Skip to content

Update VectorClass to 1.26 upstream release

Christopher Rob Jones requested to merge VectorClassUpdate into master

Changelog since last LHCb version (1.23)

FYI @graven

Projected future version (2018?):

  • use mask registers for boolean vectors of all sizes with AVX512VL
  • use C++14
  • replace long template argument lists by constant arrays

2016-12-21 version 1.26

  • added constant4ui template
  • fixed error for complexvec.h with clang
  • fixed error in vectormath_exp.h for MAX_VECTOR_SIZE < 512

2016-11-25 version 1.25

  • scatter functions
  • new functions to_float for unsigned integer vectors
  • instrset_detect function can detect AVX512VL, AVX512BW, AVX512DQ
  • functions hasF16C and hasAVX512ER for detecting instruction set extensions
  • fix bugs in horizontal_and and pow(0,0) for AVX512
  • functions improved for AVX512 and AVX512VL: pow, approx_recipr,
    approx_rsqrt
  • functions improved for AVX512DQ: 64 bit multiplication, to_double, 32 and 64 bit rotate_left, round_to_int64, truncate_to_int64
  • functions improved for AVX512ER: approx_recipr, approx_rsqrt, exponential functions

2016-10-31 version 1.24

  • fix bug in Vec8uq constructor in vectori512e.h

Merge request reports