https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115828
--- Comment #2 from Georg-Johann Lay <gjl at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #1)
> Turning off ivopts `-fno-ivopts`. avr gives:
> .L2:
> sts v+1,r19
> sts v,r18
> subi r18,2
> sbc r19,__zero_reg__
> sbiw r24, 1
> brne .L2
>
> Which seems like reasonable code.
Not really, it is using TWO counters: One that starts at 19 and decrements by
one, PLUS one that starts at 36 and decrements by 2.
Though with -fno-tree-loop-optimize -Os it is only using one counter. (Insn 32
is sub-optimal due to target issue PR115830, but apart from that you cannot do
better):
loop:
ldi r24,lo8(36) ; 26 [c=4 l=2] *movhi/4
ldi r25,0
.L2:
sts v+1,r25 ; 22 [c=4 l=4] *movhi/3
sts v,r24
sbiw r24,2 ; 23 [c=4 l=1] *addhi3_clobber/0
sbrs r25,7 ; 32 [c=4 l=2] *sbrx_branchhi
rjmp .L2
/* epilogue start */
ret ; 29 [c=0 l=1] return
