On linux ppc64 / gcc , we build with the compile flag -minsert-sched-nops=regroup_exact ; this is most likely outdated, maybe it was still useful for old Power6 machines.

See https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/RS_002f6000-and-PowerPC-Options.html
-minsert-sched-nops=scheme : This option controls which NOP insertion scheme is used during the second scheduling pass. The argument scheme takes one of the following values:
‘regroup_exact’ : Insert NOPs to force costly dependent insns into separate groups. Insert exactly as many NOPs as needed to force an insn to a new group, according to the estimated processor grouping.