[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190
--- Comment #12 from Oleg Endo ---
(In reply to Oleg Endo from comment #0)
> The following code:
>
> struct X
> {
> int a, b, c, d, e;
> };
>
> int test (X* x, unsigned int c)
> {
> int s = 0;
> unsigned int i;
> for (i = 0; i < c; ++i)
> s += x[i].b;
> return s;
> }
>
> results in:
> tst r5,r5
> bt/s.L4
> mov r5,r1
> shll2 r1
> add r5,r1
> mov.l .L9,r2
> shll2 r1
> add #-20,r1
> shlr2 r1
> mul.l r2,r1
> mov.l .L10,r2
> add #4,r4
> mov #0,r0
> sts macl,r1
> and r2,r1
> add #1,r1
> .L3:
> mov.l @r4,r2
> dt r1
> add #20,r4
> bf/s.L3
> add r2,r0
> rts
> nop
> .L4:
> rts
> mov #0,r0
> .L11:
> .align 2
> .L9:
> .long 214748365
> .L10:
> .long 1073741823
>
As of GCC 13, this still produces the same code with loop header bloat.
[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190 Richard Biener changed: What|Removed |Added Status|NEW |ASSIGNED
[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190
Oleg Endo changed:
What|Removed |Added
See Also||https://gcc.gnu.org/bugzill
||a/show_bug.cgi?id=62233,
||https://gcc.gnu.org/bugzill
||a/show_bug.cgi?id=60537
--- Comment #11 from Oleg Endo ---
And still on SH, but now with AMS optimization:
char*
test_func_00 (char* p, int c, int x)
{
do
{
*--p = (char)x;
*--p = (char)x;
*--p = (char)x;
} while (--c);
return p;
}
Results in:
mov r5,r7
mov.l .L6,r1
add r7,r7
mov r7,r2
add r5,r2
mul.l r1,r2
mov r4,r3
exts.b r6,r6
add #-3,r3
sts macl,r0
.align 2
.L2:
mov r3,r2
add #3,r2
mov.b r6,@-r2
dt r0
mov.b r6,@-r2
mov.b r6,@r3
mov r2,r3
bf/s.L2
add #-4,r3
add r5,r7
mov r4,r0
rts
sub r7,r0
.L7:
.align 2
.L6:
.long -1431655765
And with -fno-ivopts we're getting something what one would expect:
exts.b r6,r6
mov r4,r1
mov r5,r2
.align 2
.L2:
mov.b r6,@-r1
dt r2
mov.b r6,@-r1
bf/s.L2
mov.b r6,@-r1
mov r5,r0
shll2 r0
sub r0,r5
mov r4,r0
rts
add r5,r0
[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190
Alan Modra changed:
What|Removed |Added
Target|sh*-*-* arm*-*-*|sh*-*-* arm*-*-*
||powerpc*-*-*
CC||amodra at gmail dot com
--- Comment #10 from Alan Modra ---
Seen also on powerpc. Even for simple testcases like
int b[30];
void
fn1 (unsigned int n)
{
unsigned int i;
for (i = 0; i < n; i++)
b[i] = 42;
}
I see for -m32
fn1:
cmpwi 0,3,0
beq 0,.L1
slwi 9,3,2
lis 10,b-4@ha
addi 9,9,-4
la 10,b-4@l(10)
srwi 9,9,2
li 8,42
addi 9,9,1
mtctr 9
.p2align 4,,15
.L3:
stwu 8,4(10)
bdnz .L3
.L1:
blr
All of the insns marked are redundant.
For -m64 this loop doesn't even use bdnz
Loop 1 is simple:
simple exit 4 -> 5
infinite if: (expr_list:REG_DEP_TRUE (ne:SI (and:DI (plus:DI (minus:DI
(ashift:DI (reg:DI 189)
(const_int 2 [0x2]))
(reg:DI 179 [ ivtmp.7 ]))
(symbol_ref:DI ("b") [flags 0x80] ))
(const_int 3 [0x3]))
(const_int 0 [0]))
(nil))
number of iterations: (lshiftrt:DI (plus:DI (minus:DI (reg:DI 185 [ _18 ])
(reg:DI 179 [ ivtmp.7 ]))
(const_int -4 [0xfffc]))
(const_int 2 [0x2]))
upper bound: 29
likely upper bound: 29
realistic bound: -1
Doloop: Possible infinite iteration case.
Doloop: The loop is not suitable.
[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190 --- Comment #9 from Oleg Endo --- As of r233601 the issue still persists.
[Bug rtl-optimization/55190] ivopts causes loop setup bloat
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55190 --- Comment #8 from Oleg Endo olegendo at gcc dot gnu.org --- I've adjusted the title, since the issues here are not SH specific. There is also a similar PR 62233.
