Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

assorted performance optimizations #6724

Closed
wants to merge 8 commits into from
Closed

assorted performance optimizations #6724

wants to merge 8 commits into from

Conversation

thestinger
Copy link
Contributor

Passing higher alignment values gives the optimization passes more freedom since it can copy in larger chunks. This change results in rustc outputting the same post-optimization IR as clang for swaps and most copies excluding the lack of information about padding.

Code snippet:

#[inline(never)]
fn swap<T>(x: &mut T, y: &mut T) {
    util::swap(x, y);
}

Original IR (for int):

define internal fastcc void @_ZN9swap_283417_a71830ca3ed2d65d3_00E(i64*, i64*) #1 {
static_allocas:
  %2 = icmp eq i64* %0, %1
  br i1 %2, label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit, label %3

; <label>:3                                       ; preds = %static_allocas
  %4 = load i64* %0, align 1
  %5 = load i64* %1, align 1
  store i64 %5, i64* %0, align 1
  store i64 %4, i64* %1, align 1
  br label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit

_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit:  ; preds = %3, %static_allocas
  ret void
}

After #6710:

define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 1
  %3 = load i64* %1, align 1
  store i64 %3, i64* %0, align 1
  store i64 %2, i64* %1, align 1
  ret void
}

After this change:

define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 8
  %3 = load i64* %1, align 8
  store i64 %3, i64* %0, align 8
  store i64 %2, i64* %1, align 8
  ret void
}

Another example:

#[inline(never)]
fn set<T>(x: &mut T, y: T) {
    *x = y;
}

Before, with (int, int) (align 1):

define internal fastcc void @_ZN8set_282517_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 1, i1 false)
  ret void
}

After, with (int, int) (align 8):

define internal fastcc void @_ZN8set_282617_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
  ret void
}

@thestinger
Copy link
Contributor Author

Note that since #6742 landed, the parameters are also marked as noalias so the IR is strictly better than clang, but it results in pretty much the same assembly for these cases.

@graydon
Copy link
Contributor

graydon commented May 27, 2013

This is wonderful. Thanks.

bors added a commit that referenced this pull request May 27, 2013
Passing higher alignment values gives the optimization passes more freedom since it can copy in larger chunks. This change results in rustc outputting the same post-optimization IR as clang for swaps and most copies excluding the lack of information about padding.

Code snippet:

```rust
#[inline(never)]
fn swap<T>(x: &mut T, y: &mut T) {
    util::swap(x, y);
}
```

Original IR (for `int`):

```llvm
define internal fastcc void @_ZN9swap_283417_a71830ca3ed2d65d3_00E(i64*, i64*) #1 {
static_allocas:
  %2 = icmp eq i64* %0, %1
  br i1 %2, label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit, label %3

; <label>:3                                       ; preds = %static_allocas
  %4 = load i64* %0, align 1
  %5 = load i64* %1, align 1
  store i64 %5, i64* %0, align 1
  store i64 %4, i64* %1, align 1
  br label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit

_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit:  ; preds = %3, %static_allocas
  ret void
}
```

After #6710:

```llvm
define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 1
  %3 = load i64* %1, align 1
  store i64 %3, i64* %0, align 1
  store i64 %2, i64* %1, align 1
  ret void
}
```

After this change:

```llvm
define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 8
  %3 = load i64* %1, align 8
  store i64 %3, i64* %0, align 8
  store i64 %2, i64* %1, align 8
  ret void
}
```

Another example:

```rust
#[inline(never)]
fn set<T>(x: &mut T, y: T) {
    *x = y;
}
```

Before, with `(int, int)` (align 1):

```llvm
define internal fastcc void @_ZN8set_282517_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 1, i1 false)
  ret void
}
```

After, with `(int, int)` (align 8):

```llvm
define internal fastcc void @_ZN8set_282617_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
  ret void
}
```
@bors bors closed this May 27, 2013
flip1995 pushed a commit to flip1995/rust that referenced this pull request Mar 11, 2021
Refactor types lints

Ref rust-lang#6724.
As described in rust-lang#6724, `types.rs` contains many groups inside it.
In this PR, I reorganize the lints of the `types` group into their own modules.

changelog: none
flip1995 pushed a commit to flip1995/rust that referenced this pull request Mar 11, 2021
Refactor casts lint

Ref: rust-lang#6724

Changes:
1. Separate the `casts` group from the `types` group.
2. Reorganize the lints of the `casts` group into their own modules.

Notes:
1. I didn't `fix` rust-lang#6874 in order to maintain this PR as small as possible.

---
changelog: none
flip1995 pushed a commit to flip1995/rust that referenced this pull request Mar 25, 2021
Refactor unit types

Ref:  rust-lang#6724
r? `@flip1995`

Changes:
1. Extract `unit_types` from `types` group.
2. Move lints of `unit_types` to their own modules.

Notes:
Other lints of `unit_types` is still scattered around the `clippy_lints`, e.g. `result_unit_err` or `option_map_unit_fn`.
These should be addressed in another PR.

changelog: none
flip1995 pushed a commit to flip1995/rust that referenced this pull request Apr 8, 2021
Refactor types

r? `@flip1995`
This is the last PR to close rust-lang#6724 🎉
Also, this fixes rust-lang#6936.

changelog: `vec_box`: Fix FN in `const` or `static`
changelog: `linkedlist`: Fix FN in `const` or `static`
changelog: `option_option`: Fix FN in `const` or `static`
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants