https://gcc.gnu.org/bugzilla/show_bug.cgi?id=74563
--- Comment #7 from Jeffrey A. Law <law at redhat dot com> --- I was essentially thinking the same WRT a proposed patch. My concern is whether something might be passing in an unexpected rtx for operands[0] and whether or not we can get into the other pattern for MIPS16. So I've got a series of builds running with instrumentation to see if any of the cases I'm worried about triggers during a build of the usual runtime libraries.
