Skip to content

Optimize initialization of arrays using repeat expressions #43488

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

Merged
merged 5 commits into from
Aug 6, 2017
Merged
Show file tree
Hide file tree
Changes from 2 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/librustc_trans/common.rs
Original file line number Diff line number Diff line change
Expand Up @@ -372,7 +372,7 @@ pub fn const_to_uint(v: ValueRef) -> u64 {
}
}

fn is_const_integral(v: ValueRef) -> bool {
pub fn is_const_integral(v: ValueRef) -> bool {
unsafe {
!llvm::LLVMIsAConstantInt(v).is_null()
}
Expand Down
41 changes: 38 additions & 3 deletions src/librustc_trans/mir/rvalue.rs
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@
use llvm::{self, ValueRef};
use rustc::ty::{self, Ty};
use rustc::ty::cast::{CastTy, IntTy};
use rustc::ty::layout::{Layout, LayoutTyper};
use rustc::ty::layout::{self, Layout, LayoutTyper, Primitive};
use rustc::mir::tcx::LvalueTy;
use rustc::mir;
use rustc::middle::lang_items::ExchangeMallocFnLangItem;
Expand All @@ -20,7 +20,7 @@ use base;
use builder::Builder;
use callee;
use common::{self, val_ty, C_bool, C_null, C_uint};
use common::{C_integral};
use common::{C_integral, C_i32};
use adt;
use machine;
use monomorphize;
Expand Down Expand Up @@ -93,12 +93,47 @@ impl<'a, 'tcx> MirContext<'a, 'tcx> {
}

mir::Rvalue::Repeat(ref elem, ref count) => {
let dest_ty = dest.ty.to_ty(bcx.tcx());

// No need to inizialize memory of a zero-sized slice
if common::type_is_zero_size(bcx.ccx, dest_ty) {
return bcx;
}

let tr_elem = self.trans_operand(&bcx, elem);
let size = count.as_u64(bcx.tcx().sess.target.uint_type);
let size = C_uint(bcx.ccx, size);
let base = base::get_dataptr(&bcx, dest.llval);
let align = dest.alignment.to_align();

if let OperandValue::Immediate(v) = tr_elem.val {
if common::is_const_integral(v) && common::const_to_uint(v) == 0 {
let align = align.unwrap_or_else(|| bcx.ccx.align_of(tr_elem.ty));
let align = C_i32(bcx.ccx, align as i32);
let ty = type_of::type_of(bcx.ccx, dest_ty);
let size = machine::llsize_of(bcx.ccx, ty);
let fill = C_integral(Type::i8(bcx.ccx), 0, false);
base::call_memset(&bcx, base, fill, size, align, false);
return bcx;
}
}

// Use llvm.memset.p0i8.* to initialize byte arrays
let elem_layout = bcx.ccx.layout_of(tr_elem.ty).layout;
match *elem_layout {
Layout::Scalar { value: Primitive::Int(layout::I8), .. } |
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This "knows" that all scalars are immediates of LLVM type i8. I'm not sure that is always true, and it might break in the future. Can you move this to the previous if with a check that val_ty(v) == Type::i8(ccx)?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you have similar concerns for the CEnum case?
Also I'd appreciate thoughts on how to avoid duplicating the call_memset() setup code in the process.

Copy link
Contributor

@arielb1 arielb1 Aug 1, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's better to merge them and just check the LLVM type. Also use from_immediate to catch booleans too.

Layout::CEnum { discr: layout::I8, .. } => {
let align = align.unwrap_or_else(|| bcx.ccx.align_of(tr_elem.ty));
let align = C_i32(bcx.ccx, align as i32);
let fill = tr_elem.immediate();
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

use v here rather than calling immediate again.

base::call_memset(&bcx, base, fill, size, align, false);
return bcx;
}
_ => ()
}

tvec::slice_for_each(&bcx, base, tr_elem.ty, size, |bcx, llslot, loop_bb| {
self.store_operand(bcx, llslot, dest.alignment.to_align(), tr_elem);
self.store_operand(bcx, llslot, align, tr_elem);
bcx.br(loop_bb);
})
}
Expand Down
2 changes: 1 addition & 1 deletion src/librustc_trans/tvec.rs
Original file line number Diff line number Diff line change
Expand Up @@ -30,8 +30,8 @@ pub fn slice_for_each<'a, 'tcx, F>(
};

let body_bcx = bcx.build_sibling_block("slice_loop_body");
let next_bcx = bcx.build_sibling_block("slice_loop_next");
let header_bcx = bcx.build_sibling_block("slice_loop_header");
let next_bcx = bcx.build_sibling_block("slice_loop_next");

let start = if zst {
C_uint(bcx.ccx, 0usize)
Expand Down
74 changes: 74 additions & 0 deletions src/test/codegen/slice-init.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,74 @@
// Copyright 2017 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

// compile-flags: -C no-prepopulate-passes

#![crate_type = "lib"]

// CHECK-LABEL: @zero_sized_elem
#[no_mangle]
pub fn zero_sized_elem() {
// CHECK-NOT: br label %slice_loop_header{{.*}}
// CHECK-NOT: call void @llvm.memset.p0i8
let x = [(); 4];
drop(&x);
}

// CHECK-LABEL: @zero_len_array
#[no_mangle]
pub fn zero_len_array() {
// CHECK-NOT: br label %slice_loop_header{{.*}}
// CHECK-NOT: call void @llvm.memset.p0i8
let x = [4; 0];
drop(&x);
}

// CHECK-LABEL: @byte_array
#[no_mangle]
pub fn byte_array() {
// CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 7, i64 4
// CHECK-NOT: br label %slice_loop_header{{.*}}
let x = [7u8; 4];
drop(&x);
}

#[allow(dead_code)]
#[derive(Copy, Clone)]
enum Init {
Loop,
Memset,
}

// CHECK-LABEL: @byte_enum_array
#[no_mangle]
pub fn byte_enum_array() {
// CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 {{.*}}, i64 4
// CHECK-NOT: br label %slice_loop_header{{.*}}
let x = [Init::Memset; 4];
drop(&x);
}

// CHECK-LABEL: @zeroed_integer_array
#[no_mangle]
pub fn zeroed_integer_array() {
// CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 0, i64 16
// CHECK-NOT: br label %slice_loop_header{{.*}}
let x = [0u32; 4];
drop(&x);
}

// CHECK-LABEL: @nonzero_integer_array
#[no_mangle]
pub fn nonzero_integer_array() {
// CHECK: br label %slice_loop_header{{.*}}
// CHECK-NOT: call void @llvm.memset.p0i8
let x = [0x1a_2b_3c_4d_u32; 4];
drop(&x);
}