qmc v1.2.0

Added

• Added support for median lattice rules via latticecandidates and keeplattices options (thanks to @KernerM)

qmc v1.1.0

Added

• Added batching mode which allows batches of points to be computed on the cpu (thanks to @Sander80)

Changed

• Added larger generating vectors (up to 68719476767) for up to 10 dimensional integrals, see generating vector set cbcpt_cfftw2_10
• Optimised copying of result array to/from devices, results are now reduced on the device

qmc v1.0.6

• Added larger generating vectors for up to 10 dimensional integrals (cbcpt_cfftw2_10)
• Added support for cputhreads=0 (same as omitting devices={-1})

Changes:
[generatingvectors] add cbcpt_cfftw2_10
[noc[u] support cputhreads=0 (same as omitting devices={-1})

qmc v1.0.5

• Fix bug which prevents compilation with gcc 11
• Recommend use of C++14 compatible compiler

Changes:
[check] update to catch 2.13.6
[fitfunctions] set initial_parameters in NoneFunction
[docs] c++11 -> c++14

qmc v1.0.4

• Fix potential bug in least_squares routine which could lead to very poor fits when employing variance reduction
• Suppress harmless warnings from nvcc which clutter output during compilation

Changes:
[cuda-warn] attempt to suppress harmless warnings from nvcc
[least_squares] fix potentially dangerous conversion to unsigned type
[transforms] add missing __device__ specifiers
[docs] add missing flag -rdc=true

qmc v1.0.3

• Fix an error in which sometimes very large GPU kernels would silently return a result = 0 due to exceeding the available GPU memory. This issue did not affect results obtained with CPUs.

Changes:
[core] check for errors after kernel invocation

qmc v1.0.2

[types] make samples.get_x a const method
[types] fix use of std::modf for some compilers

qmc v1.0.1

• [fitfunctions] fix issue which could lead to incorrect PolySingular fit results on some systems

qmc v1.0.0

Initial public release of the qmc integrator

qmc v0.0.1-alpha

Initial pre-publication release