| 1 | =====================================================================
|
|---|
| 2 | NEW IN VERSION 13
|
|---|
| 3 | =====================================================================
|
|---|
| 4 | - (Feature) Added in the ability for XSBench to write out a binary
|
|---|
| 5 | file containing a randomized XS dataset. The code is also capable
|
|---|
| 6 | of reading in this file instead of generating a new XS dataset
|
|---|
| 7 | each time the program is run. This feature may be useful for those
|
|---|
| 8 | running in simulation environments where walltime minimization
|
|---|
| 9 | is key for logistical reasons.
|
|---|
| 10 |
|
|---|
| 11 | - Minor refactoring/reorganization of code to make the code clearer
|
|---|
| 12 | and easier to read. After many updates, the code had become a
|
|---|
| 13 | little bloated and difficult to read, so a cleanup was in order.
|
|---|
| 14 |
|
|---|
| 15 | - Removed synthetic delay injection (via dummy FLOPS or loads).
|
|---|
| 16 | These were not very useful or accurate and had not been used by
|
|---|
| 17 | anyone after the initial analyses were done with them. As they were
|
|---|
| 18 | definitely adding to the code bloat of the program, they were
|
|---|
| 19 | removed.
|
|---|
| 20 |
|
|---|
| 21 | =====================================================================
|
|---|
| 22 | NEW IN VERSION 12
|
|---|
| 23 | =====================================================================
|
|---|
| 24 | - (Bugfix) The XL and XXL runtime options didn't work correctly.
|
|---|
| 25 | The unionized energy grid overflowed the bounds of normal 4 byte
|
|---|
| 26 | integers, and actually required use of 8 byte integers.
|
|---|
| 27 |
|
|---|
| 28 | The variables "n_isotopes" and "n_gridpoints" have been refactored
|
|---|
| 29 | to 8 byte long integers. All variables that use n_isotopes and
|
|---|
| 30 | n_gridpoints as input have also been refactored to 8 byte longs.
|
|---|
| 31 |
|
|---|
| 32 | Note that a simple "patch" from version 11 to version 12 can be
|
|---|
| 33 | manually done by simply changing line 73 of GridInit.c to be a
|
|---|
| 34 | long instead of an int. The more thorough refactoring done in v12
|
|---|
| 35 | is done to "future proof" the code.
|
|---|
| 36 |
|
|---|
| 37 | =====================================================================
|
|---|
| 38 | NEW IN VERSION 11
|
|---|
| 39 | =====================================================================
|
|---|
| 40 |
|
|---|
| 41 | - Updated & greatly improved the PAPI capability of XSBench. Now
|
|---|
| 42 | events can be tallied during multi-core. See README for more
|
|---|
| 43 | info.
|
|---|
| 44 |
|
|---|
| 45 | - Added in option for thread sleep pause in between macro XS lookups.
|
|---|
| 46 | Very similar to adding dummy flops, but a little cleaner.
|
|---|
| 47 |
|
|---|
| 48 | With as small as a 0.1 ms sleep, we get linear scaling with threads.
|
|---|
| 49 | While this initially appears to confirm our initial suspicions
|
|---|
| 50 | regarding memory contention / latency problems, I think the delays
|
|---|
| 51 | resulting from the sleeps could potentially just be washing out
|
|---|
| 52 | the scaling numbers. Even with just 0.1, over 15 million lookups,
|
|---|
| 53 | the majority of the runtime (>90%) is just sleep, so scaling numbers
|
|---|
| 54 | aren't very expressive anymore. Need to implement timers that
|
|---|
| 55 | ignore the sleep parts.
|
|---|
| 56 |
|
|---|
| 57 | - Specified OpenMP schedule mode as 'dynamic'. This is the default
|
|---|
| 58 | on most systems, but now it's set explicitly since it's a lot
|
|---|
| 59 | faster than 'static' or other modes.
|
|---|
| 60 |
|
|---|
| 61 | - Added in a "benchmarking" mode, which will attempt all possible
|
|---|
| 62 | thread combinations between 1 <= nthreads <= max_threads.
|
|---|
| 63 | This helps to save considerable benchmarking time, as the
|
|---|
| 64 | data structures can be re-used between runs rather than regenerated
|
|---|
| 65 | each time. Benchmarking mode is enabled in the makefile.
|
|---|
| 66 |
|
|---|
| 67 | =====================================================================
|
|---|
| 68 | NEW IN VERSION 10
|
|---|
| 69 | =====================================================================
|
|---|
| 70 |
|
|---|
| 71 | - Changed verification mode to be more portable. The verification
|
|---|
| 72 | strategy introduced in version 9 had discrepancies on different
|
|---|
| 73 | platforms and compilers. This was due to reliance on the compiler
|
|---|
| 74 | provided rand() function producing a different series of random
|
|---|
| 75 | numbers than other implementations. Also, there were some issues
|
|---|
| 76 | with the associativity of floating point arithmetic. These issues
|
|---|
| 77 | have now all been solved, and the verification hash is consistent
|
|---|
| 78 | across all tested platforms.
|
|---|
| 79 |
|
|---|
| 80 | - Revised "XL" size parameters, as well as adding in an "XXL" size
|
|---|
| 81 | option. The XL size now uses 120GB of XS data. The XXL mode uses
|
|---|
| 82 | 252GN of XS data. More details are in the verification section of the
|
|---|
| 83 | readme.
|
|---|
| 84 |
|
|---|
| 85 | =====================================================================
|
|---|
| 86 | NEW IN VERSION 9
|
|---|
| 87 | =====================================================================
|
|---|
| 88 |
|
|---|
| 89 | - Added in new code verification mode. This can be toggled on in
|
|---|
| 90 | the makefile. When code is compiled and run, a hash of the results
|
|---|
| 91 | will be generated which can then be compared to other versions and
|
|---|
| 92 | configurations of XSBench. See readme for more details.
|
|---|
| 93 |
|
|---|
| 94 | - Moved PAPI def to makefile. Makes it easier to toggle.
|
|---|
| 95 |
|
|---|
| 96 | - Added -l command line option to set the number of cross section
|
|---|
| 97 | lookups performed by XSBench.
|
|---|
| 98 |
|
|---|
| 99 | =====================================================================
|
|---|
| 100 | NEW IN VERSION 8
|
|---|
| 101 | =====================================================================
|
|---|
| 102 |
|
|---|
| 103 | - Simplified command line interface (CLI) read in process. XSBench
|
|---|
| 104 | now supports a more traditional CLI, as follows:
|
|---|
| 105 |
|
|---|
| 106 | Usage: ./XSBench <options>
|
|---|
| 107 | Options include:
|
|---|
| 108 | -n <threads> Number of OpenMP threads to run
|
|---|
| 109 | -s <size> Size of H-M Benchmark to run (small, large, XL)
|
|---|
| 110 | -g <gridpoints> Number of gridpoints per isotope
|
|---|
| 111 | Default is equivalent to: -s large
|
|---|
| 112 |
|
|---|
| 113 | - Updated README with new CLI usage details.
|
|---|
| 114 |
|
|---|
| 115 | - Fixed several typos in the XSBench Theory PDF.
|
|---|
| 116 |
|
|---|
| 117 | =====================================================================
|
|---|
| 118 | NEW IN VERSION 7
|
|---|
| 119 | =====================================================================
|
|---|
| 120 |
|
|---|
| 121 | - Added MPI support. Multithreaded run executes on all ranks.
|
|---|
| 122 | Problem size or data is not subdivided - the exact same problem
|
|---|
| 123 | is solved in parallel by all ranks. Only MPI communication is
|
|---|
| 124 | a single reduce at the end to aggregate timing data.
|
|---|
| 125 |
|
|---|
| 126 | To enable MPi mode, simply change the MPI flag in the makefile
|
|---|
| 127 | to "MPI = yes". Make sure mpicc is available on your system.
|
|---|
| 128 |
|
|---|
| 129 | - Added in "XL" size option for a giant 277 GB energy grid. This
|
|---|
| 130 | is unlikely to fit on a single node, but is useful for
|
|---|
| 131 | experimentation purposes.
|
|---|
| 132 |
|
|---|
| 133 | - Removed "BGQ mode" CLI argument option, as it wasn't being used
|
|---|
| 134 | by anything in the code anymore.
|
|---|
| 135 |
|
|---|
| 136 | =====================================================================
|
|---|
| 137 | NEW IN VERSION 6
|
|---|
| 138 | =====================================================================
|
|---|
| 139 |
|
|---|
| 140 | - Fixed small bug in calculate_micro_xs() function. Occasionally,
|
|---|
| 141 | the index returned would be the last nuclidegridpoint for that
|
|---|
| 142 | nuclide, causing the "high" energy point to be off the end of the
|
|---|
| 143 | grid (likely into the next nuclide's energy grid). Added a check
|
|---|
| 144 | to correct for when this occurs.
|
|---|
| 145 |
|
|---|
| 146 | Note that this bug did not affect performance - only made the
|
|---|
| 147 | calculation of XS's more "correct".
|
|---|
| 148 |
|
|---|
| 149 | =====================================================================
|
|---|
| 150 | NEW IN VERSION 5
|
|---|
| 151 | =====================================================================
|
|---|
| 152 |
|
|---|
| 153 | - Added ChangeLog
|
|---|
| 154 |
|
|---|
| 155 | - Moved source code files to src/ directory.
|
|---|
| 156 |
|
|---|
| 157 | - Updated README.txt file to enhance documentation
|
|---|
| 158 |
|
|---|
| 159 | - Added significant documentation with regards to theory
|
|---|
| 160 | in the docs/XSBench_Theory.pdf file. The README.txt file is now
|
|---|
| 161 | more of a quick-start & users guide, whereas the XSBench_Theory.pdf
|
|---|
| 162 | guide covers the details and theory behind the code.
|
|---|
| 163 |
|
|---|
| 164 | =====================================================================
|
|---|
