pub struct PerformanceConfig {
pub use_avx512: bool,
pub use_avx2: bool,
pub prefetch_data: bool,
pub parallel_chunks: bool,
pub chunk_size: usize,
}Expand description
Performance configuration for hasher.
use_avx512/use_avx2 are real and wired up on std + x86_64
builds, but they mean different things depending on which method you
call:
Sha3_512Kernel::hash/update/finalize(single message): dispatch every Keccak-f[1600] permutation tosimd_avx512whenuse_avx512istrueand the CPU supports AVX-512F (falling back to scalar otherwise;use_avx2has no effect here — see below). This is not a clear win, which is whyuse_avx512defaults tofalse: measured on this crate’s owncargo bench --bench sha3_bench(this container’s Xeon), it’s within noise of scalar at 64 KiB payloads, ~3-4% slower at 1 MiB, and ~10% slower at 4 KiB. A single Keccak-f[1600] instance has no independent lanes for a 512-bit-wide register to parallelize within — π’s cross-row mixing costs real shuffle/blend instructions that scalar code doesn’t pay. Set ittruehere only if you’ve benchmarked your own workload and it actually helps.Sha3_512Kernel::hash_many(a batch of independent messages): dispatches to AVX-512×8 or AVX2×4 batched permutations insrc/simd_parallel.rs. Here the flags select a real, measured win — independent messages give the vector width genuine parallel work. See the crate-level “SIMD Status” section for full numbers; in short, AVX-512×8 measured ~5.5x over a scalar loop, AVX2×4 only ~1.2x (AVX2 has no 64-bit rotate instruction, so ρ is emulated via shift+shift+or, eating most of its 4-way parallelism).
prefetch_data, parallel_chunks, and chunk_size remain no-ops —
nothing reads them yet.
Fields§
§use_avx512: boolReal and wired up (std + x86_64 only) for both hash() and
hash_many() — see the struct-level docs above; the two methods
get very different returns on setting this true.
use_avx2: boolNo-op for hash() (AVX2 cannot accelerate a single-message
permutation — see struct docs); real, and a modest win, for
hash_many().
prefetch_data: boolNo-op today; reserved for a future prefetch optimization.
parallel_chunks: boolNo-op today; reserved for a future parallel-chunk optimization.
chunk_size: usizeNo-op today.
Trait Implementations§
Source§impl Clone for PerformanceConfig
impl Clone for PerformanceConfig
impl Copy for PerformanceConfig
Source§impl Debug for PerformanceConfig
impl Debug for PerformanceConfig
Auto Trait Implementations§
impl Freeze for PerformanceConfig
impl RefUnwindSafe for PerformanceConfig
impl Send for PerformanceConfig
impl Sync for PerformanceConfig
impl Unpin for PerformanceConfig
impl UnsafeUnpin for PerformanceConfig
impl UnwindSafe for PerformanceConfig
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more
impl<ST, DT> CastableFrom<ST, Initialized, Initialized> for DT
impl<ST, DT> CastableFrom<ST, Uninit, Uninit> for DT
Source§impl<T> CloneToUninit for Twhere
T: Clone,
impl<T> CloneToUninit for Twhere
T: Clone,
Source§impl<T> IntoEither for T
impl<T> IntoEither for T
Source§fn into_either(self, into_left: bool) -> Either<Self, Self> ⓘ
fn into_either(self, into_left: bool) -> Either<Self, Self> ⓘ
Converts
self into a Left variant of Either<Self, Self>
if into_left is true.
Converts self into a Right variant of Either<Self, Self>
otherwise. Read moreSource§fn into_either_with<F>(self, into_left: F) -> Either<Self, Self> ⓘ
fn into_either_with<F>(self, into_left: F) -> Either<Self, Self> ⓘ
Converts
self into a Left variant of Either<Self, Self>
if into_left(&self) returns true.
Converts self into a Right variant of Either<Self, Self>
otherwise. Read more