compiler-builtins subtree update

library/compiler-builtins/.github/workflows/main.yaml

@@ -1,6 +1,6 @@
 name: CI
 on:
-  push: { branches: [master] }
+  push: { branches: [main] }
   pull_request:
 
 concurrency:
@@ -89,7 +89,7 @@ jobs:
         - target: x86_64-unknown-linux-gnu
           os: ubuntu-24.04
         - target: x86_64-apple-darwin
-          os: macos-13
+          os: macos-15-intel
         - target: i686-pc-windows-msvc
           os: windows-2025
         - target: x86_64-pc-windows-msvc
@@ -239,6 +239,8 @@ jobs:
         include:
         - target: x86_64-unknown-linux-gnu
           os: ubuntu-24.04
+        - target: aarch64-unknown-linux-gnu
+          os: ubuntu-24.04-arm
     runs-on: ${{ matrix.os }}
     steps:
     - uses: actions/checkout@master
@@ -247,13 +249,13 @@ jobs:
     - name: Set up dependencies
       run: |
         sudo apt-get update
-        sudo apt-get install -y valgrind gdb libc6-dbg # Needed for iai-callgrind
+        sudo apt-get install -y valgrind gdb libc6-dbg # Needed for gungraun
         rustup update "$BENCHMARK_RUSTC" --no-self-update
         rustup default "$BENCHMARK_RUSTC"
-        # Install the version of iai-callgrind-runner that is specified in Cargo.toml
-        iai_version="$(cargo metadata --format-version=1 --features icount |
-           jq -r '.packages[] | select(.name == "iai-callgrind").version')"
-        cargo binstall -y iai-callgrind-runner --version "$iai_version"
+        # Install the version of gungraun-runner that is specified in Cargo.toml
+        gungraun_version="$(cargo metadata --format-version=1 --features icount |
+           jq -r '.packages[] | select(.name == "gungraun").version')"
+        cargo binstall -y gungraun-runner --version "$gungraun_version"
         sudo apt-get install valgrind
     - uses: Swatinem/rust-cache@v2
       with:

library/compiler-builtins/.github/workflows/publish.yaml

@@ -5,7 +5,7 @@ permissions:
   contents: write
 
 on:
-  push: { branches: [master] }
+  push: { branches: [main] }
 
 jobs:
   release-plz:

library/compiler-builtins/.github/workflows/rustc-pull.yml

@@ -17,7 +17,7 @@ jobs:
       zulip-stream-id: 219381
       zulip-topic: 'compiler-builtins subtree sync automation'
       zulip-bot-email: "[email protected]"
-      pr-base-branch: master
+      pr-base-branch: main
       branch-name: rustc-pull
     secrets:
       zulip-api-token: ${{ secrets.ZULIP_API_TOKEN }}

library/compiler-builtins/.gitignore

@@ -9,6 +9,7 @@ compiler-rt
 # Benchmark cache
 baseline-*
 iai-home
+gungraun-home
 
 # Temporary files
 *.bk

library/compiler-builtins/CONTRIBUTING.md

@@ -150,8 +150,8 @@ cargo bench --no-default-features \
 ```
 
 There are also benchmarks that check instruction count behind the `icount`
-feature. These require [`iai-callgrind-runner`] (via Cargo) and [Valgrind]
-to be installed, which means these only run on limited platforms.
+feature. These require [`gungraun-runner`] (via Cargo) and [Valgrind] to be
+installed, which means these only run on limited platforms.
 
 Instruction count benchmarks are run as part of CI to flag performance
 regresions.
@@ -163,7 +163,7 @@ cargo bench --no-default-features \
     --bench icount --bench mem_icount
 ```
 
-[`iai-callgrind-runner`]: https://crates.io/crates/iai-callgrind-runner
+[`gungraun-runner`]: https://crates.io/crates/gungraun-runner
 [Valgrind]: https://valgrind.org/
 
 ## Subtree synchronization

library/compiler-builtins/Cargo.toml

@@ -51,5 +51,6 @@ codegen-units = 1
 lto = "fat"
 
 [profile.bench]
-# Required for iai-callgrind
+# Required for gungraun
 debug = true
+strip = false

library/compiler-builtins/PUBLISHING.md

@@ -5,7 +5,7 @@ It's not great, but it works for now. PRs to improve this process would be
 greatly appreciated!
 
 1. Make sure you've got a clean working tree and it's updated with the latest
-   changes on `master`
+   changes on `main`
 2. Edit `Cargo.toml` to bump the version number
 3. Commit this change
 4. Run `git tag` to create a tag for this version

library/compiler-builtins/builtins-shim/Cargo.toml

@@ -11,7 +11,12 @@
 [package]
 name = "compiler_builtins"
 version = "0.1.160"
-authors = ["Jorge Aparicio <[email protected]>"]
+authors = [
+    "Alex Crichton <[email protected]>",
+    "Amanieu d'Antras <[email protected]>",
+    "Jorge Aparicio <[email protected]>",
+    "Trevor Gross <[email protected]>",
+]
 description = "Compiler intrinsics used by the Rust compiler."
 repository = "https://github.com/rust-lang/compiler-builtins"
 license = "MIT AND Apache-2.0 WITH LLVM-exception AND (MIT OR Apache-2.0)"

library/compiler-builtins/builtins-test/Cargo.toml

@@ -1,7 +1,6 @@
 [package]
 name = "builtins-test"
 version = "0.1.0"
-authors = ["Alex Crichton <[email protected]>"]
 edition = "2024"
 publish = false
 license = "MIT AND Apache-2.0 WITH LLVM-exception AND (MIT OR Apache-2.0)"
@@ -14,7 +13,7 @@ rand_xoshiro = "0.7"
 # To compare float builtins against
 rustc_apfloat = "0.2.3"
 # Really a dev dependency, but dev dependencies can't be optional
-iai-callgrind = { version = "0.15.2", optional = true }
+gungraun = { version = "0.17.0", optional = true }
 
 [dependencies.compiler_builtins]
 path = "../builtins-shim"
@@ -46,8 +45,8 @@ no-sys-f16-f64-convert = []
 # Skip tests that rely on f16 symbols being available on the system
 no-sys-f16 = ["no-sys-f16-f64-convert"]
 
-# Enable icount benchmarks (requires iai-callgrind and valgrind)
-icount = ["dep:iai-callgrind"]
+# Enable icount benchmarks (requires gungraun-runner and valgrind locally)
+icount = ["dep:gungraun"]
 
 # Enable report generation without bringing in more dependencies by default
 benchmarking-reports = ["criterion/plotters", "criterion/html_reports"]

library/compiler-builtins/builtins-test/benches/mem_icount.rs

@@ -1,11 +1,11 @@
-//! Benchmarks that use Callgrind (via `iai_callgrind`) to report instruction count metrics. This
+//! Benchmarks that use Callgrind (via `gungraun`) to report instruction count metrics. This
 //! is stable enough to be tested in CI.
 
 use std::hint::black_box;
 use std::{ops, slice};
 
 use compiler_builtins::mem::{memcmp, memcpy, memmove, memset};
-use iai_callgrind::{library_benchmark, library_benchmark_group, main};
+use gungraun::{library_benchmark, library_benchmark_group, main};
 
 const PAGE_SIZE: usize = 0x1000; // 4 kiB
 const MAX_ALIGN: usize = 512; // assume we may use avx512 operations one day
@@ -108,7 +108,7 @@ mod mcpy {
         ],
         setup = setup,
     )]
-    fn bench((len, mut dst, src): (usize, AlignedSlice, AlignedSlice)) {
+    fn bench_cpy((len, mut dst, src): (usize, AlignedSlice, AlignedSlice)) {
         unsafe {
             black_box(memcpy(
                 black_box(dst.as_mut_ptr()),
@@ -118,7 +118,7 @@ mod mcpy {
         }
     }
 
-    library_benchmark_group!(name = memcpy; benchmarks = bench);
+    library_benchmark_group!(name = memcpy; benchmarks = bench_cpy);
 }
 
 mod mset {
@@ -157,7 +157,7 @@ mod mset {
         ],
         setup = setup,
     )]
-    fn bench((len, mut dst): (usize, AlignedSlice)) {
+    fn bench_set((len, mut dst): (usize, AlignedSlice)) {
         unsafe {
             black_box(memset(
                 black_box(dst.as_mut_ptr()),
@@ -167,7 +167,7 @@ mod mset {
         }
     }
 
-    library_benchmark_group!(name = memset; benchmarks = bench);
+    library_benchmark_group!(name = memset; benchmarks = bench_set);
 }
 
 mod mcmp {
@@ -225,7 +225,7 @@ mod mcmp {
         ],
         setup = setup
     )]
-    fn bench((len, mut dst, src): (usize, AlignedSlice, AlignedSlice)) {
+    fn bench_cmp((len, mut dst, src): (usize, AlignedSlice, AlignedSlice)) {
         unsafe {
             black_box(memcmp(
                 black_box(dst.as_mut_ptr()),
@@ -235,7 +235,7 @@ mod mcmp {
         }
     }
 
-    library_benchmark_group!(name = memcmp; benchmarks = bench);
+    library_benchmark_group!(name = memcmp; benchmarks = bench_cmp);
 }
 
 mod mmove {
@@ -384,7 +384,7 @@ mod mmove {
         ],
         setup = setup_forward
     )]
-    fn forward((len, spread, mut buf): (usize, usize, AlignedSlice)) {
+    fn forward_move((len, spread, mut buf): (usize, usize, AlignedSlice)) {
         // Test moving from the start of the buffer toward the end
         unsafe {
             black_box(memmove(
@@ -478,7 +478,7 @@ mod mmove {
         ],
         setup = setup_backward
     )]
-    fn backward((len, spread, mut buf): (usize, usize, AlignedSlice)) {
+    fn backward_move((len, spread, mut buf): (usize, usize, AlignedSlice)) {
         // Test moving from the end of the buffer toward the start
         unsafe {
             black_box(memmove(
@@ -489,7 +489,7 @@ mod mmove {
         }
     }
 
-    library_benchmark_group!(name = memmove; benchmarks = forward, backward);
+    library_benchmark_group!(name = memmove; benchmarks = forward_move, backward_move);
 }
 
 use mcmp::memcmp;

library/compiler-builtins/builtins-test/tests/lse.rs

@@ -19,7 +19,11 @@ mod cas {
                 let mut target = expected.wrapping_add(10);
                 assert_eq!(
                     unsafe {
-                        compiler_builtins::aarch64_linux::$name::$name(expected, new, &mut target)
+                        compiler_builtins::aarch64_outline_atomics::$name::$name(
+                            expected,
+                            new,
+                            &mut target,
+                        )
                     },
                     expected.wrapping_add(10),
                     "return value should always be the previous value",
@@ -33,7 +37,11 @@ mod cas {
                 target = expected;
                 assert_eq!(
                     unsafe {
-                        compiler_builtins::aarch64_linux::$name::$name(expected, new, &mut target)
+                        compiler_builtins::aarch64_outline_atomics::$name::$name(
+                            expected,
+                            new,
+                            &mut target,
+                        )
                     },
                     expected
                 );
@@ -54,7 +62,9 @@ mod swap {
             builtins_test::fuzz_2(10000, |left: super::int_ty!($bytes), mut right| {
                 let orig_right = right;
                 assert_eq!(
-                    unsafe { compiler_builtins::aarch64_linux::$name::$name(left, &mut right) },
+                    unsafe {
+                        compiler_builtins::aarch64_outline_atomics::$name::$name(left, &mut right)
+                    },
                     orig_right
                 );
                 assert_eq!(left, right);
@@ -74,7 +84,7 @@ macro_rules! test_op {
                             let mut target = old;
                             let op: fn(super::int_ty!($bytes), super::int_ty!($bytes)) -> _ = $($op)*;
                             let expected = op(old, val);
-                            assert_eq!(old, unsafe { compiler_builtins::aarch64_linux::$name::$name(val, &mut target) }, "{} should return original value", stringify!($name));
+                            assert_eq!(old, unsafe { compiler_builtins::aarch64_outline_atomics::$name::$name(val, &mut target) }, "{} should return original value", stringify!($name));
                             assert_eq!(expected, target, "{} should store to target", stringify!($name));
                         });
                     }

library/compiler-builtins/ci/bench-icount.sh

@@ -10,45 +10,53 @@ if [ -z "$target" ]; then
     target="$host_target"
 fi
 
-iai_home="iai-home"
+# Print machine information
+uname -a
+lscpu || true
+
+gungraun_home="gungraun-home"
 
 # Use the arch as a tag to disambiguate artifacts
 tag="$(echo "$target" | cut -d'-' -f1)"
 
-# Download the baseline from master
+# Download the baseline from main
 ./ci/ci-util.py locate-baseline --download --extract --tag "$tag"
 
+# FIXME: migration from iai-named baselines to gungraun, can be dropped
+# after the first run with gungraun.
+[ -d "iai-home" ] && mv "iai-home" "$gungraun_home"
+
 # Run benchmarks once
 function run_icount_benchmarks() {
     cargo_args=(
-        "--bench" "icount"
+        "--bench" "*icount*"
         "--no-default-features"
         "--features" "unstable,unstable-float,icount"
     )
 
-    iai_args=(
-        "--home" "$(pwd)/$iai_home"
-        "--callgrind-limits=ir=5.0"
+    gungraun_args=(
+        "--home" "$(pwd)/$gungraun_home"
+        "--callgrind-limits=ir=5.0%"
         "--save-summary"
     )
 
-    # Parse `cargo_arg0 cargo_arg1 -- iai_arg0 iai_arg1` syntax
-    parsing_iai_args=0
+    # Parse `cargo_arg0 cargo_arg1 -- gungraun_arg0 gungraun_arg1` syntax
+    parsing_gungraun_args=0
     while [ "$#" -gt 0 ]; do
-        if [ "$parsing_iai_args" == "1" ]; then
-            iai_args+=("$1")
+        if [ "$parsing_gungraun_args" == "1" ]; then
+            gungraun_args+=("$1")
         elif [ "$1" == "--" ]; then
-            parsing_iai_args=1
+            parsing_gungraun_args=1
         else
             cargo_args+=("$1")
         fi
 
         shift
     done
 
-    # Run iai-callgrind benchmarks. Do this in a subshell with `&& true` to
-    # capture rather than exit on error.
-    (cargo bench "${cargo_args[@]}" -- "${iai_args[@]}") && true
+    # Run gungraun benchmarks. Do this in a subshell with `&& true` to capture
+    # rather than exit on error.
+    (cargo bench "${cargo_args[@]}" -- "${gungraun_args[@]}") && true
     exit_code="$?"
 
     if [ "$exit_code" -eq 0 ]; then
@@ -68,4 +76,4 @@ run_icount_benchmarks -- --save-baseline=hardfloat
 # Name and tar the new baseline
 name="baseline-icount-$tag-$(date -u +'%Y%m%d%H%M')-${GITHUB_SHA:0:12}"
 echo "BASELINE_NAME=$name" >>"$GITHUB_ENV"
-tar cJf "$name.tar.xz" "$iai_home"
+tar cJf "$name.tar.xz" "$gungraun_home"