https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84042
Bug ID: 84042
Summary: IVOPTS doesn't optimize int indexes on some PowerPC
code.
Product: gcc
Version: 8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: meissner at gcc dot gnu.org
Target Milestone: ---
This is a follow-on to PR target/81550.
In that bug, the following code generated more code. That in turn caused the
test to fail that was looking for a loop to be 8 instructions or less to test
the TARGET_ASM_LOOP_ALIGN_MAX_SKIP target hook:
void f(double *a, double *b, double *c, int n) {
int i;
for (i=0; i < n; i++)
a[i] = b[i] + c[i];
}
I rewrote the test to use unsigned long instead of int so that the # of
instructions is under 8, and it validates the target hook.
But we really should fix the regression.
As per PR target/81550, the original code on powerpc little endian was:
.file "loop_align.c"
.abiversion 2
.section ".text"
.align 2
.p2align 4,,15
.globl f
.type f, @function
f:
cmpwi 7,6,0
blelr 7
addi 6,6,-1
li 9,0
rldicl 6,6,0,32
addi 10,6,1
mtctr 10
.p2align 5,,31
.L3:
lfdx 0,4,9
lfdx 12,5,9
fadd 0,0,12
stfdx 0,3,9
addi 9,9,8
bdnz .L3
blr
.file "loop_align.c"
.abiversion 2
.section ".text"
.align 2
.p2align 4,,15
.globl f
.type f, @function
f:
cmpwi 7,6,0
blelr 7
addi 6,6,-1
addi 9,4,-8
rldic 6,6,3,29
addi 5,5,-8
add 4,4,6
addi 3,3,-8
.p2align 4,,15
.L3:
lfdu 0,8(9)
lfdu 12,8(5)
cmpld 7,9,4
fadd 0,0,12
stfdu 0,8(3)
beqlr 7
lfdu 0,8(9)
lfdu 12,8(5)
cmpld 7,9,4
fadd 0,0,12
stfdu 0,8(3)
bne 7,.L3
blr
The first observation is the count down loop is no longer done (i.e. no BDNZ).
Because it no longer is a count down loop, reorder blocks decided to clone the
block and have an early exit. This all makes the loop to be 12 instructions,
which causes the TARGET_ASM_LOOP_ALIGN_MAX_SKIP target hook to fail.
