Rework reserve assumes for Vec and String #119465

SUPERCILEX · 2023-12-31T04:51:54Z

Code:

pub fn zzz(v: &mut Vec<u8>, s: &[u8]) {
    v.reserve(s.len());
    v.extend_from_slice(s);
}

pub fn zzz_str(v: &mut String, s: &str) {
    v.reserve(s.len());
    v.push_str(s);
}

Before:

.section .text.tracing_tracy::zzz,"ax",@progbits
        .globl  tracing_tracy::zzz
        .p2align        4, 0x90
        .type   tracing_tracy::zzz,@function
tracing_tracy::zzz:
        .cfi_startproc
        push r15
        .cfi_def_cfa_offset 16
        push r14
        .cfi_def_cfa_offset 24
        push r12
        .cfi_def_cfa_offset 32
        push rbx
        .cfi_def_cfa_offset 40
        push rax
        .cfi_def_cfa_offset 48
        .cfi_offset rbx, -40
        .cfi_offset r12, -32
        .cfi_offset r14, -24
        .cfi_offset r15, -16
        mov rbx, rdx
        mov r15, rsi
        mov r14, rdi
        mov rax, qword ptr [rdi]
        mov r12, qword ptr [rdi + 16]
        sub rax, r12
        cmp rax, rdx
        jb .LBB20_1
        cmp rax, rbx
        jb .LBB20_3
.LBB20_4:
        mov rdi, qword ptr [r14 + 8]
        add rdi, r12
        mov rsi, r15
        mov rdx, rbx
        call qword ptr [rip + memcpy@GOTPCREL]
        add r12, rbx
        mov qword ptr [r14 + 16], r12
        add rsp, 8
        .cfi_def_cfa_offset 40
        pop rbx
        .cfi_def_cfa_offset 32
        pop r12
        .cfi_def_cfa_offset 24
        pop r14
        .cfi_def_cfa_offset 16
        pop r15
        .cfi_def_cfa_offset 8
        ret
.LBB20_1:
        .cfi_def_cfa_offset 48
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov rax, qword ptr [r14]
        mov r12, qword ptr [r14 + 16]
        sub rax, r12
        cmp rax, rbx
        jae .LBB20_4
.LBB20_3:
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov r12, qword ptr [r14 + 16]
        jmp .LBB20_4

.section .text.tracing_tracy::zzz_str,"ax",@progbits
        .globl  tracing_tracy::zzz_str
        .p2align        4, 0x90
        .type   tracing_tracy::zzz_str,@function
tracing_tracy::zzz_str:
        .cfi_startproc
        push r15
        .cfi_def_cfa_offset 16
        push r14
        .cfi_def_cfa_offset 24
        push r12
        .cfi_def_cfa_offset 32
        push rbx
        .cfi_def_cfa_offset 40
        push rax
        .cfi_def_cfa_offset 48
        .cfi_offset rbx, -40
        .cfi_offset r12, -32
        .cfi_offset r14, -24
        .cfi_offset r15, -16
        mov rbx, rdx
        mov r15, rsi
        mov r14, rdi
        mov rax, qword ptr [rdi]
        mov r12, qword ptr [rdi + 16]
        sub rax, r12
        cmp rax, rdx
        jb .LBB27_1
        cmp rax, rbx
        jb .LBB27_3
.LBB27_4:
        mov rdi, qword ptr [r14 + 8]
        add rdi, r12
        mov rsi, r15
        mov rdx, rbx
        call qword ptr [rip + memcpy@GOTPCREL]
        add r12, rbx
        mov qword ptr [r14 + 16], r12
        add rsp, 8
        .cfi_def_cfa_offset 40
        pop rbx
        .cfi_def_cfa_offset 32
        pop r12
        .cfi_def_cfa_offset 24
        pop r14
        .cfi_def_cfa_offset 16
        pop r15
        .cfi_def_cfa_offset 8
        ret
.LBB27_1:
        .cfi_def_cfa_offset 48
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov rax, qword ptr [r14]
        mov r12, qword ptr [r14 + 16]
        sub rax, r12
        cmp rax, rbx
        jae .LBB27_4
.LBB27_3:
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov r12, qword ptr [r14 + 16]
        jmp .LBB27_4

After:

.section .text.tracing_tracy::zzz_vec,"ax",@progbits
        .globl  tracing_tracy::zzz_vec
        .p2align        4, 0x90
        .type   tracing_tracy::zzz_vec,@function
tracing_tracy::zzz_vec:
        .cfi_startproc
        push r15
        .cfi_def_cfa_offset 16
        push r14
        .cfi_def_cfa_offset 24
        push r12
        .cfi_def_cfa_offset 32
        push rbx
        .cfi_def_cfa_offset 40
        push rax
        .cfi_def_cfa_offset 48
        .cfi_offset rbx, -40
        .cfi_offset r12, -32
        .cfi_offset r14, -24
        .cfi_offset r15, -16
        mov rbx, rdx
        mov r15, rsi
        mov r14, rdi
        mov rax, qword ptr [rdi]
        mov r12, qword ptr [rdi + 16]
        sub rax, r12
        cmp rax, rdx
        jb .LBB32_1
        test rbx, rbx
        js .LBB32_5
.LBB32_3:
        mov rdi, qword ptr [r14 + 8]
        add rdi, r12
        mov rax, rdi
        sub rax, r15
        neg rax
        mov rcx, rdi
        sub rcx, r15
        cmovbe rcx, rax
        cmp rcx, rbx
        jb .LBB32_6
        mov rsi, r15
        mov rdx, rbx
        call qword ptr [rip + memcpy@GOTPCREL]
        add r12, rbx
        mov qword ptr [r14 + 16], r12
        add rsp, 8
        .cfi_def_cfa_offset 40
        pop rbx
        .cfi_def_cfa_offset 32
        pop r12
        .cfi_def_cfa_offset 24
        pop r14
        .cfi_def_cfa_offset 16
        pop r15
        .cfi_def_cfa_offset 8
        ret
.LBB32_1:
        .cfi_def_cfa_offset 48
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov r12, qword ptr [r14 + 16]
        test rbx, rbx
        jns .LBB32_3
.LBB32_5:
        lea rdi, [rip + .L__unnamed_40]
        mov esi, 71
        call qword ptr [rip + core::panicking::panic_nounwind@GOTPCREL]
.LBB32_6:
        lea rdi, [rip + .L__unnamed_11]
        mov esi, 166
        call qword ptr [rip + core::panicking::panic_nounwind@GOTPCREL]

.section .text.tracing_tracy::zzz_str,"ax",@progbits
        .globl  tracing_tracy::zzz_str
        .p2align        4, 0x90
        .type   tracing_tracy::zzz_str,@function
tracing_tracy::zzz_str:
        .cfi_startproc
        push r15
        .cfi_def_cfa_offset 16
        push r14
        .cfi_def_cfa_offset 24
        push r12
        .cfi_def_cfa_offset 32
        push rbx
        .cfi_def_cfa_offset 40
        push rax
        .cfi_def_cfa_offset 48
        .cfi_offset rbx, -40
        .cfi_offset r12, -32
        .cfi_offset r14, -24
        .cfi_offset r15, -16
        mov rbx, rdx
        mov r15, rsi
        mov r14, rdi
        mov rax, qword ptr [rdi]
        mov r12, qword ptr [rdi + 16]
        sub rax, r12
        cmp rax, rdx
        jb .LBB33_1
        test rbx, rbx
        js .LBB33_5
.LBB33_3:
        mov rdi, qword ptr [r14 + 8]
        add rdi, r12
        mov rax, rdi
        sub rax, r15
        neg rax
        mov rcx, rdi
        sub rcx, r15
        cmovbe rcx, rax
        cmp rcx, rbx
        jb .LBB33_6
        mov rsi, r15
        mov rdx, rbx
        call qword ptr [rip + memcpy@GOTPCREL]
        add r12, rbx
        mov qword ptr [r14 + 16], r12
        add rsp, 8
        .cfi_def_cfa_offset 40
        pop rbx
        .cfi_def_cfa_offset 32
        pop r12
        .cfi_def_cfa_offset 24
        pop r14
        .cfi_def_cfa_offset 16
        pop r15
        .cfi_def_cfa_offset 8
        ret
.LBB33_1:
        .cfi_def_cfa_offset 48
        mov rdi, r14
        mov rsi, r12
        mov rdx, rbx
        call alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle
        mov r12, qword ptr [r14 + 16]
        test rbx, rbx
        jns .LBB33_3
.LBB33_5:
        lea rdi, [rip + .L__unnamed_40]
        mov esi, 71
        call qword ptr [rip + core::panicking::panic_nounwind@GOTPCREL]
.LBB33_6:
        lea rdi, [rip + .L__unnamed_11]
        mov esi, 166
        call qword ptr [rip + core::panicking::panic_nounwind@GOTPCREL]

rustbot · 2023-12-31T04:52:03Z

r? @m-ou-se

(rustbot has picked a reviewer for you, use r? to override)

SUPERCILEX · 2023-12-31T04:54:47Z

The optimization only works with the inlines and assumes in the top function (unless you inline(always) the intermediate functions which probably isn't desirable).

Signed-off-by: Alex Saveau <saveau.alexandre@gmail.com>

This reverts commit 1dfff2e.

bors · 2024-01-11T22:56:39Z

☔ The latest upstream changes (presumably #119864) made this pull request unmergeable. Please resolve the merge conflicts.

m-ou-se · 2024-02-15T11:34:15Z

unless you inline(always) the intermediate functions which probably isn't desirable

The functions are tiny though. Does #[inline(always)] produce any negative results?

SUPERCILEX · 2024-02-16T08:27:26Z

Maybe this is actually an LLVM bug? I just tried putting inline(always) on every single method in the call stack and it still doesn't optimize out the reserve. The only thing that works is putting an assert_unchecked inside the top-level reserve.

JohnCSimon · 2024-04-14T17:58:17Z

@SUPERCILEX

ping from triage - can you post your status on this PR? This PR has not received an update in a few months.

FYI: when a PR is ready for review, send a message containing
@rustbot ready to switch to S-waiting-on-review so the PR is in the reviewer's backlog.

Or if you're not going to continue, please close it. Thank you!

SUPERCILEX · 2024-04-14T20:39:56Z

Yeah, after all the fiddling I've done I think this is really just an LLVM bug, so I filed: #123944.

rustbot assigned m-ou-se Dec 31, 2023

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Dec 31, 2023

This comment has been minimized.

Sign in to view

SUPERCILEX force-pushed the reserve-assume branch from 496e867 to 48a433c Compare December 31, 2023 06:16

SUPERCILEX changed the title ~~Add missing reserve assume for LLVM~~ Rework reserve assumes for Vec and String Dec 31, 2023

This comment has been minimized.

Sign in to view

SUPERCILEX force-pushed the reserve-assume branch from 48a433c to d46ba1d Compare December 31, 2023 06:34

This comment has been minimized.

Sign in to view

SUPERCILEX force-pushed the reserve-assume branch from d46ba1d to 1dfff2e Compare December 31, 2023 17:03

SUPERCILEX added 4 commits January 2, 2024 09:39

Add missing reserve assumes for Vec and String

6053d5a

Signed-off-by: Alex Saveau <saveau.alexandre@gmail.com>

Add codegen tests

f94a27d

Signed-off-by: Alex Saveau <saveau.alexandre@gmail.com>

Move to unchecked_sub as suggested by @the8472

692788a

Signed-off-by: Alex Saveau <saveau.alexandre@gmail.com>

Revert "Move to unchecked_sub as suggested by @the8472"

a4b629e

This reverts commit 1dfff2e.

SUPERCILEX force-pushed the reserve-assume branch from 1dfff2e to a4b629e Compare January 2, 2024 17:39

m-ou-se added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Feb 15, 2024

SUPERCILEX mentioned this pull request Apr 14, 2024

Missed optimization with reserve and extend_from_slice #123944

Open

SUPERCILEX closed this Apr 14, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Rework reserve assumes for Vec and String #119465

Rework reserve assumes for Vec and String #119465

SUPERCILEX commented Dec 31, 2023

rustbot commented Dec 31, 2023

SUPERCILEX commented Dec 31, 2023

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

bors commented Jan 11, 2024

m-ou-se commented Feb 15, 2024

SUPERCILEX commented Feb 16, 2024

JohnCSimon commented Apr 14, 2024

SUPERCILEX commented Apr 14, 2024

Rework reserve assumes for Vec and String #119465

Rework reserve assumes for Vec and String #119465

Conversation

SUPERCILEX commented Dec 31, 2023

rustbot commented Dec 31, 2023

SUPERCILEX commented Dec 31, 2023

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

bors commented Jan 11, 2024

m-ou-se commented Feb 15, 2024

SUPERCILEX commented Feb 16, 2024

JohnCSimon commented Apr 14, 2024

SUPERCILEX commented Apr 14, 2024