omp reduce is broken on Win7 64b msvc 2008

[GoogleCodeExporter]

c:\dev\hg\thrust-tot\testing>scons MSVC_VERSION=9.0 backend=omp 
tests=TestReduce -j4
scons: Reading SConscript files ...
scons: done reading SConscript files.

scons: warning: you do not seem to have the pywin32 extensions installed;
        parallel (-j) builds may not work reliably with open Python files.
File "C:\Python27\Scripts\scons.py", line 190, in <module>
scons: Building targets ...
nvcc.exe -o debug.obj -c -arch=sm_10 --x=c++ -Xcompiler 
-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_OMP -Xcompiler 
-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -Xcompiler /Ox -Xcompiler /openmp 
-Xcompiler /DNOMINMAX -Xcompiler /wd4503 -Xcompiler /bigobj -I 
c:\dev\hg\thrust-tot -I "C:\Program Files\NV
IDIA GPU Computing Toolkit\CUDA\v4.1\include" -I 
C:/dev/tbb40_20120201oss/include -I C:\dev\hg\thrust-tot\testing debug.cu
nvcc.exe -o reduce.obj -c -arch=sm_10 --x=c++ -Xcompiler 
-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_OMP -Xcompiler 
-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -Xcompiler /Ox -Xcompiler /openmp 
-Xcompiler /DNOMINMAX -Xcompiler /wd4503 -Xcompiler /bigobj -I 
c:\dev\hg\thrust-tot -I "C:\Program Files\N
VIDIA GPU Computing Toolkit\CUDA\v4.1\include" -I 
C:/dev/tbb40_20120201oss/include -I C:\dev\hg\thrust-tot\testing reduce.cu
debug.cu
nvcc.exe -o reduce_by_key.obj -c -arch=sm_10 --x=c++ -Xcompiler 
-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_OMP -Xcompiler 
-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -Xcompiler /Ox -Xcompiler /openmp 
-Xcompiler /DNOMINMAX -Xcompiler /wd4503 -Xcompiler /bigobj -I 
c:\dev\hg\thrust-tot -I "C:\Program
Files\NVIDIA GPU Computing Toolkit\CUDA\v4.1\include" -I 
C:/dev/tbb40_20120201oss/include -I C:\dev\hg\thrust-tot\testing 
reduce_by_key.cu
nvcc.exe -o reduce_large.obj -c -arch=sm_10 --x=c++ -Xcompiler 
-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_OMP -Xcompiler 
-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -Xcompiler /Ox -Xcompiler /openmp 
-Xcompiler /DNOMINMAX -Xcompiler /wd4503 -Xcompiler /bigobj -I 
c:\dev\hg\thrust-tot -I "C:\Program F
iles\NVIDIA GPU Computing Toolkit\CUDA\v4.1\include" -I 
C:/dev/tbb40_20120201oss/include -I C:\dev\hg\thrust-tot\testing reduce_large.cu
reduce.cu
reduce_by_key.cu
reduce_large.cu
nvcc.exe -o testframework.obj -c -arch=sm_10 --x=c++ -Xcompiler 
-DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_OMP -Xcompiler 
-DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -Xcompiler /Ox -Xcompiler /openmp 
-Xcompiler /DNOMINMAX -Xcompiler /wd4503 -Xcompiler /bigobj -I 
c:\dev\hg\thrust-tot -I "C:\Program
Files\NVIDIA GPU Computing Toolkit\CUDA\v4.1\include" -I 
C:/dev/tbb40_20120201oss/include -I C:\dev\hg\thrust-tot\testing 
testframework.cu
testframework.cu
link /nologo /OUT:tester.exe "/LIBPATH:C:\Program Files\NVIDIA GPU Computing 
Toolkit\CUDA\v4.1\lib\x64" /LIBPATH:C:\dev\tbb40_20120201oss\lib\intel64\vc9 
cudart.lib VCOMP.lib debug.obj reduce.obj reduce_by_key.obj reduce_large.obj 
testframework.obj
scons: done building targets.

c:\dev\hg\thrust-tot\testing>tester
Running 13 unit tests.
F......F...FF
================================================================
FAILURE: TestReduce
[reduce.cu:51] values are not equal: 1485082345 -762754360 [type='int']
================================================================
FAILURE: TestReduceSimpleDevice
[reduce.cu:30] values are not equal: -17442 2 [type='short']
================================================================
FAILURE: TestReduceWithLargeTypes
[reduce_large.cu:20] values are not equal. [type='struct FixedVector<int,1>']
================================================================
FAILURE: TestReduceWithOperator
[reduce.cu:104] values are not equal: 7 9 [type='unsigned int']
================================================================
Totals: 4 failures, 0 known failures, 0 errors, and 9 passes.
Time:  0 minutes

c:\dev\hg\thrust-tot\testing>

Original issue reported on code.google.com by jaredhoberock on 4 Mar 2012 at 2:47

[GoogleCodeExporter]

I believe the second decomposition [1] is computed incorrectly, as the second 
reduce_intervals is summing off the end of the array.

ARB Nathan

[1] 
http://code.google.com/p/thrust/source/browse/thrust/system/omp/detail/reduce.in
l#48

Original comment by jaredhoberock on 4 Mar 2012 at 3:29

[GoogleCodeExporter]

This issue was closed by revision 0eafb5be657a.

Original comment by jaredhoberock on 5 Mar 2012 at 12:14

• Changed state: Fixed