Scalasca: ChangeLog

"Fossies" - the Fresh Open Source Software Archive

Member "scalasca-2.6/ChangeLog" (19 Apr 2021, 7200 Bytes) of package /linux/misc/scalasca-2.6.tar.gz:

As a special service "Fossies" has tried to format the requested text file into HTML format (style: standard) with prefixed line numbers. Alternatively you can here view or download the uninterpreted source code file. See also the last Fossies "Diffs" side-by-side code changes report for "ChangeLog":

2.3.1_vs_2.4.

1 ------------------- Released version 2.6 ----------------------------- 2 3 * Build system improvements: 4 - Auto-detect Cray XC platforms with ARM CPUs, supporting Cray, 5 ARM, and GCC compilers 6 - Added support for Clang and AMD AOCC compilers 7 - Updated support for Spectrum MPI 8 9 * Automatic trace analyzer changes & improvements: 10 - Revised "Early Reduce" wait state definition. 11 - Added calculation of "Early Reduce" delay costs. 12 - Fixed various delay cost calculation and propagation issues. 13 - Fixed various inconsistencies between wait-state and root-cause 14 analysis. 15 - Made POSIX threads analysis consistent with Score-P by avoiding 16 thread function stub call paths underneath 'pthread_create'. 17 This also fixes a deadlock when analyzing traces containing 18 "orphaned threads". 19 20 * Measurement nexus (scan) changes: 21 - Added preset mode for multi-run measurements with a preset for 22 POP analysis requirements as an use case. 23 - Added support for multiple file systems in SCAN_TRACE_FILESYS 24 by using a colon separated list of paths. 25 26 * Analysis report postprocessing changes: 27 - Add metric hierarchies for CUDA, OpenCL, and OpenACC. 28 (NOTE: The trace analysis still only supports host-side events!) 29 - Renamed '-c' command-line option of 'square' to '-C' for running 30 sanity checks on newly created reports. 31 - Added new '-c' command-line option to 'square' to allow specifying 32 the number of counters considered during report scoring (for 33 consistency with 'scorep-score'). 34 - Added new '-x' command-line option to 'square' to allow passing 35 options directly through to 'scorep-score'. 36 - Avoid unnecessary aggregation/postprocessing of reports with 37 multi-run experiments. 38 39 * Substantial code cleanup. 40 41 42 ------------------- Released version 2.5 ----------------------------- 43 44 * Support for 45 - Score-P v5.0, incl. virtual process/thread topologies 46 47 * Automatic trace analyzer changes & improvements: 48 - Various fixes and improvements in timestamp correction algorithm. 49 - Fixed 'Late Receiver' instance tracking. 50 - Slightly improved analysis report data collation. 51 52 * Added support for multi-run experiments. 53 54 * Code refactoring and various bug fixes. 55 56 * Improved user documentation: 57 - Revised User Guide including command reference. 58 - Added man pages. 59 60 61 ------------------- Released version 2.4 ----------------------------- 62 63 * Support for 64 - Cube v4.4 65 66 * Build system improvements: 67 - Fix build issues with compilers defaulting to C++11 or higher 68 (e.g., Intel 2017, PGI 17). 69 - Fix build issues with PGI 16+ compilers (pgCC no longer available) 70 - Fix build issues on Cray systems, now also properly taking 71 CRAYPE_LINK_TYPE setting into account 72 73 * Automatic trace analyzer changes & improvements: 74 - Fix rare crash/deadlock in critical-path/delay analysis while 75 analyzing MPI persistent communication. 76 - Improved memory management. 77 - Improved handling of OTF2 traces in SIONlib containers. 78 - Improved trace reading times, especially at scale. 79 - Fixed detection of wait states in active-target synchronization 80 based on EPIK traces 81 82 * Code refactoring and various bug fixes. 83 84 85 ------------------ Released version 2.3.1 ---------------------------- 86 87 * Build system improvements: 88 - Fixed build issue with GCC 6.1. 89 - Fixed build issue on the Intel Xeon Phi platform. 90 91 92 ------------------- Released version 2.3 ----------------------------- 93 94 * Support for 95 - Score-P v2.0 96 - OTF2 v2.0 97 98 * Automatic trace analyzer changes & improvements: 99 - Experimental support for Score-P traces collected using 100 sampling (see OPEN_ISSUES for limitations). 101 102 * Improved analysis report postprocessing: 103 - Revised metric hierarchies (organization, metric naming, etc). 104 - Suppress calculation of performance properties that are 105 only relevant for unused parallel programming models. 106 107 * Performance property documentation fixes & improvements. 108 109 * Build system improvements. 110 111 * Code refactoring and various bug fixes. 112 113 114 ------------------- Released version 2.2.2 --------------------------- 115 116 * Platform support: 117 - Fixed a build issue on the Intel Xeon Phi platform. 118 - Improved support for the 'ibrun' launcher. 119 120 * Automatic trace analyzer changes & improvements: 121 - Worked around rare run-time issue with MVAPICH2. 122 123 124 ------------------- Released version 2.2.1 --------------------------- 125 126 * Platform support: 127 - Added build system support for Power8/Linux. 128 - Added build system support for 64-bit ARM/Linux (AArch64). 129 - Prefer linking static over dynamic Cube/OTF2 libraries on 130 Fujitsu K/FX10/FX100. 131 132 * Automatic trace analyzer changes & improvements: 133 - Fixed delay-cost propagation through OpenMP barrier wait states. 134 - Various algorithmic optimizations reducing overall analysis 135 time for traces of multi-threaded applications: 136 ~ Improved memory management. 137 ~ Improved trace preprocessing. 138 ~ Improved timestamp correction. 139 140 * Code refactoring and various bug fixes. 141 142 143 ------------------- Released version 2.2 ----------------------------- 144 145 * Support for 146 - Score-P v1.4 147 - OTF2 v1.5, incl. full SIONlib support (if configured) 148 - Cube v4.3 149 150 * Platform support: 151 - Added support for Intel Xeon Phi, native mode only. 152 - Added support for Fujitsu FX100 (thanks to T. Nakamura, 153 Fujitsu Ltd). 154 155 * Automatic trace analyzer changes & improvements: 156 - Added basic support for POSIX threads. 157 - Added basic support for OpenMP tasking. 158 - Added lock contention analysis (OpenMP & POSIX threads). 159 - Added root-cause/delay analysis (MPI & OpenMP). 160 - New command-line options '--[no-]rootcause'. 161 162 * Code refactoring and various bug fixes. 163 164 165 ------------------- Released version 2.1 ----------------------------- 166 167 * Support for 168 - Score-P v1.3 169 - OTF2 v1.4 170 171 * Platform support: 172 - Added support for Fujitsu FX10 & K computer. 173 - Improved support for Cray systems. 174 175 * Automatic trace analyzer changes & improvements: 176 - Added Critical-path analysis. 177 - Improved Late Receiver detection. 178 - New command-line options '--[no-]critical-path' and '--single-pass'. 179 - Fixed crash in data collation when number of OpenMP threads varied 180 among MPI processes. 181 182 * Code refactoring and various small bug fixes. 183 184 * Initial version of updated User Guide (still work in progress). 185 186 187 ------------------- Released version 2.0 ----------------------------- 188 189 * Support for 190 - Score-P v1.2 191 - OTF2 v1.2 192 - Cube v4.2 193 194 * New build system based on GNU autotools. 195 196 * Significant amount of code refactoring. 197 198 * Automatic trace analyzer changes & improvements: 199 - Support for arbitrary deep system trees. 200 - Improved performance of timestamp correction. 201 - Pattern instance tracking and statistics are now enabled by 202 default. 203 - New command-line options '--verbose', '--[no-]time-correct', 204 and '--[no-]statistics'. 205 - Limited backward-compatibility support for handling existing 206 traces in EPILOG format generated by Scalasca v1.