RELEASE_NOTES 7.32 KB
Newer Older
1 2 3 4 5 6
0.5.2 (July 13, 2016)

Summer of CODES was another huge success!
This update adds the commit function to the CODES models, representing an
update to the latest version of ROSS (d3bdc07)

7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.5.1 (June 09, 2016)

network:
==========

corrected link latency calculation in dragonfly model

printf argument mismatch in dragonfly model

refactors to the torus bandwidth calculation to mirror that of the other
  networks (no functional change)

more robust type conversions in dragonfly (int sizes -> uint64_t)

Jonathan Jenkins's avatar
Jonathan Jenkins committed
21 22 23
removed the redundant and obsolete MPI replay simulator
  (modelnet-mpi-wrklds.c). The proper version to use is modelnet-mpi-replay.c

24 25 26
----------

0.5.0 (May 24, 2016)
27 28 29 30 31 32 33

general:
==========

codes-base and codes-net have been combined into a single project (now at
https://xgitlab.cels.anl.gov/codes/codes).

Jonathan Jenkins's avatar
Jonathan Jenkins committed
34
updated to ROSS revision d9cef53.
35 36 37 38 39 40 41 42

fixed a large number of warnings across the codebase.

networks:
==========

addition of the SlimFly network topology, corresponding to the Wolfe et al.
  paper "Modeling a Million-node Slim Fly Network using Parallel Discrete-event
43
  Simulation", at SIGSIM-PADS'16. See README.slimfly.txt
44
  (src/networks/model-net/doc).
45 46 47

modelnet now supports sampling at regular intervals. Dragonfly LPs can
  currently make use of this - others can be added based on demand. See
48
  (src/network-workloads/README_synthetic.txt).
49

50 51 52
dragonfly and torus network models credit-based flow control has been updated.
    Dragonfly model's adaptive routing algorithms have been updated. For details,
    see paper "Enabling Parallel Simulation of Large-Scale HPC Network Systems",
53
    M.Mubarak et al., at IEEE Trans. on Parallel and Distributed Systems (TPDS).
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
allow 0-byte messages in model-net.

enable "local" model-net messages (for LPs sharing the same model-net endpoint -
  approximates a zero-copy ownership pass of the payload)

the model_net_event family of functions now return a token value which must
  be passed to model_net_event_rc2.

workloads:
==========

concurrent workload support added to workload generators and the MPI
  simulation layer. See codes-jobmap.h and the modified codes-workload.h.
  Thanks to Xu Yang for the partial contributions. Not all workload generators
  support concurrent workloads at this time.

scripts for generating job allocations specific to the torus and dragonfly
  topologies. See scripts/allocation_gen. Thanks to Xu Yang for the
  contributions.

multiple fixes to the MPI simulation layer and DUMPI workload generator.

concurrent workload support and more flexible rank mappings in the MPI
  simulation layer. Thanks to Xu Yang for the initial code.

removed scalatrace workload generator, which never made it to a usable state.

a new checkpoint IO workload generator has been added, based on the Daly paper
  "A higher order estimate of the optimum checkpoint interval for restart
84 85
  dumps" at Future Generation Computing Systems 2004. See README.codes-workload
  (src/workload).
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

utilities:
==========

fixes to rc-stack - memory no longer leaks in sequential mode, and optimistic
  debug mode is now supported.

added a more performance-sensitive function, codes_mapping_get_lp_info2, to
  codes-mapping (passes const pointers around instead of copying strings)

added a "mapping context" API for better controlling implicit LP-LP mappings,
  including modelnet, local storage model, and resource. See
  codes/codes-mapping-context.h and added functions in the mentioned LPs for
  details.

formalized a callback mechanism for CODES, replacing previous ad-hoc
  methods of passing control from LPs to arbitrary other LPs. The request and
  local storage model LP APIs have been changed to use this mechanism, and
  model-net has additional APIs to use user-provided mapping contexts. See
  codes/codes-callback.h for the API and tests/resource-test.c for advanced
  usage.

deprecations:
==========

model_net_event_rc (use model_net_event_rc2, which will eventually be renamed
  to model_net_event_rc)

codes_event_new (define your own bounds-checking macro if need be)

116
----------
117

118 119 120 121 122 123 124 125 126 127
0.4.1 (September 30, 2015)

general:
==========

fix compatibility with recent ROSS releases

----------

0.4.0 (May 6, 2015)
128

129 130
codes-base

131 132
general:
==========
133

134
significant source reorganization / refactoring
135

136
refactor some private headers out of the public eye
137

138 139 140 141
dead code removal

documentation:
==========
142

143
improved example_heterogeneous example program
144

Jonathan Jenkins's avatar
Jonathan Jenkins committed
145
added configuration to example_heterogeneous showing two torus networks
146

147 148
reorganized files to prevent name collisions on OSX. Top-level docs other than
  copyright now in doc directory
149

150 151 152 153
additions to best practice document

configurator:
==========
154

155
more stable file format for configurator output
156

157
ignore unrelated parameters passed into filter_configs
158

159 160 161 162
handle empty cfields in configurator

workloads:
==========
163

164
combined network and IO workload APIs into a single one
165

166
adding dumpi workload support in codes-workload-dump utility
167

168
workload dump utility option cleanup
169

170
renamed "bgp" workload generator to "iolang", significant cleanups
171

172
put network workload ops in workload dump util
173

174
removing one of the dumpi libraries from the build. It was generating some unwanted dumpi files.
175

176 177 178 179
network workload API more fleshed out

utilities:
==========
180

Jonathan Jenkins's avatar
Jonathan Jenkins committed
181
configuration bug fixes for larger LP type counts
182

183
resource LP annotation mapping hooks
184

185
local storage model API switch to use annotations
186

187
better configuration error handling
188

189
hedge against precision loss in codes_local_latency (see codes.h)
190

191 192 193
use a different RNG than default for codes_local_latency
- prevents addition/removal of codes_local_latency calls from poisoning RNG
  stream of calling model
194

195 196
added simple GVT-aware stack with garbage collection (see rc-stack.h)

197 198
codes-net

Jonathan Jenkins's avatar
Jonathan Jenkins committed
199 200
general:
==========
201

Jonathan Jenkins's avatar
Jonathan Jenkins committed
202
cleanup of much of the code base
203

Jonathan Jenkins's avatar
Jonathan Jenkins committed
204
more informative error for failure to find modelnet lps
205

Jonathan Jenkins's avatar
Jonathan Jenkins committed
206
removed redundant include directory on install (was 'install/codes/codes/*.h')
Jonathan Jenkins's avatar
Jonathan Jenkins committed
207 208 209

documentation:
==========
210

Jonathan Jenkins's avatar
Jonathan Jenkins committed
211 212
reorganized files to prevent name collisions on OSX. Top-level docs other than
  copyright now in doc directory
213

Jonathan Jenkins's avatar
Jonathan Jenkins committed
214
updated code documentation
215

Jonathan Jenkins's avatar
Jonathan Jenkins committed
216
fix linker error in certain cases with codes-base
217

Jonathan Jenkins's avatar
Jonathan Jenkins committed
218 219 220 221 222
tweaked config error handling


networks:
==========
223
fix to loggp latency calculation when using "receive queue"
224

Jonathan Jenkins's avatar
Jonathan Jenkins committed
225
made torus lps agnostic to groups and aware of annotations
226

Jonathan Jenkins's avatar
Jonathan Jenkins committed
227
miscellaneous fixes to dragonfly model
228

Jonathan Jenkins's avatar
Jonathan Jenkins committed
229 230
updates to simplep2p: support for having different latency/bw at sender &
  receiver end. See src/models/networks/model-net/doc/README.simplep2p.txt
231

Jonathan Jenkins's avatar
Jonathan Jenkins committed
232
minor fixes to usage of quickhash in replay tool
233

Jonathan Jenkins's avatar
Jonathan Jenkins committed
234
fixed RNG reverse computation bug in loggp
235

Jonathan Jenkins's avatar
Jonathan Jenkins committed
236 237 238 239
fixed swapped arguments in round-robin scheduler causing short circuit

workloads:
==========
240

Jonathan Jenkins's avatar
Jonathan Jenkins committed
241
minor changes to dumpi trace config files
242

Jonathan Jenkins's avatar
Jonathan Jenkins committed
243
resolving minor bug with reverse computation in dumpi traces
244

Jonathan Jenkins's avatar
Jonathan Jenkins committed
245
Updating network trace code to use the combined workload API
246

Jonathan Jenkins's avatar
Jonathan Jenkins committed
247
Adding synthetic traffic patterns (currently with dragonfly model)
248

Jonathan Jenkins's avatar
Jonathan Jenkins committed
249
Adding network workload test program for debugging
250

Jonathan Jenkins's avatar
Jonathan Jenkins committed
251
Updating MPI wait/wait_all code in replay tool
252

253 254 255
----------

0.3.0 (November 7, 2014)
256

257 258
codes-base

259 260
Initial "official" release. Against previous repository revisions, this release
includes more complete documentation.
261 262 263

codes-net

Jonathan Jenkins's avatar
Jonathan Jenkins committed
264 265 266 267
Initial "official" release. Against previous repository revisions, this release
includes more complete documentation and a rename of the "simplewan" model to
the "simplep2p" (simple point-to-point) model to more accurately represent
what it's modeling.