[llvm] [SROA] Unfold gep of index select (PR #80983)

[Artem Belevich via llvm-commits]

================
@@ -158,11 +158,20 @@ bb:
 
 define i32 @test_select_idx_memcpy(i1 %c, ptr %p) {
----------------
Artem-B wrote:

> The point of this test is to show that the select unfolding works correctly even if the alloca is splittable, but non-promotable.

The test does all that now. 

My point is that the test itself has `malloc` in the title, yet we're talking about `alloca` and `select` unfolding. The fact that we're doing a useful change right now makes me question what the test was supposed to test previously. IR was left essentially unchanged, so it's hard to infer the intent from that, other than memcpy is involved, somehow.  We're probably not testing that, whatever that was, now.

Id you do not think that it would make sense to test alloca spolitting vs select folding separately, then the test could use a different name, reflecting its purpose now.

https://github.com/llvm/llvm-project/pull/80983