Skip to content

Add more manual aarch64 load/store intrinsic tests #2086

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.

Sign up for GitHub

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
folkertdev merged 2 commits into from
Apr 17, 2026
Merged

Add more manual aarch64 load/store intrinsic tests #2086

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f0e3285
Add round trip tests for v{ld,st}{2,3,4} & v{ld,st}{2,3,4} lane intri…
adamgemmell Apr 2, 2026
2d49c7b
Generate some svldff1 tests now that the qemu bug has been fixed
adamgemmell Apr 15, 2026
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
1 change: 1 addition & 0 deletions .github/workflows/main.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -348,6 +348,7 @@ jobs:
- name: Run miri tests
env:
TARGET: "aarch64-unknown-linux-gnu"
RUSTFLAGS: "-Ctarget-cpu=neoverse-v3"
run: |
# read filters and join them with a space.
FILTERS=$(cat aarch64-miri-tests.txt | tr '\n' ' ')
Expand Down
4 changes: 2 additions & 2 deletions crates/core_arch/src/aarch64/neon/generated.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12189,7 +12189,7 @@ pub unsafe fn vld3q_lane_p64<const LANE: i32>(a: *const p64, b: poly64x2x3_t) ->
#[rustc_legacy_const_generics(2)]
#[stable(feature = "neon_intrinsics", since = "1.59.0")]
pub unsafe fn vld3q_lane_s8<const LANE: i32>(a: *const i8, b: int8x16x3_t) -> int8x16x3_t {
static_assert_uimm_bits!(LANE, 3);
static_assert_uimm_bits!(LANE, 4);
unsafe extern "unadjusted" {
#[cfg_attr(
any(target_arch = "aarch64", target_arch = "arm64ec"),
Expand Down Expand Up @@ -12571,7 +12571,7 @@ pub unsafe fn vld4q_lane_f64<const LANE: i32>(a: *const f64, b: float64x2x4_t) -
#[rustc_legacy_const_generics(2)]
#[stable(feature = "neon_intrinsics", since = "1.59.0")]
pub unsafe fn vld4q_lane_s8<const LANE: i32>(a: *const i8, b: int8x16x4_t) -> int8x16x4_t {
static_assert_uimm_bits!(LANE, 3);
static_assert_uimm_bits!(LANE, 4);
unsafe extern "unadjusted" {
#[cfg_attr(
any(target_arch = "aarch64", target_arch = "arm64ec"),
Expand Down
205 changes: 127 additions & 78 deletions crates/core_arch/src/aarch64/neon/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1037,6 +1037,7 @@ mod tests {
macro_rules! wide_store_load_roundtrip_fp16 {
($( $name:ident $args:tt);* $(;)?) => {
$(
#[cfg_attr(miri, ignore)]
#[simd_test(enable = "neon,fp16")]
#[cfg(not(target_arch = "arm64ec"))]
unsafe fn $name() {
Expand All @@ -1055,13 +1056,13 @@ mod tests {
test_vld1q_f16_x3(f16, 24, float16x8x3_t, vst1q_f16_x3, vld1q_f16_x3);
test_vld1q_f16_x4(f16, 32, float16x8x4_t, vst1q_f16_x4, vld1q_f16_x4);

test_vld2_f16_x2(f16, 8, float16x4x2_t, vst2_f16, vld2_f16);
test_vld2_f16_x3(f16, 12, float16x4x3_t, vst3_f16, vld3_f16);
test_vld2_f16_x4(f16, 16, float16x4x4_t, vst4_f16, vld4_f16);
test_vld2_f16(f16, 8, float16x4x2_t, vst2_f16, vld2_f16);
test_vld3_f16(f16, 12, float16x4x3_t, vst3_f16, vld3_f16);
test_vld4_f16(f16, 16, float16x4x4_t, vst4_f16, vld4_f16);

test_vld2q_f16_x2(f16, 16, float16x8x2_t, vst2q_f16, vld2q_f16);
test_vld3q_f16_x3(f16, 24, float16x8x3_t, vst3q_f16, vld3q_f16);
test_vld4q_f16_x4(f16, 32, float16x8x4_t, vst4q_f16, vld4q_f16);
test_vld2q_f16(f16, 16, float16x8x2_t, vst2q_f16, vld2q_f16);
test_vld3q_f16(f16, 24, float16x8x3_t, vst3q_f16, vld3q_f16);
test_vld4q_f16(f16, 32, float16x8x4_t, vst4q_f16, vld4q_f16);
}

macro_rules! wide_store_load_roundtrip_aes {
Expand Down Expand Up @@ -1195,101 +1196,149 @@ mod tests {
}

wide_store_load_roundtrip_neon! {
test_vld2_f32_x2(f32, 4, float32x2x2_t, vst2_f32, vld2_f32);
test_vld2_f32_x3(f32, 6, float32x2x3_t, vst3_f32, vld3_f32);
test_vld2_f32_x4(f32, 8, float32x2x4_t, vst4_f32, vld4_f32);
test_vld2_f32(f32, 4, float32x2x2_t, vst2_f32, vld2_f32);
test_vld3_f32(f32, 6, float32x2x3_t, vst3_f32, vld3_f32);
test_vld4_f32(f32, 8, float32x2x4_t, vst4_f32, vld4_f32);

test_vld2q_f32_x2(f32, 8, float32x4x2_t, vst2q_f32, vld2q_f32);
test_vld3q_f32_x3(f32, 12, float32x4x3_t, vst3q_f32, vld3q_f32);
test_vld4q_f32_x4(f32, 16, float32x4x4_t, vst4q_f32, vld4q_f32);
test_vld2q_f32(f32, 8, float32x4x2_t, vst2q_f32, vld2q_f32);
test_vld3q_f32(f32, 12, float32x4x3_t, vst3q_f32, vld3q_f32);
test_vld4q_f32(f32, 16, float32x4x4_t, vst4q_f32, vld4q_f32);

test_vld2_f64_x2(f64, 2, float64x1x2_t, vst2_f64, vld2_f64);
test_vld2_f64_x3(f64, 3, float64x1x3_t, vst3_f64, vld3_f64);
test_vld2_f64_x4(f64, 4, float64x1x4_t, vst4_f64, vld4_f64);
test_vld2_f64(f64, 2, float64x1x2_t, vst2_f64, vld2_f64);
test_vld3_f64(f64, 3, float64x1x3_t, vst3_f64, vld3_f64);
test_vld4_f64(f64, 4, float64x1x4_t, vst4_f64, vld4_f64);

test_vld2q_f64_x2(f64, 4, float64x2x2_t, vst2q_f64, vld2q_f64);
test_vld3q_f64_x3(f64, 6, float64x2x3_t, vst3q_f64, vld3q_f64);
test_vld4q_f64_x4(f64, 8, float64x2x4_t, vst4q_f64, vld4q_f64);
test_vld2q_f64(f64, 4, float64x2x2_t, vst2q_f64, vld2q_f64);
test_vld3q_f64(f64, 6, float64x2x3_t, vst3q_f64, vld3q_f64);
test_vld4q_f64(f64, 8, float64x2x4_t, vst4q_f64, vld4q_f64);

test_vld2_s8_x2(i8, 16, int8x8x2_t, vst2_s8, vld2_s8);
test_vld2_s8_x3(i8, 24, int8x8x3_t, vst3_s8, vld3_s8);
test_vld2_s8_x4(i8, 32, int8x8x4_t, vst4_s8, vld4_s8);
test_vld2_s8(i8, 16, int8x8x2_t, vst2_s8, vld2_s8);
test_vld3_s8(i8, 24, int8x8x3_t, vst3_s8, vld3_s8);
test_vld4_s8(i8, 32, int8x8x4_t, vst4_s8, vld4_s8);

test_vld2q_s8_x2(i8, 32, int8x16x2_t, vst2q_s8, vld2q_s8);
test_vld3q_s8_x3(i8, 48, int8x16x3_t, vst3q_s8, vld3q_s8);
test_vld4q_s8_x4(i8, 64, int8x16x4_t, vst4q_s8, vld4q_s8);
test_vld2q_s8(i8, 32, int8x16x2_t, vst2q_s8, vld2q_s8);
test_vld3q_s8(i8, 48, int8x16x3_t, vst3q_s8, vld3q_s8);
test_vld4q_s8(i8, 64, int8x16x4_t, vst4q_s8, vld4q_s8);

test_vld2_s16_x2(i16, 8, int16x4x2_t, vst2_s16, vld2_s16);
test_vld2_s16_x3(i16, 12, int16x4x3_t, vst3_s16, vld3_s16);
test_vld2_s16_x4(i16, 16, int16x4x4_t, vst4_s16, vld4_s16);
test_vld2_s16(i16, 8, int16x4x2_t, vst2_s16, vld2_s16);
test_vld3_s16(i16, 12, int16x4x3_t, vst3_s16, vld3_s16);
test_vld4_s16(i16, 16, int16x4x4_t, vst4_s16, vld4_s16);

test_vld2q_s16_x2(i16, 16, int16x8x2_t, vst2q_s16, vld2q_s16);
test_vld3q_s16_x3(i16, 24, int16x8x3_t, vst3q_s16, vld3q_s16);
test_vld4q_s16_x4(i16, 32, int16x8x4_t, vst4q_s16, vld4q_s16);
test_vld2q_s16(i16, 16, int16x8x2_t, vst2q_s16, vld2q_s16);
test_vld3q_s16(i16, 24, int16x8x3_t, vst3q_s16, vld3q_s16);
test_vld4q_s16(i16, 32, int16x8x4_t, vst4q_s16, vld4q_s16);

test_vld2_s32_x2(i32, 4, int32x2x2_t, vst2_s32, vld2_s32);
test_vld2_s32_x3(i32, 6, int32x2x3_t, vst3_s32, vld3_s32);
test_vld2_s32_x4(i32, 8, int32x2x4_t, vst4_s32, vld4_s32);
test_vld2_s32(i32, 4, int32x2x2_t, vst2_s32, vld2_s32);
test_vld3_s32(i32, 6, int32x2x3_t, vst3_s32, vld3_s32);
test_vld4_s32(i32, 8, int32x2x4_t, vst4_s32, vld4_s32);

test_vld2q_s32_x2(i32, 8, int32x4x2_t, vst2q_s32, vld2q_s32);
test_vld3q_s32_x3(i32, 12, int32x4x3_t, vst3q_s32, vld3q_s32);
test_vld4q_s32_x4(i32, 16, int32x4x4_t, vst4q_s32, vld4q_s32);
test_vld2q_s32(i32, 8, int32x4x2_t, vst2q_s32, vld2q_s32);
test_vld3q_s32(i32, 12, int32x4x3_t, vst3q_s32, vld3q_s32);
test_vld4q_s32(i32, 16, int32x4x4_t, vst4q_s32, vld4q_s32);

test_vld2_s64_x2(i64, 2, int64x1x2_t, vst2_s64, vld2_s64);
test_vld2_s64_x3(i64, 3, int64x1x3_t, vst3_s64, vld3_s64);
test_vld2_s64_x4(i64, 4, int64x1x4_t, vst4_s64, vld4_s64);
test_vld2_s64(i64, 2, int64x1x2_t, vst2_s64, vld2_s64);
test_vld3_s64(i64, 3, int64x1x3_t, vst3_s64, vld3_s64);
test_vld4_s64(i64, 4, int64x1x4_t, vst4_s64, vld4_s64);

test_vld2q_s64_x2(i64, 4, int64x2x2_t, vst2q_s64, vld2q_s64);
test_vld3q_s64_x3(i64, 6, int64x2x3_t, vst3q_s64, vld3q_s64);
test_vld4q_s64_x4(i64, 8, int64x2x4_t, vst4q_s64, vld4q_s64);
test_vld2q_s64(i64, 4, int64x2x2_t, vst2q_s64, vld2q_s64);
test_vld3q_s64(i64, 6, int64x2x3_t, vst3q_s64, vld3q_s64);
test_vld4q_s64(i64, 8, int64x2x4_t, vst4q_s64, vld4q_s64);

test_vld2_u8_x2(u8, 16, uint8x8x2_t, vst2_u8, vld2_u8);
test_vld2_u8_x3(u8, 24, uint8x8x3_t, vst3_u8, vld3_u8);
test_vld2_u8_x4(u8, 32, uint8x8x4_t, vst4_u8, vld4_u8);
test_vld2_u8(u8, 16, uint8x8x2_t, vst2_u8, vld2_u8);
test_vld3_u8(u8, 24, uint8x8x3_t, vst3_u8, vld3_u8);
test_vld4_u8(u8, 32, uint8x8x4_t, vst4_u8, vld4_u8);

test_vld2q_u8_x2(u8, 32, uint8x16x2_t, vst2q_u8, vld2q_u8);
test_vld3q_u8_x3(u8, 48, uint8x16x3_t, vst3q_u8, vld3q_u8);
test_vld4q_u8_x4(u8, 64, uint8x16x4_t, vst4q_u8, vld4q_u8);
test_vld2q_u8(u8, 32, uint8x16x2_t, vst2q_u8, vld2q_u8);
test_vld3q_u8(u8, 48, uint8x16x3_t, vst3q_u8, vld3q_u8);
test_vld4q_u8(u8, 64, uint8x16x4_t, vst4q_u8, vld4q_u8);

test_vld2_u16_x2(u16, 8, uint16x4x2_t, vst2_u16, vld2_u16);
test_vld2_u16_x3(u16, 12, uint16x4x3_t, vst3_u16, vld3_u16);
test_vld2_u16_x4(u16, 16, uint16x4x4_t, vst4_u16, vld4_u16);
test_vld2_u16(u16, 8, uint16x4x2_t, vst2_u16, vld2_u16);
test_vld3_u16(u16, 12, uint16x4x3_t, vst3_u16, vld3_u16);
test_vld4_u16(u16, 16, uint16x4x4_t, vst4_u16, vld4_u16);

test_vld2q_u16_x2(u16, 16, uint16x8x2_t, vst2q_u16, vld2q_u16);
test_vld3q_u16_x3(u16, 24, uint16x8x3_t, vst3q_u16, vld3q_u16);
test_vld4q_u16_x4(u16, 32, uint16x8x4_t, vst4q_u16, vld4q_u16);
test_vld2q_u16(u16, 16, uint16x8x2_t, vst2q_u16, vld2q_u16);
test_vld3q_u16(u16, 24, uint16x8x3_t, vst3q_u16, vld3q_u16);
test_vld4q_u16(u16, 32, uint16x8x4_t, vst4q_u16, vld4q_u16);

test_vld2_u32_x2(u32, 4, uint32x2x2_t, vst2_u32, vld2_u32);
test_vld2_u32_x3(u32, 6, uint32x2x3_t, vst3_u32, vld3_u32);
test_vld2_u32_x4(u32, 8, uint32x2x4_t, vst4_u32, vld4_u32);
test_vld2_u32(u32, 4, uint32x2x2_t, vst2_u32, vld2_u32);
test_vld3_u32(u32, 6, uint32x2x3_t, vst3_u32, vld3_u32);
test_vld4_u32(u32, 8, uint32x2x4_t, vst4_u32, vld4_u32);

test_vld2q_u32_x2(u32, 8, uint32x4x2_t, vst2q_u32, vld2q_u32);
test_vld3q_u32_x3(u32, 12, uint32x4x3_t, vst3q_u32, vld3q_u32);
test_vld4q_u32_x4(u32, 16, uint32x4x4_t, vst4q_u32, vld4q_u32);
test_vld2q_u32(u32, 8, uint32x4x2_t, vst2q_u32, vld2q_u32);
test_vld3q_u32(u32, 12, uint32x4x3_t, vst3q_u32, vld3q_u32);
test_vld4q_u32(u32, 16, uint32x4x4_t, vst4q_u32, vld4q_u32);

test_vld2_u64_x2(u64, 2, uint64x1x2_t, vst2_u64, vld2_u64);
test_vld2_u64_x3(u64, 3, uint64x1x3_t, vst3_u64, vld3_u64);
test_vld2_u64_x4(u64, 4, uint64x1x4_t, vst4_u64, vld4_u64);
test_vld2_u64(u64, 2, uint64x1x2_t, vst2_u64, vld2_u64);
test_vld3_u64(u64, 3, uint64x1x3_t, vst3_u64, vld3_u64);
test_vld4_u64(u64, 4, uint64x1x4_t, vst4_u64, vld4_u64);

test_vld2q_u64_x2(u64, 4, uint64x2x2_t, vst2q_u64, vld2q_u64);
test_vld3q_u64_x3(u64, 6, uint64x2x3_t, vst3q_u64, vld3q_u64);
test_vld4q_u64_x4(u64, 8, uint64x2x4_t, vst4q_u64, vld4q_u64);
test_vld2q_u64(u64, 4, uint64x2x2_t, vst2q_u64, vld2q_u64);
test_vld3q_u64(u64, 6, uint64x2x3_t, vst3q_u64, vld3q_u64);
test_vld4q_u64(u64, 8, uint64x2x4_t, vst4q_u64, vld4q_u64);

test_vld2_p8_x2(p8, 16, poly8x8x2_t, vst2_p8, vld2_p8);
test_vld2_p8_x3(p8, 24, poly8x8x3_t, vst3_p8, vld3_p8);
test_vld2_p8_x4(p8, 32, poly8x8x4_t, vst4_p8, vld4_p8);
test_vld2_p8(p8, 16, poly8x8x2_t, vst2_p8, vld2_p8);
test_vld3_p8(p8, 24, poly8x8x3_t, vst3_p8, vld3_p8);
test_vld4_p8(p8, 32, poly8x8x4_t, vst4_p8, vld4_p8);

test_vld2q_p8_x2(p8, 32, poly8x16x2_t, vst2q_p8, vld2q_p8);
test_vld3q_p8_x3(p8, 48, poly8x16x3_t, vst3q_p8, vld3q_p8);
test_vld4q_p8_x4(p8, 64, poly8x16x4_t, vst4q_p8, vld4q_p8);
test_vld2q_p8(p8, 32, poly8x16x2_t, vst2q_p8, vld2q_p8);
test_vld3q_p8(p8, 48, poly8x16x3_t, vst3q_p8, vld3q_p8);
test_vld4q_p8(p8, 64, poly8x16x4_t, vst4q_p8, vld4q_p8);

test_vld2_p16_x2(p16, 8, poly16x4x2_t, vst2_p16, vld2_p16);
test_vld2_p16_x3(p16, 12, poly16x4x3_t, vst3_p16, vld3_p16);
test_vld2_p16_x4(p16, 16, poly16x4x4_t, vst4_p16, vld4_p16);
test_vld2_p16(p16, 8, poly16x4x2_t, vst2_p16, vld2_p16);
test_vld3_p16(p16, 12, poly16x4x3_t, vst3_p16, vld3_p16);
test_vld4_p16(p16, 16, poly16x4x4_t, vst4_p16, vld4_p16);

test_vld2q_p16_x2(p16, 16, poly16x8x2_t, vst2q_p16, vld2q_p16);
test_vld3q_p16_x3(p16, 24, poly16x8x3_t, vst3q_p16, vld3q_p16);
test_vld4q_p16_x4(p16, 32, poly16x8x4_t, vst4q_p16, vld4q_p16);
test_vld2q_p16(p16, 16, poly16x8x2_t, vst2q_p16, vld2q_p16);
test_vld3q_p16(p16, 24, poly16x8x3_t, vst3q_p16, vld3q_p16);
test_vld4q_p16(p16, 32, poly16x8x4_t, vst4q_p16, vld4q_p16);
}

macro_rules! lane_wide_store_load_roundtrip {
($elem_ty:ty, $len:expr, $idx:expr, $vec_ty:ty, $store:ident, $load:ident) => {
let vals: [$elem_ty; $len] = crate::array::from_fn(|i| i as $elem_ty);
let a: $vec_ty = transmute(vals);
let mut tmp = [0 as $elem_ty; 4];
$store::<$idx>(tmp.as_mut_ptr().cast(), a);
let r: $vec_ty = $load::<$idx>(tmp.as_ptr().cast(), a);
let out: [$elem_ty; $len] = transmute(r);
assert_eq!(out, vals);
};
}

macro_rules! lane_wide_store_load_roundtrip_neon {
($( $name:ident $args:tt);* $(;)?) => {
$(
#[cfg_attr(miri, ignore)]
#[simd_test(enable = "neon")]
unsafe fn $name() {
lane_wide_store_load_roundtrip! $args;
}
)*
};
}

lane_wide_store_load_roundtrip_neon! {
test_vld2q_lane_s8(i8, 32, 15, int8x16x2_t, vst2q_lane_s8, vld2q_lane_s8);
test_vld3q_lane_s8(i8, 48, 15, int8x16x3_t, vst3q_lane_s8, vld3q_lane_s8);
test_vld4q_lane_s8(i8, 64, 15, int8x16x4_t, vst4q_lane_s8, vld4q_lane_s8);

test_vld2q_lane_u8(u8, 32, 15, uint8x16x2_t, vst2q_lane_u8, vld2q_lane_u8);
test_vld3q_lane_u8(u8, 48, 15, uint8x16x3_t, vst3q_lane_u8, vld3q_lane_u8);
test_vld4q_lane_u8(u8, 64, 15, uint8x16x4_t, vst4q_lane_u8, vld4q_lane_u8);

test_vld2_lane_s64(i64, 2, 0, int64x1x2_t, vst2_lane_s64, vld2_lane_s64);
test_vld3_lane_s64(i64, 3, 0, int64x1x3_t, vst3_lane_s64, vld3_lane_s64);
test_vld4_lane_s64(i64, 4, 0, int64x1x4_t, vst4_lane_s64, vld4_lane_s64);
test_vld2q_lane_s64(i64, 4, 1, int64x2x2_t, vst2q_lane_s64, vld2q_lane_s64);
test_vld3q_lane_s64(i64, 6, 1, int64x2x3_t, vst3q_lane_s64, vld3q_lane_s64);
test_vld4q_lane_s64(i64, 8, 1, int64x2x4_t, vst4q_lane_s64, vld4q_lane_s64);

test_vld2_lane_u64(u64, 2, 0, uint64x1x2_t, vst2_lane_u64, vld2_lane_u64);
test_vld3_lane_u64(u64, 3, 0, uint64x1x3_t, vst3_lane_u64, vld3_lane_u64);
test_vld4_lane_u64(u64, 4, 0, uint64x1x4_t, vst4_lane_u64, vld4_lane_u64);
test_vld2q_lane_u64(u64, 4, 1, uint64x2x2_t, vst2q_lane_u64, vld2q_lane_u64);
test_vld3q_lane_u64(u64, 6, 1, uint64x2x3_t, vst3q_lane_u64, vld3q_lane_u64);
test_vld4q_lane_u64(u64, 8, 1, uint64x2x4_t, vst4q_lane_u64, vld4q_lane_u64);
}
}

Expand Down
Loading