https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88834
Bug ID: 88834
Summary: [SVE] Poor addressing mode choices for LD2 and ST2
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: enhancement
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: rsandifo at gcc dot gnu.org
Target Milestone: ---
Compiling this code with -O3 -march=armv8-a+sve:
void
f (int *restrict x, int *restrict y, int *restrict z, int n)
{
for (int i = 0; i < n; i += 2)
{
x[i] = y[i] + z[i];
x[i + 1] = y[i + 1] - z[i + 1];
}
}
gives:
f:
.LFB0:
.cfi_startproc
cmp w3, 0
ble .L1
sub w4, w3, #1
cntw x3
ptrue p1.s, all
lsr w4, w4, 1
add w4, w4, 1
whilelo p0.s, xzr, x4
.p2align 3,,7
.L3:
ld2w {z4.s - z5.s}, p0/z, [x1]
ld2w {z2.s - z3.s}, p0/z, [x2]
add z0.s, z4.s, z2.s
sub z1.s, z5.s, z3.s
st2w {z0.s - z1.s}, p0, [x0]
incb x1, all, mul #2
whilelo p0.s, x3, x4
incb x0, all, mul #2
incb x2, all, mul #2
incw x3
ptest p1, p0.b
bne .L3
.L1:
ret
.cfi_endproc
Rather than have one INCB per address, we should have a single IV
that tracks the index, something like:
ld2w {z4.s - z5.s}, p0/z, [x1, x4, lsl #2]
ld2w {z2.s - z3.s}, p0/z, [x2, x4, lsl #2]
add z0.s, z4.s, z2.s
sub z1.s, z5.s, z3.s
st2w {z0.s - z1.s}, p0, [x0, x4, lsl #2]
incw x4, all, mul #2 // or inch
I think this will need work in both the target code and ivopts.
