Skip to main content Crate fastalp Copy item path Source pub use header::ChunkType ;pub use header::read_count ;pub use capi ::*;capibitpack capi capiheader arr_8 Constructs an 8-element array by applying an expression to index 0..8.
通过对索引 0..8 分别求值构造 8 元素数组（完全内联，消除手动 +1, +2, +3） match_pack_23 Dispatches bit-width 1..=20, 24, 28, 32 to monomorphized chunk packing function.
统一分发 1..=20, 24, 28, 32 位宽至单态化 8 元素块打包内核 unroll_8 Unrolls a block 8 times with $idx bound to 0..8.
将逻辑按索引 0..8 重复展开 8 次顺序执行 write_4 Writes 4 evaluated elements to consecutive raw pointer memory *($dst).add(k) = expr(k).
向连续裸指针内存顺序写入 4 个计算结果（局部绑定 base 指针） write_8 Writes 8 evaluated elements to consecutive raw pointer memory *($dst).add(k) = expr(k).
向连续裸指针内存顺序写入 8 个计算结果（局部绑定 base 指针，杜绝表达式重复求值） AlpParams Compact ALP encoding and decoding parameters.
封装 ALP 核心参数 (exp, fac, bit_width, use_div)，消除多参数冗余传递并统一头部打包与类型判断 BestParams Optimal parameters discovered by sampling
采样与最优系数选择结果 Encoder Stateful encoder that caches optimal parameters and scratch buffers across adjacent chunks.
状态化编码器：在连续数据块编码时复用已探测的最优参数与内部工作缓冲区，消除重复采样开销与内存分配。 Error CHUNK_SIZE Standard chunk size for typical ALP blocks.
典型 ALP 标准块大小 (1024) CHUNK_SIZE_1024 AlpFloat ALP floating-point abstraction trait (unifies zero-cost f32 and f64 compression).
ALP 浮点数抽象特征（统一 f32 与 f64 零成本编解码） compress Generic floating-point slice compression.
通用压缩浮点数切片 compress_delta Generic floating-point slice compression enforcing Delta differential mode.
强制使用 Delta 差分模式压缩浮点数切片 compress_delta_into Floating-point compression with enforced Delta differential encoding.
强制使用 Delta 一阶差分模式压缩浮点数组并直接写入 dst 缓冲区 compress_into Generic floating-point compression writing directly into dst buffer.
通用压缩浮点数组并直接写入 dst 缓冲区（自适应选择 FOR 或 Delta 差分模式） count Reads the element count from the compressed data header in O(1) time without decompressing the payload. decompress Generic floating-point slice decompression.
通用解压浮点数切片 decompress_into Generic floating-point decompression into dst buffer.
通用解压浮点数组至 dst 缓冲区（自动分发 RAW、标准 FOR 与 Delta 差分块） decompress_into_raw ⚠ Generic floating-point decompression directly into raw pointer memory.
通用解压浮点数组至裸指针内存（零堆分配、零内存拷贝，避免未初始化切片构造） decompress_into_slice Generic floating-point decompression into destination slice.
通用解压浮点数组至目标切片（零堆分配、零内存拷贝） max_compressed_size Computes the maximum possible compressed buffer size in bytes for count values of type F.
Guaranteed to never overflow even in the worst-case uncompressible fallback. Result