Commit Graph

15658 Commits

Author SHA1 Message Date
Vladislav Vinogradov
1be1a28920 move CUDA core tests to core module 2014-12-23 17:42:49 +03:00
Vladislav Vinogradov
53862687d5 rename CudaMem -> HostMem to better reflect its purpose 2014-12-23 17:42:49 +03:00
Vladislav Vinogradov
9210d8e542 move allocMatFromBuf function to farneback.cpp:
* it is the only place, where it is used
* no need to make this function public
2014-12-23 17:42:49 +03:00
Vladislav Vinogradov
1d82aecf45 minor reorganization for CUDA doxygen groups:
move main CUDA group to modules/core/cuda.hpp
2014-12-23 17:42:20 +03:00
Vladislav Vinogradov
b5ab82fdbd mark old CUDA device layer as deprecated and remove it from doxygen documentation
add a note to use new cudev module as a replacement
2014-12-23 17:42:14 +03:00
Vladislav Vinogradov
68e08bbecd fix null stream initialization for multi-gpu systems 2014-12-23 17:41:24 +03:00
Vladislav Vinogradov
05d40946f3 move StackAllocator to cpp file
it is internal class, no need to export it
2014-12-23 17:41:24 +03:00
Vladislav Vinogradov
7ed38b97c3 fix cuda::BufferPool deinitialization
The deinitialization of BufferPool internal objects is controled by global
object, but it depends on other global objects, which leads to errors
caused by undefined deinitialization order of global objects.

I merge global objects initialization into single class, which performs
initialization and deinitialization in correct order.
2014-12-23 17:41:24 +03:00
Vladislav Vinogradov
b75b0c822d install test data on Windows platform too 2014-12-23 17:40:41 +03:00
Alexander Alekhin
864ec5ef45 IPPICV: don't use full paths in dependencies 2014-12-23 17:23:35 +03:00
Vadim Pisarevsky
fd6ef87c32 Merge pull request #3529 from jet47:fix-linux-install 2014-12-23 13:38:23 +00:00
Vadim Pisarevsky
95ec92994d Merge pull request #3536 from mshabunin:doxygen-intro 2014-12-23 13:36:40 +00:00
Jiri Drbalek
d71e001736 fix crash when sample point out of image boundaries 2014-12-22 21:01:17 +00:00
Maksim Shabunin
06c2a70c49 Fixed some mistakes 2014-12-22 17:21:37 +03:00
Maksim Shabunin
637b615e08 Tutorial: documenting OpenCV 2014-12-22 15:51:37 +03:00
Vadim Pisarevsky
d9f159a554 Merge pull request #3513 from mshabunin:compat-30 2014-12-22 11:58:01 +00:00
Vadim Pisarevsky
c0005fd293 Merge pull request #3520 from JoeHowse:master 2014-12-22 11:14:29 +00:00
Vadim Pisarevsky
f12bd999bf Merge pull request #3524 from jet47:fix-cuda-warnings 2014-12-22 10:58:07 +00:00
Vadim Pisarevsky
a1df295079 Merge pull request #3525 from jet47:fix-cudev-tests 2014-12-22 10:57:07 +00:00
Vadim Pisarevsky
7b20ce4952 Merge pull request #3490 from oresths:symmcolumnsmall_fix 2014-12-22 10:44:47 +00:00
Vadim Pisarevsky
432546e4c4 Merge pull request #3512 from vins31:OpenNi2_AsusXtion 2014-12-22 10:39:42 +00:00
Vadim Pisarevsky
700a388173 Merge pull request #3499 from StevenPuttemans:fix_2432 2014-12-22 10:29:31 +00:00
Vadim Pisarevsky
5fea331d42 Merge pull request #3510 from boaz001:feature-4057 2014-12-22 10:27:33 +00:00
Vadim Pisarevsky
1ab551487d Merge pull request #3516 from ana-GT:openni2_defaultMode 2014-12-22 10:26:30 +00:00
Vadim Pisarevsky
060d67517a Merge pull request #3518 from wangyan42164:ocl_cascade_detect 2014-12-22 10:25:47 +00:00
Vadim Pisarevsky
199f1aec2e Merge pull request #3519 from fvgoto:patch-1 2014-12-22 10:25:00 +00:00
Vadim Pisarevsky
35d730bf2b Merge pull request #3528 from ilya-lavrenov:update_android_cmake 2014-12-22 10:22:20 +00:00
Vladislav Vinogradov
ec33c4ae36 increase epsilons for tests due to different optimizations (IPP vs CUDA, float vs double) 2014-12-22 11:48:45 +03:00
Vladislav Vinogradov
25f33a7e30 update cudev color conversions according to the latest changes in CPU code 2014-12-22 11:48:45 +03:00
Vladislav Vinogradov
48c9c24da6 disable -Wshadow warning for CUDA modules:
it is generated by CUDA headers and we can't fix it
2014-12-22 11:48:19 +03:00
orestis
fffe2464cd Change DescriptorExtractor_ORB regression test
to compensate for neon ieee754 non-compliancy.
Also changed the comparison between max valid and calculated distance to
make the error message more accurate (in case curMaxDist == maxDist)
2014-12-21 21:27:03 +02:00
orestis
9811a739b0 Change gaussianBlur5x5 perf test epsilon
Set it 1 instead of 0.001, as is already done in gaussianBlur3x3. That
will allow integer destination matrices that are not exactly the same,
but very close to the expected result, to pass the test.
2014-12-20 17:14:21 +02:00
orestis
9c6da03504 SymmRowSmallVec_32f 1x5 asymm
NEON speedup: 2.31x
Auto-vect speedup: 2.26x

Test kernel: [-0.9432, -1.1528, 0, 1.1528, 0.9432]
2014-12-19 22:51:42 +02:00
orestis
13c0855114 SymmRowSmallVec_32f 1x5
NEON speedup: 2.36x
Auto-vect speedup: 2.36x

Test kernel: [0.1, 0.2408, 0.3184, 0.2408, 0.1]
2014-12-19 22:47:06 +02:00
orestis
ed0ce48179 SymmColumnVec_32f16s asymm
NEON speedup: 9.46x
Auto-vect speedup: 1x

Test kernel: [-0.9432, -1.1528, 0, 1.1528, 0.9432]
2014-12-19 22:44:39 +02:00
orestis
a2a131799f SymmColumnVec_32f16s
NEON speedup: 8.64x
Auto-vect speedup: 1x

Test kernel: [0.1, 0.2408, 0.3184, 0.2408, 0.1]
2014-12-19 22:42:31 +02:00
orestis
37e018454d SymmColumnSmallVec_32s16s 3x1 asymm
NEON speedup: 2.12x
Auto-vect speedup: 1.01x

Test kernel: [-2, 0, 2]
2014-12-19 22:40:55 +02:00
orestis
4443d6b0a1 SymmColumnSmallVec_32s16s [-1, 0, 1]
NEON speedup: 3.27x
Auto-vect speedup: 1.01x
2014-12-19 22:37:52 +02:00
orestis
99e782e62c SymmColumnSmallVec_32s16s 3x1
NEON speedup: 1.75x
Auto-vect speedup: 1x
2014-12-19 22:36:46 +02:00
orestis
33dfeb85be SymmColumnSmallVec_32s16s [3, 10, 3] Scharr
NEON speedup: 2.04x
Auto-vect speedup: 1x
2014-12-19 22:35:52 +02:00
orestis
61a7f48bf4 SymmColumnSmallVec_32s16s [1, -2, 1]
NEON speedup: 2.75x
Auto-vect speedup: 1.01x
2014-12-19 22:34:11 +02:00
orestis
4f906372e2 SymmColumnSmallVec_32s16s [1, 2, 1]
NEON speedup: 2.66x
Auto-vect speedup: 1x
2014-12-19 22:33:11 +02:00
orestis
80a0364465 SymmColumnVec_32s8u asymm
NEON speedup: 2.95x
Auto-vect speedup: 1x

Test kernel: [-0.9432, -1.1528, 0, 1.1528, 0.9432]
2014-12-19 22:29:54 +02:00
orestis
4f5916f12d SymmColumnVec_32s8u
NEON speedup: 1.96x
Auto-vect speedup: 1x

Test kernel: [0.0708, 0.2445, 0.3694, 0.2445, 0.0708]
2014-12-19 22:26:41 +02:00
orestis
1fb966dc61 SymmRowSmallVec_8u32s 1x5 asymm
NEON speedup: 3.14x
Auto-vect speedup: 1.6x

Test kernel: [-5, -2, 0, 2, 5]
2014-12-19 22:23:09 +02:00
orestis
2e7b9a2c0f SymmRowSmallVec_8u32s 1x3 asymmetric
NEON speedup: 1.95x
Auto-vect speedup: 1.17x

Test kernel: [-2, 0, 2]
2014-12-19 22:15:37 +02:00
orestis
969a218057 SymmRowSmallVec_8u32s [-1, 0, 1]
NEON speedup: 1.84x
Auto-vect speedup: 1.2x
2014-12-19 22:11:52 +02:00
orestis
c0019a42e4 SymmRowSmallVec_8u32s 1x5 general
NEON speedup: 3.86x
Auto-vect speedup: 1.67x

Test kernel: [0.0708, 0.2445, 0.3694, 0.2445, 0.0708]
2014-12-19 22:10:58 +02:00
orestis
cb48d7798d SymmRowSmallVec_8u32s 1x3 general
NEON speedup: 2.56x
Auto-vect speedup: 1.26x

Test kernel: [1, 3, 1]
2014-12-19 22:08:23 +02:00
orestis
0ba3b06efd SymmRowSmallVec_8u32s [1, 2, 1]
NEON speedup: 4.1
Auto-vect speedup: 2.25
2014-12-19 22:02:57 +02:00