[llvm] [CodeGen] Combine two loops in SloIndexes.cpp file (PR #127631)

[via llvm-commits]

================
@@ -1145,18 +1158,31 @@ entry:
 }
 
 define arm_aapcs_vfpcc <8 x half> @shuffle3_f16(<8 x half> %src) {
-; CHECK-LABEL: shuffle3_f16:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmovx.f16 s5, s3
-; CHECK-NEXT:    vmovx.f16 s6, s1
-; CHECK-NEXT:    vmovx.f16 s4, s0
-; CHECK-NEXT:    vins.f16 s1, s0
-; CHECK-NEXT:    vins.f16 s6, s4
-; CHECK-NEXT:    vins.f16 s5, s3
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s7, s1
-; CHECK-NEXT:    vmov q0, q1
-; CHECK-NEXT:    bx lr
+; CHECK-LV-LABEL: shuffle3_f16:
----------------
Rifet-c wrote:

The output reordered because the loops were merged -> previously we were removing all the indexes first and then, after it was completely done, we were correcting them altogether from the start (second loop), now these to processes are going in parallel, hence the indexes may be different.
As far as I understood, changes in output are more of a reordering, rather than real changes, but I might've overlooked something, hence I can investigate the changes a bit deeper if you want. What do you say?

https://github.com/llvm/llvm-project/pull/127631