https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102018
--- Comment #5 from Torbjorn SVENSSON <azoff at gcc dot gnu.org> --- (In reply to Christophe Lyon from comment #4) > Rather than a cost issue, it's probably because you use +fp.dp (which > -mcpu=cortex-m7 does too), which enables the double-precision FPU. cortex-m4 > does not support it, so the cmpdf_trap_vfp pattern is not enabled > (TARGET_VFP_DOUBLE is false). Ah, right. Does this mean that we are missing some condition on the existing patterns and/or missing a new pattern?
