Bug: 148898578

Clone this repo:
  1. dba5619 Upgrade FXdiv to 561254d968e5679460e6a0a743206410284d9f46 am: 74c9d9ef6a by Haibo Huang · 4 months ago master android-r-beta-2 android-r-beta-3
  2. 74c9d9e Upgrade FXdiv to 561254d968e5679460e6a0a743206410284d9f46 by Haibo Huang · 4 months ago
  3. 561254d Update README by Marat Dukhan · 4 months ago
  4. 62d7d30 Disable inline assembly by default by Marat Dukhan · 4 months ago
  5. 8e02fe6 Indicate cloberred condition flags in inline assembly by Marat Dukhan · 4 months ago

FXdiv

MIT License Build Status

Header-only library for division via fixed-point multiplication by inverse

On 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.

Features

  • Integer division for uint32_t, uint64_t, and size_t
  • Header-only library, no installation or build required
  • Compatible with C99, C++, OpenCL, and CUDA
  • Uses platform-specific compiler intrinsics for optimal performance
  • Covered with unit tests and microbenchmarks

Example

#include <fxdiv.h>

/* Division of array by a constant: reference implementation */
void divide_array_c(size_t length, uint32_t array[], uint32_t divisor) {
  for (size_t i = 0; i < length; i++) {
    array[i] /= divisor;
  }
}

/* Division of array by a constant: implementation with FXdiv */
void divide_array_fxdiv(size_t length, uint32_t array[], uint32_t divisor) {
  const struct fxdiv_divisor_uint32_t precomputed_divisor =
    fxdiv_init_uint32_t(divisor);
  for (size_t i = 0; i < length; i++) {
    array[i] = fxdiv_quotient_uint32_t(array[i], precomputed_divisor);
  }
}

Status

Currently working features:

Platformuint32_tuint64_tsize_t
x86-64 gccWorksWorksWorks
x86-64 clangWorksWorksWorks
x86-64 MSVCWorksWorksWorks
x86 gccWorksWorksWorks
x86 clangWorksWorksWorks
x86 MSVCWorksWorksWorks
ARMv7 gccWorksWorksWorks
ARMv7 clangWorksWorksWorks
ARMv7 MSVC*CompilesCompilesCompiles
ARM64 gccWorksWorksWorks
ARM64 clangWorksWorksWorks
ARM64 MSVC*CompilesCompilesCompiles
PPC64 gccWorksWorksWorks
WAsm clangWorksWorksWorks
Asm.js clangWorksWorksWorks
PNaCl clangWorksWorksWorks
CUDAUntestedUntestedUntested
OpenCLUntestedUntestedUntested

*ARMv7 and ARM64 builds with MSVC are presumed to work, but were only verified to compile successfully

References

  • 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