The xvcvdpsp instruction currently only writes the result to the lower 32-bits of the destination field. It should write the same result to the upper and lower 32-bits of the result.
Created attachment 118975 [details] fix for the xvcvdpsp instruction The attached patch was committed to fix the issue. commit 91bd56be4153cb5cf6d1fa80ade4b2432ff81ab0 Author: Carl Love <carll@us.ibm.com> Date: Fri Mar 22 12:39:32 2019 -0500
Created attachment 118976 [details] version 2, a line got left out of the previous patch Updated patch, caught the missing line before pushing patch to mainline. commit 30a24515f09f9118c469ae11c9d574da729c97ed Author: Carl Love <carll@us.ibm.com> Date: Fri Mar 22 12:42:27 2019 -0500
No regressions found by automated regression testing. Closing.