Skip to content

Make SIMD stores safe #238

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
Shnatsel merged 1 commit into from
May 29, 2026
Merged

Make SIMD stores safe #238

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
192 changes: 24 additions & 168 deletions fearless_simd/src/generated/avx2.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -128,13 +128,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f32x4(self, a: f32x4<Self>, dest: &mut [f32; 4usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f32,
dest.as_mut_ptr(),
4usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f32x4(self, a: u8x16<Self>) -> f32x4<Self> {
Expand Down Expand Up @@ -434,13 +428,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i8x16(self, a: i8x16<Self>, dest: &mut [i8; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i8,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i8x16(self, a: u8x16<Self>) -> i8x16<Self> {
Expand Down Expand Up @@ -664,13 +652,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u8x16(self, a: u8x16<Self>, dest: &mut [u8; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u8,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u8x16(self, a: u8x16<Self>) -> u8x16<Self> {
Expand Down Expand Up @@ -987,13 +969,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i16x8(self, a: i16x8<Self>, dest: &mut [i16; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i16,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i16x8(self, a: u8x16<Self>) -> i16x8<Self> {
Expand Down Expand Up @@ -1192,13 +1168,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u16x8(self, a: u16x8<Self>, dest: &mut [u16; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u16,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u16x8(self, a: u8x16<Self>) -> u16x8<Self> {
Expand Down Expand Up @@ -1490,13 +1460,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i32x4(self, a: i32x4<Self>, dest: &mut [i32; 4usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i32,
dest.as_mut_ptr(),
4usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i32x4(self, a: u8x16<Self>) -> i32x4<Self> {
Expand Down Expand Up @@ -1697,13 +1661,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u32x4(self, a: u32x4<Self>, dest: &mut [u32; 4usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u32,
dest.as_mut_ptr(),
4usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u32x4(self, a: u8x16<Self>) -> u32x4<Self> {
Expand Down Expand Up @@ -1998,13 +1956,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f64x2(self, a: f64x2<Self>, dest: &mut [f64; 2usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f64,
dest.as_mut_ptr(),
2usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f64x2(self, a: u8x16<Self>) -> f64x2<Self> {
Expand Down Expand Up @@ -2315,13 +2267,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f32x8(self, a: f32x8<Self>, dest: &mut [f32; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f32,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f32x8(self, a: u8x32<Self>) -> f32x8<Self> {
Expand Down Expand Up @@ -2684,13 +2630,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i8x32(self, a: i8x32<Self>, dest: &mut [i8; 32usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i8,
dest.as_mut_ptr(),
32usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i8x32(self, a: u8x32<Self>) -> i8x32<Self> {
Expand Down Expand Up @@ -2997,13 +2937,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u8x32(self, a: u8x32<Self>, dest: &mut [u8; 32usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u8,
dest.as_mut_ptr(),
32usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u8x32(self, a: u8x32<Self>) -> u8x32<Self> {
Expand Down Expand Up @@ -3425,13 +3359,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i16x16(self, a: i16x16<Self>, dest: &mut [i16; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i16,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i16x16(self, a: u8x32<Self>) -> i16x16<Self> {
Expand Down Expand Up @@ -3719,13 +3647,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u16x16(self, a: u16x16<Self>, dest: &mut [u16; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u16,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u16x16(self, a: u8x32<Self>) -> u16x16<Self> {
Expand Down Expand Up @@ -4133,13 +4055,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i32x8(self, a: i32x8<Self>, dest: &mut [i32; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i32,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i32x8(self, a: u8x32<Self>) -> i32x8<Self> {
Expand Down Expand Up @@ -4401,13 +4317,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u32x8(self, a: u32x8<Self>, dest: &mut [u32; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u32,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u32x8(self, a: u8x32<Self>) -> u32x8<Self> {
Expand Down Expand Up @@ -4778,13 +4688,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f64x4(self, a: f64x4<Self>, dest: &mut [f64; 4usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f64,
dest.as_mut_ptr(),
4usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f64x4(self, a: u8x32<Self>) -> f64x4<Self> {
Expand Down Expand Up @@ -5171,13 +5075,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f32x16(self, a: f32x16<Self>, dest: &mut [f32; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f32,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f32x16(self, a: u8x64<Self>) -> f32x16<Self> {
Expand Down Expand Up @@ -5594,13 +5492,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i8x64(self, a: i8x64<Self>, dest: &mut [i8; 64usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i8,
dest.as_mut_ptr(),
64usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i8x64(self, a: u8x64<Self>) -> i8x64<Self> {
Expand Down Expand Up @@ -5872,13 +5764,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u8x64(self, a: u8x64<Self>, dest: &mut [u8; 64usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u8,
dest.as_mut_ptr(),
64usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u8x64(self, a: u8x64<Self>) -> u8x64<Self> {
Expand Down Expand Up @@ -6339,13 +6225,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i16x32(self, a: i16x32<Self>, dest: &mut [i16; 32usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i16,
dest.as_mut_ptr(),
32usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i16x32(self, a: u8x64<Self>) -> i16x32<Self> {
Expand Down Expand Up @@ -6626,13 +6506,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u16x32(self, a: u16x32<Self>, dest: &mut [u16; 32usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u16,
dest.as_mut_ptr(),
32usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u16x32(self, a: u8x64<Self>) -> u16x32<Self> {
Expand Down Expand Up @@ -7097,13 +6971,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_i32x16(self, a: i32x16<Self>, dest: &mut [i32; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const i32,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_i32x16(self, a: u8x64<Self>) -> i32x16<Self> {
Expand Down Expand Up @@ -7380,13 +7248,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_u32x16(self, a: u32x16<Self>, dest: &mut [u32; 16usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const u32,
dest.as_mut_ptr(),
16usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_u32x16(self, a: u8x64<Self>) -> u32x16<Self> {
Expand Down Expand Up @@ -7824,13 +7686,7 @@ impl Simd for Avx2 {
}
#[inline(always)]
fn store_array_f64x8(self, a: f64x8<Self>, dest: &mut [f64; 8usize]) -> () {
unsafe {
core::ptr::copy_nonoverlapping(
(&raw const a.val.0) as *const f64,
dest.as_mut_ptr(),
8usize,
);
}
crate::transmute::checked_transmute_store(a.val.0, dest);
}
#[inline(always)]
fn cvt_from_bytes_f64x8(self, a: u8x64<Self>) -> f64x8<Self> {
Expand Down
Loading
Loading