Merge #5354

5354: Add opt-in mimalloc feature r=matklad a=ivan

This adds a `mimalloc` feature to use [mimalloc](https://github.com/microsoft/mimalloc) via [mimalloc_rust](https://github.com/purpleprotocol/mimalloc_rust), and a corresponding `cargo xtask install --server --mimalloc`.

In my tests on Linux, mimalloc seems to run consistently faster than jemalloc and uses only slightly more memory in `analysis-stats` on chalk. Also, builds with mimalloc produce a binary 3MB smaller than builds with jemalloc.

A summary of `env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/` runs on NixOS master on an Intel 4790K in VMware Workstation:

<table>
<tr>
<td></td><td>self-reported time</td><td>elapsed time</td><td>max RSS</td>
</tr>
<tr><td>glibc 2.30 run 1</td><td>225.1 sec</td><td>3:46.91</td><td>1075208</td></tr>
<tr><td>glibc 2.30 run 2</td><td>228.4 sec</td><td>3:50.13</td><td>1074996</td></tr>
<tr><td>jemalloc run 1</td><td>201.8 sec</td><td>3:23.03</td><td>1055960</td></tr>
<tr><td>jemalloc run 2</td><td>199.2 sec</td><td>3:20.41</td><td>1065040</td></tr>
<tr><td>mimalloc run 1</td><td>188.6 sec</td><td>3:09.77</td><td>1105584</td></tr>
<tr><td>mimalloc run 2</td><td>185.0 sec</td><td>3:06.23</td><td>1108132</td></tr>
<tr><td>mimalloc + lto run 1</td><td>160.7 sec</td><td>2:41.80</td><td>1106076</td></tr>
<tr><td>mimalloc + lto run 2</td><td>162.2 sec</td><td>2:43.31</td><td>1104268</td></tr>
</tr>
</table>

I included an `lto = true; codegen-units = 1` run out of curiosity, this PR doesn't enable it.

<details>
<summary>analysis-stats benchmark runs</summary>

## default

```
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.10s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 333.880345ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 10.176299461s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 214.968806927s, 0b allocated 0b resident
Total: 225.145114417s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 225.34
        System time (seconds): 1.49
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:46.91
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1075208
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 6
        Minor (reclaiming a frame) page faults: 294711
        Voluntary context switches: 365
        Involuntary context switches: 3273
        Swaps: 0
        File system inputs: 2904
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.10s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 332.711598ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 9.895020518s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 218.5001697s, 0b allocated 0b resident
Total: 228.39519833s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 228.26
        System time (seconds): 1.75
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:50.13
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1074996
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 9
        Minor (reclaiming a frame) page faults: 294748
        Voluntary context switches: 330
        Involuntary context switches: 1561
        Swaps: 0
        File system inputs: 12608
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
```

## jemalloc
```
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.11s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 356.090374ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 10.176550183s, 439mb allocated 465mb resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 191.607201827s, 993mb allocated 1036mb resident
Total: 201.783937913s, 993mb allocated 1036mb resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 201.07
        System time (seconds): 1.89
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:23.03
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1055960
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 357755
        Voluntary context switches: 240
        Involuntary context switches: 1889
        Swaps: 0
        File system inputs: 256
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.10s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 317.917622ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 9.902142185s, 439mb allocated 463mb resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 189.295317017s, 993mb allocated 1046mb resident
Total: 199.197555943s, 993mb allocated 1046mb resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 198.64
        System time (seconds): 1.67
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:20.41
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1065040
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 369013
        Voluntary context switches: 243
        Involuntary context switches: 2835
        Swaps: 0
        File system inputs: 0
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
```

## mimalloc
```
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.12s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 332.116806ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 9.796643695s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 178.82132362s, 0b allocated 0b resident
Total: 188.617975605s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 187.70
        System time (seconds): 1.97
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:09.77
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1105584
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 296481
        Voluntary context switches: 222
        Involuntary context switches: 1868
        Swaps: 0
        File system inputs: 256
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.13s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 320.046776ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 9.287690124s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 175.710939697s, 0b allocated 0b resident
Total: 184.998640033s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 184.38
        System time (seconds): 1.81
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 3:06.23
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1108132
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 297055
        Voluntary context switches: 374
        Involuntary context switches: 2374
        Swaps: 0
        File system inputs: 0
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
```

## mimalloc + lto
```
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.11s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 369.600196ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 7.572726834s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 153.090899101s, 0b allocated 0b resident
Total: 160.663635235s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 160.01
        System time (seconds): 1.70
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 2:41.80
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1106076
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 1
        Minor (reclaiming a frame) page faults: 296610
        Voluntary context switches: 209
        Involuntary context switches: 2798
        Swaps: 0
        File system inputs: 8
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
# env time -v cargo run --release -p rust-analyzer -- analysis-stats ../chalk/
    Finished release [optimized] target(s) in 0.10s
     Running `target/release/rust-analyzer analysis-stats ../chalk/`
[ERROR ra_project_model] cyclic dependency chalk-integration -> chalk-engine
[ERROR ra_project_model] cyclic dependency chalk-recursive -> chalk-integration
[ERROR ra_project_model] cyclic dependency chalk-solve -> chalk-integration
Database loaded 334.630658ms
Crates in this dir: 11
Total modules found: 159
Total declarations: 2631
Total functions: 1947
Item Collection: 7.71699197s, 0b allocated 0b resident
Total expressions: 57094
Expressions of unknown type: 2938 (5%)
Expressions of partially unknown type: 2427 (4%)
Type mismatches: 232
Inference: 154.50351318s, 0b allocated 0b resident
Total: 162.220513775s, 0b allocated 0b resident
        Command being timed: "cargo run --release -p rust-analyzer -- analysis-stats ../chalk/"
        User time (seconds): 161.52
        System time (seconds): 1.74
        Percent of CPU this job got: 99%
        Elapsed (wall clock) time (h:mm:ss or m:ss): 2:43.31
        Average shared text size (kbytes): 0
        Average unshared data size (kbytes): 0
        Average stack size (kbytes): 0
        Average total size (kbytes): 0
        Maximum resident set size (kbytes): 1104268
        Average resident set size (kbytes): 0
        Major (requiring I/O) page faults: 0
        Minor (reclaiming a frame) page faults: 296183
        Voluntary context switches: 200
        Involuntary context switches: 1666
        Swaps: 0
        File system inputs: 0
        File system outputs: 0
        Socket messages sent: 0
        Socket messages received: 0
        Signals delivered: 0
        Page size (bytes): 4096
        Exit status: 0
```
</details>

Co-authored-by: Ivan Kozik <[email protected]>

2 files changed, 25 insertions, 6 deletions

diff --git a/xtask/src/install.rs b/xtask/src/install.rs
index 9ba77a3aa..a0dc0c9c2 100644
--- a/xtask/src/install.rs
+++ b/xtask/src/install.rs
@@ -19,7 +19,13 @@ pub enum ClientOpt {
 }
 
 pub struct ServerOpt {
-    pub jemalloc: bool,
+    pub malloc: Malloc,
+}
+
+pub enum Malloc {
+    System,
+    Jemalloc,
+    Mimalloc,
 }
 
 impl InstallCmd {
@@ -130,8 +136,12 @@ fn install_server(opts: ServerOpt) -> Result<()> {
         )
     }
 
-    let jemalloc = if opts.jemalloc { "--features jemalloc" } else { "" };
-    let res = run!("cargo install --path crates/rust-analyzer --locked --force {}", jemalloc);
+    let malloc_feature = match opts.malloc {
+        Malloc::System => "",
+        Malloc::Jemalloc => "--features jemalloc",
+        Malloc::Mimalloc => "--features mimalloc",
+    };
+    let res = run!("cargo install --path crates/rust-analyzer --locked --force {}", malloc_feature);
 
     if res.is_err() && old_rust {
         eprintln!(
diff --git a/xtask/src/main.rs b/xtask/src/main.rs
index f447613d4..399ff7204 100644
--- a/xtask/src/main.rs
+++ b/xtask/src/main.rs
@@ -14,7 +14,7 @@ use pico_args::Arguments;
 use xtask::{
     codegen::{self, Mode},
     dist::run_dist,
-    install::{ClientOpt, InstallCmd, ServerOpt},
+    install::{ClientOpt, InstallCmd, Malloc, ServerOpt},
     not_bash::pushd,
     pre_commit, project_root,
     release::{PromoteCmd, ReleaseCmd},
@@ -46,6 +46,7 @@ FLAGS:
         --client-code    Install only VS Code plugin
         --server         Install only the language server
         --jemalloc       Use jemalloc for server
+        --mimalloc       Use mimalloc for server
     -h, --help           Prints help information
         "
                 );
@@ -61,13 +62,21 @@ FLAGS:
                 return Ok(());
             }
 
-            let jemalloc = args.contains("--jemalloc");
+            let malloc = match (args.contains("--jemalloc"), args.contains("--mimalloc")) {
+                (false, false) => Malloc::System,
+                (true, false) => Malloc::Jemalloc,
+                (false, true) => Malloc::Mimalloc,
+                (true, true) => {
+                    eprintln!("error: Cannot use both `--jemalloc` and `--mimalloc`");
+                    return Ok(());
+                }
+            };
 
             args.finish()?;
 
             InstallCmd {
                 client: if server { None } else { Some(ClientOpt::VsCode) },
-                server: if client_code { None } else { Some(ServerOpt { jemalloc }) },
+                server: if client_code { None } else { Some(ServerOpt { malloc }) },
             }
             .run()
         }