rustc_main/
main.rs

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
// We need this feature as it changes `dylib` linking behavior and allows us to link to `rustc_driver`.
#![feature(rustc_private)]

// A note about jemalloc: rustc uses jemalloc when built for CI and
// distribution. The obvious way to do this is with the `#[global_allocator]`
// mechanism. However, for complicated reasons (see
// https://github.com/rust-lang/rust/pull/81782#issuecomment-784438001 for some
// details) that mechanism doesn't work here. Also, we must use a consistent
// allocator across the rustc <-> llvm boundary, and `#[global_allocator]`
// wouldn't provide that.
//
// Instead, we use a lower-level mechanism. rustc is linked with jemalloc in a
// way such that jemalloc's implementation of `malloc`, `free`, etc., override
// the libc allocator's implementation. This means that Rust's `System`
// allocator, which calls `libc::malloc()` et al., is actually calling into
// jemalloc.
//
// A consequence of not using `GlobalAlloc` (and the `tikv-jemallocator` crate
// provides an impl of that trait, which is called `Jemalloc`) is that we
// cannot use the sized deallocation APIs (`sdallocx`) that jemalloc provides.
// It's unclear how much performance is lost because of this.
//
// As for the symbol overrides in `main` below: we're pulling in a static copy
// of jemalloc. We need to actually reference its symbols for it to get linked.
// The two crates we link to here, `std` and `rustc_driver`, are both dynamic
// libraries. So we must reference jemalloc symbols one way or another, because
// this file is the only object code in the rustc executable.
//
// NOTE: if you are reading this comment because you want to set a custom `global_allocator` for
// benchmarking, consider using the benchmarks in the `rustc-perf` collector suite instead:
// https://github.com/rust-lang/rustc-perf/blob/master/collector/README.md#profiling
//
// NOTE: if you are reading this comment because you want to replace jemalloc with another allocator
// to compare their performance, see
// https://github.com/rust-lang/rust/commit/b90cfc887c31c3e7a9e6d462e2464db1fe506175#diff-43914724af6e464c1da2171e4a9b6c7e607d5bc1203fa95c0ab85be4122605ef
// for an example of how to do so.

fn main() {
    // See the comment at the top of this file for an explanation of this.
    #[cfg(feature = "jemalloc")]
    {
        use std::os::raw::{c_int, c_void};

        #[used]
        static _F1: unsafe extern "C" fn(usize, usize) -> *mut c_void = jemalloc_sys::calloc;
        #[used]
        static _F2: unsafe extern "C" fn(*mut *mut c_void, usize, usize) -> c_int =
            jemalloc_sys::posix_memalign;
        #[used]
        static _F3: unsafe extern "C" fn(usize, usize) -> *mut c_void = jemalloc_sys::aligned_alloc;
        #[used]
        static _F4: unsafe extern "C" fn(usize) -> *mut c_void = jemalloc_sys::malloc;
        #[used]
        static _F5: unsafe extern "C" fn(*mut c_void, usize) -> *mut c_void = jemalloc_sys::realloc;
        #[used]
        static _F6: unsafe extern "C" fn(*mut c_void) = jemalloc_sys::free;

        // On OSX, jemalloc doesn't directly override malloc/free, but instead
        // registers itself with the allocator's zone APIs in a ctor. However,
        // the linker doesn't seem to consider ctors as "used" when statically
        // linking, so we need to explicitly depend on the function.
        #[cfg(target_os = "macos")]
        {
            extern "C" {
                fn _rjem_je_zone_register();
            }

            #[used]
            static _F7: unsafe extern "C" fn() = _rjem_je_zone_register;
        }
    }

    rustc_driver::main()
}