Auto merge of #41418 - hirschenberger:prefetch-intrinsic, r=nagisa
Adding support for the llvm `prefetch` intrinsic Optimize `slice::binary_search` by using prefetching.
This commit is contained in:
commit
668e698bba
@ -564,8 +564,55 @@ extern "rust-intrinsic" {
|
||||
pub fn atomic_umax_rel<T>(dst: *mut T, src: T) -> T;
|
||||
pub fn atomic_umax_acqrel<T>(dst: *mut T, src: T) -> T;
|
||||
pub fn atomic_umax_relaxed<T>(dst: *mut T, src: T) -> T;
|
||||
|
||||
/// The `prefetch` intrinsic is a hint to the code generator to insert a prefetch instruction
|
||||
/// if supported; otherwise, it is a noop.
|
||||
/// Prefetches have no effect on the behavior of the program but can change its performance
|
||||
/// characteristics.
|
||||
///
|
||||
/// The `locality` argument must be a constant integer and is a temporal locality specifier
|
||||
/// ranging from (0) - no locality, to (3) - extremely local keep in cache
|
||||
#[cfg(not(stage0))]
|
||||
pub fn prefetch_read_data<T>(data: *const T, locality: i32);
|
||||
/// The `prefetch` intrinsic is a hint to the code generator to insert a prefetch instruction
|
||||
/// if supported; otherwise, it is a noop.
|
||||
/// Prefetches have no effect on the behavior of the program but can change its performance
|
||||
/// characteristics.
|
||||
///
|
||||
/// The `locality` argument must be a constant integer and is a temporal locality specifier
|
||||
/// ranging from (0) - no locality, to (3) - extremely local keep in cache
|
||||
#[cfg(not(stage0))]
|
||||
pub fn prefetch_write_data<T>(data: *const T, locality: i32);
|
||||
/// The `prefetch` intrinsic is a hint to the code generator to insert a prefetch instruction
|
||||
/// if supported; otherwise, it is a noop.
|
||||
/// Prefetches have no effect on the behavior of the program but can change its performance
|
||||
/// characteristics.
|
||||
///
|
||||
/// The `locality` argument must be a constant integer and is a temporal locality specifier
|
||||
/// ranging from (0) - no locality, to (3) - extremely local keep in cache
|
||||
#[cfg(not(stage0))]
|
||||
pub fn prefetch_read_instruction<T>(data: *const T, locality: i32);
|
||||
/// The `prefetch` intrinsic is a hint to the code generator to insert a prefetch instruction
|
||||
/// if supported; otherwise, it is a noop.
|
||||
/// Prefetches have no effect on the behavior of the program but can change its performance
|
||||
/// characteristics.
|
||||
///
|
||||
/// The `locality` argument must be a constant integer and is a temporal locality specifier
|
||||
/// ranging from (0) - no locality, to (3) - extremely local keep in cache
|
||||
#[cfg(not(stage0))]
|
||||
pub fn prefetch_write_instruction<T>(data: *const T, locality: i32);
|
||||
}
|
||||
|
||||
// Empty bootstrap implementations for stage0 compilation
|
||||
#[cfg(stage0)]
|
||||
pub fn prefetch_read_data<T>(_data: *const T, _locality: i32) { /* EMPTY */ }
|
||||
#[cfg(stage0)]
|
||||
pub fn prefetch_write_data<T>(_data: *const T, _locality: i32) { /* EMPTY */ }
|
||||
#[cfg(stage0)]
|
||||
pub fn prefetch_read_instruction<T>(_data: *const T, _locality: i32) { /* EMPTY */ }
|
||||
#[cfg(stage0)]
|
||||
pub fn prefetch_write_instruction<T>(_data: *const T, _locality: i32) { /* EMPTY */ }
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
|
||||
pub fn atomic_fence();
|
||||
|
@ -942,6 +942,7 @@ fn declare_intrinsic(ccx: &CrateContext, key: &str) -> Option<ValueRef> {
|
||||
ifn!("llvm.x86.seh.recoverfp", fn(i8p, i8p) -> i8p);
|
||||
|
||||
ifn!("llvm.assume", fn(i1) -> void);
|
||||
ifn!("llvm.prefetch", fn(i8p, t_i32, t_i32, t_i32) -> void);
|
||||
|
||||
if ccx.sess().opts.debuginfo != NoDebugInfo {
|
||||
ifn!("llvm.dbg.declare", fn(Type::metadata(ccx), Type::metadata(ccx)) -> void);
|
||||
|
@ -255,7 +255,18 @@ pub fn trans_intrinsic_call<'a, 'tcx>(bcx: &Builder<'a, 'tcx>,
|
||||
}
|
||||
C_nil(ccx)
|
||||
},
|
||||
|
||||
"prefetch_read_data" | "prefetch_write_data" |
|
||||
"prefetch_read_instruction" | "prefetch_write_instruction" => {
|
||||
let expect = ccx.get_intrinsic(&("llvm.prefetch"));
|
||||
let (rw, cache_type) = match name {
|
||||
"prefetch_read_data" => (0, 1),
|
||||
"prefetch_write_data" => (1, 1),
|
||||
"prefetch_read_instruction" => (0, 0),
|
||||
"prefetch_write_instruction" => (1, 0),
|
||||
_ => bug!()
|
||||
};
|
||||
bcx.call(expect, &[llargs[0], C_i32(ccx, rw), llargs[1], C_i32(ccx, cache_type)], None)
|
||||
},
|
||||
"ctlz" | "cttz" | "ctpop" | "bswap" |
|
||||
"add_with_overflow" | "sub_with_overflow" | "mul_with_overflow" |
|
||||
"overflowing_add" | "overflowing_sub" | "overflowing_mul" |
|
||||
|
@ -133,6 +133,14 @@ pub fn check_intrinsic_type<'a, 'tcx>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
|
||||
],
|
||||
tcx.mk_nil())
|
||||
}
|
||||
"prefetch_read_data" | "prefetch_write_data" |
|
||||
"prefetch_read_instruction" | "prefetch_write_instruction" => {
|
||||
(1, vec![tcx.mk_ptr(ty::TypeAndMut {
|
||||
ty: param(0),
|
||||
mutbl: hir::MutImmutable
|
||||
}), tcx.types.i32],
|
||||
tcx.mk_nil())
|
||||
}
|
||||
"drop_in_place" => {
|
||||
(1, vec![tcx.mk_mut_ptr(param(0))], tcx.mk_nil())
|
||||
}
|
||||
|
75
src/test/codegen/prefetch.rs
Normal file
75
src/test/codegen/prefetch.rs
Normal file
@ -0,0 +1,75 @@
|
||||
// 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"]
|
||||
#![feature(core_intrinsics)]
|
||||
|
||||
use std::intrinsics::{prefetch_read_data, prefetch_write_data,
|
||||
prefetch_read_instruction, prefetch_write_instruction};
|
||||
|
||||
#[no_mangle]
|
||||
pub fn check_prefetch_read_data(data: &[i8]) {
|
||||
unsafe {
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 0, i32 1)
|
||||
prefetch_read_data(data.as_ptr(), 0);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 1, i32 1)
|
||||
prefetch_read_data(data.as_ptr(), 1);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 2, i32 1)
|
||||
prefetch_read_data(data.as_ptr(), 2);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 3, i32 1)
|
||||
prefetch_read_data(data.as_ptr(), 3);
|
||||
}
|
||||
}
|
||||
|
||||
#[no_mangle]
|
||||
pub fn check_prefetch_write_data(data: &[i8]) {
|
||||
unsafe {
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 0, i32 1)
|
||||
prefetch_write_data(data.as_ptr(), 0);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 1, i32 1)
|
||||
prefetch_write_data(data.as_ptr(), 1);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 2, i32 1)
|
||||
prefetch_write_data(data.as_ptr(), 2);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 3, i32 1)
|
||||
prefetch_write_data(data.as_ptr(), 3);
|
||||
}
|
||||
}
|
||||
|
||||
#[no_mangle]
|
||||
pub fn check_prefetch_read_instruction(data: &[i8]) {
|
||||
unsafe {
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 0, i32 0)
|
||||
prefetch_read_instruction(data.as_ptr(), 0);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 1, i32 0)
|
||||
prefetch_read_instruction(data.as_ptr(), 1);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 2, i32 0)
|
||||
prefetch_read_instruction(data.as_ptr(), 2);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 0, i32 3, i32 0)
|
||||
prefetch_read_instruction(data.as_ptr(), 3);
|
||||
}
|
||||
}
|
||||
|
||||
#[no_mangle]
|
||||
pub fn check_prefetch_write_instruction(data: &[i8]) {
|
||||
unsafe {
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 0, i32 0)
|
||||
prefetch_write_instruction(data.as_ptr(), 0);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 1, i32 0)
|
||||
prefetch_write_instruction(data.as_ptr(), 1);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 2, i32 0)
|
||||
prefetch_write_instruction(data.as_ptr(), 2);
|
||||
// CHECK: call void @llvm.prefetch(i8* %{{.*}}, i32 1, i32 3, i32 0)
|
||||
prefetch_write_instruction(data.as_ptr(), 3);
|
||||
}
|
||||
}
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user