https://gcc.gnu.org/bugzilla/show_bug.cgi?id=49678
ktkachov at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ktkachov at gcc dot gnu.org
--- Comment #3 from ktkachov at gcc dot gnu.org ---
The problem here is indeed sched1.
For the targets where it's failing we generate:
mul r3, r0, r3
strd r4, [sp, #-8]!
umull r4, r5, r0, r2
mla r2, r2, r1, r3
mov r0, r4
add r5, r2, r5
mov r1, r5
ldrd r4, [sp]
add sp, sp, #8
bx lr
but when we compile with -fno-schedule-insns we get the optimal:
mul r3, r0, r3
mla r3, r2, r1, r3
umull r0, r1, r0, r2
add r1, r3, r1
bx lr
