duckstation

duckstation, but archived from the revision just before upstream changed it to a proprietary software project, this version is the libre one
git clone https://git.neptards.moe/u3shit/duckstation.git
Log | Files | Refs | README | LICENSE

README.md (9819B)


      1 xxHash - Extremely fast hash algorithm
      2 ======================================
      3 
      4 <!-- TODO: Update. -->
      5 xxHash is an Extremely fast Hash algorithm, running at RAM speed limits.
      6 It successfully completes the [SMHasher](https://code.google.com/p/smhasher/wiki/SMHasher) test suite
      7 which evaluates collision, dispersion and randomness qualities of hash functions.
      8 Code is highly portable, and hashes are identical on all platforms (little / big endian).
      9 
     10 |Branch      |Status   |
     11 |------------|---------|
     12 |master      | [![Build Status](https://travis-ci.org/Cyan4973/xxHash.svg?branch=master)](https://travis-ci.org/Cyan4973/xxHash?branch=master) |
     13 |dev         | [![Build Status](https://travis-ci.org/Cyan4973/xxHash.svg?branch=dev)](https://travis-ci.org/Cyan4973/xxHash?branch=dev) |
     14 
     15 
     16 
     17 Benchmarks
     18 -------------------------
     19 
     20 The benchmark uses SMHasher speed test, compiled with Visual 2010 on a Windows Seven 32-bit box.
     21 The reference system uses a Core 2 Duo @3GHz
     22 
     23 
     24 | Name          |   Speed            | Quality | Author            |
     25 |---------------|--------------------|:-------:|-------------------|
     26 | [xxHash]      | 5.4 GB/s           |   10    | Y.C.              |
     27 | MurmurHash 3a | 2.7 GB/s           |   10    | Austin Appleby    |
     28 | SBox          | 1.4 GB/s           |    9    | Bret Mulvey       |
     29 | Lookup3       | 1.2 GB/s           |    9    | Bob Jenkins       |
     30 | CityHash64    | 1.05 GB/s          |   10    | Pike & Alakuijala |
     31 | FNV           | 0.55 GB/s          |    5    | Fowler, Noll, Vo  |
     32 | CRC32         | 0.43 GB/s &dagger; |    9    |                   |
     33 | MD5-32        | 0.33 GB/s          |   10    | Ronald L.Rivest   |
     34 | SHA1-32       | 0.28 GB/s          |   10    |                   |
     35 
     36 [xxHash]: https://www.xxhash.com
     37 
     38 Note &dagger;: SMHasher's CRC32 implementation is known to be slow. Faster implementations exist.
     39 
     40 Q.Score is a measure of quality of the hash function.
     41 It depends on successfully passing SMHasher test set.
     42 10 is a perfect score.
     43 Algorithms with a score < 5 are not listed on this table.
     44 
     45 A more recent version, XXH64, has been created thanks to [Mathias Westerdahl](https://github.com/JCash),
     46 which offers superior speed and dispersion for 64-bit systems.
     47 Note however that 32-bit applications will still run faster using the 32-bit version.
     48 
     49 SMHasher speed test, compiled using GCC 4.8.2, on Linux Mint 64-bit.
     50 The reference system uses a Core i5-3340M @2.7GHz
     51 
     52 | Version    | Speed on 64-bit  | Speed on 32-bit  |
     53 |------------|------------------|------------------|
     54 | XXH64      | 13.8 GB/s        |  1.9 GB/s        |
     55 | XXH32      |  6.8 GB/s        |  6.0 GB/s        |
     56 
     57 This project also includes a command line utility, named `xxhsum`, offering similar features to `md5sum`,
     58 thanks to [Takayuki Matsuoka](https://github.com/t-mat)'s contributions.
     59 
     60 
     61 ### License
     62 
     63 The library files `xxhash.c` and `xxhash.h` are BSD licensed.
     64 The utility `xxhsum` is GPL licensed.
     65 
     66 
     67 ### New hash algorithms
     68 
     69 Starting with `v0.7.0`, the library includes a new algorithm named `XXH3`,
     70 which is able to generate 64 and 128-bit hashes.
     71 
     72 The new algorithm is much faster than its predecessors for both long and small inputs,
     73 which can be observed in the following graphs:
     74 
     75 ![XXH3, bargraph](https://user-images.githubusercontent.com/750081/61976096-b3a35f00-af9f-11e9-8229-e0afc506c6ec.png)
     76 
     77 ![XXH3, latency, random size](https://user-images.githubusercontent.com/750081/61976089-aedeab00-af9f-11e9-9239-e5375d6c080f.png)
     78 
     79 To access these new prototypes, one needs to unlock their declaration, using the build macro `XXH_STATIC_LINKING_ONLY`.
     80 
     81 The algorithm is currently in development, meaning its return values might still change in future versions.
     82 However, the API is stable, and can be used in production, typically for ephemeral
     83 data (produced and consumed in same session).
     84 
     85 `XXH3`'s return values will be finalized upon reaching `v0.8.0`.
     86 
     87 
     88 ### Build modifiers
     89 
     90 The following macros can be set at compilation time to modify libxxhash's behavior. They are all disabled by default.
     91 
     92 - `XXH_INLINE_ALL`: Make all functions `inline`, with implementations being directly included within `xxhash.h`.
     93                     Inlining functions is beneficial for speed on small keys.
     94                     It's _extremely effective_ when key length is expressed as _a compile time constant_,
     95                     with performance improvements being observed in the +200% range .
     96                     See [this article](https://fastcompression.blogspot.com/2018/03/xxhash-for-small-keys-impressive-power.html) for details.
     97                     Note: there is no need to compile an `xxhash.o` object file in this case.
     98 - `XXH_NO_INLINE_HINTS`: By default, xxHash uses tricks like `__attribute__((always_inline))` and `__forceinline` to try and improve performance at the cost of code size. Defining this to 1 will mark all internal functions as `static`, allowing the compiler to decide whether to inline a function or not. This is very useful when optimizing for the smallest binary size, and it is automatically defined when compiling with `-O0`, `-Os`, `-Oz`, or `-fno-inline` on GCC and Clang. This may also increase performance depending on the compiler and the architecture.
     99 - `XXH_REROLL`: Reduces the size of the generated code by not unrolling some loops. Impact on performance may vary, depending on the platform and the algorithm.
    100 - `XXH_ACCEPT_NULL_INPUT_POINTER`: if set to `1`, when input is a `NULL` pointer,
    101                                    xxHash'd result is the same as a zero-length input
    102                                    (instead of a dereference segfault).
    103                                    Adds one branch at the beginning of the hash.
    104 - `XXH_FORCE_MEMORY_ACCESS`: The default method `0` uses a portable `memcpy()` notation.
    105                              Method `1` uses a gcc-specific `packed` attribute, which can provide better performance for some targets.
    106                              Method `2` forces unaligned reads, which is not standards compliant, but might sometimes be the only way to extract better read performance.
    107                              Method `3` uses a byteshift operation, which is best for old compilers which don't inline `memcpy()` or big-endian systems without a byteswap instruction
    108 - `XXH_CPU_LITTLE_ENDIAN`: By default, endianess is determined at compile time.
    109                            It's possible to skip auto-detection and force format to little-endian, by setting this macro to 1.
    110                             Setting it to 0 forces big-endian.
    111 - `XXH_PRIVATE_API`: same impact as `XXH_INLINE_ALL`.
    112                      Name underlines that XXH_* symbols will not be exported.
    113 - `XXH_NAMESPACE`: Prefixes all symbols with the value of `XXH_NAMESPACE`.
    114                     Useful to evade symbol naming collisions,
    115                     in case of multiple inclusions of xxHash's source code.
    116                     Client applications can still use the regular function name,
    117                     as symbols are automatically translated through `xxhash.h`.
    118 - `XXH_STATIC_LINKING_ONLY`: gives access to the state declaration for static allocation.
    119                              Incompatible with dynamic linking, due to risks of ABI changes.
    120 - `XXH_NO_LONG_LONG`: removes support for XXH3 and XXH64 for targets without 64-bit support.
    121 - `XXH_IMPORT`: MSVC specific: should only be defined for dynamic linking, as it prevents linkage errors.
    122 
    123 
    124 ### Building xxHash - Using vcpkg
    125 
    126 You can download and install xxHash using the [vcpkg](https://github.com/Microsoft/vcpkg) dependency manager:
    127 
    128     git clone https://github.com/Microsoft/vcpkg.git
    129     cd vcpkg
    130     ./bootstrap-vcpkg.sh
    131     ./vcpkg integrate install
    132     ./vcpkg install xxhash
    133 
    134 The xxHash port in vcpkg is kept up to date by Microsoft team members and community contributors. If the version is out of date, please [create an issue or pull request](https://github.com/Microsoft/vcpkg) on the vcpkg repository.
    135 
    136 
    137 ### Example
    138 
    139 Calling xxhash 64-bit variant from a C program:
    140 
    141 ```C
    142 #include "xxhash.h"
    143 
    144     (...)
    145     XXH64_hash_t hash = XXH64(buffer, size, seed);
    146 }
    147 ```
    148 
    149 Using streaming variant is more involved, but makes it possible to provide data incrementally:
    150 ```C
    151 #include "stdlib.h"   /* abort() */
    152 #include "xxhash.h"
    153 
    154 
    155 XXH64_hash_t calcul_hash_streaming(FileHandler fh)
    156 {
    157     /* create a hash state */
    158     XXH64_state_t* const state = XXH64_createState();
    159     if (state==NULL) abort();
    160 
    161     size_t const bufferSize = SOME_SIZE;
    162     void* const buffer = malloc(bufferSize);
    163     if (buffer==NULL) abort();
    164 
    165     /* Initialize state with selected seed */
    166     XXH64_hash_t const seed = 0;   /* or any other value */
    167     if (XXH64_reset(state, seed) == XXH_ERROR) abort();
    168 
    169     /* Feed the state with input data, any size, any number of times */
    170     (...)
    171     while ( /* any condition */ ) {
    172         size_t const length = get_more_data(buffer, bufferSize, fh);   
    173         if (XXH64_update(state, buffer, length) == XXH_ERROR) abort();
    174         (...)
    175     }
    176     (...)
    177 
    178     /* Get the hash */
    179     XXH64_hash_t const hash = XXH64_digest(state);
    180 
    181     /* State can be re-used; in this example, it is simply freed  */
    182     free(buffer);
    183     XXH64_freeState(state);
    184 
    185     return hash;
    186 }
    187 ```
    188 
    189 
    190 ### Other programming languages
    191 
    192 Aside from the C reference version,
    193 xxHash is also available in many different programming languages,
    194 thanks to many great contributors.
    195 They are [listed here](https://www.xxhash.com/#other-languages).
    196 
    197 
    198 ### Branch Policy
    199 
    200 > - The "master" branch is considered stable, at all times.
    201 > - The "dev" branch is the one where all contributions must be merged
    202     before being promoted to master.
    203 >   + If you plan to propose a patch, please commit into the "dev" branch,
    204       or its own feature branch.
    205       Direct commit to "master" are not permitted.