Rollup merge of #125165 - Oneirical:pgo-branch-weights, r=jieyouxu

Migrate `run-make/pgo-branch-weights` to `rmake`

Part of #121876 and the associated [Google Summer of Code project](https://blog.rust-lang.org/2024/05/01/gsoc-2024-selected-projects.html).

This is a scary one and I expect things to break. Set as draft, because this isn't ready.

- [x] There is this comment here, which suggests the test is excluded from the testing process due to a platform specific issue? I can't see anything here that would cause this test to not run...
> // FIXME(mati865): MinGW GCC miscompiles compiler-rt profiling library but with Clang it works
// properly. Since we only have GCC on the CI ignore the test for now."

EDIT: This is specific to Windows-gnu.

- [x] The Makefile has this line:
```
ifneq (,$(findstring x86,$(TARGET)))
COMMON_FLAGS=-Clink-args=-fuse-ld=gold
```
I honestly can't tell whether this is checking if the target IS x86, or IS NOT. EDIT: It's checking if it IS x86.

- [x] I don't know why the Makefile was trying to pass an argument directly in the Makefile instead of setting that "aaaaaaaaaaaa2bbbbbbbbbbbb2bbbbbbbbbbbbbbbbcc" input as a variable in the Rust program directly. I changed that, let me know if that was wrong.

- [x] Trying to rewrite `cat "$(TMPDIR)/interesting.ll" | "$(LLVM_FILECHECK)" filecheck-patterns.txt` resulted in some butchery. For starters, in `tools.mk`, LLVM_FILECHECK corrects its own backslashes on Windows distributions, but there is no further mention of it, so I assume this is a preset environment variable... but is it really? Then, the command itself uses a Standard Input and a passed input file as an argument simultaneously, according to the [documentation](https://llvm.org/docs/CommandGuide/FileCheck.html#synopsis).

try-job: x86_64-mingw
try-job: i686-mingw

Author: matthiaskrgr

src/tools/run-make-support/src/lib.rs

@@ -6,7 +6,7 @@
 pub mod cc;
 pub mod clang;
 pub mod diff;
-pub mod llvm_readobj;
+pub mod llvm;
 pub mod run;
 pub mod rustc;
 pub mod rustdoc;
@@ -23,8 +23,8 @@ pub use wasmparser;
 pub use cc::{cc, extra_c_flags, extra_cxx_flags, Cc};
 pub use clang::{clang, Clang};
 pub use diff::{diff, Diff};
-pub use llvm_readobj::{llvm_readobj, LlvmReadobj};
-pub use run::{run, run_fail};
+pub use llvm::{llvm_profdata, llvm_readobj, llvm_filecheck, LlvmProfdata, LlvmReadobj, LlvmFilecheck};
+pub use run::{run, run_fail, run_with_args};
 pub use rustc::{aux_build, rustc, Rustc};
 pub use rustdoc::{bare_rustdoc, rustdoc, Rustdoc};
 

src/tools/run-make-support/src/llvm.rs

@@ -0,0 +1,153 @@
+use std::env;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Output};
+use std::io::{BufReader, Read, Write};
+use std::fs::File;
+
+use crate::handle_failed_output;
+
+/// Construct a new `llvm-readobj` invocation. This assumes that `llvm-readobj` is available
+/// at `$LLVM_BIN_DIR/llvm-readobj`.
+pub fn llvm_readobj() -> LlvmReadobj {
+    LlvmReadobj::new()
+}
+
+/// Construct a new `llvm-profdata` invocation. This assumes that `llvm-profdata` is available
+/// at `$LLVM_BIN_DIR/llvm-profdata`.
+pub fn llvm_profdata() -> LlvmProfdata {
+    LlvmProfdata::new()
+}
+
+/// Construct a new `llvm-filecheck` invocation. This assumes that `llvm-filecheck` is available
+/// at `$LLVM_FILECHECK`.
+pub fn llvm_filecheck() -> LlvmFilecheck {
+    LlvmFilecheck::new()
+}
+
+/// A `llvm-readobj` invocation builder.
+#[derive(Debug)]
+pub struct LlvmReadobj {
+    cmd: Command,
+}
+
+/// A `llvm-profdata` invocation builder.
+#[derive(Debug)]
+pub struct LlvmProfdata {
+    cmd: Command,
+}
+
+/// A `llvm-filecheck` invocation builder.
+#[derive(Debug)]
+pub struct LlvmFilecheck {
+    cmd: Command,
+}
+
+crate::impl_common_helpers!(LlvmReadobj);
+
+/// Generate the path to the bin directory of LLVM.
+pub fn llvm_bin_dir() -> PathBuf {
+    let llvm_bin_dir = env::var("LLVM_BIN_DIR")
+    .expect("`LLVM_BIN_DIR` not specified, but this is required to find `llvm-readobj`");
+    PathBuf::from(llvm_bin_dir)
+}
+
+impl LlvmReadobj {
+    /// Construct a new `llvm-readobj` invocation. This assumes that `llvm-readobj` is available
+    /// at `$LLVM_BIN_DIR/llvm-readobj`.
+    pub fn new() -> Self {
+        let llvm_readobj = llvm_bin_dir().join("llvm-readobj");
+        let cmd = Command::new(llvm_readobj);
+        Self { cmd }
+    }
+
+    /// Provide an input file.
+    pub fn input<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        self.cmd.arg(path.as_ref());
+        self
+    }
+
+    /// Pass `--file-header` to display file headers.
+    pub fn file_header(&mut self) -> &mut Self {
+        self.cmd.arg("--file-header");
+        self
+    }
+
+    /// Get the [`Output`][::std::process::Output] of the finished process.
+    #[track_caller]
+    pub fn command_output(&mut self) -> Output {
+        self.cmd.output().expect("failed to get output of finished process")
+    }
+}
+
+impl LlvmProfdata {
+    /// Construct a new `llvm-profdata` invocation. This assumes that `llvm-profdata` is available
+    /// at `$LLVM_BIN_DIR/llvm-profdata`.
+    pub fn new() -> Self {
+        let llvm_profdata = llvm_bin_dir().join("llvm-profdata");
+        let cmd = Command::new(llvm_profdata);
+        Self { cmd }
+    }
+
+    /// Provide an input file.
+    pub fn input<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        self.cmd.arg("-o");
+        self.cmd.arg(path.as_ref());
+        self
+    }
+
+    /// Specify the output file path.
+    pub fn output<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        self.cmd.arg(path.as_ref());
+        self
+    }
+
+    /// Take several profile data files generated by PGO instrumentation and merge them
+    /// together into a single indexed profile data file.
+    pub fn merge(&mut self) -> &mut Self {
+        self.cmd.arg("merge");
+        self
+    }
+
+    /// Get the [`Output`][::std::process::Output] of the finished process.
+    #[track_caller]
+    pub fn command_output(&mut self) -> Output {
+        self.cmd.output().expect("failed to get output of finished process")
+    }
+}
+
+impl LlvmFilecheck {
+    /// Construct a new `llvm-filecheck` invocation. This assumes that `llvm-filecheck` is available
+    /// at `$LLVM_FILECHECK`.
+    pub fn new() -> Self {
+        let llvm_filecheck = env::var("LLVM_FILECHECK").expect("LLVM_FILECHECK env var not specified");
+        let cmd = Command::new(llvm_filecheck);
+        Self { cmd }
+    }
+
+    /// Pipe a file into standard input containing patterns that will be matched against the .patterns(path) call.
+    pub fn stdin<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        let file = File::open(path).unwrap();
+        let reader = BufReader::new(file);
+        let byte_vec = read_bytes(reader).expect("failed to read bytes of standard input");
+        let byte_slice = byte_vec.as_slice();
+        self.cmd.stdin(std::process::Stdio::piped());
+        let mut child = self.cmd.spawn().unwrap();
+        let mut stdin = child.stdin.take().unwrap();
+        stdin.write_all(byte_slice).unwrap();
+        stdin.flush().unwrap();
+        child.wait_with_output().unwrap();
+        self
+    }
+
+    /// Provide the patterns that need to be matched.
+    pub fn patterns<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        self.cmd.arg(path.as_ref());
+        self
+    }
+}
+
+fn read_bytes<R: Read>(mut reader: R) -> Result<Vec<u8>, std::io::Error> {
+    let mut buffer = Vec::new();
+    reader.read_to_end(&mut buffer)?;
+    Ok(buffer)
+}

src/tools/run-make-support/src/llvm_readobj.rs

@@ -6,12 +6,17 @@ use crate::is_windows;
 
 use super::{bin_name, handle_failed_output};
 
-fn run_common(name: &str) -> (Command, Output) {
+fn run_common(name: &str, args: Option<&[&str]>) -> (Command, Output) {
     let mut bin_path = PathBuf::new();
     bin_path.push(env::var("TMPDIR").unwrap());
     bin_path.push(&bin_name(name));
     let ld_lib_path_envvar = env::var("LD_LIB_PATH_ENVVAR").unwrap();
     let mut cmd = Command::new(bin_path);
+    if let Some(args) = args {
+        for arg in args {
+            cmd.arg(arg);
+        }
+    }
     cmd.env(&ld_lib_path_envvar, {
         let mut paths = vec![];
         paths.push(PathBuf::from(env::var("TMPDIR").unwrap()));
@@ -43,7 +48,20 @@ pub fn run(name: &str) -> Output {
     let caller_location = std::panic::Location::caller();
     let caller_line_number = caller_location.line();
 
-    let (cmd, output) = run_common(name);
+    let (cmd, output) = run_common(name, None);
+    if !output.status.success() {
+        handle_failed_output(&cmd, output, caller_line_number);
+    }
+    output
+}
+
+/// Run a built binary with one or more argument(s) and make sure it succeeds.
+#[track_caller]
+pub fn run_with_args(name: &str, args: &[&str]) -> Output {
+    let caller_location = std::panic::Location::caller();
+    let caller_line_number = caller_location.line();
+
+    let (cmd, output) = run_common(name, Some(args));
     if !output.status.success() {
         handle_failed_output(&cmd, output, caller_line_number);
     }
@@ -56,7 +74,7 @@ pub fn run_fail(name: &str) -> Output {
     let caller_location = std::panic::Location::caller();
     let caller_line_number = caller_location.line();
 
-    let (cmd, output) = run_common(name);
+    let (cmd, output) = run_common(name, None);
     if output.status.success() {
         handle_failed_output(&cmd, output, caller_line_number);
     }

src/tools/run-make-support/src/run.rs

@@ -131,6 +131,24 @@ impl Rustc {
         self
     }
 
+    /// Specify directory path used for profile generation
+    pub fn profile_generate<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        let mut arg = OsString::new();
+        arg.push("-Cprofile-generate=");
+        arg.push(path.as_ref());
+        self.cmd.arg(&arg);
+        self
+    }
+
+    /// Specify directory path used for profile usage
+    pub fn profile_use<P: AsRef<Path>>(&mut self, path: P) -> &mut Self {
+        let mut arg = OsString::new();
+        arg.push("-Cprofile-use=");
+        arg.push(path.as_ref());
+        self.cmd.arg(&arg);
+        self
+    }
+
     /// Specify error format to use
     pub fn error_format(&mut self, format: &str) -> &mut Self {
         self.cmd.arg(format!("--error-format={format}"));

src/tools/run-make-support/src/rustc.rs

@@ -193,7 +193,6 @@ run-make/pass-linker-flags/Makefile
 run-make/pass-non-c-like-enum-to-c/Makefile
 run-make/pdb-alt-path/Makefile
 run-make/pdb-buildinfo-cl-cmd/Makefile
-run-make/pgo-branch-weights/Makefile
 run-make/pgo-gen-lto/Makefile
 run-make/pgo-gen-no-imp-symbols/Makefile
 run-make/pgo-gen/Makefile

src/tools/tidy/src/allowed_run_make_makefiles.txt

@@ -0,0 +1,48 @@
+// This test generates an instrumented binary - a program which
+// will keep track of how many times it calls each function, a useful
+// feature for optimization. Then, an argument (aaaaaaaaaaaa2bbbbbbbbbbbb2bbbbbbbbbbbbbbbbcc)
+// is passed into the instrumented binary, which should react with a number of function calls
+// fully known in advance. (For example, the letter 'a' results in calling f1())
+
+// If the test passes, the expected function call count was added to the use-phase LLVM-IR.
+// See https://github.com/rust-lang/rust/pull/66631
+
+//@ needs-profiler-support
+//@ ignore-cross-compile
+
+// (This test has problems generating profdata on mingw. This could use further investigation.)
+//@ ignore-windows-gnu
+
+use run_make_support::{
+    llvm_filecheck, llvm_profdata, run_with_args, rustc, rustdoc, target, tmp_dir,
+};
+use std::fs;
+
+fn main() {
+    let path_prof_data_dir = tmp_dir().join("prof_data_dir");
+    let path_merged_profdata = path_prof_data_dir.join("merged.profdata");
+    rustc().input("opaque.rs").run();
+    fs::create_dir_all(&path_prof_data_dir);
+    rustc()
+        .input("interesting.rs")
+        .profile_generate(&path_prof_data_dir)
+        .opt()
+        .codegen_units(1)
+        .run();
+    rustc().input("main.rs").profile_generate(&path_prof_data_dir).opt().run();
+    run_with_args("main", &["aaaaaaaaaaaa2bbbbbbbbbbbb2bbbbbbbbbbbbbbbbcc"]);
+    llvm_profdata()
+        .merge()
+        .output(&path_merged_profdata)
+        .input(path_prof_data_dir)
+        .command_output();
+    rustc()
+        .input("interesting.rs")
+        .profile_use(path_merged_profdata)
+        .opt()
+        .codegen_units(1)
+        .emit("llvm-ir")
+        .run();
+    let interesting_ll = tmp_dir().join("interesting.ll");
+    llvm_filecheck().patterns("filecheck-patterns.txt").stdin(interesting_ll);
+}