MARQOV issueshttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues2022-02-24T12:20:34Zhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/199complex thingies2022-02-24T12:20:34ZFlorian Gothcomplex thingiesenable complex support for our HDF5 support classes.enable complex support for our HDF5 support classes.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/79improve parallel progress output2022-02-04T12:55:52ZManuel Schrauthimprove parallel progress outputDue to the parallel processes simultaneously spamming the terminal, the "progress bar" is messed up. Ideas for temporary solutions:
- individual icons for parallel processes
- threads write individual log files (this was in the old code...Due to the parallel processes simultaneously spamming the terminal, the "progress bar" is messed up. Ideas for temporary solutions:
- individual icons for parallel processes
- threads write individual log files (this was in the old code anyways)
- threads send their progress status to some kind of host class, which manages the output (@fgoth is this even possibly, technically?)
- other ideas?https://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/195Logging2022-02-04T07:24:20ZFlorian GothLoggingWe need some improved kind of logging...
Here's a first draft of a class:
https://git.physik.uni-wuerzburg.de/fgoth/flogWe need some improved kind of logging...
Here's a first draft of a class:
https://git.physik.uni-wuerzburg.de/fgoth/flogFlorian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/196RGG seg fault2022-02-01T14:41:13ZManuel SchrauthRGG seg faultthe Ising on RGG simulation throw a segfault on the cluster, whereas on my desktop everything works. Perhaps some version incompatibility ...the Ising on RGG simulation throw a segfault on the cluster, whereas on my desktop everything works. Perhaps some version incompatibility ...https://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/193implement rgg2022-01-26T12:53:17ZManuel Schrauthimplement rggimplement the random geometric graphimplement the random geometric graphManuel SchrauthManuel Schrauthhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/192Squueze out more performance from Regular Lattice2022-01-20T14:07:40ZFlorian GothSquueze out more performance from Regular Latticetemplatifying the dimensionality we can et rid of std::vector allocations and deallocations.templatifying the dimensionality we can et rid of std::vector allocations and deallocations.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/189speed comparison2022-01-14T17:05:52ZManuel Schrauthspeed comparisonThis is the reference for my current simulations: [hasenbusch2011.pdf](/uploads/045b3d2ca3bda66343d167829ebd0952/hasenbusch2011.pdf)
According to Appendix A.2, about `2.5e8 measurements` on a `64^3` lattice took less than 4 CPU years on...This is the reference for my current simulations: [hasenbusch2011.pdf](/uploads/045b3d2ca3bda66343d167829ebd0952/hasenbusch2011.pdf)
According to Appendix A.2, about `2.5e8 measurements` on a `64^3` lattice took less than 4 CPU years on a single core of a Quad-Core AMD Opteron Processor 2378 running at 2.4 GHz. Let's say it was **one** year for this lattice size:
```
nemcs = 2.5e8
hours = 365*24
-> nemcs/hour = 28.500
```
Now our timing results on the new part of the ITPA cluster (without hyper threading) are:
```
nemcs = 10.000
hours = 9.1
-> nemcs/hour = 1100
```
Which is about a factor 25x slower ...Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/183heisenberg hpc2022-01-13T14:54:44ZManuel Schrauthheisenberg hpcI abuse this issue for posting results of mid-scale Heisenberg simulationsI abuse this issue for posting results of mid-scale Heisenberg simulationshttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/187Phi4 model issues2022-01-13T14:54:07ZManuel SchrauthPhi4 model issuesmaxima are shifted to higher beta as the lattice size is increased, why? hopefully something trivial ...
![phi](/uploads/5be70c76929ec3dd81ac03a16b81fb86/phi.png)maxima are shifted to higher beta as the lattice size is increased, why? hopefully something trivial ...
![phi](/uploads/5be70c76929ec3dd81ac03a16b81fb86/phi.png)https://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/186improve C++2022-01-07T11:37:05ZFlorian Gothimprove C++Flo wants to improve the C++ use of MARQOV...Flo wants to improve the C++ use of MARQOV...Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/184fix test with eigen2021-12-09T14:52:09ZFlorian Gothfix test with eigenOn "some" PCs, most notably my desktop the test of MARQOV with eigen fails. I found out that this due to
an incorrect use of the {} syntax for initialization.
StateVector{0} is a constructor call. In the case of Eigen this leads to the ...On "some" PCs, most notably my desktop the test of MARQOV with eigen fails. I found out that this due to
an incorrect use of the {} syntax for initialization.
StateVector{0} is a constructor call. In the case of Eigen this leads to the creation of a vector of a specified size with 0.
StateVector{} is initialization, which for aggregate types(e.g. structs, arrays) performs zero initialization.
The last is also the behaviour that we intended to be used.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/180hdf5 issue on mpi16 nodes2021-12-08T16:07:17ZManuel Schrauthhdf5 issue on mpi16 nodesMARQOV crashes with these messages on the ITPA cluster mpi16 nodes
[error.txt](/uploads/2cd5520414c3d71e4b014a39d2b8c494/error.txt)
maybe version to old? in this case, cmake should complain already though, doesn't it?MARQOV crashes with these messages on the ITPA cluster mpi16 nodes
[error.txt](/uploads/2cd5520414c3d71e4b014a39d2b8c494/error.txt)
maybe version to old? in this case, cmake should complain already though, doesn't it?Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/181deactivate PT2021-12-03T11:28:09ZManuel Schrauthdeactivate PT@fgoth what is the "canonical" way to deactivate the parallel tempering? I'm about to run a number of MARQOVdemo instances which obviously crashes due to statespaces of different size (coming from regularlatticeloop)@fgoth what is the "canonical" way to deactivate the parallel tempering? I'm about to run a number of MARQOVdemo instances which obviously crashes due to statespaces of different size (coming from regularlatticeloop)https://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/179dedicated place for config files2021-11-18T19:35:21ZManuel Schrauthdedicated place for config fileshave something like a "present working directory" where config files can be stored independently of the build directoryhave something like a "present working directory" where config files can be stored independently of the build directoryhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/177Wolff: improve memory management2021-11-15T09:17:18ZFlorian GothWolff: improve memory managementWe can reuse the cluster size of the previous step as a predictor for the size of the next step. Then we can alleviate the growing process of std::vector.We can reuse the cluster size of the previous step as a predictor for the size of the next step. Then we can alleviate the growing process of std::vector.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/178cmake issue2021-11-10T17:53:57ZManuel Schrauthcmake issuecmake won't run through on some machines (our workstations / itpa cluster). It complains about policies and SWIG
```
Found SWIG: /usr/bin/swig3.0 (found suitable version "3.0.10", minimum required is "3.0")
CMake Error at src/CMakeLists...cmake won't run through on some machines (our workstations / itpa cluster). It complains about policies and SWIG
```
Found SWIG: /usr/bin/swig3.0 (found suitable version "3.0.10", minimum required is "3.0")
CMake Error at src/CMakeLists.txt:27 (cmake_policy):
Policy "CMP0086" is not known to this version of CMake.
CMake Error at src/CMakeLists.txt:28 (cmake_policy):
Policy "CMP0078" is not known to this version of CMake.
-- SWIG found: /usr/bin/swig3.0.
-- Found PythonLibs: /usr/lib/x86_64-linux-gnu/libpython3.5m.so (found version "3.5.3")
-- Python found, building Python bindings
CMake Error at src/CMakeLists.txt:38 (swig_add_library):
Unknown CMake command "swig_add_library".
-- Configuring incomplete, errors occurred!
```
On my personal laptop and on JULIA everything works fine. @fgoth any idea how this can be fixed?Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/174Adapt optimized Metropolis to wolff2021-11-07T18:44:50ZFlorian GothAdapt optimized Metropolis to wolff#19 should be adapted to the Wolff algorithm.#19 should be adapted to the Wolff algorithm.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/171Lattice parameters in HDF52021-10-28T14:37:09ZFlorian GothLattice parameters in HDF5Find a way to dump lattice parameters into the HDF5 container.
By default dump the size. Make a user-definable template/function that can be specialized for each lattice by the user.Find a way to dump lattice parameters into the HDF5 container.
By default dump the size. Make a user-definable template/function that can be specialized for each lattice by the user.https://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/19use new Exponential of FAU2021-10-28T14:28:26ZFlorian Gothuse new Exponential of FAUlet's try the new exponential of the FAU.let's try the new exponential of the FAU.Florian GothFlorian Gothhttps://git.physik.uni-wuerzburg.de/marqov/marqov/-/issues/175drawing real random numbers2021-10-28T14:28:08ZFlorian Gothdrawing real random numbersOn the bottom of the page here, there is an alternative formula for conversion of random integers to random reals:
https://prng.di.unimi.it/
and here:
https://prng.di.unimi.it/random_real.cOn the bottom of the page here, there is an alternative formula for conversion of random integers to random reals:
https://prng.di.unimi.it/
and here:
https://prng.di.unimi.it/random_real.c