• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# FXdiv
2
3[![MIT License](https://img.shields.io/badge/License-MIT%20License-blue.svg)](https://github.com/Maratyszcza/FXdiv/blob/master/LICENSE)
4[![Build Status](https://img.shields.io/travis/Maratyszcza/FXdiv.svg)](https://travis-ci.org/Maratyszcza/FXdiv)
5
6
7Header-only library for division via fixed-point multiplication by inverse
8
9On modern CPUs and GPUs integer division is several times slower than multiplication. FXdiv implements an algorithm to replace an integer division with a multiplication and two shifts. This algorithm improves performance when an application performs repeated divisions by the same divisor.
10
11## Features
12
13- Integer division for `uint32_t`, `uint64_t`, and `size_t`
14- Header-only library, no installation or build required
15- Compatible with C99, C++, OpenCL, and CUDA
16- Uses platform-specific compiler intrinsics for optimal performance
17- Covered with unit tests and microbenchmarks
18
19## Example
20
21```c
22#include <fxdiv.h>
23
24/* Division of array by a constant: reference implementation */
25void divide_array_c(size_t length, uint32_t array[], uint32_t divisor) {
26  for (size_t i = 0; i < length; i++) {
27    array[i] /= divisor;
28  }
29}
30
31/* Division of array by a constant: implementation with FXdiv */
32void divide_array_fxdiv(size_t length, uint32_t array[], uint32_t divisor) {
33  const struct fxdiv_divisor_uint32_t precomputed_divisor =
34    fxdiv_init_uint32_t(divisor);
35  for (size_t i = 0; i < length; i++) {
36    array[i] = fxdiv_quotient_uint32_t(array[i], precomputed_divisor);
37  }
38}
39```
40
41## Status
42
43Currently working features:
44
45| Platform        | uint32_t | uint64_t | size_t   |
46| --------------- |:--------:|:--------:|:--------:|
47| x86-64 gcc      | Works    | Works    | Works    |
48| x86-64 clang    | Works    | Works    | Works    |
49| x86-64 MSVC     | Works    | Works    | Works    |
50| x86 gcc         | Works    | Works    | Works    |
51| x86 clang       | Works    | Works    | Works    |
52| x86 MSVC        | Works    | Works    | Works    |
53| ARMv7 gcc       | Works    | Works    | Works    |
54| ARMv7 clang     | Works    | Works    | Works    |
55| ARMv7 MSVC*     | Compiles | Compiles | Compiles |
56| ARM64 gcc       | Works    | Works    | Works    |
57| ARM64 clang     | Works    | Works    | Works    |
58| ARM64 MSVC*     | Compiles | Compiles | Compiles |
59| PPC64 gcc       | Works    | Works    | Works    |
60| WAsm clang      | Works    | Works    | Works    |
61| Asm.js clang    | Works    | Works    | Works    |
62| PNaCl clang     | Works    | Works    | Works    |
63| CUDA            | Untested | Untested | Untested |
64| OpenCL          | Untested | Untested | Untested |
65
66*ARMv7 and ARM64 builds with MSVC are presumed to work, but were only verified to compile successfully
67
68## References
69
70- Granlund, Torbjörn, and Peter L. Montgomery. "Division by invariant integers using multiplication." In ACM SIGPLAN Notices, vol. 29, no. 6, pp. 61-72. ACM, 1994. Available: [gmplib.org/~tege/divcnst-pldi94.pdf](https://gmplib.org/~tege/divcnst-pldi94.pdf)
71