Commit message (Collapse) | Author | Age | ||
---|---|---|---|---|
... | ||||
* | CarpetProlongateTest: add tests for higher order restriction | Roland Haas | 2013-06-10 | |
| | | | | test data generated from 9748e3a83336a9ccae73a66481530f8049e8c78d | |||
* | CarpetLib: Re-implement part of bboxset2 for pre-C++11 compilers | Erik Schnetter | 2013-06-08 | |
| | | | | Provide alternative, less elegant implementations of some of bboxset2 that is accepted by pre-C++11 compilers. | |||
* | CarpetLib: Whitespace change | Erik Schnetter | 2013-06-08 | |
| | ||||
* | CarpetLib: Don't use run-together right angle brackets in templates | Erik Schnetter | 2013-06-08 | |
| | ||||
* | CarpetLib: Instantiate some templates that are necessary on some systems | Erik Schnetter | 2013-06-08 | |
| | ||||
* | CarpetLib: Provide manual look-alikes for certain C++11 features | Erik Schnetter | 2013-06-08 | |
| | | | | | Provide "forall" based on Boost's FOREACH. Provide shared_ptr based on Boost's shared_ptr. | |||
* | Requirements: Do not mark current timelevel as invalid after time level cycling | Erik Schnetter | 2013-05-26 | |
| | | | | The current time level retains its current value during time level cycling. | |||
* | CarpetLib: Disable bboxset2 by default | Erik Schnetter | 2013-05-26 | |
| | | | | Change from CARPET_NO_BBOXSET2 and CARPET_BBOXSET2 to CARPET_ENABLE_BBOXSET2 (build it) and CARPET_USE_BBOXSET2 (use it). | |||
* | Carpet: Output list of compute nodes only in verbose mode | Erik Schnetter | 2013-05-26 | |
| | ||||
* | Carpet: Remove unused, broken code | Erik Schnetter | 2013-05-26 | |
| | ||||
* | Carpet: Disable automatic synchronisation after restriction | Erik Schnetter | 2013-05-26 | |
| | | | | Restricted grid functions have boundary conditions applied anyway (if used correctly), which involves synchronisation (if done correctly). An explicit synchronisation is thus only overhead. | |||
* | Carpet: Provide global instead of process-local timer output | Erik Schnetter | 2013-05-26 | |
| | | | | | | Reduce timer values across all processes before output, reporting both maximum and average. TimerNode::getGlobalTime: new function | |||
* | Carpet: Use fabs() instead of abs() | Erik Schnetter | 2013-05-26 | |
| | | | | Some compilers don't like abs() for floating point numbers. | |||
* | CarpetIOF5: Implement lower-dimensional output (i.e. 1d lines, 2d slices) | Erik Schnetter | 2013-05-26 | |
| | ||||
* | Requirements: Check return value of CCTK_TraverseString | Erik Schnetter | 2013-05-26 | |
| | ||||
* | Requirements: Enforce 80 character line length limit | Erik Schnetter | 2013-05-26 | |
| | ||||
* | Requirements: Re-order #include statements | Erik Schnetter | 2013-05-26 | |
| | ||||
* | TestBBoxSet2: New thorn | Erik Schnetter | 2013-05-26 | |
| | ||||
* | LoopControl: Implement automatic performance improvements | Erik Schnetter | 2013-05-26 | |
| | | | | | | | | | | Redesigned internal data structures. This touches many parts of the code. Keep statistics about past loop iterations. Output statistics when terminating. Use these statistics to dynamically choose looping parameters. Allow disabling this feature, e.g. for benchmarks. | |||
* | CarpetLib: New class bboxset2 | Erik Schnetter | 2013-05-26 | |
| | | | | | | | | | | Rename bboxset to bboxset1. Implement new class bboxset2, which uses a different internal datastructure than bboxset1. Make bboxset a typedef for either bboxset1 (default) or bboxset2, as selected by the compile-time macro CARPET_BBOXSET2. Disable all bboxset2 code if CARPET_NO_BBOXSET2 is given, since bboxset2 uses newer C++ constructs not available on some older compilers. | |||
* | CarpetLib: New function bbox::anti_contracted_for | Erik Schnetter | 2013-05-26 | |
| | | | | anti_contracted_for is the exact opposite of contracted for. It is similar to expanded_for, but may return different results in corner cases. | |||
* | CarpetLib: Replace more uses of operator% by calls to imod | Erik Schnetter | 2013-05-26 | |
| | ||||
* | CarpetLib: New function bbox::is_poison to check for poison | Erik Schnetter | 2013-05-26 | |
| | ||||
* | CarpetLib: Remove superfluous check that all indices are non-negative | Erik Schnetter | 2013-05-26 | |
| | ||||
* | CarpetLib: Use imod instead of % operator in bbox class | Erik Schnetter | 2013-05-26 | |
| | | | | This avoids rounding problems near zero. | |||
* | CarpetLib: Use C++ standard methods to compare containers | Erik Schnetter | 2013-05-26 | |
| | | | | | Remove CarpetLib's "equals" method. Implement equal_to, less, greater, greater_equal, less_equal, not_equal for vect and bbox classes. | |||
* | CarpetLib: Correct vect::count | Erik Schnetter | 2013-05-26 | |
| | ||||
* | CarpetLib: Provide rint, round, trunc, and lrint for vectors | Erik Schnetter | 2013-05-26 | |
| | ||||
* | CarpetIO*: New parameter CarpetIO*::out_group_separator | Erik Schnetter | 2013-05-26 | |
| | | | | out_group_separator chooses the string by which thorn name and group name are separated in file names. The default is "::" for backward compatibility. This parameter only affects output where CarpetIO*::one_file_per_group is set; otherwise, the thorn name does not appear in the file name. | |||
* | Reduce time spent in cycling timelevels if there is only one timelevel. | Peter Diener | 2013-05-24 | |
| | | | | | | | | | Avoid calling routine that swaps pointers to different timelevels if there is only one timelevel active for a grid variable. This is an performance improvement if there are a large number of grid variables (as in the case of large vectors of grid variables). Signed-off-by: Peter Diener <diener@cct.lsu.edu> | |||
* | CarpetLib: Provide idiv and imod implementations for vect<CCTK_REAL,...> | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Remove now outdated definitions of idiv and imod | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Correct some type declarations for 4d grid structures | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Add missing #include files to cacheinfo.cc | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Disable unused refluxing face descriptions | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Add routines for integer division that always round down | Erik Schnetter | 2013-05-20 | |
| | | | | Standard integer division truncates, which can lead to problems if an intermediat result is negative. | |||
* | CarpetLib: Correct isnorma() for complex numbers | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Don't define static_assert | Erik Schnetter | 2013-05-20 | |
| | | | | This should now be done by the flesh, using autoconf. | |||
* | CarpetLib: Don't explicitly #include cctk_Config.h | Erik Schnetter | 2013-05-20 | |
| | | | | This is done by cctk.h automatically. | |||
* | CarpetLib: More verbose error output if sanity check fails | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Only check cache line size if it is used | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetLib: Change meaning of vect::count to count true (non-zero) elements | Erik Schnetter | 2013-05-20 | |
| | | | | Change count() from counting all elements, which is already done by size(), to counting true (non-zero) elements, which is what Fortran's count() does as well. | |||
* | CarpetLib: Rename vect::zip to vect::vzip | Erik Schnetter | 2013-05-20 | |
| | | | | vect::map is already called vect::vmap. | |||
* | CarpetLib: Add constructors to create a vector from a low-dimensional vector ↵ | Erik Schnetter | 2013-05-20 | |
| | | | | and a scalar | |||
* | CarpetLib: Add some comments explaining implementation | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetIOASCII: Indentation change | Erik Schnetter | 2013-05-20 | |
| | ||||
* | CarpetInterp2: Convert some loop bounds to int | Erik Schnetter | 2013-05-20 | |
| | | | | The Cray compiler requests this for OpenMP. | |||
* | CarpetIOF5: Do not output irrelevant "Writing simulation metadata" messages | Erik Schnetter | 2013-05-19 | |
| | ||||
* | CarpetIOF5: Correct calculating the I/O processor numbers | Erik Schnetter | 2013-05-19 | |
| | ||||
* | LoopControl: disable smt by default | Roland Haas | 2013-05-12 | |
| | | | | | We currently have issues with this switch changing results of computation when used in combination with openmp and using all cores. |