Skip to content

Commit

Permalink
Rollup merge of rust-lang#118264 - lukas-code:optimized-draining, r=t…
Browse files Browse the repository at this point in the history 
…he8472

Optimize `VecDeque::drain` for (half-)open ranges

The most common use cases of `VecDeque::drain` consume either the entire queue or elements from the front or back.[^1] This PR makes these operations faster by optimizing the generated code of the destructor of the drain:

* `.drain(..)` is now the same as `.clear()`.
* `.drain(n..)` is now (almost[^2]) the same as `.truncate(n)`.
* `.drain(..n)` is now an efficient "advance" function. This operation is not provided by a dedicated function and optimizing it is my main motivation for this PR.

Previously, all of these cases generated a function call to the destructor of the `DropGuard`, emitting a lot of unused machine code as well as unnecessary branches and loads/stores of stack variables.

There are no algorithmic changes in this PR, but it simplifies the code enough to allow LLVM to recognize the special cases and optimize accordingly. Most notably, it allows elimination of the rather large [`wrap_copy`] function.

Some [rudimentary microbenchmarks][benches] show a performance improvement of **~3x-4x** on my machine for the special cases and roughly equal performance for the general case.

Best reviewed commit by commit.

[^1]: source: GitHub code search: [full range `drain(..)` = 7.5k results][full], [from front `drain(..n)` = 3.2k results][front], [from back `drain(n..)` = 1.6k results][back], [from middle `drain(n..m)` = <500 results][middle]

[^2]: `.drain(0..)` and `.clear()` reset the head to 0, but `.truncate(0)` does not.

[full]: https://github.com/search?type=code&q=%2FVecDeque%28.%7C%5Cn%29%2B%5C.drain%5C%280%3F%5C.%5C.%5C%29%2F+lang%3ARust
[front]: https://github.com/search?type=code&q=%2FVecDeque%28.%7C%5Cn%29%2B%5C.drain%5C%280%3F%5C.%5C.%5B%5E%29%5D.*%5C%29%2F+lang%3ARust
[back]: https://github.com/search?type=code&q=%2FVecDeque%28.%7C%5Cn%29%2B%5C.drain%5C%28%5B%5E0%5D.*%5C.%5C.%5C%29%2F+lang%3ARust
[middle]: https://github.com/search?type=code&q=%2FVecDeque%28.%7C%5Cn%29%2B%5C.drain%5C%28%5B%5E0%5D.*%5C.%5C.%5B%5E%29%5D.*%5C%29%2F+lang%3ARust
[`wrap_copy`]: https://github.com/rust-lang/rust/blob/4fd68eb47bad1c121417ac4450b2f0456150db86/library/alloc/src/collections/vec_deque/mod.rs#L262-L391
[benches]: https://gist.github.com/lukas-code/c97bd707d074c4cc31f241edbc7fd2a2

<details>
<summary>generated assembly</summary>

before:
```asm
clear:
	sub rsp, 40
	mov rax, qword ptr [rdi + 24]
	mov qword ptr [rdi + 24], 0
	mov qword ptr [rsp], rdi
	mov qword ptr [rsp + 8], rax
	xorps xmm0, xmm0
	movups xmmword ptr [rsp + 16], xmm0
	mov qword ptr [rsp + 32], rax
	test rax, rax
	je .LBB1_2
	mov rcx, qword ptr [rdi]
	mov rdx, qword ptr [rdi + 16]
	xor esi, esi
	cmp rdx, rcx
	cmovae rsi, rcx
	sub rdx, rsi
	mov rsi, rcx
	sub rsi, rdx
	lea rdi, [rdx + rax]
	cmp rsi, rax
	cmovb rdi, rcx
	sub rdi, rdx
	mov qword ptr [rsp + 16], rdi
	mov qword ptr [rsp + 32], 0
.LBB1_2:
	mov rdi, rsp
	call core::ptr::drop_in_place<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<i32,alloc::alloc::Global>>
	add rsp, 40
	ret

truncate:
	mov rax, qword ptr [rdi + 24]
	sub rax, rsi
	jbe .LBB2_2
	sub rsp, 40
	mov qword ptr [rdi + 24], rsi
	mov qword ptr [rsp], rdi
	mov qword ptr [rsp + 8], rax
	mov rcx, qword ptr [rdi]
	mov rdx, qword ptr [rdi + 16]
	add rdx, rsi
	xor edi, edi
	cmp rdx, rcx
	cmovae rdi, rcx
	mov qword ptr [rsp + 24], 0
	sub rdx, rdi
	mov rdi, rcx
	sub rdi, rdx
	lea r8, [rdx + rax]
	cmp rdi, rax
	cmovb r8, rcx
	sub rsi, rdx
	add rsi, r8
	mov qword ptr [rsp + 16], rsi
	mov qword ptr [rsp + 32], 0
	mov rdi, rsp
	call core::ptr::drop_in_place<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<i32,alloc::alloc::Global>>
	add rsp, 40

advance:
	mov rcx, qword ptr [rdi + 24]
	mov rax, rcx
	sub rax, rsi
	jbe .LBB3_1
	sub rsp, 40
	mov qword ptr [rdi + 24], 0
	mov qword ptr [rsp], rdi
	mov qword ptr [rsp + 8], rsi
	mov qword ptr [rsp + 16], 0
	mov qword ptr [rsp + 24], rax
	mov qword ptr [rsp + 32], rsi
	test rsi, rsi
	je .LBB3_6
	mov rax, qword ptr [rdi]
	mov rcx, qword ptr [rdi + 16]
	xor edx, edx
	cmp rcx, rax
	cmovae rdx, rax
	sub rcx, rdx
	mov rdx, rax
	sub rdx, rcx
	lea rdi, [rcx + rsi]
	cmp rdx, rsi
	cmovb rdi, rax
	sub rdi, rcx
	mov qword ptr [rsp + 16], rdi
	mov qword ptr [rsp + 32], 0
.LBB3_6:
	mov rdi, rsp
	call core::ptr::drop_in_place<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<i32,alloc::alloc::Global>>
	add rsp, 40
	ret
.LBB3_1:
	test rcx, rcx
	je .LBB3_3
	mov qword ptr [rdi + 24], 0
.LBB3_3:
	mov qword ptr [rdi + 16], 0
	ret

remove:
	sub rsp, 40
	cmp rdx, rsi
	jb .LBB4_5
	mov rax, qword ptr [rdi + 24]
	mov rcx, rax
	sub rcx, rdx
	jb .LBB4_6
	mov qword ptr [rdi + 24], rsi
	mov qword ptr [rsp], rdi
	sub rdx, rsi
	mov qword ptr [rsp + 8], rdx
	mov qword ptr [rsp + 16], rsi
	mov qword ptr [rsp + 24], rcx
	mov qword ptr [rsp + 32], rdx
	je .LBB4_4
	mov rax, qword ptr [rdi]
	mov rcx, qword ptr [rdi + 16]
	add rcx, rsi
	xor edi, edi
	cmp rcx, rax
	cmovae rdi, rax
	sub rcx, rdi
	mov rdi, rax
	sub rdi, rcx
	lea r8, [rcx + rdx]
	cmp rdi, rdx
	cmovb r8, rax
	sub rsi, rcx
	add rsi, r8
	mov qword ptr [rsp + 16], rsi
	mov qword ptr [rsp + 32], 0
.LBB4_4:
	mov rdi, rsp
	call core::ptr::drop_in_place<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<i32,alloc::alloc::Global>>
	add rsp, 40
	ret
.LBB4_5:
	lea rax, [rip + .L__unnamed_2]
	mov rdi, rsi
	mov rsi, rdx
	mov rdx, rax
	call qword ptr [rip + core::slice::index::slice_index_order_fail@GOTPCREL]
.LBB4_6:
	lea rcx, [rip + .L__unnamed_2]
	mov rdi, rdx
	mov rsi, rax
	mov rdx, rcx
	call qword ptr [rip + core::slice::index::slice_end_index_len_fail@GOTPCREL]

core::ptr::drop_in_place<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<i32,alloc::alloc::Global>>:
	push rbp
	push r15
	push r14
	push r13
	push r12
	push rbx
	sub rsp, 24
	mov rsi, qword ptr [rdi + 32]
	test rsi, rsi
	je .LBB0_2
	mov rax, qword ptr [rdi + 16]
	add rsi, rax
	jb .LBB0_45
.LBB0_2:
	mov r13, qword ptr [rdi]
	mov rbp, qword ptr [rdi + 8]
	mov rbx, qword ptr [r13 + 24]
	lea r12, [rbx + rbp]
	mov r15, qword ptr [rdi + 24]
	lea rsi, [r15 + r12]
	test rbx, rbx
	je .LBB0_10
	test r15, r15
	je .LBB0_42
	cmp rbx, r15
	jbe .LBB0_12
	mov r14, qword ptr [r13]
	mov rax, qword ptr [r13 + 16]
	add r12, rax
	xor ecx, ecx
	cmp r12, r14
	mov rdx, r14
	cmovb rdx, rcx
	sub r12, rdx
	add rbx, rax
	cmp rbx, r14
	cmovae rcx, r14
	sub rbx, rcx
	mov rcx, rbx
	sub rcx, r12
	je .LBB0_42
	mov rdi, qword ptr [r13 + 8]
	mov rax, rcx
	add rax, r14
	cmovae rax, rcx
	mov r8, r14
	sub r8, r12
	mov rcx, r14
	sub rcx, rbx
	mov rdx, r15
	sub rdx, r8
	mov qword ptr [rsp + 16], rsi
	jbe .LBB0_18
	cmp rax, r15
	jae .LBB0_24
	mov rdx, r15
	sub rdx, r8
	shl rdx, 2
	cmp r15, rcx
	jbe .LBB0_30
	sub r8, rcx
	mov qword ptr [rsp], rdi
	mov rax, qword ptr [rsp]
	lea rdi, [rax + 4*r8]
	mov rsi, qword ptr [rsp]
	mov qword ptr [rsp + 8], rcx
	mov r15, r8
	call qword ptr [rip + memmove@GOTPCREL]
	sub r14, r15
	mov rax, qword ptr [rsp]
	lea rsi, [rax + 4*r14]
	shl r15, 2
	mov rdi, qword ptr [rsp]
	mov rdx, r15
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, qword ptr [rsp]
	lea rsi, [rdi + 4*r12]
	lea rdi, [rdi + 4*rbx]
	mov r15, qword ptr [rsp + 8]
	jmp .LBB0_36
.LBB0_10:
	test r15, r15
	je .LBB0_17
	mov rax, qword ptr [r13]
	sub rsi, rbp
	add rbp, qword ptr [r13 + 16]
	xor ecx, ecx
	cmp rbp, rax
	cmovae rcx, rax
	sub rbp, rcx
	mov qword ptr [r13 + 16], rbp
	jmp .LBB0_43
.LBB0_12:
	mov rdx, qword ptr [r13 + 16]
	mov r15, qword ptr [r13]
	lea rax, [rdx + rbp]
	xor ecx, ecx
	cmp rax, r15
	cmovae rcx, r15
	mov r12, rax
	sub r12, rcx
	mov rcx, r12
	sub rcx, rdx
	je .LBB0_41
	mov rdi, qword ptr [r13 + 8]
	mov rax, rcx
	add rax, r15
	cmovae rax, rcx
	mov r8, r15
	sub r8, rdx
	mov rcx, r15
	sub rcx, r12
	mov r14, rbx
	sub r14, r8
	mov qword ptr [rsp + 16], rsi
	jbe .LBB0_21
	cmp rax, rbx
	jae .LBB0_26
	mov qword ptr [rsp], rdx
	mov rdx, rbx
	sub rdx, r8
	shl rdx, 2
	cmp rbx, rcx
	jbe .LBB0_32
	sub r8, rcx
	mov rbx, rdi
	lea rdi, [rdi + 4*r8]
	mov rsi, rbx
	mov qword ptr [rsp + 8], rcx
	mov r14, r8
	call qword ptr [rip + memmove@GOTPCREL]
	sub r15, r14
	lea rsi, [rbx + 4*r15]
	shl r14, 2
	mov rdi, rbx
	mov rdx, r14
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, rbx
	mov rax, qword ptr [rsp]
	lea rsi, [rbx + 4*rax]
	lea rdi, [rbx + 4*r12]
	mov rbx, qword ptr [rsp + 8]
	jmp .LBB0_40
.LBB0_17:
	xorps xmm0, xmm0
	movups xmmword ptr [r13 + 16], xmm0
	jmp .LBB0_44
.LBB0_18:
	mov r14, r15
	sub r14, rcx
	jbe .LBB0_28
	cmp rax, r15
	jae .LBB0_33
	lea rax, [rcx + r12]
	sub r15, rcx
	lea rsi, [rdi + 4*rax]
	shl r15, 2
	mov r14, rdi
	mov rdx, r15
	mov r15, rcx
	jmp .LBB0_31
.LBB0_21:
	mov r14, rbx
	sub r14, rcx
	jbe .LBB0_29
	cmp rax, rbx
	jae .LBB0_34
	lea rax, [rcx + rdx]
	sub rbx, rcx
	lea rsi, [rdi + 4*rax]
	shl rbx, 2
	mov r14, rdi
	mov r15, rdx
	mov rdx, rbx
	mov rbx, rcx
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r14
	lea rsi, [r14 + 4*r15]
	lea rdi, [r14 + 4*r12]
	jmp .LBB0_40
.LBB0_24:
	sub r15, rcx
	jbe .LBB0_35
	sub rcx, r8
	mov qword ptr [rsp + 8], rcx
	lea rsi, [rdi + 4*r12]
	mov r12, rdi
	lea rdi, [rdi + 4*rbx]
	lea rdx, [4*r8]
	mov r14, r8
	call qword ptr [rip + memmove@GOTPCREL]
	add r14, rbx
	lea rdi, [r12 + 4*r14]
	mov rbx, qword ptr [rsp + 8]
	lea rdx, [4*rbx]
	mov rsi, r12
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r12
	lea rsi, [r12 + 4*rbx]
	jmp .LBB0_36
.LBB0_26:
	sub rbx, rcx
	jbe .LBB0_37
	sub rcx, r8
	lea rsi, [rdi + 4*rdx]
	mov r15, rdi
	lea rdi, [rdi + 4*r12]
	lea rdx, [4*r8]
	mov r14, rcx
	mov qword ptr [rsp], r8
	call qword ptr [rip + memmove@GOTPCREL]
	add r12, qword ptr [rsp]
	lea rdi, [r15 + 4*r12]
	lea rdx, [4*r14]
	mov rsi, r15
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r15
	lea rsi, [r15 + 4*r14]
	jmp .LBB0_40
.LBB0_28:
	lea rsi, [rdi + 4*r12]
	lea rdi, [rdi + 4*rbx]
	jmp .LBB0_36
.LBB0_29:
	lea rsi, [rdi + 4*rdx]
	lea rdi, [rdi + 4*r12]
	jmp .LBB0_40
.LBB0_30:
	lea rax, [r8 + rbx]
	mov r14, rdi
	lea rdi, [rdi + 4*rax]
	mov rsi, r14
	mov r15, r8
.LBB0_31:
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r14
	lea rsi, [r14 + 4*r12]
	lea rdi, [r14 + 4*rbx]
	jmp .LBB0_36
.LBB0_32:
	lea rax, [r12 + r8]
	mov rbx, rdi
	lea rdi, [rdi + 4*rax]
	mov rsi, rbx
	mov r14, r8
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, rbx
	mov rax, qword ptr [rsp]
	lea rsi, [rbx + 4*rax]
	jmp .LBB0_38
.LBB0_33:
	lea rsi, [rdi + 4*r12]
	mov r15, rdi
	lea rdi, [rdi + 4*rbx]
	lea rdx, [4*rcx]
	mov rbx, rcx
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r15
	add rbx, r12
	lea rsi, [r15 + 4*rbx]
	mov r15, r14
	jmp .LBB0_36
.LBB0_34:
	lea rsi, [rdi + 4*rdx]
	mov rbx, rdi
	lea rdi, [rdi + 4*r12]
	mov r15, rdx
	lea rdx, [4*rcx]
	mov r12, rcx
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, rbx
	add r12, r15
	lea rsi, [rbx + 4*r12]
	jmp .LBB0_39
.LBB0_35:
	lea rsi, [rdi + 4*r12]
	mov r14, rdi
	lea rdi, [rdi + 4*rbx]
	mov r12, rdx
	lea rdx, [4*r8]
	mov r15, r8
	call qword ptr [rip + memmove@GOTPCREL]
	add r15, rbx
	mov rsi, r14
	lea rdi, [r14 + 4*r15]
	mov r15, r12
.LBB0_36:
	shl r15, 2
	mov rdx, r15
	call qword ptr [rip + memmove@GOTPCREL]
	mov rsi, qword ptr [rsp + 16]
	jmp .LBB0_42
.LBB0_37:
	lea rsi, [rdi + 4*rdx]
	mov rbx, rdi
	lea rdi, [rdi + 4*r12]
	lea rdx, [4*r8]
	mov r15, r8
	call qword ptr [rip + memmove@GOTPCREL]
	add r12, r15
	mov rsi, rbx
.LBB0_38:
	lea rdi, [rbx + 4*r12]
.LBB0_39:
	mov rbx, r14
.LBB0_40:
	shl rbx, 2
	mov rdx, rbx
	call qword ptr [rip + memmove@GOTPCREL]
	mov r15, qword ptr [r13]
	mov rax, qword ptr [r13 + 16]
	add rax, rbp
	mov rsi, qword ptr [rsp + 16]
.LBB0_41:
	xor ecx, ecx
	cmp rax, r15
	cmovae rcx, r15
	sub rax, rcx
	mov qword ptr [r13 + 16], rax
.LBB0_42:
	sub rsi, rbp
.LBB0_43:
	mov qword ptr [r13 + 24], rsi
.LBB0_44:
	add rsp, 24
	pop rbx
	pop r12
	pop r13
	pop r14
	pop r15
	pop rbp
	ret
.LBB0_45:
	lea rdx, [rip + .L__unnamed_1]
	mov rdi, rax
	call qword ptr [rip + core::slice::index::slice_index_order_fail@GOTPCREL]
```

after:
```asm
clear:
	movups xmmword ptr [rdi + 16], xmm0
	ret

truncate:
	cmp qword ptr [rdi + 24], rsi
	jbe .LBB2_4
	test rsi, rsi
	jne .LBB2_3
	mov qword ptr [rdi + 16], 0
.LBB2_3:
	mov qword ptr [rdi + 24], rsi
.LBB2_4:
	ret

advance:
	mov rcx, qword ptr [rdi + 24]
	mov rax, rcx
	sub rax, rsi
	jbe .LBB3_1
	mov rcx, qword ptr [rdi]
	add rsi, qword ptr [rdi + 16]
	xor edx, edx
	cmp rsi, rcx
	cmovae rdx, rcx
	sub rsi, rdx
	mov qword ptr [rdi + 16], rsi
	mov qword ptr [rdi + 24], rax
	ret
.LBB3_1:
	test rcx, rcx
	je .LBB3_3
	mov qword ptr [rdi + 24], 0
.LBB3_3:
	mov qword ptr [rdi + 16], 0
	ret

remove:
	push rbp
	push r15
	push r14
	push r13
	push r12
	push rbx
	push rax
	mov r15, rsi
	mov r14, rdx
	sub r14, rsi
	jb .LBB4_9
	mov rbx, rdi
	mov r12, qword ptr [rdi + 24]
	mov r13, r12
	sub r13, rdx
	jb .LBB4_10
	mov qword ptr [rbx + 24], r15
	mov rbp, r12
	sub rbp, r14
	test r15, r15
	je .LBB4_4
	cmp rbp, r15
	jne .LBB4_11
.LBB4_4:
	cmp r12, r14
	jne .LBB4_6
.LBB4_5:
	mov qword ptr [rbx + 16], 0
	jmp .LBB4_8
.LBB4_11:
	mov rdi, rbx
	mov rsi, r14
	mov rdx, r15
	mov rcx, r13
	call <<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<T,A> as core::ops::drop::Drop>::drop::copy_data
	cmp r12, r14
	je .LBB4_5
.LBB4_6:
	cmp r13, r15
	jbe .LBB4_8
	mov rax, qword ptr [rbx]
	add r14, qword ptr [rbx + 16]
	xor ecx, ecx
	cmp r14, rax
	cmovae rcx, rax
	sub r14, rcx
	mov qword ptr [rbx + 16], r14
.LBB4_8:
	mov qword ptr [rbx + 24], rbp
	add rsp, 8
	pop rbx
	pop r12
	pop r13
	pop r14
	pop r15
	pop rbp
	ret
.LBB4_9:
	lea rax, [rip + .L__unnamed_1]
	mov rdi, r15
	mov rsi, rdx
	mov rdx, rax
	call qword ptr [rip + core::slice::index::slice_index_order_fail@GOTPCREL]
.LBB4_10:
	lea rax, [rip + .L__unnamed_1]
	mov rdi, rdx
	mov rsi, r12
	mov rdx, rax
	call qword ptr [rip + core::slice::index::slice_end_index_len_fail@GOTPCREL]

<<alloc::collections::vec_deque::drain::Drain<T,A> as core::ops::drop::Drop>::drop::DropGuard<T,A> as core::ops::drop::Drop>::drop::copy_data:
	push rbp
	push r15
	push r14
	push r13
	push r12
	push rbx
	push rax
	mov r14, rsi
	cmp rdx, rcx
	jae .LBB0_1
	mov r12, qword ptr [rdi]
	mov rax, qword ptr [rdi + 16]
	add r14, rax
	xor ecx, ecx
	cmp r14, r12
	cmovae rcx, r12
	sub r14, rcx
	mov r15, rdx
	mov r13, r14
	mov r14, rax
	mov rcx, r13
	sub rcx, r14
	je .LBB0_18
.LBB0_4:
	mov rdi, qword ptr [rdi + 8]
	mov rax, rcx
	add rax, r12
	cmovae rax, rcx
	mov rbx, r12
	sub rbx, r14
	mov rcx, r12
	sub rcx, r13
	mov rbp, r15
	sub rbp, rbx
	jbe .LBB0_5
	cmp rax, r15
	jae .LBB0_12
	mov rdx, r15
	sub rdx, rbx
	shl rdx, 2
	cmp r15, rcx
	jbe .LBB0_16
	sub rbx, rcx
	mov rbp, rdi
	lea rdi, [rdi + 4*rbx]
	mov r15, qword ptr [rip + memmove@GOTPCREL]
	mov rsi, rbp
	mov qword ptr [rsp], rcx
	call r15
	sub r12, rbx
	lea rsi, [4*r12]
	add rsi, rbp
	shl rbx, 2
	mov rdi, rbp
	mov rdx, rbx
	call r15
	mov rdi, rbp
	lea rsi, [4*r14]
	add rsi, rbp
	lea rdi, [4*r13]
	add rdi, rbp
	mov r15, qword ptr [rsp]
	jmp .LBB0_7
.LBB0_1:
	mov r15, rcx
	add r14, rdx
	mov r12, qword ptr [rdi]
	mov r13, qword ptr [rdi + 16]
	add r14, r13
	xor eax, eax
	cmp r14, r12
	mov rcx, r12
	cmovb rcx, rax
	sub r14, rcx
	add r13, rdx
	cmp r13, r12
	cmovae rax, r12
	sub r13, rax
	mov rcx, r13
	sub rcx, r14
	jne .LBB0_4
.LBB0_18:
	add rsp, 8
	pop rbx
	pop r12
	pop r13
	pop r14
	pop r15
	pop rbp
	ret
.LBB0_5:
	mov rbx, r15
	sub rbx, rcx
	jbe .LBB0_6
	cmp rax, r15
	jae .LBB0_9
	lea rax, [rcx + r14]
	sub r15, rcx
	lea rsi, [rdi + 4*rax]
	shl r15, 2
	mov rbx, rdi
	mov rdx, r15
	mov r15, rcx
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, rbx
	lea rsi, [rbx + 4*r14]
	lea rdi, [rbx + 4*r13]
	jmp .LBB0_7
.LBB0_12:
	sub r15, rcx
	jbe .LBB0_13
	sub rcx, rbx
	lea rsi, [rdi + 4*r14]
	mov r12, rdi
	lea rdi, [rdi + 4*r13]
	lea rdx, [4*rbx]
	mov r14, qword ptr [rip + memmove@GOTPCREL]
	mov rbp, rcx
	call r14
	add rbx, r13
	lea rdi, [r12 + 4*rbx]
	lea rdx, [4*rbp]
	mov rsi, r12
	call r14
	mov rdi, r12
	lea rsi, [r12 + 4*rbp]
	jmp .LBB0_7
.LBB0_6:
	lea rsi, [rdi + 4*r14]
	lea rdi, [rdi + 4*r13]
	jmp .LBB0_7
.LBB0_16:
	lea rax, [rbx + r13]
	mov r15, rdi
	lea rdi, [rdi + 4*rax]
	mov rsi, r15
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r15
	lea rsi, [r15 + 4*r14]
	lea rdi, [r15 + 4*r13]
	mov r15, rbx
	jmp .LBB0_7
.LBB0_9:
	lea rsi, [rdi + 4*r14]
	mov r15, rdi
	lea rdi, [rdi + 4*r13]
	lea rdx, [4*rcx]
	mov r12, rcx
	call qword ptr [rip + memmove@GOTPCREL]
	mov rdi, r15
	add r12, r14
	lea rsi, [r15 + 4*r12]
	mov r15, rbx
	jmp .LBB0_7
.LBB0_13:
	lea rsi, [rdi + 4*r14]
	mov r14, rdi
	lea rdi, [rdi + 4*r13]
	lea rdx, [4*rbx]
	call qword ptr [rip + memmove@GOTPCREL]
	add rbx, r13
	mov rsi, r14
	lea rdi, [r14 + 4*rbx]
	mov r15, rbp
.LBB0_7:
	shl r15, 2
	mov rdx, r15
	add rsp, 8
	pop rbx
	pop r12
	pop r13
	pop r14
	pop r15
	pop rbp
	jmp qword ptr [rip + memmove@GOTPCREL]
```

</details>
  • Loading branch information
@GuillaumeGomez
GuillaumeGomez authored Feb 15, 2024
2 parents cbddf31 + da1c7f6 commit 9c78654
Show file tree
Hide file tree
Showing 2 changed files with 182 additions and 56 deletions.
170 changes: 114 additions & 56 deletions library/alloc/src/collections/vec_deque/drain.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,8 @@ pub struct Drain<
drain_len: usize,
// index into the logical array, not the physical one (always lies in [0..deque.len))
idx: usize,
// number of elements after the drain range
tail_len: usize,
// number of elements remaining after dropping the drain
new_len: usize,
remaining: usize,
// Needed to make Drain covariant over T
_marker: PhantomData<&'a T>,
Expand All @@ -41,12 +41,12 @@ impl<'a, T, A: Allocator> Drain<'a, T, A> {
drain_len: usize,
) -> Self {
let orig_len = mem::replace(&mut deque.len, drain_start);
let tail_len = orig_len - drain_start - drain_len;
let new_len = orig_len - drain_len;
Drain {
deque: NonNull::from(deque),
drain_len,
idx: drain_start,
tail_len,
new_len,
remaining: drain_len,
_marker: PhantomData,
}
Expand Down Expand Up @@ -79,7 +79,7 @@ impl<T: fmt::Debug, A: Allocator> fmt::Debug for Drain<'_, T, A> {
f.debug_tuple("Drain")
.field(&self.drain_len)
.field(&self.idx)
.field(&self.tail_len)
.field(&self.new_len)
.field(&self.remaining)
.finish()
}
Expand All @@ -95,9 +95,26 @@ impl<T, A: Allocator> Drop for Drain<'_, T, A> {
fn drop(&mut self) {
struct DropGuard<'r, 'a, T, A: Allocator>(&'r mut Drain<'a, T, A>);

let guard = DropGuard(self);

if mem::needs_drop::<T>() && guard.0.remaining != 0 {
unsafe {
// SAFETY: We just checked that `self.remaining != 0`.
let (front, back) = guard.0.as_slices();
// since idx is a logical index, we don't need to worry about wrapping.
guard.0.idx += front.len();
guard.0.remaining -= front.len();
ptr::drop_in_place(front);
guard.0.remaining = 0;
ptr::drop_in_place(back);
}
}

// Dropping `guard` handles moving the remaining elements into place.
impl<'r, 'a, T, A: Allocator> Drop for DropGuard<'r, 'a, T, A> {
#[inline]
fn drop(&mut self) {
if self.0.remaining != 0 {
if mem::needs_drop::<T>() && self.0.remaining != 0 {
unsafe {
// SAFETY: We just checked that `self.remaining != 0`.
let (front, back) = self.0.as_slices();
Expand All @@ -108,70 +125,111 @@ impl<T, A: Allocator> Drop for Drain<'_, T, A> {

let source_deque = unsafe { self.0.deque.as_mut() };

let drain_start = source_deque.len();
let drain_len = self.0.drain_len;
let drain_end = drain_start + drain_len;

let orig_len = self.0.tail_len + drain_end;
let new_len = self.0.new_len;

if T::IS_ZST {
// no need to copy around any memory if T is a ZST
source_deque.len = orig_len - drain_len;
source_deque.len = new_len;
return;
}

let head_len = drain_start;
let tail_len = self.0.tail_len;
let head_len = source_deque.len; // #elements in front of the drain
let tail_len = new_len - head_len; // #elements behind the drain

match (head_len, tail_len) {
(0, 0) => {
source_deque.head = 0;
source_deque.len = 0;
}
(0, _) => {
source_deque.head = source_deque.to_physical_idx(drain_len);
source_deque.len = orig_len - drain_len;
}
(_, 0) => {
source_deque.len = orig_len - drain_len;
}
_ => unsafe {
if head_len <= tail_len {
source_deque.wrap_copy(
source_deque.head,
source_deque.to_physical_idx(drain_len),
head_len,
);
source_deque.head = source_deque.to_physical_idx(drain_len);
source_deque.len = orig_len - drain_len;
// Next, we will fill the hole left by the drain with as few writes as possible.
// The code below handles the following control flow and reduces the amount of
// branches under the assumption that `head_len == 0 || tail_len == 0`, i.e.
// draining at the front or at the back of the dequeue is especially common.
//
// H = "head index" = `deque.head`
// h = elements in front of the drain
// d = elements in the drain
// t = elements behind the drain
//
// Note that the buffer may wrap at any point and the wrapping is handled by
// `wrap_copy` and `to_physical_idx`.
//
// Case 1: if `head_len == 0 && tail_len == 0`
// Everything was drained, reset the head index back to 0.
// H
// [ . . . . . d d d d . . . . . ]
// H
// [ . . . . . . . . . . . . . . ]
//
// Case 2: else if `tail_len == 0`
// Don't move data or the head index.
// H
// [ . . . h h h h d d d d . . . ]
// H
// [ . . . h h h h . . . . . . . ]
//
// Case 3: else if `head_len == 0`
// Don't move data, but move the head index.
// H
// [ . . . d d d d t t t t . . . ]
// H
// [ . . . . . . . t t t t . . . ]
//
// Case 4: else if `tail_len <= head_len`
// Move data, but not the head index.
// H
// [ . . h h h h d d d d t t . . ]
// H
// [ . . h h h h t t . . . . . . ]
//
// Case 5: else
// Move data and the head index.
// H
// [ . . h h d d d d t t t t . . ]
// H
// [ . . . . . . h h t t t t . . ]

// When draining at the front (`.drain(..n)`) or at the back (`.drain(n..)`),
// we don't need to copy any data. The number of elements copied would be 0.
if head_len != 0 && tail_len != 0 {
join_head_and_tail_wrapping(source_deque, drain_len, head_len, tail_len);
// Marking this function as cold helps LLVM to eliminate it entirely if
// this branch is never taken.
// We use `#[cold]` instead of `#[inline(never)]`, because inlining this
// function into the general case (`.drain(n..m)`) is fine.
// See `tests/codegen/vecdeque-drain.rs` for a test.
#[cold]
fn join_head_and_tail_wrapping<T, A: Allocator>(
source_deque: &mut VecDeque<T, A>,
drain_len: usize,
head_len: usize,
tail_len: usize,
) {
// Pick whether to move the head or the tail here.
let (src, dst, len);
if head_len < tail_len {
src = source_deque.head;
dst = source_deque.to_physical_idx(drain_len);
len = head_len;
} else {
source_deque.wrap_copy(
source_deque.to_physical_idx(head_len + drain_len),
source_deque.to_physical_idx(head_len),
tail_len,
);
source_deque.len = orig_len - drain_len;
src = source_deque.to_physical_idx(head_len + drain_len);
dst = source_deque.to_physical_idx(head_len);
len = tail_len;
};

unsafe {
source_deque.wrap_copy(src, dst, len);
}
},
}
}
}
}

let guard = DropGuard(self);
if guard.0.remaining != 0 {
unsafe {
// SAFETY: We just checked that `self.remaining != 0`.
let (front, back) = guard.0.as_slices();
// since idx is a logical index, we don't need to worry about wrapping.
guard.0.idx += front.len();
guard.0.remaining -= front.len();
ptr::drop_in_place(front);
guard.0.remaining = 0;
ptr::drop_in_place(back);
if new_len == 0 {
// Special case: If the entire dequeue was drained, reset the head back to 0,
// like `.clear()` does.
source_deque.head = 0;
} else if head_len < tail_len {
// If we moved the head above, then we need to adjust the head index here.
source_deque.head = source_deque.to_physical_idx(drain_len);
}
source_deque.len = new_len;
}
}

// Dropping `guard` handles moving the remaining elements into place.
}
}

Expand Down
68 changes: 68 additions & 0 deletions tests/codegen/vecdeque-drain.rs
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,68 @@
// Check that draining at the front or back doesn't copy memory.

// compile-flags: -O
// ignore-debug: the debug assertions get in the way

#![crate_type = "lib"]

use std::collections::VecDeque;

// CHECK-LABEL: @clear
// CHECK-NOT: call
// CHECK-NOT: br
// CHECK: getelementptr inbounds
// CHECK-NEXT: call void @llvm.memset
// CHECK-NEXT: ret void
#[no_mangle]
pub fn clear(v: &mut VecDeque<i32>) {
v.drain(..);
}

// CHECK-LABEL: @truncate
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK-NOT: br
// CHECK: ret void
#[no_mangle]
pub fn truncate(v: &mut VecDeque<i32>, n: usize) {
if n < v.len() {
v.drain(n..);
}
}

// CHECK-LABEL: @advance
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK: br
// CHECK-NOT: call
// CHECK-NOT: br
// CHECK: ret void
#[no_mangle]
pub fn advance(v: &mut VecDeque<i32>, n: usize) {
if n < v.len() {
v.drain(..n);
} else {
v.clear();
}
}

// CHECK-LABEL: @remove
// CHECK: call
#[no_mangle]
pub fn remove(v: &mut VecDeque<i32>, a: usize, b: usize) {
v.drain(a..b);
}

0 comments on commit 9c78654

Please sign in to comment.