I compile libreoffice with 'make' from the top-level and: Dynamic formual compiler: unhandled token: unhandled opcode at :0 Dynamic formual compiler: unhandled token: unhandled opcode at :0 Dynamic formual compiler: unhandled token: unhandled opcode at :0 Dynamic formual compiler: unhandled token: unhandled opcode at :0 Dynamic formual compiler: unhandled token: unhandled opcode at :0 opencl-test.cxx:3420:Assertion Test name: ScOpenclTest::testFinancialXirrFormula double equality assertion failed - Expected: 1.04286406930622e-16 - Actual : 1.40411934087164e-17 - Delta : 1.04286406930622e-20 Failures !!! Run: 222 Failure total: 1 Failures: 1 Errors: 0 Error: a unit test failed, please do one of: export DEBUGCPPUNIT=TRUE # for exception catching export CPPUNITTRACE="gdb --args" # for interactive debugging on Linux export CPPUNITTRACE="\"[full path to devenv.exe]\" /debugexe" # for interactive debugging in Visual Studio export VALGRIND=memcheck # for memory checking and retry using: make CppunitTest_sc_opencl_test make[1]: *** [/data/opt/libreoffice/master/workdir/CppunitTest/sc_opencl_test.test] Error 1 make: *** [build] Error 2 I'm running openSUSE 13.1 32bit (ie. x86 mode) on x86_64 hardware. I attach clinfo & hwinfo. HTH.
Created attachment 95743 [details] clinfo.txt
Created attachment 95744 [details] hwinfo.txt
AnalysisAddIn::getXirr() in scaddins/source/analysis/financial.cxx implements this. This is a numerically very unstable algorithm similar to our goal seek function.
Kohei Yoshida committed a patch related to this issue. It has been pushed to "master": http://cgit.freedesktop.org/libreoffice/core/commit/?id=ca25cdcb398f200ce5d53286eea02d49d8d1c5f4 fdo#76140: Reimplement XIRR OpenCL test. The patch should be included in the daily builds available at http://dev-builds.libreoffice.org/daily/ in the next 24-48 hours. More information about daily builds can be found at: http://wiki.documentfoundation.org/Testing_Daily_Builds Affected users are encouraged to test the fix and report feedback.
I tried to understand why Calc's results are often different than Excel's then quickly gave up. The algorithm used in XIRR is much like goal seek's and is pretty much a black box. Instead, I just focused on the OpenCL calculation results against Calc's internal interpreter, which is what the original test did anyway. Hopefully my above commit will improve the situation.
Michael said this fixes for him. I'll mark it fixed.
Kohei Yoshida committed a patch related to this issue. It has been pushed to "libreoffice-4-2": http://cgit.freedesktop.org/libreoffice/core/commit/?id=a6a51ef9dd991e11c90317f4d6c29dfb310e1a17&h=libreoffice-4-2 fdo#76140: Reimplement XIRR OpenCL test. It will be available in LibreOffice 4.2.6. The patch should be included in the daily builds available at http://dev-builds.libreoffice.org/daily/ in the next 24-48 hours. More information about daily builds can be found at: http://wiki.documentfoundation.org/Testing_Daily_Builds Affected users are encouraged to test the fix and report feedback.