https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90424
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> --- One complication is that V2SFmode isn't valid on the target so at least lowpart sets of V4SFmode are not easily possible (stupid MMX?), for V8QImode we get a corresponding integer mode which works in the end. The thing is that on GIMPLE we are already restricting the destination vector type to supported ones when triggering SSA rewrite via BIT_INSERT_EXPR. This also makes writing target independent testcases for this hard :/
