https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83084
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> --- The sched1 dump differences are: // rs6000_sched_reorder : // rs6000_variable_issue (more = 7) = 0 -;; 0--> b 1: i10031 [scratch]=unspec[[scratch]] 36 :DU_both_power8,LSU_power8:GENERAL_REGS+0(0)VSX_REGS+0(0)CR_REGS+0(0)SPECIAL_REGS+0(0):model 0 +;; 0--> b 1: i10032 r138=unspec/v[[r131]] 1 :DU_both_power8,LU_or_LSU_power8:GENERAL_REGS+1(1)VSX_REGS+0(0)CR_REGS+0(0)SPECIAL_REGS+0(0) +rescanning insn with uid = 8. // rs6000_sched_reorder2 : // rs6000_sched_reorder : // rs6000_variable_issue (more = 7) = 0 -;; 1--> b 1: i10032 r138=unspec/v[[r131]] 1 :DU_both_power8,LU_or_LSU_power8:GENERAL_REGS+1(1)VSX_REGS+0(0)CR_REGS+0(0)SPECIAL_REGS+0(0):model 1 +;; 1--> b 1: i10031 [scratch]=unspec[[scratch]] 36 :DU_both_power8,LSU_power8:@GENERAL_REGS+0(0)@VSX_REGS+0(0)@CR_REGS+0(0)@SPECIAL_REGS+0(0):model 0 +;; 1--> b 1: i 8 loc clobber :nothing:GENERAL_REGS+0(0)VSX_REGS+0(0)CR_REGS+0(0)SPECIAL_REGS+0(0) // rs6000_sched_reorder2 :
