Ivan Stojiljkovic
be660a2e16
Performance profiler
...
The performance profiler supersedes perf tracer (controlled by SGDParams::m_perfTraceLevel).
Few events were added to cover the most important parts of mini-batch iteration. New events can be easily added for more detailed profiling.
2016-12-29 19:35:52 +01:00
Zhou Wang
4126a4b2ef
minor changes
2016-12-10 15:33:45 +01:00
Zhou Wang
9628b7087c
add conditional build for swig csharp
2016-12-08 14:13:22 +01:00
Qiwei Ye
af52a3cf57
adding warring for Windows build
2016-11-12 13:20:31 +08:00
KeDengMS
143433ae36
More fixes to warnings
2016-11-05 19:12:38 -07:00
KeDengMS
77e794a09d
Fix typo
2016-11-05 18:43:24 -07:00
KeDengMS
f1cf493f28
Apply NoOpt to all
2016-11-05 18:40:31 -07:00
KeDengMS
00a097dc65
Increase linker speed for no_opt build in CTNKv2LibraryDll
2016-11-05 18:37:43 -07:00
Alexey Reznichenko
e986d27c13
Add protobuf-based serialization
2016-10-25 01:00:38 +02:00
KeDengMS
d7e02247a8
Add Release_NoOpt flavor for python debugging
2016-10-14 14:21:23 -07:00
Eldar Akchurin
22fc37dab0
Bumping mkl version in configs
2016-09-26 11:55:39 +02:00
Nikos Karampatziakis
002e920bc5
Update CNTK.Cpp.props to CUDA8;
...
Add atomicAdd fix by DanielMerget
2016-08-25 15:57:17 -07:00
Zhou Wang
86ec025f32
Use mt19937 instead of rand()
2016-08-23 19:24:45 +02:00
Wolfgang Manousek
40d247841a
remove reference to acml math library
2016-08-17 10:04:56 +02:00
Eldar Akchurin
3eb2c5ee7a
Addressing code review comments
2016-06-30 10:46:41 +02:00
Eldar Akchurin
447edb9765
Adding support for OpenCV 3.1
2016-06-30 10:46:41 +02:00
Mark Hillebrand
595d4828b3
Math.vcxproj: tune dependency checking
2016-06-16 10:44:51 +02:00
Mark Hillebrand
35cb5738e7
Fix copy
2016-06-14 18:29:56 +02:00
Mark Hillebrand
52ae07279d
CNTK custom MKL support
2016-06-14 17:39:24 +02:00
Mark Hillebrand
71d0c194ff
.vcxproj: start sharing settings, starting with static libraries
2016-04-26 17:02:01 +02:00
Mark Hillebrand
93e331a5db
Windows: build support for different CUDA versions (7.0 and 7.5)
...
Set %CNTK_CUDA_VERSION% to 7.0 or 7.5 to select a specific CUDA version.
Otherwise, highest available CUDA version will be chosen.
2016-04-05 14:55:42 +02:00
Mark Hillebrand
1b29d2fd69
.vcxproj: let intermediate output go into a common directory
...
Fixes #12 .
2016-02-10 09:10:07 +01:00
Mark Hillebrand
2c2a721fa8
CNTK.Cpp.props: define properties to distinguish build flavors and use them
2016-02-10 09:05:31 +01:00
Mark Hillebrand
ea55623979
Introduce CNTK.Cpp.props and load in every .vcxproj
2016-02-10 09:05:31 +01:00