[Bug target/86011] Inefficient code generated for ldivmod with constant value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86011
Andrew Pinski changed:
What|Removed |Added
Status|NEW |RESOLVED
Target Milestone|--- |11.0
Resolution|--- |FIXED
--- Comment #4 from Andrew Pinski ---
Fixed in GCC 11:
lsrsr1, r2, #30
push{r4, lr}
orr r1, r1, r3, lsl #2
bic lr, r1, #-1073741824
bic r1, r2, #-1073741824
add r1, r1, lr
asrsr4, r3, #31
add r3, r1, r3, lsr #28
and r1, r4, #137
add r3, r3, r1
movwr1, #57025
movtr1, 13617
bic r4, r4, #75
umull lr, r1, r1, r3
mov lr, #77
lsrsr1, r1, #4
mls r3, lr, r1, r3
movwr1, #14469
movtr1, 16171
add r3, r3, r4
str r3, [r0, #4]
subsr2, r2, r3
umull r2, r1, r2, r1
str r2, [r0]
pop {r4, pc}
For:
#define N 7723
struct foo { long a, b; };
struct foo test(long long x)
{
return (struct foo){x / N, x % N};
}
GCC 11+ Produces:
push{r4, lr}
mov r4, r0
mov r1, r3
mov r0, r2
movsr3, #0
movwr2, #7723
bl __aeabi_ldivmod
strdr0, r2, [r4]
mov r0, r4
pop {r4, pc}
Which is exactly what you want too.
[Bug target/86011] Inefficient code generated for ldivmod with constant value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86011 Patrick Oppenlander changed: What|Removed |Added Version|8.1.0 |9.2.0 --- Comment #3 from Patrick Oppenlander --- Still present in 9.2.0.
[Bug target/86011] Inefficient code generated for ldivmod with constant value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86011 Steven Bosscher changed: What|Removed |Added Status|WAITING |NEW
[Bug target/86011] Inefficient code generated for ldivmod with constant value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86011
--- Comment #2 from Patrick Oppenlander ---
Sure,
# cat test.c
struct foo { long a, b; };
struct foo test(long long x)
{
return (struct foo){x / 77, x % 77};
}
# gcc --version
gcc (GCC) 8.1.0
Copyright (C) 2018 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
# gcc -c -O2 test.c
# objdump -d test.o
test.o: file format elf32-littlearm
Disassembly of section .text:
:
0: e92d40d0push{r4, r6, r7, lr}
4: e1a06002mov r6, r2
8: e1a07003mov r7, r3
c: e1a04000mov r4, r0
10: e3a0204dmov r2, #77 ; 0x4d
14: e3a03000mov r3, #0
18: e1a6mov r0, r6
1c: e1a01007mov r1, r7
20: ebfebl 0 <__aeabi_ldivmod>
24: e1a01007mov r1, r7
28: e3a0204dmov r2, #77 ; 0x4d
2c: e3a03000mov r3, #0
30: e584str r0, [r4]
34: e1a6mov r0, r6
38: ebfebl 0 <__aeabi_ldivmod>
3c: e1a4mov r0, r4
40: e5842004str r2, [r4, #4]
44: e8bd80d0pop {r4, r6, r7, pc}
Looks like the same problem is still there.
[Bug target/86011] Inefficient code generated for ldivmod with constant value
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86011 Richard Biener changed: What|Removed |Added Keywords||missed-optimization Target||arm Status|UNCONFIRMED |WAITING Last reconfirmed||2018-06-01 Component|other |target Ever confirmed|0 |1 --- Comment #1 from Richard Biener --- I think this was fixed with GCC 8. Can you check?
