- Oct 19, 2020
-
-
Bertrand Bellenot authored
-
- Sep 30, 2020
-
-
Enrico Guiraud authored
Co-authored-by:
Stefan Wunsch <stefan.wunsch@cern.ch>
-
Enrico Guiraud authored
-
- Jul 24, 2020
-
-
Lorenzo Moneta authored
* Fix computation of Minos errrors in Minuit2 when parameters are bounded. A bug was introduced when the conditions "new minimum found" was added in Minos. (ROOT-10854) Add some fixes also the message reporting in Minuit2. Restore usage of ROOT message reporting system when MInuit2 is built within ROOT * Add a protection in TMinuit to avoid a 0/0 division. Fix ROOT-10835 This can happen when parameter limit is 0 and parameter value at minimum is 0 and a Minos error analysis is performed for that parameter. * Fix running Minos error when a new minimum is found. Fix the handling of the situation of Minos error in Minuit2 when a new minimum is found. Add a function in the Minimizer to get the status code of Minos to flag when a new minimum is found Fix function returning minimum values in Minuit2Minimizer in case of multiple runs. The minimum values are now copied in the Mininuit2Minimizer local vector when the minimization is run * Update Fitter::CalculateMinosError when a new minimum is found in Minos Recompute the minos error for the previous run parameter in case a new minimum was found * Use fabs instead of abs. Fix problem reported by CI builds * small fixes in documentation as suggested by Stephan's review * Revert changes in the Minuit2 CMAKE file for enabling Minuit2 error and warning messages. These changes will be done in a separate commit and a new PR.
-
- May 18, 2020
-
-
Lorenzo Moneta authored
* Fix case when ROOT is running with MT and TVIrtualFitter pointer to last fitter is not present * Fix multithread fitting when computing a bin intengral or re-normalizing by the bin volume (case of Thnsparse fit). The fix is applied for both least-square and binned likelihood fit Improve stressHistoFit test. Add option "-t" to run in multi-thread mode. Add the option X for fitting to make sure chi2 method is run Now all stresshistofit tests pass in multi-thread mode. Note that when computing integral in multi-thread only the GuassIntegrator of mathcore is thread safe for 1D. Also the AdaptiveIntegratorMultiDim, used in the multi-dim case, is thread safe. * Remove non-thread safe access to BinData::BinEdge also in other fitting functions and replace them with the thread safe version
-
- May 14, 2020
-
-
Mattias Ellert authored
-
- May 11, 2020
-
-
Lorenzo Moneta authored
The warnings are coming when including the Vc/Vc header : warning: ‘V2 Vc_1::Vector<T, Vc_1::VectorAbi::Scalar>::reinterpretCast() const’ is deprecated: use reinterpret_components_cast instead [-Wdeprecated-declarations] Add a pragma to suppress -Wdeprecated-declarations in Types.h
-
Sergey Linev authored
Allow to use some of them outside ROOT - genetic/minuit2 In tests for mathcore/mathmore do not use find_package(ROOT) and its variables
-
- May 06, 2020
-
-
Olivier Couet authored
-
- May 05, 2020
-
-
Lorenzo Moneta authored
Add a new page (doc/Matrix.md)describing TMatrix combining the Web module documentation with the pre-existing ones. Reference the new matrix page from the TMatrix,TMatrixTBase, TMatrixTSym and TVectorT refeence class documentation
-
- May 04, 2020
-
-
Sergey Linev authored
-
- May 01, 2020
-
-
Oksana Shadura authored
* Update gtest to latest version 1.10.0 More info: https://github.com/google/googletest/releases/tag/release-1.10.0 * Adjust cmake variables to be used by Ninja of build byproducts * Replace deprecated functions for googletest 1.10 release * Add missing gtest_main target * Fix locations for gtest and gtest_main targets * Update CMake configuration for gtest tests Tests were failing with next error: /usr/bin/ld: ../../../googletest-prefix/src/googletest-build/lib/libgtest_main.a(gtest_main.cc.o): in function testing::InitGoogleTest(int*, char**) /usr/bin/ld: gtest_main.cc:(.text.startup+0x2f): undefined reference to testing::UnitTest::Run() * Patch by Bertrand Bellenot: Update build configuration for Windows * We build googletest in Release mode (for Debug names of libraries should be different and require special treatment) * [formating] Add EOL in CMakeLists.txt
-
- Apr 30, 2020
-
-
Lorenzo Moneta authored
-
Lorenzo Moneta authored
This is not the Poisson probability discrete distribution (Poisson pmf), but a it is some other function that is used in some part of RooFit (e.g. Asimov data sets)
-
Lorenzo Moneta authored
It was returning by mistake exp(-mu) a very large value !! Now TMath uses ROOT::Math::poisson_pdf(n,mu) that returns a NaN ( a quiet NaN) in this case Avoid returning signaling Nan also for the case n > 0 and mu < 0
-
- Apr 27, 2020
-
-
Sergey Linev authored
-
Sergey Linev authored
-
- Apr 23, 2020
-
-
Sergey Linev authored
-
Stephan Hageboeck authored
The tests for TFoam ended up in the wrong location, since add_subdirectory is not setting the required property.
-
- Apr 22, 2020
-
-
Stephan Hageboeck authored
[ROOT-10277] When creating a lot of datasets in RooFit, TFoam eventually crashed, since it was using a TRefArray, which does some global UID magic that overflows. Replacing the TRefArray by a simple vector of indices fixes that. This is possible, since the TRefArray only points to active elements in another collection, so simply storing the index in that collection is sufficient. A ROOT-6.20 TFoam was created on lxplus, and written to a file. Reading this file, and generating the 10 next random numbers is tested now.
-
Sergey Linev authored
Try to exclude as much as possible ${CMAKE_BUILD_DIR)/include from includes paths. By this one much better control library dependencies - includes from other ROOT libraries not "visible" during the build. Several generated files placed first to ${CMAKE_BUILD_DIR)/ginclude and then copied to include. Dictionary generation still uses only ${CMAKE_BUILD_DIR)/include Co-authored-by:
Axel Naumann <Axel.Naumann@cern.ch>
-
Lorenzo Moneta authored
Remove the .gitignore file which can cause the remaining of duplicate files in local repo after having build a standalone minuit2 tarball Update the documentation of building minui2 tarballs
-
Lorenzo Moneta authored
MathMore headers
-
Lorenzo Moneta authored
(e.g. PdfFuncMathCore inside of PdfFunc.h) The files PdfFunc.h, ProbFunc.h and QuantFunc.h should not also be used when building the mathcore dictionary
-
- Apr 21, 2020
-
-
Sergey Linev authored
-
- Apr 16, 2020
-
-
Alexander Voigt authored
expansion in terms of Chebyshev polynomials. Compare with [Yudell L. Luke: Mathematical functions and their approximations, Academic Press Inc., New York 1975, p.67, Table 3.12]: 0.42996693560813697204 0.40975987533077105847 -0.01858843665014591965 0.00145751084062267855 -0.00014304184442340049 0.00001588415541879553 -0.00000190784959386583 0.00000024195180854165 -0.00000003193341274252 0.00000000434545062677 -0.00000000060578480118 0.00000000008612097799 -0.00000000001244331660 0.00000000000182255696 -0.00000000000027006766 0.00000000000004042209 -0.00000000000000610325 0.00000000000000092863 -0.00000000000000014226 0.00000000000000002193 -0.00000000000000000340 0.00000000000000000053 -0.00000000000000000008 0.00000000000000000001 On IEEE-conforming platforms this change does not have any numerical effect, because the two floating point numbers double x = 0.40975987533077105; double y = 0.40975987533077106; have the same binary representation at double precision.
-
- Apr 11, 2020
-
-
Vassil Vassilev authored
The intent of this code seemed to be avoiding a CINT issue. Now cling can handle such constructs. This fixes an issue with runtime_cxxmodules on OSX sdk where even when NDEBUG is defined the assert still checks the identifier.
-
- Apr 09, 2020
-
-
Vassil Vassilev authored
This fixes a nightly build issue on OSX with -Druntime_cxxmodules=On by default. There rootcling still needs to parse the assert statement even if NDEBUG is defined.
-
- Apr 08, 2020
-
-
Vassil Vassilev authored
This fixes a nightly build issue on OSX with -Druntime_cxxmodules=On by default. There rootcling still needs to parse the assert statement even if NDEBUG is defined.
-
- Apr 06, 2020
-
-
Enrico Guiraud authored
TThreadExecutor does not activate implicit multi-threading (ROOT::IsImplicitMTEnabled() is false after constructing a TThreadExecutor) but it does change the size of the thread-pool, which is not only used by implicit multi-threading features, but also by TThreadExecutor and TTreeProcessorMT. So the thread pool is not the "ImplicitMTPool" but it's really ROOT's one and only pool of threads, so we prefer ROOT::GetThreadPoolSize to GetImplicitMTPoolSize.
-
Lorenzo Moneta authored
-
surya2191997 authored
-
Sergey Linev authored
In many places incldues are not required for the dictionary generation, therefore target_include_directories can be used
-
- Mar 26, 2020
-
-
Lorenzo Moneta authored
-
Lorenzo Moneta authored
std::numeric_limits<epsilon> is 2 larger than eps computed before in Minuit2. Use then *4 instead of *8.
-
Lorenzo Moneta authored
suggested by Hans Dembinski.
-
- Mar 16, 2020
-
-
Lorenzo Moneta authored
When calling Minimize the global prnit level is synchronized to the local one and then re-set to its original value Apply this to all Builder implementations (Fumili, Simplex, VM (Migrad) ). ScanBuilder is not touched since it is not printing any output
-
- Mar 10, 2020
-
-
Sergey Linev authored
-
- Mar 02, 2020
-
-
Sergey Linev authored
-
- Feb 28, 2020
-
-
Lorenzo Moneta authored
-