https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119876
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Depends on| |119920
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
With the patch I attached to PR 119920 which does the factoring in ifconv we
get:
```
.L2:
vmovdqa32 b(%rax), %zmm0
addq $64, %rax
vpcmpd $6, %zmm3, %zmm0, %k1 // compare
vmovdqa32 %zmm1, %zmm0 // set zmm0 to 2s
vpsrld $31, %zmm2, %zmm0{%k1} // blends in 1
vpaddd c-64(%rax), %zmm0, %zmm0 // add
vmovdqa32 %zmm0, a-64(%rax)
cmpq $3968, %rax
jne .L2
```
Or rather:
```
<bb 3> [local count: 665719666]:
# ivtmp.36_60 = PHI <ivtmp.36_64(3), 0(2)>
vect__1.13_49 = MEM <vector(16) int> [(int *)&b + ivtmp.36_60 * 1];
mask__4.14_50 = vect__1.13_49 > { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0 };
vect__28.15_51 = .VCOND_MASK (mask__4.14_50, { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1 }, { 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2 });
vect_pretmp_6.10_46 = MEM <vector(16) int> [(int *)&c + ivtmp.36_60 * 1];
vect_iftmp.16_52 = vect_pretmp_6.10_46 + vect__28.15_51;
MEM <vector(16) int> [(int *)&a + ivtmp.36_60 * 1] = vect_iftmp.16_52;
ivtmp.36_64 = ivtmp.36_60 + 64;
if (ivtmp.36_64 != 3968)
goto <bb 3>; [98.39%]
else
goto <bb 4>; [1.61%]
```
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119920
[Bug 119920] Missed vectorization for conditioned adds
