https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-DftiComputeBackwards-with-DFTI-NUMBER-OF-TRANSFORMS-gt-1-can/m-p/1476138#M34456 <P>Hi,</P><P><BR /></P><P>Thanks for posting in Intel Communities.</P><P><BR /></P><P>Based on this, and our discussion about creating a priority ticket if needed, I am going ahead and closing this case for now.</P><P>Please feel free to contact us further if you need any assistance from our side.&nbsp;</P><P><BR /></P><P>Thanks &amp; Regards,</P><P>Varsha</P><BR /> Thu, 13 Apr 2023 04:52:22 GMT VarshaS_Intel 2023-04-13T04:52:22Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-DftiComputeBackwards-with-DFTI-NUMBER-OF-TRANSFORMS-gt-1-can/m-p/1474408#M34432 <P>I have a testcase that demonstrates DftiComputeBackwards producing slightly different results when calling in a loop vs setting DFTI_NUMBER_OF_TRANSFORMS to do multiple transforms in 1 pass. This is causing unexpected results for our use of MKL. For some reason the problem disappears when AVX512 mode is enabled. See reproducer testcase for details.</P> Sat, 08 Apr 2023 00:06:34 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-DftiComputeBackwards-with-DFTI-NUMBER-OF-TRANSFORMS-gt-1-can/m-p/1474408#M34432 swalker 2023-04-08T00:06:34Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-DftiComputeBackwards-with-DFTI-NUMBER-OF-TRANSFORMS-gt-1-can/m-p/1476138#M34456 <P>Hi,</P><P><BR /></P><P>Thanks for posting in Intel Communities.</P><P><BR /></P><P>Based on this, and our discussion about creating a priority ticket if needed, I am going ahead and closing this case for now.</P><P>Please feel free to contact us further if you need any assistance from our side.&nbsp;</P><P><BR /></P><P>Thanks &amp; Regards,</P><P>Varsha</P><BR /> Thu, 13 Apr 2023 04:52:22 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-DftiComputeBackwards-with-DFTI-NUMBER-OF-TRANSFORMS-gt-1-can/m-p/1476138#M34456 VarshaS_Intel 2023-04-13T04:52:22Z