https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-custom-dll-conflicting-with-DirectX-libraries/m-p/905618#M11612 We found the solution: managed DirectX by default sets the FPU precision to float! This was causing the driver routine to fail. Dont know if this is a bug in Lapack or not. SYEVR is described to be a "relatively robust" method. IMO the driver should at least report a meaningfull errror - but it just gives the error from the failed internal routine for creation of eigenvectors, which reflects a wrong argument number (-6). <BR /><BR />Anyway, after preventing DX to alter the precision, everything works as expected again. <BR /><BR /><BR /> Wed, 13 Aug 2008 00:43:51 GMT cpatt 2008-08-13T00:43:51Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-custom-dll-conflicting-with-DirectX-libraries/m-p/905617#M11611 I'm calling mkl_custom.dll (ia32, cdecl, sequential, predefined lapack exports) from within a C# class library via PInvoke. Everything works fine so far. <BR />As soon as my application also loads the managed DirectX dependencies (loaded from within Direct3D libraries) and creates some plots with them, calling MKL functions afterwards give strange results. <BR />I tested and confirmed this reproducable for 'dsyevr' with jobz = 'V'. On return, the info parameter gives -6, which means "lda - parameter was wrong" - but it's not.<BR /><BR />Running on WinXP, VS2005, <BR />MKL 10.0.3.021, Manifest embedded into mkl_custom, reference to msvcr80.dll . <BR />Microsoft.DirectX.dll - 5.4.0 <BR />Microsoft.DirectX.Direct3D.dll - 9.5.132<BR />Microsoft.DirectX.Direct3DX.dll - 5.4.0<BR /><BR />Any ideas? May be the runtime loaded by DX somehow conflicts with the one used by MKL? <BR /><BR />BTW: the problem does not occour, if I FIRST load DirectX and MKL afterwards (first time only). But as soon as I load DX after having loaded MKL already, the MKL function gives wrong results again.<BR /> Mon, 11 Aug 2008 16:10:42 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-custom-dll-conflicting-with-DirectX-libraries/m-p/905617#M11611 cpatt 2008-08-11T16:10:42Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-custom-dll-conflicting-with-DirectX-libraries/m-p/905618#M11612 We found the solution: managed DirectX by default sets the FPU precision to float! This was causing the driver routine to fail. Dont know if this is a bug in Lapack or not. SYEVR is described to be a "relatively robust" method. IMO the driver should at least report a meaningfull errror - but it just gives the error from the failed internal routine for creation of eigenvectors, which reflects a wrong argument number (-6). <BR /><BR />Anyway, after preventing DX to alter the precision, everything works as expected again. <BR /><BR /><BR /> Wed, 13 Aug 2008 00:43:51 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-custom-dll-conflicting-with-DirectX-libraries/m-p/905618#M11612 cpatt 2008-08-13T00:43:51Z