Implement llvm.x86.addcarry.32 by refactoring llvm.x86.addcarry.64 to make it generic

This commit is contained in:
Eduardo Sánchez Muñoz 2023-09-21 20:08:52 +02:00
parent d6b30b88d0
commit 356039985e
2 changed files with 43 additions and 14 deletions

View File

@ -5,7 +5,6 @@ use rustc_target::spec::abi::Abi;
use crate::*;
use helpers::bool_to_simd_element;
use helpers::convert::Truncate as _;
use shims::foreign_items::EmulateByNameResult;
mod sse;
@ -26,23 +25,30 @@ pub(super) trait EvalContextExt<'mir, 'tcx: 'mir>:
// Prefix should have already been checked.
let unprefixed_name = link_name.as_str().strip_prefix("llvm.x86.").unwrap();
match unprefixed_name {
"addcarry.64" if this.tcx.sess.target.arch == "x86_64" => {
// Computes u8+u64+u64, returning tuple (u8,u64) comprising the output carry and truncated sum.
// Used to implement the `_addcarry_u32` and `_addcarry_u64` functions.
// Computes u8+uX+uX (uX is u32 or u64), returning tuple (u8,uX) comprising
// the output carry and truncated sum.
"addcarry.32" | "addcarry.64" => {
if unprefixed_name == "addcarry.64" && this.tcx.sess.target.arch != "x86_64" {
return Ok(EmulateByNameResult::NotSupported);
}
let [c_in, a, b] = this.check_shim(abi, Abi::Unadjusted, link_name, args)?;
let c_in = this.read_scalar(c_in)?.to_u8()?;
let a = this.read_scalar(a)?.to_u64()?;
let b = this.read_scalar(b)?.to_u64()?;
let a = this.read_immediate(a)?;
let b = this.read_immediate(b)?;
#[allow(clippy::arithmetic_side_effects)]
// adding two u64 and a u8 cannot wrap in a u128
let wide_sum = u128::from(c_in) + u128::from(a) + u128::from(b);
#[allow(clippy::arithmetic_side_effects)] // it's a u128, we can shift by 64
let (c_out, sum) = ((wide_sum >> 64).truncate::<u8>(), wide_sum.truncate::<u64>());
let (sum, overflow1) = this.overflowing_binary_op(mir::BinOp::Add, &a, &b)?;
let (sum, overflow2) = this.overflowing_binary_op(
mir::BinOp::Add,
&sum,
&ImmTy::from_uint(c_in, a.layout),
)?;
#[allow(clippy::arithmetic_side_effects)] // adding two bools into a u8
let c_out = u8::from(overflow1) + u8::from(overflow2);
let c_out_field = this.project_field(dest, 0)?;
this.write_scalar(Scalar::from_u8(c_out), &c_out_field)?;
let sum_field = this.project_field(dest, 1)?;
this.write_scalar(Scalar::from_u64(sum), &sum_field)?;
this.write_scalar(Scalar::from_u8(c_out), &this.project_field(dest, 0)?)?;
this.write_immediate(*sum, &this.project_field(dest, 1)?)?;
}
name if name.starts_with("sse.") => {

View File

@ -1,3 +1,24 @@
#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
mod x86 {
#[cfg(target_arch = "x86")]
use core::arch::x86 as arch;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64 as arch;
fn adc(c_in: u8, a: u32, b: u32) -> (u8, u32) {
let mut sum = 0;
// SAFETY: There are no safety requirements for calling `_addcarry_u32`.
// It's just unsafe for API consistency with other intrinsics.
let c_out = unsafe { arch::_addcarry_u32(c_in, a, b, &mut sum) };
(c_out, sum)
}
pub fn main() {
assert_eq!(adc(1, 1, 1), (0, 3));
assert_eq!(adc(3, u32::MAX, u32::MAX), (2, 1));
}
}
#[cfg(target_arch = "x86_64")]
mod x86_64 {
use core::arch::x86_64 as arch;
@ -17,6 +38,8 @@ mod x86_64 {
}
fn main() {
#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
x86::main();
#[cfg(target_arch = "x86_64")]
x86_64::main();
}