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

Rollup merge of #133053 - liushuyu:simd-test-x86-baseline-fix, r=workingjubilee

Browse source 
tests: Fix the SIMD FFI tests with certain x86 configuration

This pull request fixes the SIMD FFI tests with certain x86 configurations by gating the SSE2 intrinsic behind the `sse2` feature gate. A generic LLVM intrinsic that is easy to un-fuse on those platforms is added to compensate for those platforms.
This commit is contained in:
Jubilee 2024-11-14 17:55:27 -08:00 • committed by GitHub
commit efe2c44269
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 4 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
tests/run-make/simd-ffi/simd.rs
View file

@ -25,7 +25,7 @@ pub struct i32x4([i32; 4]);
extern "C" { extern "C" {
// _mm_sll_epi32 // _mm_sll_epi32
#[cfg(any(target_arch = "x86", target_arch = "x86-64"))] #[cfg(all(any(target_arch = "x86", target_arch = "x86-64"), target_feature = "sse2"))]
#[link_name = "llvm.x86.sse2.psll.d"] #[link_name = "llvm.x86.sse2.psll.d"]
fn integer(a: i32x4, b: i32x4) -> i32x4; fn integer(a: i32x4, b: i32x4) -> i32x4;
@ -38,15 +38,13 @@ extern "C" {
#[link_name = "llvm.aarch64.neon.maxs.v4i32"] #[link_name = "llvm.aarch64.neon.maxs.v4i32"]
fn integer(a: i32x4, b: i32x4) -> i32x4; fn integer(a: i32x4, b: i32x4) -> i32x4;
// just some substitute foreign symbol, not an LLVM intrinsic; so // Use a generic LLVM intrinsic to do type checking on other platforms
// we still get type checking, but not as detailed as (ab)using
// LLVM.
#[cfg(not(any( #[cfg(not(any(
target_arch = "x86", all(any(target_arch = "x86", target_arch = "x86-64"), target_feature = "sse2"),
target_arch = "x86-64",
target_arch = "arm", target_arch = "arm",
target_arch = "aarch64" target_arch = "aarch64"
)))] )))]
#[link_name = "llvm.smax.v4i32"]
fn integer(a: i32x4, b: i32x4) -> i32x4; fn integer(a: i32x4, b: i32x4) -> i32x4;
} }