added env variables to makefiles

added env variables to makefiles

environment files

added nvvp files

added env variables to makefiles

2000x2000 image by default

added cuo files into exclusion list

calling the right version

lense type dispatcher enabled for hand coded avx version

inserting guard to allow linking with the lenstool library

added possibility to link with the lenstool library using the -D__WITH_LIBTOOL…

added improved version of lense derivative selection

update

added function to call the correct lense type gradient function correctly

useful functions

reverting to original configuration

interface change update

interface cleanup + SIS implementation for SOA

deleted binary file

interface change update

updated library + gradient benchmark with lense 81 auto-vectorized version

clean repo

added chi benchmark + general updates

update

bug fix, precision test passed

gradent benchmark using jauzac input

updated makefile to create library

added sources for the different versions of the derivative

updated (ie working) AVX512F version

added toggle to remove benchmarks baselines

added fine level threading using openmp

AVX512F (KNL) version added

using native flag for the compilation architecture

addind the unix timestamp to the directory file

removed binary file

jauziac benchmark implementated, avx and scalar version (3x for now).

update with scalar and vector (avx) version

better output

major overhaul of the benchmark to include real-life data from jauzac paper

removed unecessary old file

setup header file

adding new jauzac benchmark

adding vector operators for the intel compilers to ensure backward compatibility

hand optimized simd math operators

added hand optimized inverse and sqrt operations + cleanup

gradient file header

corrected executable name

higher precision in result

vectorization update with RCP + newton raphson methode to increase the precision

hand vectorized version added: 50% faster than the scalar one probably because…

new values

big test comparison

timer added

full loop expension

added inlined first version of the kernel

update

modified makefile to take compiler into account + small update

sperated main and compute kernels

authorship

overhaul: new approach with better register usage

authorship + overhaul: better use of the registers and better prefetechers

added permuted 8x4 dgemm kernel

update

authorship + better prefetechers

new kernels update

better prefetchers + use of more registers

added unrolled 4x4 permutation code

adding .gitignore

update

dgemm for neon - initial commit