This is an old revision of the document!
Table of Contents
News log
SVN changelog
------------------------------------------------------------------------
r182 | massimot | 2014-08-13 23:53:35 +0200 (Wed, 13 Aug 2014) | 3 lines
added the pool evolution pattern into the high-level pattern set (doxygen)
------------------------------------------------------------------------
r181 | massimot | 2014-08-13 23:31:25 +0200 (Wed, 13 Aug 2014) | 3 lines
makefile fix
------------------------------------------------------------------------
r180 | aldinuc | 2014-08-13 01:55:00 +0200 (Wed, 13 Aug 2014) | 1 line
------------------------------------------------------------------------
r179 | aldinuc | 2014-08-13 01:53:56 +0200 (Wed, 13 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r178 | aldinuc | 2014-08-12 01:23:03 +0200 (Tue, 12 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r177 | aldinuc | 2014-08-11 18:22:33 +0200 (Mon, 11 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r176 | aldinuc | 2014-08-11 17:25:41 +0200 (Mon, 11 Aug 2014) | 2 lines
Allocator documentation
------------------------------------------------------------------------
r175 | aldinuc | 2014-08-11 02:29:34 +0200 (Mon, 11 Aug 2014) | 2 lines
Substantial cleaning of documentation
------------------------------------------------------------------------
r174 | aldinuc | 2014-08-10 17:43:18 +0200 (Sun, 10 Aug 2014) | 3 lines
Cleaning
------------------------------------------------------------------------
r173 | aldinuc | 2014-08-10 16:26:53 +0200 (Sun, 10 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r172 | massimot | 2014-08-10 10:03:33 +0200 (Sun, 10 Aug 2014) | 4 lines
- fixed a bug in the init_data_static
- cleanup
------------------------------------------------------------------------
r171 | aldinuc | 2014-08-10 02:36:25 +0200 (Sun, 10 Aug 2014) | 2 lines
Doxygen major update (not yet complete)
------------------------------------------------------------------------
r170 | aldinuc | 2014-08-10 02:35:27 +0200 (Sun, 10 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r169 | aldinuc | 2014-08-10 02:33:20 +0200 (Sun, 10 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r168 | massimot | 2014-08-09 17:38:12 +0200 (Sat, 09 Aug 2014) | 3 lines
moved name
------------------------------------------------------------------------
r167 | aldinuc | 2014-08-09 17:23:39 +0200 (Sat, 09 Aug 2014) | 1 line
------------------------------------------------------------------------
r166 | aldinuc | 2014-08-09 17:20:50 +0200 (Sat, 09 Aug 2014) | 2 lines
minor fix
------------------------------------------------------------------------
r165 | aldinuc | 2014-08-09 16:07:35 +0200 (Sat, 09 Aug 2014) | 2 lines
Documentation
------------------------------------------------------------------------
r164 | aldinuc | 2014-08-09 16:06:30 +0200 (Sat, 09 Aug 2014) | 2 lines
more tests
------------------------------------------------------------------------
r163 | massimot | 2014-08-09 11:47:16 +0200 (Sat, 09 Aug 2014) | 3 lines
fixed 2 problems appearing when the collector is removed in the task-farm pattern
------------------------------------------------------------------------
r162 | aldinuc | 2014-08-07 16:24:48 +0200 (Thu, 07 Aug 2014) | 2 lines
Cleaning
------------------------------------------------------------------------
r161 | aldinuc | 2014-08-07 16:20:29 +0200 (Thu, 07 Aug 2014) | 2 lines
QT-mandebrot example (beta) added
------------------------------------------------------------------------
r160 | massimot | 2014-08-05 21:59:23 +0200 (Tue, 05 Aug 2014) | 4 lines
- fixed a problem that happens (sometimes) when the collector is removed in the farm pattern
- minor fixes in mapping_utils
------------------------------------------------------------------------
r159 | aldinuc | 2014-08-05 01:38:45 +0200 (Tue, 05 Aug 2014) | 2 lines
Bowtie2 scripts - for compatibility with BT2
------------------------------------------------------------------------
r158 | massimot | 2014-08-02 09:16:24 +0200 (Sat, 02 Aug 2014) | 2 lines
restored
------------------------------------------------------------------------
r157 | massimot | 2014-08-02 09:15:16 +0200 (Sat, 02 Aug 2014) | 3 lines
mess with svn
------------------------------------------------------------------------
r156 | massimot | 2014-08-01 15:31:57 +0200 (Fri, 01 Aug 2014) | 1 line
------------------------------------------------------------------------
r155 | massimot | 2014-08-01 11:39:10 +0200 (Fri, 01 Aug 2014) | 4 lines
removed test_dotprod from CMakeLists
------------------------------------------------------------------------
r154 | aldinuc | 2014-07-31 23:17:14 +0200 (Thu, 31 Jul 2014) | 2 lines
Minor cmake fix
------------------------------------------------------------------------
r153 | aldinuc | 2014-07-31 23:03:55 +0200 (Thu, 31 Jul 2014) | 2 lines
Minor cmake fix
------------------------------------------------------------------------
r152 | aldinuc | 2014-07-31 22:59:22 +0200 (Thu, 31 Jul 2014) | 1 line
------------------------------------------------------------------------
r151 | aldinuc | 2014-07-31 19:09:31 +0200 (Thu, 31 Jul 2014) | 2 lines
Bowtie-2.0.6-FF. The Fastflow version of the Bowtie 2 sequence aligner. Faster than the original.
------------------------------------------------------------------------
r150 | massimot | 2014-07-31 11:00:32 +0200 (Thu, 31 Jul 2014) | 3 lines
added a very simple test showing how to use the MDF pattern
------------------------------------------------------------------------
r149 | aldinuc | 2014-07-30 18:08:17 +0200 (Wed, 30 Jul 2014) | 2 lines
Default std now c++11
------------------------------------------------------------------------
r148 | aldinuc | 2014-07-30 18:07:33 +0200 (Wed, 30 Jul 2014) | 2 lines
Cosmetic fixes and Cmakefile update
------------------------------------------------------------------------
r147 | massimot | 2014-07-30 18:00:25 +0200 (Wed, 30 Jul 2014) | 3 lines
fixed compilation problems
------------------------------------------------------------------------
r146 | aldinuc | 2014-07-30 17:06:23 +0200 (Wed, 30 Jul 2014) | 2 lines
Clang compliance (minor): data structure padding
------------------------------------------------------------------------
r145 | massimot | 2014-07-30 17:01:00 +0200 (Wed, 30 Jul 2014) | 3 lines
removed
------------------------------------------------------------------------
r144 | massimot | 2014-07-30 16:34:20 +0200 (Wed, 30 Jul 2014) | 5 lines
fixed bug in parallel_for_internals.hpp (missing iniziatization of workersspinwait static_scheduling member variables)
------------------------------------------------------------------------
r143 | massimot | 2014-07-30 10:23:09 +0200 (Wed, 30 Jul 2014) | 3 lines
removed ff_matmul_map
------------------------------------------------------------------------
r142 | massimot | 2014-07-30 10:09:18 +0200 (Wed, 30 Jul 2014) | 3 lines
cleanup
------------------------------------------------------------------------
r141 | massimot | 2014-07-30 09:40:42 +0200 (Wed, 30 Jul 2014) | 3 lines
new ff_Map pattern based on ParallelForReduce
------------------------------------------------------------------------
r140 | massimot | 2014-07-30 09:39:05 +0200 (Wed, 30 Jul 2014) | 3 lines
cleanup
------------------------------------------------------------------------
r139 | massimot | 2014-07-28 16:16:40 +0200 (Mon, 28 Jul 2014) | 3 lines
added licence file and added an header to each file containing the licence statement
------------------------------------------------------------------------
r138 | massimot | 2014-07-24 19:02:20 +0200 (Thu, 24 Jul 2014) | 3 lines
first implementation of the ParallelForPipeReudce pattern
------------------------------------------------------------------------
r137 | massimot | 2014-07-24 18:59:58 +0200 (Thu, 24 Jul 2014) | 3 lines
updated performance tests for the function_min app
------------------------------------------------------------------------
r136 | massimot | 2014-07-24 18:59:21 +0200 (Thu, 24 Jul 2014) | 3 lines
new tests
------------------------------------------------------------------------
r135 | massimot | 2014-07-17 12:10:58 +0200 (Thu, 17 Jul 2014) | 4 lines
fixed compilation problem of the test_pool*.cpp tests
------------------------------------------------------------------------
r134 | massimot | 2014-07-15 15:27:01 +0200 (Tue, 15 Jul 2014) | 3 lines
new performance tests after bug fix
------------------------------------------------------------------------
r133 | massimot | 2014-07-15 14:57:30 +0200 (Tue, 15 Jul 2014) | 3 lines
fixed bug: abs-->fabs
------------------------------------------------------------------------
r132 | massimot | 2014-07-14 14:50:46 +0200 (Mon, 14 Jul 2014) | 3 lines
added a new application using the PoolEvolution pattern
------------------------------------------------------------------------
r131 | massimot | 2014-07-09 10:19:31 +0200 (Wed, 09 Jul 2014) | 4 lines
poolEvolution interface improved
ParallelFor improved (also some more info has been added at the beginning of the parallel_for.hpp file)
------------------------------------------------------------------------
r130 | massimot | 2014-06-25 14:50:15 +0200 (Wed, 25 Jun 2014) | 5 lines
- added spinWait in the pool pattern
- introduced threadPause in the parallel for
------------------------------------------------------------------------
r129 | massimot | 2014-06-19 15:52:05 +0200 (Thu, 19 Jun 2014) | 3 lines
added pool test
------------------------------------------------------------------------
r128 | massimot | 2014-06-10 09:53:35 +0200 (Tue, 10 Jun 2014) | 7 lines
- separated parallel_for_internals.hpp and parallel_for.hpp
- added the possibility to control the grain size in the static scheduling
policy of the ParallelFor(Reduce) pattern(s)
- ff_node::eosnotify interface changed:
virtual void eosnotify(int id=-1) --> virtual void eosnotify(ssize_t id=-1)
------------------------------------------------------------------------
r127 | aldinuc | 2014-05-14 00:32:46 +0200 (Wed, 14 May 2014) | 2 lines
Doxygen fixes
------------------------------------------------------------------------
r126 | aldinuc | 2014-05-10 16:55:57 +0200 (Sat, 10 May 2014) | 2 lines
MacOS fixes
------------------------------------------------------------------------
r125 | aldinuc | 2014-05-10 16:55:36 +0200 (Sat, 10 May 2014) | 2 lines
MacOS fixes
------------------------------------------------------------------------
r124 | aldinuc | 2014-05-10 16:55:13 +0200 (Sat, 10 May 2014) | 2 lines
MacOS fixes
------------------------------------------------------------------------
r123 | aldinuc | 2014-05-07 15:15:43 +0200 (Wed, 07 May 2014) | 1 line
Windows warnings fixes (types)
------------------------------------------------------------------------
r122 | massimot | 2014-05-05 11:49:51 +0200 (Mon, 05 May 2014) | 4 lines
- poolEvolution improvements
- added parallel_for_idx
------------------------------------------------------------------------
r121 | massimot | 2014-04-30 18:07:24 +0200 (Wed, 30 Apr 2014) | 4 lines
added the parallel_for_thid and
parallel_reduce_thid methods both in the ParallelFor and in the ParallelReduce classes
------------------------------------------------------------------------
r120 | aldinuc | 2014-04-30 17:12:46 +0200 (Wed, 30 Apr 2014) | 1 line
Fixes for MSVC 2013: All examples not depending from third-party libraries (boost, opencv) compile. Still many warnings related to int and long size on win64 platform.
------------------------------------------------------------------------
r119 | aldinuc | 2014-04-30 16:56:22 +0200 (Wed, 30 Apr 2014) | 2 lines
Fixes for clang 5.1.
------------------------------------------------------------------------
r118 | aldinuc | 2014-04-30 13:56:11 +0200 (Wed, 30 Apr 2014) | 2 lines
Minor string fix
------------------------------------------------------------------------
r117 | aldinuc | 2014-04-30 12:41:36 +0200 (Wed, 30 Apr 2014) | 1 line
MSVS 2013 compilation partially fixed. parallel_for.hpp still not working.
------------------------------------------------------------------------
r116 | massimot | 2014-04-30 10:39:43 +0200 (Wed, 30 Apr 2014) | 2 lines
fixes compiling with CUDA 5
------------------------------------------------------------------------
r115 | massimot | 2014-04-30 09:48:33 +0200 (Wed, 30 Apr 2014) | 2 lines
fix
------------------------------------------------------------------------
r114 | massimot | 2014-04-30 09:39:54 +0200 (Wed, 30 Apr 2014) | 5 lines
added the poolEvolutionPattern.
minor changes to stencilReduceCUDA
addedd some tests
------------------------------------------------------------------------
r113 | massimot | 2014-04-15 08:40:35 +0200 (Tue, 15 Apr 2014) | 3 lines
minor changes to test_multi_input8
------------------------------------------------------------------------
r112 | massimot | 2014-04-14 18:33:02 +0200 (Mon, 14 Apr 2014) | 2 lines
new simple multi-input test added
------------------------------------------------------------------------
r111 | massimot | 2014-04-08 16:28:44 +0200 (Tue, 08 Apr 2014) | 4 lines
- fixed compilation problem with the Barrier
- new tests for stencilReduceCUDA
- moved stencil.hpp to stencilReduce.hpp
------------------------------------------------------------------------
r110 | massimot | 2014-03-24 18:22:53 +0100 (Mon, 24 Mar 2014) | 3 lines
changed FF_MAP
------------------------------------------------------------------------
r109 | massimot | 2014-03-24 16:35:50 +0100 (Mon, 24 Mar 2014) | 3 lines
CMakeLists fixes
------------------------------------------------------------------------
r108 | massimot | 2014-03-24 16:10:54 +0100 (Mon, 24 Mar 2014) | 3 lines
typos
------------------------------------------------------------------------
r107 | massimot | 2014-03-24 15:36:29 +0100 (Mon, 24 Mar 2014) | 2 lines
fixes for APPLE
------------------------------------------------------------------------
r106 | massimot | 2014-03-24 15:12:26 +0100 (Mon, 24 Mar 2014) | 3 lines
removed test file
------------------------------------------------------------------------
r105 | massimot | 2014-03-24 15:09:31 +0100 (Mon, 24 Mar 2014) | 3 lines
Old version of Salt & Pepper denoiser application (spd-denoiser) removed
------------------------------------------------------------------------
r104 | massimot | 2014-03-24 15:08:12 +0100 (Mon, 24 Mar 2014) | 3 lines
Salt & Pepper and Gaussian denoiser - new version with CUDA support
------------------------------------------------------------------------
r103 | massimot | 2014-03-23 16:57:30 +0100 (Sun, 23 Mar 2014) | 5 lines
- moved to 2.0.4
- added stencilReduceCUDA.hpp
- added CUDA tests
------------------------------------------------------------------------
r102 | massimot | 2014-03-21 22:49:08 +0100 (Fri, 21 Mar 2014) | 8 lines
- many changes, among these:
* introduced GO_OUT: a new reserved message (see config.hpp)
* optimized ParallelFor and ParallelReduce computation when executed inside
a serial loop
* changed the Barrier implementation (spinBarrier still the same)
* FF_EOS_NOFREEZE no longer propagated over the output channel
------------------------------------------------------------------------
r101 | aldinuc | 2014-03-21 12:58:24 +0100 (Fri, 21 Mar 2014) | 1 line
------------------------------------------------------------------------
r100 | aldinuc | 2014-03-14 01:24:42 +0100 (Fri, 14 Mar 2014) | 2 lines
Minor fix
------------------------------------------------------------------------
r99 | aldinuc | 2014-03-14 01:24:06 +0100 (Fri, 14 Mar 2014) | 1 line
------------------------------------------------------------------------
r98 | massimot | 2014-03-12 10:51:13 +0100 (Wed, 12 Mar 2014) | 3 lines
minor fixes
------------------------------------------------------------------------
r97 | massimot | 2014-03-10 15:21:20 +0100 (Mon, 10 Mar 2014) | 18 lines
Version 2.0.3 released, see Changelog.txt for the changes. In the following
some major enhancements:
- a new lock-free task scheduling for the ParallelFor has been implemented.
- defined ParallelFor and ParallelForReduce classes
- removed initial underscore in the preprocessor conditional guards in
order to be compliant to the C++ standard
- added the possibility to freeze and thaw all farm threads and then
restart a lower number of workers in a farm having a collector thread.
(see tests/test_stopstartall.cpp)
- fixed possible node termination problem (in case the freezing flag was set)
- ff_send_out_to modified in order to better work with ondemand scheduling
policy
- changed some int types with size_t, moved to svector whenever possible
- more possible combinations of farm, pipeline and feedback using
ff_minode and ff_monode
- better multi-input and multi-output support (ff_minode ff_monode)
------------------------------------------------------------------------
r96 | aldinuc | 2014-02-27 22:59:17 +0100 (Thu, 27 Feb 2014) | 2 lines
Added n of workers in wt-mandelbrot example
------------------------------------------------------------------------
r95 | massimot | 2014-02-20 10:53:01 +0100 (Thu, 20 Feb 2014) | 5 lines
- added the possibility to freeze and thaw all farm threads and then
restart a lower number of workers in a farm having a collector thread.
- other minor changes
------------------------------------------------------------------------
r94 | massimot | 2014-02-10 09:14:06 +0100 (Mon, 10 Feb 2014) | 3 lines
new iterations stealing strategy implemented (experimental)
------------------------------------------------------------------------
r93 | massimot | 2014-02-08 14:42:08 +0100 (Sat, 08 Feb 2014) | 5 lines
- fixed an issue in the ofarm skeleton
- PARFORREDUCE optimized for the case when it is called multiple times
in a sequential loop
------------------------------------------------------------------------
r92 | massimot | 2014-02-06 16:34:04 +0100 (Thu, 06 Feb 2014) | 4 lines
instroduced an optimization in the PARFOR pattern for the case when more threads than cores are used
------------------------------------------------------------------------
r91 | massimot | 2014-02-06 15:07:52 +0100 (Thu, 06 Feb 2014) | 3 lines
fixed icc warning
------------------------------------------------------------------------
r90 | massimot | 2014-02-05 16:34:56 +0100 (Wed, 05 Feb 2014) | 5 lines
- avoided to call scheduletask in the ff_send_out_to
- minor fix for the scheduletask method
------------------------------------------------------------------------
r89 | massimot | 2014-02-03 17:56:55 +0100 (Mon, 03 Feb 2014) | 3 lines
added TBB code in test_parfor_unbalanced.cpp
------------------------------------------------------------------------
r88 | massimot | 2014-02-03 15:36:50 +0100 (Mon, 03 Feb 2014) | 6 lines
- added the runtests.sh scripts
now all tests have default parameters and 'make test' can be used
for a quick check
- fixed an issue for multi input configurations
------------------------------------------------------------------------
r87 | massimot | 2014-02-03 09:36:04 +0100 (Mon, 03 Feb 2014) | 3 lines
unbalanced for test case added
------------------------------------------------------------------------
r86 | massimot | 2014-02-03 09:35:27 +0100 (Mon, 03 Feb 2014) | 5 lines
- fixed possible node termination problem (in case the freezing flag was set)
- commented out the optimization heuristic in the parallel_for pattern
since it does not work well with high unbalanced loop iterations
------------------------------------------------------------------------
r85 | massimot | 2014-01-30 18:25:26 +0100 (Thu, 30 Jan 2014) | 3 lines
fixed a possible issue in the parallel_for/reduce pattern
------------------------------------------------------------------------
r84 | massimot | 2014-01-30 09:39:02 +0100 (Thu, 30 Jan 2014) | 3 lines
fix for adapting the macro-dataflow pattern to the new multi-input interface
------------------------------------------------------------------------
r83 | massimot | 2014-01-29 19:03:25 +0100 (Wed, 29 Jan 2014) | 13 lines
- FF_PARFOR_DONE FF_PARFORREDUCE_DONE macros changed
- parallel_for/reduce: fixed some issues and optimized the case when it is
used with just 1 thread
- ff_send_out_to modified in order to better work with ondemand scheduling
policy
- cleaning gt code (not finished)
- cleaning lb code (not finished)
TODO: more testing!
------------------------------------------------------------------------
r82 | massimot | 2014-01-27 11:05:58 +0100 (Mon, 27 Jan 2014) | 4 lines
first commit of the newer 2.0.3 version
see the Changelog file for the list of changes
------------------------------------------------------------------------
r81 | aldinuc | 2014-01-02 15:15:20 +0100 (Thu, 02 Jan 2014) | 3 lines
Cosmetic changes to make Clang++ happy
------------------------------------------------------------------------
r80 | aldinuc | 2014-01-01 19:52:14 +0100 (Wed, 01 Jan 2014) | 1 line
PARFOR minor fixes for MSVS
------------------------------------------------------------------------
r79 | aldinuc | 2014-01-01 19:51:23 +0100 (Wed, 01 Jan 2014) | 1 line
MAP compliance improved (MSVS)
------------------------------------------------------------------------
r78 | aldinuc | 2013-12-29 02:40:11 +0100 (Sun, 29 Dec 2013) | 2 lines
C++11 compilation warning
------------------------------------------------------------------------
r77 | aldinuc | 2013-12-28 18:42:23 +0100 (Sat, 28 Dec 2013) | 2 lines
Fixes ro compile C++11 code under MSVC10
------------------------------------------------------------------------
r76 | aldinuc | 2013-12-28 17:29:34 +0100 (Sat, 28 Dec 2013) | 2 lines
Windows cmake instruction fixed
------------------------------------------------------------------------
r75 | massimot | 2013-12-18 19:45:59 +0100 (Wed, 18 Dec 2013) | 6 lines
- some cosmetic changes
- ff_buffernode introduced and fixed some issues with ff_monode
- more low-level examples showing how to build "complex" graph of nodes
------------------------------------------------------------------------
r74 | massimot | 2013-12-16 12:40:17 +0100 (Mon, 16 Dec 2013) | 3 lines
tested all d-test using the new ZeroMQ stable release (4.0.3), minor fixes to some tests
------------------------------------------------------------------------
r73 | massimot | 2013-12-12 16:33:57 +0100 (Thu, 12 Dec 2013) | 4 lines
Better estimation of the bounded and unbounded SPSC queue length.
Thanks to M. Moniruzzaman (moniruzzaman@hlrs.de) for the patch.
------------------------------------------------------------------------
r72 | aldinuc | 2013-12-06 18:24:59 +0100 (Fri, 06 Dec 2013) | 2 lines
Detection of c++11 features supported by the compiler for the cmake
------------------------------------------------------------------------
r71 | massimot | 2013-11-30 08:55:38 +0100 (Sat, 30 Nov 2013) | 3 lines
moved broadcast_task from protected to public in the loadbalancer
------------------------------------------------------------------------
r70 | massimot | 2013-11-27 10:57:09 +0100 (Wed, 27 Nov 2013) | 3 lines
moved all_gather from protected to public in the ff_gatherer class
------------------------------------------------------------------------
r69 | massimot | 2013-11-23 11:39:42 +0100 (Sat, 23 Nov 2013) | 4 lines
- fixed a bug in the macro data-flow (mdf) pattern
which occasionally causes deadlock
------------------------------------------------------------------------
r68 | massimot | 2013-11-10 11:55:31 +0100 (Sun, 10 Nov 2013) | 9 lines
- general enhancement (tested only for Linux at moment)
fixed some problems with the current version in the SWPS3 example
(thanks to Maurizio Drocco)
- introduced the ff_mdf, stencil2D and ff_graphsearch patterns (experimental)
- added ff_chol_mdf (macro-dataflow implementation of the Cholesky algorithm)
- fixed a bug in the pipeline when used as an accelerator
- comments cleanup
------------------------------------------------------------------------
r67 | massimot | 2013-10-29 14:46:14 +0100 (Tue, 29 Oct 2013) | 5 lines
- ff_pipe enhancment
- moved ff_send_out to public methods in the ff_node
- ff_farm now accepts also lambdas as worker functions (using std::function)
------------------------------------------------------------------------
r66 | massimot | 2013-10-28 11:36:48 +0100 (Mon, 28 Oct 2013) | 6 lines
enhancement:
- more on ParallelFor pattern
- added the possibility to freeze and thaw one single worker thread
in a task-farm skeleton (see tests/test_stopstartthreads2.cpp)
------------------------------------------------------------------------
r65 | massimot | 2013-09-26 12:02:37 +0200 (Thu, 26 Sep 2013) | 3 lines
Small fix for ARM compilation using gcc. Thanks to David Dight.
------------------------------------------------------------------------
r64 | massimot | 2013-09-11 15:04:20 +0200 (Wed, 11 Sep 2013) | 3 lines
minor fix: nanasleep-based code is currently available only on Linux systems
------------------------------------------------------------------------
r63 | massimot | 2013-09-08 21:30:16 +0200 (Sun, 08 Sep 2013) | 16 lines
First check in of the 2.0.2 release. Main changes:
- introduced FF_ESAVER (Energy SAVER) compiler flag (experimental) (thanks to Mehdi Goli )
- ff_relax now uses nanosleep instead of usleep
- first full porting to ARM processor (thanks to Mauro Mulatero)
- added new interfaces for the ff_farm
- added a new construct ff_pipe only for C++11
- fixed some problems with cmake compilation
- added ff_send_out_to as a method of the ff_loadbalancer class
- added parallel_for.hpp which implements 'OpenMP-like' parallel_for
- added cleanup_workers method in the farm skeleton
- removed fallback function in the farm emitter
- removed many warnings when compiling with the Intel compiler
- fixed get_my_id() problem for pipeline
------------------------------------------------------------------------
r62 | massimot | 2013-04-24 10:47:19 +0200 (Wed, 24 Apr 2013) | 6 lines
- removed lots of warning when compiling with Intel's compiler
- matmult example reworked
- added the test_MISD test for showing how to implement a simple
MISD farm skeleton with fully programmable tasks scheduler
------------------------------------------------------------------------
r61 | massimot | 2013-04-13 11:33:40 +0200 (Sat, 13 Apr 2013) | 3 lines
release major update
------------------------------------------------------------------------
r60 | massimot | 2013-04-13 11:14:22 +0200 (Sat, 13 Apr 2013) | 3 lines
removed old denoiser app
------------------------------------------------------------------------
r59 | massimot | 2013-04-09 15:04:39 +0200 (Tue, 09 Apr 2013) | 3 lines
added experimental directory and moved ofed into experimental/OFED folder
------------------------------------------------------------------------
r58 | massimot | 2013-04-09 11:00:22 +0200 (Tue, 09 Apr 2013) | 3 lines
update to release 2.0.1
------------------------------------------------------------------------
r57 | massimot | 2013-01-02 17:41:23 +0100 (Wed, 02 Jan 2013) | 3 lines
minor fix in the init_thread_affinity function
------------------------------------------------------------------------
r56 | massimot | 2012-12-31 14:19:29 +0100 (Mon, 31 Dec 2012) | 4 lines
- added support for thread affinity on windows platforms
- moved to 2.0.1 version
------------------------------------------------------------------------
r55 | massimot | 2012-12-27 19:21:38 +0100 (Thu, 27 Dec 2012) | 9 lines
- Distributed version ported and tested on Win7 64-bit.
- little fix on the allocator.hpp, freesegment now release aligned memory
to avoid stack corruption on Win platform that requires symmetric
primitives for aligned memory allocation and deallocation.
- Winsock.h (windows.h) and winsock2.h incompatbility issue partially solved.
<ff/dnode.h> should be included before other includes to avoid the problem.
It includes winsock2.h that is required by zeromq.
------------------------------------------------------------------------
r54 | massimot | 2012-12-16 10:37:42 +0100 (Sun, 16 Dec 2012) | 4 lines
fixed memory leak in the CLHSpinLock
added missing fixedsize parameter in the ff_ofarm constructor
------------------------------------------------------------------------
r53 | massimot | 2012-12-09 08:40:19 +0100 (Sun, 09 Dec 2012) | 4 lines
Fixed compilation problems for Windows platform: Windows 7 64 bit using Visual Studio 10
cmake -G "Visual Studio 10 Win64" ..
------------------------------------------------------------------------
r52 | massimot | 2012-12-08 08:57:48 +0100 (Sat, 08 Dec 2012) | 5 lines
added 2 new papers:
- 2012_dff_CoreGridWS presents the distributed features of FastFlow (Europar 2012, CoreGrid Workshop)
- 2012_ff-Tilera_PDP presents the porting of FastFlow on Tilera's Tile64 many-core architecture (PDP 2013)
------------------------------------------------------------------------
r51 | massimot | 2012-12-07 18:51:48 +0100 (Fri, 07 Dec 2012) | 3 lines
fixed cmake problem with spd-denoiser (video version)
------------------------------------------------------------------------
r50 | massimot | 2012-12-06 21:06:16 +0100 (Thu, 06 Dec 2012) | 4 lines
more doxygen,
changed interface for ff_ofarm
------------------------------------------------------------------------
r49 | aldinuc | 2012-12-05 22:57:49 +0100 (Wed, 05 Dec 2012) | 1 line
------------------------------------------------------------------------
r48 | massimot | 2012-12-02 22:22:48 +0100 (Sun, 02 Dec 2012) | 3 lines
fixed bug in ff_ofarm when using multiple freeze and wake-up operations
------------------------------------------------------------------------
r47 | massimot | 2012-11-23 08:42:53 +0100 (Fri, 23 Nov 2012) | 6 lines
- added class ff_dinout that allows to have a single dnode with both
input and output external channels.
- added the class ff_ofarm to implement the ordered farm template
- fixed problem in gt.hpp class (thanks to Daniele De Sensi)
------------------------------------------------------------------------
r46 | massimot | 2012-11-12 08:33:15 +0100 (Mon, 12 Nov 2012) | 4 lines
- added distributed map test (dmap)
- some cleanups to the dnode and svector classes
------------------------------------------------------------------------
r45 | massimot | 2012-11-10 15:57:21 +0100 (Sat, 10 Nov 2012) | 3 lines
fixed compilation problem for test_mpmc
------------------------------------------------------------------------
r44 | massimot | 2012-11-09 19:57:58 +0100 (Fri, 09 Nov 2012) | 6 lines
- newer version of the cwc simulator example
- small bug fixed
- introduced the compilation flag NO_DEFAULT_MAPPING
- more class documentation
------------------------------------------------------------------------
r43 | massimot | 2012-10-31 16:27:26 +0100 (Wed, 31 Oct 2012) | 8 lines
- added the map template
- added the possibility to have multi-input Emitter in a farm template
- added the class threadMapper (default thread pinning implemented)
- implemented spinBarrier
- more debugging
------------------------------------------------------------------------
r42 | aldinuc | 2012-10-12 23:05:45 +0200 (Fri, 12 Oct 2012) | 2 lines
Minor fix
------------------------------------------------------------------------
r41 | massimot | 2012-08-28 06:18:12 +0200 (Tue, 28 Aug 2012) | 5 lines
- more debugging for distributed systems
- added some simple tests
- spd_denoiser updated
------------------------------------------------------------------------
r40 | massimot | 2012-07-18 08:47:43 +0200 (Wed, 18 Jul 2012) | 4 lines
added some comments in the dnode class
bug fix for the test11_pipe test
------------------------------------------------------------------------
r39 | massimot | 2012-07-10 06:53:10 +0200 (Tue, 10 Jul 2012) | 3 lines
typo fixed, committed correct version of the tests
------------------------------------------------------------------------
r38 | massimot | 2012-07-08 19:34:05 +0200 (Sun, 08 Jul 2012) | 8 lines
Moving to the new version 2.0.0. This is the first commit.
Lots of changes and improvements (at least we hope so)
take a look at the Changelog file.
This first version has not been fully tested on OSX and
Windows.
------------------------------------------------------------------------
r37 | massimot | 2011-08-23 19:38:40 +0200 (Tue, 23 Aug 2011) | 3 lines
versione 1.1.1, see Changelog
------------------------------------------------------------------------
r36 | aldinuc | 2011-04-22 17:02:15 +0200 (Fri, 22 Apr 2011) | 3 lines
Readme update
------------------------------------------------------------------------
r35 | massimot | 2011-04-22 15:38:39 +0200 (Fri, 22 Apr 2011) | 2 lines
added missing files
------------------------------------------------------------------------
r34 | massimot | 2011-04-22 15:36:38 +0200 (Fri, 22 Apr 2011) | 3 lines
FastFlow-1.1.0 release, see the Changelog.txt file for full details.
------------------------------------------------------------------------
r33 | massimot | 2011-04-10 20:21:32 +0200 (Sun, 10 Apr 2011) | 4 lines
- MPMC reworked (removed DCAS)
- minor fixes and cleanup
------------------------------------------------------------------------
r32 | aldinuc | 2011-04-08 00:56:29 +0200 (Fri, 08 Apr 2011) | 2 lines
Few bug fixed in the windows version
------------------------------------------------------------------------
r31 | aldinuc | 2011-04-03 15:47:09 +0200 (Sun, 03 Apr 2011) | 2 lines
Bug-fix for cwc example onto windos platform
------------------------------------------------------------------------
r30 | massimot | 2011-04-01 19:37:55 +0200 (Fri, 01 Apr 2011) | 4 lines
upgrade to v.1.0.9 with many improvments and some fixes
first version with fully functional Windows porting
------------------------------------------------------------------------
r29 | massimot | 2011-01-07 12:10:41 +0100 (Fri, 07 Jan 2011) | 3 lines
minor fixes and cleanup
------------------------------------------------------------------------
r28 | massimot | 2010-12-29 23:19:43 +0100 (Wed, 29 Dec 2010) | 5 lines
"Single-Producer/Single-Consumer Queue on Shared Cache Multi-Core Systems"
by Massimo Torquati. Technical Report: TR-10-20, Computer Science Department
University of Pisa Italy
------------------------------------------------------------------------
r27 | massimot | 2010-12-29 23:14:21 +0100 (Wed, 29 Dec 2010) | 4 lines
- lots of fixes and improvements
- added Multi-Producers/Multi-Consumers queue implementation (MSqueue by Michael and Scott)
------------------------------------------------------------------------
r26 | massimot | 2010-11-22 11:04:37 +0100 (Mon, 22 Nov 2010) | 9 lines
- added the first experimental version of thread's mapping (pinning) APIs
- 'make test' improved (cmake)
- multipush method added to the uSWSR_Ptr_Buffer queue (experimental)
- added the method get_channel_id() in the emitter (lb.hpp) and collector
(gt.hpp) class so now it is possible to know from which worker thread
an imput task is coming from.
- other minor fixes
------------------------------------------------------------------------
r25 | massimot | 2010-10-03 18:34:07 +0200 (Sun, 03 Oct 2010) | 2 lines
added the Quicksort example
------------------------------------------------------------------------
r24 | massimot | 2010-09-27 16:22:19 +0200 (Mon, 27 Sep 2010) | 2 lines
added cmake compilation support for the fibonacci example
------------------------------------------------------------------------
r23 | massimot | 2010-09-26 22:47:54 +0200 (Sun, 26 Sep 2010) | 2 lines
added the fibonacci example.
------------------------------------------------------------------------
r22 | massimot | 2010-09-01 19:39:38 +0200 (Wed, 01 Sep 2010) | 2 lines
fixed problem with Qt version for qt-mandelbrot
------------------------------------------------------------------------
r21 | massimot | 2010-09-01 18:58:47 +0200 (Wed, 01 Sep 2010) | 3 lines
minor fixes
added Changelog.txt file
------------------------------------------------------------------------
r20 | massimot | 2010-09-01 12:11:07 +0200 (Wed, 01 Sep 2010) | 4 lines
qt-mandelbrot bug fix
minor changes
more checks for cmake
------------------------------------------------------------------------
r19 | massimot | 2010-08-28 16:01:16 +0200 (Sat, 28 Aug 2010) | 13 lines
First commit of the new 1.0.0 version.
Here are some of the things done:
- some bugs fixed
- unbounded SWSR queue improved (removed all locks)
- all .hpp files moved into the ff directory
- more tests
- cmake support (thanks to Fedor Sakharov)
- improved the accelerator sturcture (added FF_EOS_NOFREEZE tag)
- added the management of second level streams
- added the 'stop' method in the farm and pipeline skeleton
------------------------------------------------------------------------
r18 | massimot | 2010-05-20 17:35:09 +0200 (Thu, 20 May 2010) | 2 lines
added missing include stdint.h
------------------------------------------------------------------------
r17 | massimot | 2010-05-02 16:59:38 +0200 (Sun, 02 May 2010) | 2 lines
cleaned TODO file
------------------------------------------------------------------------
r16 | massimot | 2010-05-02 16:56:46 +0200 (Sun, 02 May 2010) | 6 lines
- fixed ffStats when run_then_freeze is called multiple time.
- inhibited the possibility to add Emitter's fallback function
when master-worker configuration is used (this limitation
will be removed in the near future)
------------------------------------------------------------------------
r15 | massimot | 2010-04-20 08:36:21 +0200 (Tue, 20 Apr 2010) | 2 lines
removed memory leak in the farm destructor method
------------------------------------------------------------------------
r14 | massimot | 2010-04-15 14:26:10 +0200 (Thu, 15 Apr 2010) | 2 lines
ops! typo
------------------------------------------------------------------------
r13 | massimot | 2010-04-14 19:55:58 +0200 (Wed, 14 Apr 2010) | 2 lines
fixed posix_memalign problem with "old" OSX systems
------------------------------------------------------------------------
r12 | massimot | 2010-03-24 08:51:17 +0100 (Wed, 24 Mar 2010) | 2 lines
fixed strict-aliasing warning
------------------------------------------------------------------------
r11 | massimot | 2010-03-22 19:12:19 +0100 (Mon, 22 Mar 2010) | 2 lines
moved to LGPLv3
------------------------------------------------------------------------
r10 | massimot | 2010-03-22 18:14:49 +0100 (Mon, 22 Mar 2010) | 9 lines
RC2 second commit before version release:
- introduced the ubuffer.hpp file containing a new and fast unbounded buffer
implementation based on the SWSR_Ptr_Buffer
- introduced the possibility to choose at compile time between bounded and unbounded buffer
in the implementation of the point-to-point channel (SWSR_Ptr_Buffer vs uSWSR_Ptr_Buffer)
- the allocator has been revisited and improved
- minor cosmetic changes
------------------------------------------------------------------------
r9 | massimot | 2010-03-04 22:00:30 +0100 (Thu, 04 Mar 2010) | 9 lines
RC2 first commit
- added some applications (matmul, cholesky, pbzip2)
- remuved ticks_wait in emitter and collector, added usleep
- fixed some bugs and some things missing (thanks to Sune Mai)
- allocator statistics revisited
- added broadcast_task
- minor code cleanup
------------------------------------------------------------------------
r8 | massimot | 2010-02-03 17:26:35 +0100 (Wed, 03 Feb 2010) | 9 lines
new fastflow version 1.0.0 release candidate 1
* some more tests
* some more applications
* improved interface
* improved memory allocator
* new accelerator mode (beta version)
* minor fixes and cleanup
------------------------------------------------------------------------
r7 | aldinuc | 2010-01-02 17:19:21 +0100 (Sat, 02 Jan 2010) | 2 lines
Farm in accelerator configuration added (ff_farm::offload). One test added (test_acc_1.cpp).
------------------------------------------------------------------------
r6 | massimot | 2009-12-03 15:41:34 +0100 (Thu, 03 Dec 2009) | 2 lines
added nqueens example
------------------------------------------------------------------------
r5 | massimot | 2009-11-16 23:13:09 +0100 (Mon, 16 Nov 2009) | 9 lines
new version (0.9.7):
- new farm interface
- pipeline skeleton
- ff_send_out function introduced
- possibility to build arbitrary graphs of pipeline and farms
- mandelbrot set application
- fixed some bugs
------------------------------------------------------------------------
r4 | aldinuc | 2009-11-10 16:11:13 +0100 (Tue, 10 Nov 2009) | 1 line
------------------------------------------------------------------------
r3 | aldinuc | 2009-11-10 15:59:32 +0100 (Tue, 10 Nov 2009) | 1 line
------------------------------------------------------------------------
r2 | massimot | 2009-10-19 11:44:05 +0200 (Mon, 19 Oct 2009) | 6 lines
* moved to new version (0.6.1)
* added swps3 example app.
* minor code cleanup
* added memory barrier into buffer.hpp
------------------------------------------------------------------------
r1 | massimot | 2009-10-15 17:11:25 +0200 (Thu, 15 Oct 2009) | 1 line
new import
------------------------------------------------------------------------
News log
- Aug 2014 Reference manual and new tests.
- Jul 2014 Bowtie2-ff and image restoration applications.
- Jul 2014 Several new high-level patterns: ParallelFor, ParallelForReduce, Pool, MDF. New syntax for high-level patterns.
- Jun 2014 FastFlow layers major update.
- Mar 2014 Stencil-Reduce pattern for CUDA GPGPUs.
- Jan 2014 FastFlow 2.0.4 has been committed to the Sourceforge SVN.Lot of new features. See Changelog.txt
- Jan 2014 FastFlow 2.0.3 has been committed to the Sourceforge SVN.Lot of new features. See Changelog.txt
- 28 Oct 2013 ParallelFor pattern available in the FastFlow 2.0.2 version.
- 28 Oct 2013 FastFlow 2.0.2 is in the Sourceforge SVN. Soon we are going to release a new tarball.
- 23 Oct 2013 FastFlow presented by M. Torquati and M. Danelutto at CERN Concurrency forum meeting webinars.
- 13 Apr 2013 FastFlow 2.0.1 has been committed to the Sourceforge SVN. Soon we are going to put on Sourceforge a new tarball.
- 10 Jul 2012 FastFlow 2.0 is now on sourceforge SVN. It includes the distributed version and several new applications (including denoiser). The release is almost stable for Linux. Several features are still beta for non-linux platforms.
- 26 June 2012 FastFlow is now working on iOS 5.x (iPhone/Ipad). Looking for people willing to test it with a real world application (e.g. a game). Here a screenshot.
- 18 May 2012 FastFlow is going to support clusters of SMPs. Distributed version is now under test. Still working on GPUs.
- 16 May 2012 The uSPSC component of FastFlow, i.e. the unbound wait-free SPSC queue will be presented at EuroPar 2012 in Rhodes. It is fast.
- 25 Mar 2012 Prof. Marco Danelutto included FastFlow in the teaching material of the Distributed systems: paradigms and models course held within the Master on Computer Science and Networking at University of Pisa, Italy. He also wrote a programming primer. Please send to us any comment, question to improve it.
- 05 Jan 2012 Working on GPUs We are working on the integration of FastFlow self-offloading with GPUs. We have some encouraging preliminary results. Stay tuned.
- 05 Jan 2011 FastFlow is now part of the IMPACT project (Innovative Methods for Particle Colliders at the Terascale, founded by Compagnia di San Paolo, with CERN, through a competition with 11% of acceptance rate). We are working at a parallel, high-performance simulation tool for QCD background modeling (e.g. for Higgs boson detection).
- 30 May 2011 FastFlow has won the HPC Advisory Council best project Award (Spring 2011, assigned twice a year). To be announced at International Supercomputing 2011.
- 24 May 2011 FastFlow is now part of the ParaPhrase EU-STREP/FWP7 project: Parallel Patterns for Adaptive Heterogeneous Multicore Systems (starting Oct. 2011).
- 23 Apr. 2011 FastFlow 1.1 is now available. Many new features (see changelog file within tarball). Now working on almost any recent 32/64bit Linux, 32/64bit MacOS 10.4,10.5,10.6, 32/64bit Windows XP and 7 (with Visual Studio). Cmake now generate Makefile, Xcode and Visual Studio projects.
- 25 Mar. 2011 First Windows native port is now ready (FastFlow 1.0.9, available on sourceforge svn), currently tested on Windows 7 x64 with Visual Studio 10. Looking for beta-testers, write us if you are interested in trying it.
- 17 Mar. 2011 — TR-11-16: Porting Decision Tree Building and Pruning Algorithms to Multicore using FastFlow. To the best of our knowledge, the first implementation of parallel pruning of decision trees in data mining literature.
- 7 Jan. 2011 — We are preparing to reverse many new features in the new version of FastFlow; some of them are already in the Sourceforge SVN. Also, we got interesting new applications: take a look to Edge-preserving image denoiser and its impressive performance.
- 15 Oct. 2010 — We started a FAQ page. Any further question or comment is welcome.
- 12 Oct. 2010 — A new paper and a new presentation are available on the website.
- 1 Oct. 2010 — Due to the many requests we have begun to work to a FastFlow programming tutorial. A complete tutorial is not yet available, however, if you would like try FastFlow, we suggest to 1) Understand the architecture looking to this website, papers and talk slides available on this site. 2) Look at the tutorial page in this web site. 3) Start playing with examples on tests directory; the tests are designed to be progressive in complexity. 3) Write to fastflow[AT]di.unito.it in case of any problem.
- 1 Oct. 2010 — It seems we will have a windows version soon. Stay tuned.
- 7 Sep. 2010 — A porting onto Windows/Visual Studio platform is ongoing (designed, development ongoing). We are looking for developers and beta testers.
- 1 Sep. 2010 — FastFlow 1.0 is now ready. Working on any Linux (i386, x86_64) and any MacOS > 10.2 (PPC, i368, x86_64).
- 01/09/10 - New major release 1.0.0 (download it here).
- 27/06/10 - FastFlow stable release 1.0 is going to be released. It incorporates many new features and performance improvements. With the new version we succeeded to squeeze some speedup for very fine-grained tasks (down to 9 nanoseconds on a dual-quad-core @2.2GHz, well don't expect optimal speedup at this grain, but bear in mind the suggested size for TBB is around 5-10 microseconds
). - 21/05/10 - TR-10-11: Porting Decision Tree Algorithms to Multicore using FastFlow is now available (accepted at ECML PKDD 2010, 16% acceptance rate).
- 29/03/10 - StochKit-FF parallel version is now running (alpha quality, some already planned performance optimization are not yet implemented). It is not yet in the public SVN, but it can be distributed on demand for testing. Just write us if you would like to try it.
- 22/03/10 - FastFlow 1.0.0 RC2 tarball is now available for download on sourceforge.
- 07/03/10 - Fortcoming: StochKit-FF: A FastFlow-enabled (parallel version for multicore) of StochKit the popular stochastic simulation framework developed at UC Santa Barbara. The FastFlow-enabled version will exploit the novel “selective memory” technique.
- 25/02/10 - A new FastFlow-related technique (a.k.a. self-offloading) is available in TR-10-03 [ADK10], where we present a further evolution of FastFlow enabling programmers to offload part of their workload on a dynamically created software accelerator running on unused CPUs. The offloaded function can be easily derived from pre-existing sequential code. We emphasize in particular the effective trade-off between human productivity and execution efficiency of the approach.
- 17/02/10 - Take a look here at the very interesting Tito Ingargiola's blog post on lock-free programming and about the use of FastFlow framework.
- 05/02/10 - Major site updates: architecture and tutorial pages updated. New applications-and-performances page.
- 03/02/10 - FastFlow 1.0.0 RC1 tarball is now available on on sourceforge (download)
- 08/01/10 - Fortcoming: fastflow accelerator, a brand new way for the semi-automatic (yet efficient) parallelisation of existing sequential C/C++ applications. Available soon.
- 30/12/09 - FastFlow site has a new FAQ section. Check it out. Please send us any question via email or sourceforce forum.
- 16/12/09 - We are going to release a new version improving performance of applications in dual-core processors using adaptive implementation templates. No performance reduction for 4/8/16-cores (already available from sourceforge project SVN). New examples under testing: Fast N-queens applications and QT+fastflow Mandelbrot. Working on allocator (preliminary tests: faster than Hoard and TBB).
- 10/11/09 - FastFlow paper to appear at PDP 2010 has been published on this page (draft, comparison against TBB, OpenMP, Cilk).
- 10/11/09 - FastFlow paper appeared at ParCo 2009 has been published on this page (comparison against POSIX thread).
ffnamespace/changelog.1408025328.txt.gz · Last modified: 2014/08/14 16:08 by aldinuc
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 3.0 Unported
