stringzilla 5.0.2

Search, hash, sort, fingerprint, and fuzzy-match strings faster via SWAR, SIMD, and GPGPU
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78

[package]
authors = ["Ash Vardanian <1983160+ashvardanian@users.noreply.github.com>"]
categories = [
    "text-processing",
    "hardware-support",
    "no-std",
    "wasm",
    "external-ffi-bindings",
]
description = "Search, hash, sort, fingerprint, and fuzzy-match strings faster via SWAR, SIMD, and GPGPU"
documentation = "https://docs.rs/stringzilla"
edition = "2021"
homepage = "https://ashvardanian.com/posts/stringzilla/"
include = [
    "rust/**",
    "c/**",
    "include/**",
    "forkunion/include/**",
    "probes/**",
    "build.rs",
]
keywords = ["simd", "search", "retrieval", "hash", "sort"]
license = "Apache-2.0"
name = "stringzilla"
publish = true
repository = "https://github.com/ashvardanian/StringZilla"
version = "5.0.2"

[lib]
name = "stringzilla"
path = "rust/lib.rs"

[features]
default = ["std", "dynamic-dispatch"]
std = []
# Runtime SIMD dispatch: compile every ISA tier and pick the best one at load via the dispatch table
# (mirrors CMake's `stringzilla_shared`). Disable (`--no-default-features`) to resolve the tier at
# compile time instead - the analog of the header-only `stringzilla_header`.
cpus = [
    "std",            # std is required for multi-threaded backend
    "allocator-api2",
    "stringtape",
    "dep:forkunion",  # Compiles and links the thread-pool runtime; StringZillas only consumes its C API
] # Multi-threaded CPU backend (StringZillas)
cuda = ["std", "cpus"] # CUDA GPU backend (includes multi-threaded CPU backend)
dynamic-dispatch = []
rocm = ["std", "cpus"] # ROCm GPU backend (includes multi-threaded CPU backend)

[dependencies]
allocator-api2 = { version = "0.4", optional = true }
forkunion = { version = "3.0", optional = true }
stringtape = { version = "2.4.2", optional = true }

[build-dependencies]
cc = "1.2.47"

[lints.clippy]
# Catch platform-specific type issues like `c_char` differences
cast-sign-loss = "warn"
# Catch potential FFI issues
not-unsafe-ptr-arg-deref = "warn"
# Catch undefined behavior with pointers
invalid_null_ptr_usage = "warn"
# Catch transmute issues
transmute_ptr_to_ptr = "warn"

[lints.rust]
# Catch ABI mismatches in FFI
improper_ctypes = "warn"
# Catch platform-specific code that might fail on other targets
unexpected_cfgs = "warn"

[profile.release]
# Align the optimized build with the CMake and Python builds at -O2; `build.rs` forwards this `OPT_LEVEL`
# to `cc` for the bundled C/C++ sources. -O3's aggressive inlining bloats the heavy `utf8_uncased`
# translation unit (~3 MB object, ~44s to compile) with no measured runtime win. `dev` keeps the default
# -O0 so clean builds stay fast.
opt-level = 2