-
Notifications
You must be signed in to change notification settings - Fork 21
/
CHANGES
507 lines (325 loc) · 25.4 KB
/
CHANGES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
* 2020.08.25 - [gfursin] fixed a bug when adding a new dataset
* 2020.01.31 - [gfursin] improved error message in program module.
* 2019.07.09 - [gfursin] propagate compiler_tags specifically for "get_compiler_version" subaction based on request here:
https://groups.google.com/forum/#!topic/collective-knowledge/Lu-kQCSrhIE
* 2019.05.26 - [gfursin] added 2 flags to ck run program
--record_deps={file to record resolved dependencies}
--run_batch_name={use this filename instead of randomly generated one
to run a workload}
* 2019.04.28 - [gfursin] added automatic installation of MILEPOST repo
(see https://github.com/ctuning/ck/issues/100)
* 2019.03.30 - [gfursin] changed -lm to %CK_EXTRA_LIB_M% in package:lib-rtl-xopenme
* 2019.03.08 - [gfursin] added prototype of a flag --use_clang_opt to ck compile program
* 2019.03.05 - [gfursin] added link to http://ReuseResearch.com
* 2019.02.08 - [gfursin] improved problem reporting in third-party program workflows (pipelines)
* 2018.12.02 - [gfursin] added sorting of run_cmds in program meta (using integer "sort")
* 2018.11.28 - [gfursin] added info about failed program workflows (how to report)
* 2018.11.26 - Added APIs to listed modules
* 2018.11.21 - [gfursin] added automatic redirect of some APIs
* 2018.10.31 - [gfursin] minor fix when redirecting git revision to null on Linux
(see https://github.com/ctuning/reproduce-milepost-project/issues/8).
* 2018.07.17 - [gfursin] added support for new website when generating lists of components
* 2018.07.15 - [gfursin] added recording of MILEPOST features to JSON file:
ck pipeline program:{program_name} --speed --no_run --milepost --milepost_out_file=milepost-features.json
* 2018.06.21 - [gfursin] when adding new dataset CK will now ask for tags and a file
[gfursin] merged patch from Stuart (see issue #30).
* 2018.06.17 - [gfursin] added --debug_run_cmd to substitute original command line in program with a user one
* 2018.06.06 - [gfursin] automatically generate list of all programs for http://cknowledge.org/shared-programs.html
* 2018.05.27 - [sztaylor] propogate "target" to platform-specific modules (detect info, set freq, etc) - see issue #12.
* 2018.05.24 - [gfursin] provided prototype of templates when adding new programs (ck add program --template)
* 2018.05.20 - added "image classification (simulation)" program.species to differentiate
benchmarking on real image classification vs random values
(needed for some ASPLOS'18 ReQuEST submissions)
* 2018.05.03 - added env (CK_TARGET_PATH) to program to get an access to machine specific files
(to support ReQuEST AI benchmark submission with Apache AVRO)
* 2018.04.07 - added support for RPC remote machine (ck add machine) - useful for accessing TVM/MXNET
* 2018.03.09 - improved cross-compilation for Windows for programs
* 2018.02.22 - added "skip_compiler_vars" to program module to support Java programs
* 2018.02.19 - added experiment.view:request-default to visualize
raw experiments from ReQuEST tournaments
- added program.species:image.classification
for image classification ReQuEST @ ASPLOS'18 tournament
* 2018.02.12 - fixed error reporting in program pipeline
if there is a problem loading pre/post processing python scripts
(noted for some ReQuEST submissions)
* 2018.02.09 - dumping tmp-dvdt-prof-<stat repetition number>.json - Flavio needs it
to show better statistics during DNN optimization
* 2018.01.01 - added possibility to add $<< to program sources
* 2017.12.14 - fixed minor issue with CT_REPEAT_MAIN (env is not updated in some cases)
* 2017.11.20 - added possibility to turn on/off specific cores during crowd-tuning
* 2017.11.17 - added extra_ld_vars to deps to be able to add multiple libs for linking during program compilation
* 2017.11.13 - now reading program output with console encoding rather than UTF-8
- added "ck find_output program:{some program UOA}" to find all outputs for validation
* 2017.11.09 - added ondemand CPU/GPU freq setting after experiments to "calm" system
to avoid overheating on embedded/IoT devices
- print LWS and GWS for OpenCL kernel profiles
* 2017.11.08 - added script:ctuning.process.dvdt-prof to process raw OpenCL stats from dvdt prof
* 2017.10.12 - added dataset file description
* 2017.10.11 - improved program.output validation recording - new outputs are recoded to local
unless specified explicitly by --output_validation_repo to avoid polluting existing project repos
* 2017.10.10 - added flag --pipeline_flags to directly update pipeline in "ck autotune pipeline"
* 2017.10.09 - added flag --program_output_uoa to "ck run program" and "ck pipeline program"
to be able to check output from a different program
(i.e. create reference output for similar groups of programs)
* 2017.10.05 - added ck pipeline program:xyz --valgrind
* 2017.10.04 - added unified_memory check when detecting OpenCL
* 2017.10.03 - asking for dvdt-prof2
* 2017.10.02 - fixing minor bug when sending multiple data set files to Android device
- allow to send only one data set file out of many
* 2017.09.28 - print OpenCL kernel timing from DVDT prof after "ck benchmark program --dvdt_prof"
- added support to collect Mali GPU hardware counters if available
* 2017.09.27 - important fix in autotuning to merge env + choices.env
may break things - need to be more tested
* 2017.09.24 - added program tags to program.output and user-friendly name to entry
(can be viewed via ck search program.output --print_name)
* 2017.09.23 - added comparison of numerical outputs with thresholds
(to collaboratively check numerical stability)
* 2017.09.22 - fixed output check if dataset json file present but keys are overloaded from cmd via env ...
* 2017.09.21 - added support to cache and reuse dependencies by tags (for automating crowdbenchmarking)
* 2017.09.20 * added support to load dataset files to run_envs (for example to specify tensorshape via env)
* added support to validate output depending on env vars (batch size, randomize, etc)
* fixing error when copying program to a non-existen repo
* 2017.09.19 * minor output improvement
* 2017.08.29 * added reduce_bug option to localize bugs during compiler fuzzing (reducing complexity)
* 2017.08.28 * fixing ck replay experiment (proper rebuilding of choices)!
* 2017.08.27 * added experiment.view:fuzzing-basic to view results from crowd-fuzzing various SW such as compilers
* 2017.08.13 * added --shared_solution_cid= (-Ocid-uid1-uid2) to reuse shared solutions in program pipelines
* 2017.07.30 * added prepare_table_with_results program
(generate HTML and Tex with optimization results - useful for interactive optimization reports)
* 2017.07.28 * added demo to reduce solution complexity (pruning compiler flags)
* 2017.07.15 * improved iterative complexity reduction
* added skip_best_base_flag to program pipeline (to allow exploration of different base optimization levels with replay)
* 2017.07.14 * fixed complexity reduction of solutions (pruning)
* 2017.05.18 - allow sub-set of command line keys (useful for DNN autotuning)
- added program:custom-autotuning-wrapper as a dummy wrapper for customized autotuning plugins
* 2017.05.04 - added description of GCC 6.3.0, 7.0.0, 7.1.0
* 2017.05.03 - added distance calculation by MILEPOST features (reproducing my past work)
* 2017.05.02 - added description of MILEPOST program features
* 2017.04.24 - fixed minor bug during Caffe autotuning on Android devices
- removing unnecessary compilation env during workload run
* 2017.04.20 - moved LMDB package to ck-env repo
- fixed bug with pre-processing script when running on remote device (should run on host)
* 2017.04.16 - fixed program crowdtuning when multiple target machines are connected
* 2017.04.15 - moved many generic packages to ck-env repo
* 2017.04.06 - updated pipeline.cmd to create customizable autotuning scenarios
and execute external command lines (CK just updates autotuning variables)
- added 'loop-with-next' and 'parallel-loop-with-next' autotuning modes
(thanks to Flavio Vella)
- fix detection of OpenCL on ARM32 devices (-pie flag missing)
* 2017.03.28 - improved processing of env vars via progrma meta
- added customization scripts before compilation
* 2017.03.26 - added package:tool-bazel-*-universal (Linux, Windows, possibly MacOS)
- added package:env-msys2-20161025 (Windows)
* 2017.03.25 - added package:jdk-8u121-universal (Linux, Windows, partially MacOS)
* 2017.03.23 - moved package:jdk-8u111-linux , lib-libjpeg , tool-bazel-0.4.1 , tool-bazel-0.4.2 from ck-tensorflow repo here
* 2017.03.18 - added package:tool-cmake-3.5.2-universal (for backwards compatibility with Android on Windows)
* 2017.03.16 - added package:tool-cmake-3.7.2-universal
* 2017.03.02 - when running program added support to add env vars from deps by OS tag
- useful to set up vars for Android such as VIENNACL_CACHE_PATH
* 2017.03.01 - added xopenme_get_timer function to xOpenME
- added compile -D vars with host and target OS (-DCK_HOST_OS_NAME_{CK OS name} -DCK_HOST_OS_NAME2_{CK OS name 2} -DCK_TARGET_OS_NAME_{CK OS name} -DCK_TARGET_OS_NAME2_{CK OS name2})
* 2017.02.21 - improving ck benchmark program (passing more CMD params)
* 2017.01.09 - added LLVM 3.9.0 package for Mac OS
* 2017.01.29 - added "XOPENME_TIME_FILE" var to change tmp-ck-timer.json output file
(for example for parallel processing of Caffe classification)
* 2017.01.26 - added --deps.KEY=UOA
* 2017.01.17 - added OpenCL 1.1 stubs. Also added tags for v1.2 and v1.1 for stabs to specialize deps ...
* 2017.01.14 - improved replay support (for OpenCL kernels from Lift/CLBlast during crowd-tuning)
* 2017.01.13 - changing data_name when copying program ...
* 2017.01.08 - a few fixes to support Caffe on Windows and Android
* 2017.01.05 - added '--skip_exec' to skip execution of a program but still run postprocessing - useful for debugging
* 2017.01.03 - added '--safe' mode to check environment (will first search for CK packages and only then detecting soft)
* 2016.12.30 * do not copy files with the same size to remote device (Android)
useful not to copy large DNN models to the remote device
* 2016.12.29 * major update to support compilation of Caffe deps for Android fully via CK
* added pre-process script for compilation (mainly to customize various ARM features for Android)
* 2016.12.27 * added Android support for protobuf 3.1.0 and trunk (with patch) to make Caffe work on Android
* 2016.12.23 * now checks if program supports a given host and target (similar to package)
* 2016.12.21 * major update (added run-time SW deps handling)
* REVERTED (updating env directly from deps (experimental)) ...
* 2016.12.20 * adding to program module pruning of command lines by deps and vice versa ...
* 2016.12.15 * now pass dataset_uoa, dataset_file, dataset_path to pre and post processing scripts
(useful for GEMMBENCH and universal tuning)
* 2016.12.13 * improved prunning of choices in raw experiments
* 2016.12.12 * added "ck benchmark program:XYZ" to automatically set up env, freq, and run autotune with 1 iteration
* updated autotuning view
* 2016.12.11 * added output check (for now binary comparison)
* sorted machine choices (host is first one to be able to press Enter to select it)
* added support to select GPGPU to program pipeline
* 2016.12.10 * added gflags-2.2.0-linux (to be able to install stable gflags if it is not installed)
* added glog-0.3.4-linux
* added protobuf-3.1.0-linux
* fixed selection of OpenCL/CUDA devices for workloads with GPGPU requirement
(to fix them during crowd-tuning/crowd-optimization)
* 2016.12.09 * added protobuf-3.0.0 for Android (otherwise latest trunk doesn't compile for Android)
* 2016.12.05 * recording program version (to distinguish species during crowd-tuning)
* 2016.11.30 * improved gprof output (for RPi project)
* 2016.11.29 * added gflags 2.1.2 for Android (otherwise trunk is not compiling anymore)
* 2016.11.28 * improved GPGPU detection and added --sudo
(do not fail if OpenCL lib not found; print no CK-enabled GPGPU devices found)
* 2016.11.20 * added possibility to change characteristics directly from the postprocess script
* 2016.11.18 * redirected git stderr output to null to avoid misunderstandings during compilation
* 2016.11.16 * added clang 3.9.0 as installable package for Linux, Android
* 2016.11.14 * added support to copy extra files to Android
(useful for CrystaX NDK to copy libcrystax.so)
* added support for both CrystaX and Google NDK
* 2016.11.10 * added gflags package for Android
* added glog package for Android (with patch)
* added lmdb package for Android
* added opencv package for Android (moved to ck-math)
* added Google's protobuf for Linux and Android
* 2016.11.04 * making os.getcwd() safer
* 2016.10.12 * fixing order of libs (https://github.com/ctuning/ck/issues/67)
* 2016.09.27 * adding support for remote execution unified across all platforms
* 2016.09.26 * changing 'device' to 'machine'
* 2016.09.25 * added packages "tool-show-cuda-devices" and "tool-show-opencl-devices"
* 2016.09.23 * added selection of GPGPU target to 'ck detect platform.gpgpu' widget
* 2016.09.22 * added 'device' support to program pipeline (with selector)
* added adb soft dependency to android remote hosts
* 2016.09.21 * added support for --params in module "program" to pass parameters
to pre and post processing scripts, for example to customize ARM Workload Automation
* 2016.09.19 * added support for --target in module "program" (see module "device" in ck-env)
* 2016.09.15 * added program meta "run_via_third_party":"yes" - to run via WA on Anrdoid, for example.
* 2016.09.05 * changed soft functions 'detect' to 'internal_detect' and 'check' to 'detect'
* 2016.09.04 * fixed selection of wrong dataset when extra dataset tags are present
* added --env.{KEY} to pipeline program
* 2016.08.08 * added Git info for OpenME
* 2016.07.20 * added support for CUDA compute 6.x (Pascal) and 7.x (Volta)
* 2016.07.13 * added 'add_file_to' dataset -> to add dataset files and register them in the entry meta
* 2016.07.08 * added pre processing script via CK (with params)
* 2016.06.22 * added possibility to run postprocess script via CK (i.e. with the possibility to reuse all CK functions)
* 2016.06.20 * improving execution on remote device (using .sh instead of CMD which may become too long and fail)
* 2016.06.08 * passing custom autotuner meta from program to custom autotuning script
* 2016.05.30 * support for new CK web server
* added possibility to run internal reference autotuning iterations
when checking pre-existing solutions (useful for experiment replay/reproducibility)
* 2016.05.29 - added 'exchange_locally' mode to GPGPU detection and info exchange
- added support to replay experiments even when customized autotuner ...
* 2016.05.27 - added support to print various files after run
* 2016.05.18 - calculating compute capability for GPGPU ...
- passing more internal vars to customized autotuner
- add possibility to collect raw results of all autotuning iterations
to be processed externally (useful for customized crowd-tuning)
- moving machine learning based and customizer autotuning
to external plugins
* 2016.05.17 - adding '--no_vars' to skip default -D variables and hence use default vars from the sources
* 2016.05.16 - adding 'program.behavior' module to crowd-benchmark/crowd-tune program behavior
across diverse data sets and hardware
* 2016.05.13 - moving platform.gpgpu here from ck-env repository to prepare for CUDA/OpenCL crowd-tuning
* 2016.05.06 - adding dummy JULIA language meta info (optimizations), after MIT visit
* 2016.04.28 - major update to packages and environment to automate detection of already
installed software (for Windows, Linux and partially Android).
- adding compiler_tags to force compiler type (llvm,icc,gcc,microsoft,ctuning-cc,...)
* 2016.04.19 - fixing bug with parallel-random choice selection (during autotuning);
found when working on OpenCL crowdtuning for BLAS ...
- added possibility to use customized autotuner (directly changes pipeline)
* 2016.04.15 - moving platform.init entries to ck-env (as it is needed for environment, i.e. to support Windows)
* 2016.03.25 - adding support to extract MILEPOST/cTuning cc static program features (legacy)
* 2016.03.21 - adding universal OpenCL stubs package (1.2) targeting Android devices
* 2016.03.07 * adding support for pre processing scripts (improving OpenCL/CUDA/MPI autotuning)
* adding extra tags to select datasets (useful for mobile phone crowdtuning to select small datasets)
* 2016.02.26 * adding preliminary support to crowdsource program optimization using Android mobile phones
* 2016.02.17 * adding dimension reduction to pipeline
(leaving only influential optimizations + inverting and turning off all if needed)
very important for proper collaborative machine learning
* changed platform.accelerator -> platform.gpu
* 2016.02.13 * improving checking of pre-existing solutions. Started implementing automatic pruning of choices ...
* 2016.02.12 * adding support to check pre-selection solutions (found during crowdtuning)
during autotuning (preparation for clustering and optimization reactions)
* recording characteristics to pre-selected solutions
* 2016.02.10 * adding ARM and general CPU specific flags for GCC and LLVM
* changing arch-specific to cpu-specific tags
* 2016.02.09 * improving GCC compiler flag description
* adding support to explore compiler flags with various tags (boolean, parametric, arch-specific)
* 2016.02.08 * preparing LLVM descriptions
* fixing detection of the most close compiler description (for autotuning/crowdtuning)
* adding aggregation of failed cases (to automatically/manually detect compiler bugs)
* 2016.02.05 * started adding pipeline key descriptions (user-friendly info)
* various small fixes and enhancements for experiment replay
* 2016.02.04 * improving search for the most close compiler description (version)
* adding GCC v5.3 and v6.0 descriptions
* 2016.02.03 * adding detected compiler description to choices
* improving compiler optimization flag web viewer
* 2016.02.01 * added mode to skip recording pipeline/desc info to experiment
(to avoid sending huge files between servers during optimization crowdsourcing)
* 2016.01.29 * general improvements
* adding calculation of multi-dimensional improvements
* adding universal conditions on solutions during auto-tuning
(such as improve energy without sacrificing execution time,
or improve execution time without increasing code size ...)
* 2016.01.15 * returning added or deleted points/sub-points during autotuning
(for crowd-tuning)
* Adding margins when comparing values on Pareto frontier
(used to monitor only significant speedups for when crowdsourcing autotuning
particularly across mobile devices)
* Fixing small bug in RTL for older MILEPOST project extracted codelets
(to crowd-tune them and them use to speed up optimization and modeling
large apps based on these codelets - see our papers)
* 2016.01.11 * adding support to select program, cmd, dataset and dataset file randomly
to support program multi-objective crowdtuning
* adding 'ck clean_tmp program:*' to clean tmp dirs in all program entries
(to clean up results of crowd-tuning)
* 2016.01.04 * adding crowdtune action -> redirecting to "ck crowdsource program.optimization"
* 2015.12.31 * adding dummy "program.output" to check program output during benchmarking/tuning
* 2015.12.30 * adding "ask_enter_after_each_iteration" - if 'yes', ask to press Enter after each iteration
- useful for debugging, analyzing Pareto and checking speedups
* fixing bug with feature selection for Pareto
* 2015.12.15 * adding "force_pipeline_update" key to properly update pipelines (workflows) during replay
* 2015.12.13 * adding boolean keys "compilation_success_bool" and "run_success_bool" (True/False)
to autotuning pipeline to speed up predictive analytics
* recording extra info for reproducibility in flat array (pipeline_state
with statistical repetition number and with pipeline fail and reason)
* 2015.12.12 * failing autotuning pipeline if post-processing scripts fails
* 2015.12.07 * fixing compiler flag autotuning demo (small API change)
* fixing a bug in scatter graphs (found when plotting results of multi-objective optimization)
* 2015.12.06 * adding 'local_platform' key to be able to use local platform parameters
(useful to retarget experiments from another machine)
* 2015.11.23 * adding algorithm module dummy (future work)
* 2015.11.22 * fixing compile flag autotuning demos;
* fixing a few bugs in random compiler flag autotuning (iterative compilation);
* 2015.11.20 * fixing problem with return code when compiling programs, if programs are not from GIT repo
* 2015.11.16 * adding correct selection of default parameters for the first iteration of random autotuning
(noticed during dgemmbench OpenCL autotuning)
* 2015.11.10 * adding "sleep" parameter to autotune pipeline (set delay between iterations)
* fixing error in exhaustive exploration when using "loop" with "choices"
* adding extra_dataset_files to dataset module to support more files
when sending to Android (such as OpenCL autotuning via mobile phones)
* 2015.10.28 * adding CK_PROG_COMPILER_VARS when compiling via script ...
* adding "git rev-parse HEAD" during compilation to check
GIT hash for the compiled program (useful to debug and explain
unexpected behavior when reproducing experiments)
* 2015.10.27 * adding 'cp' to copy programs and update 'backup_data_uid' (fixing issue #41)
* 2015.10.22 * adding support to compile via scripts
(useful for PENCILCC from CARP project and LLVM LNT)
* adding "add_extra_env_for_compilation" in compilation meta
(useful for PENCILBENCH)
* 2015.09.25 * fixing xopenme.h for Windows
* 2015.09.22 * fixing pipeline update when dataset is changed (i.e. updating related dataset files)
(reported by Anton)
* taking statistical repeat from pipeline if exists (to properly replay experiments)
* 2015.09.10 * moved 'script:platform.init' to 'platform.init:*'
* 2015.09.09 * adding 'skip_stat_analysis' to compact recorded data when stat analysis is not needed
(useful for CLSmith)
* adding function 'import_all_files' in 'dataset' to automatically register all dataset files
* 2015.09.08 * improving dataset file selector (suggested by Anton)
* making text to select UIDs less ambiguous (asked by Anton)
* fixing program calibration problem for slambench (detected by Anton)
* if iterations in pipeline == -1, infinite loop (suggested by Andrei Lascu)
* 2015.09.02 * turning off 'state check' (frequency change during experiments) by default
in autotuning demos
* 2015.09.01 * adding timeout for program compilation and execution
(useful for research in computer engineering where
generated code may be buggy and result in infinite loops,
i.e. during autotuning or compiler testing and bug detection)
* providing more info about compilation and execution failures
* adding possibility to record only failed cases
(useful to crowdsource testing compilers/programs/os/architecture for bugs)
* 2015.08.12 * adding 'auto' tag to identify/detect automatically extracted compiler descriptions
* adding parsing of reserved key $#ck_take_from_{CID}$# - for example, used to calculate accuracy for autotuning in slambench
* adding possibility to reverse key order in Pareto frontier filter during multi-objective autotuning
(useful for FPS in slambench, for example)
* 2015.08.11 * showing compiler flags in HTML (when viewing entries)
* fixing some wrong compiler flag descriptions
* 2015.08.10 * added GCC 5.2.0 CK package (will be downloaded and installed)
* 2015.08.09 * add extra dict when recording experiments (useful to record name and subview_uoa, for example)
* fixing a few bugs in experiment viewing