mathieu/os-rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Rollup merge of #133674 - scottmcm:chain-carrying-add, r=Amanieu

Browse source 
Fix chaining `carrying_add`s

Something about the MIR lowering for `||` ended up breaking this, but it's fixed by changing the code to use `|` instead.

I also added an assembly test to ensure it *keeps* being [`adc`](https://www.felixcloutier.com/x86/adc).

cc https://github.com/rust-lang/rust/issues/85532#issuecomment-2495119815, which noticed this.
This commit is contained in:
Matthias Krüger 2024-12-01 14:30:11 +01:00 • committed by GitHub
commit 2f00feb616
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 34 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
library/core/src/num/uint_macros.rs
View file

@ -2354,7 +2354,7 @@ macro_rules! uint_impl {
// to generate optimal code for now, and LLVM doesn't have an equivalent intrinsic
let (a, b) = self.overflowing_add(rhs);
let (c, d) = a.overflowing_add(carry as $SelfT);
(c, b || d)
(c, b | d)
}
/// Calculates `self` + `rhs` with a signed `rhs`.

33
tests/assembly/x86_64-bigint-add.rs Normal file
View file

@ -0,0 +1,33 @@
//@ only-x86_64
//@ assembly-output: emit-asm
//@ compile-flags: --crate-type=lib -O -C target-cpu=x86-64-v4
//@ compile-flags: -C llvm-args=-x86-asm-syntax=intel
#![no_std]
#![feature(bigint_helper_methods)]
// This checks that the `carrying_add` implementation successfully chains, to catch
// issues like <https://github.com/rust-lang/rust/issues/85532#issuecomment-2495119815>
// This forces the ABI to avoid the windows-vs-linux ABI differences.
// CHECK-LABEL: bigint_chain_carrying_add:
#[no_mangle]
pub unsafe extern "sysv64" fn bigint_chain_carrying_add(
dest: *mut u64,
src1: *const u64,
src2: *const u64,
n: usize,
mut carry: bool,
) -> bool {
// CHECK: mov [[TEMP:r..]], qword ptr [rsi + 8*[[IND:r..]] + 8]
// CHECK: adc [[TEMP]], qword ptr [rdx + 8*[[IND]] + 8]
// CHECK: mov qword ptr [rdi + 8*[[IND]] + 8], [[TEMP]]
// CHECK: mov [[TEMP]], qword ptr [rsi + 8*[[IND]] + 16]
// CHECK: adc [[TEMP]], qword ptr [rdx + 8*[[IND]] + 16]
// CHECK: mov qword ptr [rdi + 8*[[IND]] + 16], [[TEMP]]
for i in 0..n {
(*dest.add(i), carry) = u64::carrying_add(*src1.add(i), *src2.add(i), carry);
}
carry
}