https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117312
--- Comment #9 from H. Peter Anvin <hpa at zytor dot com> --- So this sounds like it would solve additional problems, which may very well make it worthwhile. I just want to reiterate that for the inline assembly case specifically, just doing the "heavy hammer" thing of disabling the redzone (which looks to me to be what the flags intrinsics currently do) is almost certainly just fine; I don't think this is common enough to justify a ton of extra work, nor that the performance penalty of losing the redzone is too hideous.
