rust-lang / portable-simd

The testing ground for the future of portable SIMD in Rust

rust-lang/portable-simd Issues

When is it available in the stable version of rust?
Closed a month ago6
Seemless softwere/hardwere computation interoperation
Updated 2 months ago1
migration pitfall from packed_simd: PartialOrd behaves differently?
Updated 2 months ago4
Bad codegen for bitwise OR/AND masks
Updated 2 months ago4
Saturating casts for integers, like `_mm_packus_epi16` and `_mm_packs_epi16`
Updated 2 months ago5
Bit operations on float
Updated 2 months ago8
Bounds check is not eliminated
Updated 2 months ago5
UB for certain transmutes could be easier to detect
Closed 2 months ago
API to specify max number of elements
Updated 2 months ago5
Missing SIMD version of `round_ties_even`
Updated 2 months ago1
from_bitmask_vector on big-endian calls simd_select_bitmask with the mask at the wrong end of the byte
Closed 3 months ago1
MulAdd and MulAddAssign
Updated 4 months ago1
Move intrinsic declarations to core and verify behavior
Closed 4 months ago1
Simd::load implementation does not match documentation
Closed 4 months ago
Critical issues before stabilization
Updated 4 months ago21
Find-first-set style operations directly on masks
Closed 5 months ago2
Ensure `load`s of vector types have `!noundef`
Updated 5 months ago
Support `MaybeUninit` vectors
Updated 6 months ago9
Support for `_mm_maddubs_epi16` and `_mm_maddubs_epi16` and similar?
Updated 7 months ago2
non-power-of-2 Simd types have wrong size
Updated 7 months ago42
Poor performance on u64 and i64
Updated 7 months ago5
don't use `read_unaligned` to read from an array casted to `*const Simd`
Updated 7 months ago1
Missing Neg for unsigned types
Closed 7 months ago7
Not using WASM bitmask / all_true instructions
Closed 9 months ago4
`wasm32-unknown-unknown` doesn't use simd instructions (at least, for one particular example, but plausibly more)
Updated 9 months ago3
How to `Simd::cast` now that it is removed
Closed 9 months ago6
`cast_ptr` is willing to cast between mut and non-mut pointers
Closed a year ago3
Looking for "blend" methods.
Closed a year ago2
ensure array/slice load/store functions actually optimize to LLVM IR vector `load`/`store` instructions
Updated a year ago
Support avx512 bitmasks with dynamic feature detection
Updated a year ago3
explicit simd store -- opposite of Simd::from_slice
Closed a year ago2
Weird codegen with avx512 feature and masks > 8 lanes
Updated a year ago6
LLVM error when using f64x8.scatter_select
Updated a year ago3
Float to integer casts on x86 produce scalar instructions
Updated a year ago2
Simd-using functions sometimes scalarize after inlining, even if they use vector ops on their own
Updated a year ago8
gather with i32 indexes
Updated a year ago11
ChaCha20 as a possible addition to examples area
Closed a year ago9
Transpose methods?
Updated a year ago8
change gather/scatter mask sizes to match data size not index/address size since that matches x86_64
Updated a year ago
Suboptimal codegen for mask8x32 all
Updated a year ago4
Should not use transmute to cast between pointers and integers
Closed a year ago10
`simd_swizzle` plays not well with type inference
Updated a year ago1
missing mask8x4
Closed a year ago1
Trait "From<Mask<T, LANES>> for Simd<T, LANES>" is currently not implemented
Closed 2 years ago2
Should the full_mask / bitmask distinction work when used via stdlib?
Closed 2 years ago3
Update docs
Closed 2 years ago2
Mask::test/test_unchecked not getting fully inlined
Closed 2 years ago3
Using Simd interleave leads to wrong result for Lanes = 1
Closed 2 years ago1
Arbitrary bitselect
Closed 2 years ago7
Conversion from arch types
Closed 2 years ago6