Version 4.0
- Multiple speed improvements, GPU. See the bitbucket page (https://bitbucket.org/feynmanIntegrals/fiesta)
Version 3.4
- A version that can be compiled under Max OS X
Version 3.3
- Some bug fixes eliminating rare crashes.
- Shipped with Cuba 3.3 and kyotocabinet
- make dep to install dependences
- IntegrationCommand[] to learn the command used to create the out database
Version 3.2:
- Multiple bug fixes related to SDExpandASY with negative indices or PreResolve
- The out databased is locked (to prevent some loss of results)
- New CubaBatch option (and vectorization of the algorithm)
Version 3.1:
- PreResolve option starts working with indices different from 1
- ConservativeComplexShift option
- NegativeTermsHandling switched to Auto (which means no handling for Complex or Asy)
- Extra argument for SDExpandASY to pass info to the code on signs of parameters (for PreResolve and OnlyPrepareRegions)
- No pure native mode any longer (CIntegrate)
- Switching to Cuba3.3.
Version 3.0:
An absolutly new version with multiple new features such as
- integration in physical regions
- new asymptotic expansion methods
- MPI-parallelization
- multiple speed and memory improvements
- total reconstruction: the integrator is completely separated and can be used by other programs
Version 2.5:
comes with a new package of sources and binaries
- fixed a number of bugs that prevented SDExpand from working (asymptotics)
- significant speedup and memory economy in some cases
Version 2.4:
comes with a new package of sources and binaries
- significant speedup and memory economy on expression preparation stage
- the code can restart CIntegrate links in case they crash
- bug fix - the integrator could crash in some rare cases (highly degenerate functions)
- bug fix - slightly wrong presicion estimate in some rare cases
Version 2.3:
comes with a new package of sources and binaries
- fixed a bug for non-integer indices
- fixed warnings for Mathematica 7 and 8
- new option for square resolution: AdvancedSquares3 (works better on threshhold
)
- new non-obligatory argument for SDEvaluateDirect to provide a list of delta-fu
nctions
- better debugging with IntegrationDebug
- new sector decomposition strategies STRATEGY_KU0, STRATEGY_KU, STRATEGY_KU2 (K
aneko, Ueda).
KU2 is slowest, but less sectors. KU0 is fastest, but more sectors
- new boolean option PolesMultiplicity, if true, forces the SDAnalyze mode to pr
ovide poles multiplicities as well
- bug fix - intrepolation near the origin did not actually work as expected,
resulting in integration errors in some specific situations
- bug fix - integration crashed when one of the exponents exceeded 13312
Version 2.0.0 (28.11.2009)
- improved treatement of numerical instabilities
- improves parallelization
- new integrators
- Speer sectors
- more general classes of integrals
To summarize, it as a completely new project much more powerfull than the initial one. For details please refer to .....
Version 1.2.1 (28.04.2009)
- Small fix for the UF function (the code could still fail in case of non-rational (say, 0.1) kinematic invariants), thanks to Valery Yundin
Version 1.2.0 (28.03.2009)
- Parallelization of the Mathematica part (only under Mathematica 7). To use it, set NumberOfSubkernels>1
- Bug fix - the code failed on some integrals equal to zero
- Bug fix - the STRATEGY_S failed on some special integrals
- Bug fix - the code could fail in case of non-rational (say, 0.1) kinematic invariants
- new option - ReturnErrorWithBrackets - if True, the errors in the results are returned not like pm57 but like pm[57]
- the code is distributed with new CIntegrate binaries that have a possibility to choose different integration methods. As an addition to the original Vegas integrator of Fortran, the Cuba library by Thomas Hahn algorithms can be used. If you wish to compile the binaries yourself, please use the latest Cuba library (version 1.5 released 06.03.2009) and run export CFLAGS="-O3"; ./configure when building the Cuba library.
Version 1.1.0 (31.12.2008)
- STRATEGY_SS added (Speer sectors)
- a possibility to use new binaries with different integrators (the binaries are still in preparation)
- STRATEGY_S is prevented from freezing, in case of very long time requirements a failback to STRATEGY_A is performed
- bug fix for IfCut with UsingC=False
- bug fix - extra error messages removed with UsingC=False
- new option: d0 (4 by default) standing for the dimension
Version 1.0.0: first public release (22.08.2008)
Plans:
- add complex numbers to the CIntegrate
- more efficient parallelization with expression preparation, a possibility to use multiple Mathematica kernels (Mathematica 7)
- extra integrators