https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101276
--- Comment #3 from Hongtao.liu <crazylht at gmail dot com> --- (In reply to Richard Biener from comment #2) > It might slow down things when data is encrypted in a loop and the error > state > is accumulated and checked afterwards? This is the exact case that the user mentioned, by the time the state is checked, idata has been already written to odata. They hope when intructions failed to encrpt idata, 0 would used for odata.
