do, but debug-assertions doesn't.
This commit is contained in:
Julian Wollersberger 2020-10-04 15:09:06 +02:00 committed by Joshua Nelson
parent 476dd40f63
commit 9a46f17fab
1 changed files with 13 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
src/profiling.md
View File

@ -21,7 +21,7 @@ Depending on what you're trying to measure, there are several different approach
eg. `cargo -Ztimings build`. eg. `cargo -Ztimings build`.
You can use this flag on the compiler itself with `CARGOFLAGS="-Ztimings" ./x.py build` You can use this flag on the compiler itself with `CARGOFLAGS="-Ztimings" ./x.py build`
## Optimizing rustc's self-compile-times with cargo-llvm-lines ## Optimizing rustc's bootstrap times with `cargo-llvm-lines`
Using [cargo-llvm-lines](https://github.com/dtolnay/cargo-llvm-lines) you can count the Using [cargo-llvm-lines](https://github.com/dtolnay/cargo-llvm-lines) you can count the
number of lines of LLVM IR across all instantiations of a generic function. number of lines of LLVM IR across all instantiations of a generic function.
@ -38,8 +38,8 @@ cargo install cargo-llvm-lines
RUSTFLAGS="--emit=llvm-ir" ./x.py build --stage 0 compiler/rustc RUSTFLAGS="--emit=llvm-ir" ./x.py build --stage 0 compiler/rustc
# Single crate, eg. rustc_middle # Single crate, eg. rustc_middle
cargo llvm-lines --files ./build/x86_64-unknown-linux-gnu/stage0-rustc/x86_64-unknown-linux-gnu/debug/deps/rustc_middle* > llvm-lines-middle.txt cargo llvm-lines --files ./build/x86_64-unknown-linux-gnu/stage0-rustc/x86_64-unknown-linux-gnu/debug/deps/rustc_middle-a539a639bdab6513.ll > llvm-lines-middle.txt
# Whole compiler at once # Specify all crates of the compiler. (Relies on the glob support of your shell.)
cargo llvm-lines --files ./build/x86_64-unknown-linux-gnu/stage0-rustc/x86_64-unknown-linux-gnu/debug/deps/*.ll > llvm-lines.txt cargo llvm-lines --files ./build/x86_64-unknown-linux-gnu/stage0-rustc/x86_64-unknown-linux-gnu/debug/deps/*.ll > llvm-lines.txt
``` ```
@ -72,7 +72,7 @@ you will be compiling rustc _a lot_.
I recommend changing a few settings in `config.toml` to make it bearable: I recommend changing a few settings in `config.toml` to make it bearable:
``` ```
[rust] [rust]
# A debug build takes _a fourth_ as long on my machine, # A debug build takes _a third_ as long on my machine,
# but compiling more than stage0 rustc becomes unbearably slow. # but compiling more than stage0 rustc becomes unbearably slow.
optimize = false optimize = false
@ -81,15 +81,14 @@ incremental = false
# We won't be running it, so no point in compiling debug checks. # We won't be running it, so no point in compiling debug checks.
debug = false debug = false
# Caution: This changes the output of llvm-lines. # Using a single codegen unit gives less output, but is slower to compile.
# Using a single codegen unit gives more accurate output, but is slower to compile. codegen-units = 0 # num_cpus
# Changing it to the number of cores on my machine increased the output
# from 3.5GB to 4.1GB and decreased compile times from 5½ min to 4 min.
codegen-units = 1
#codegen-units = 0 # num_cpus
``` ```
What I'm still not sure about is if inlining in MIR optimizations affect llvm-lines. The llvm-lines output is affected by several options.
The output with `-Zmir-opt-level=0` and `-Zmir-opt-level=1` is the same, `optimize = false` increases it from 2.1GB to 3.5GB and `codegen-units = 0` to 4.1GB.
but it feels like that some functions that show up at the top should be to small
to have such a high impact. Inlining should only happens in LLVM though. MIR optimizations have little impact. Compared to the default `RUSTFLAGS="-Zmir-opt-level=1"`,
level 0 adds 0.3GB and level 2 removes 0.2GB.
Inlining currently only happens in LLVM, but this might change in the future.