DAG: Use phi in alloca constant case to create virtual registers (#130254)

This is a follow up from 39bf765,
for the other case handled here. We would create CopyToReg marked
as uniform, even though the end phi would need to use VGPRs due
to another divergent input. There's no directly observable change in
the final output of the new test, but it does hit this case.

Author: arsenm

llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

@@ -12045,7 +12045,7 @@ SelectionDAGBuilder::HandlePHINodesInSuccessorBlocks(const BasicBlock *LLVMBB) {
           assert(isa<AllocaInst>(PHIOp) &&
                  FuncInfo.StaticAllocaMap.count(cast<AllocaInst>(PHIOp)) &&
                  "Didn't codegen value into a register!??");
-          Reg = FuncInfo.CreateRegs(PHIOp);
+          Reg = FuncInfo.CreateRegs(&PN);
           CopyValueToVirtualRegister(PHIOp, Reg);
         }
       }

llvm/test/CodeGen/AMDGPU/copy-to-reg-frameindex.ll

@@ -35,3 +35,52 @@ done:
   store i32 %extract.0, ptr addrspace(1) %out, align 4
   ret void
 }
+
+; When creating registers for %divergent.alloca.phi, we should report
+; the CopyToReg as divergent values (not uniform just because the
+; alloca is uniform)
+define void @phi_with_alloca_and_divergent_copy_to_reg(ptr addrspace(5) %divergent.private, ptr addrspace(1) %out, i32 %a, i32 %b, i32 %c) {
+; CHECK-LABEL: phi_with_alloca_and_divergent_copy_to_reg:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v7, v2
+; CHECK-NEXT:    v_mov_b32_e32 v6, v1
+; CHECK-NEXT:    s_mov_b64 s[4:5], 0
+; CHECK-NEXT:    v_lshrrev_b32_e64 v2, 6, s32
+; CHECK-NEXT:  .LBB1_1: ; %loop
+; CHECK-NEXT:    ; =>This Inner Loop Header: Depth=1
+; CHECK-NEXT:    v_mov_b32_e32 v1, v2
+; CHECK-NEXT:    v_lshl_add_u32 v2, v3, 2, v1
+; CHECK-NEXT:    buffer_store_dword v3, v2, s[0:3], 0 offen
+; CHECK-NEXT:    v_add_u32_e32 v2, 1, v3
+; CHECK-NEXT:    v_cmp_lt_u32_e32 vcc, 15, v2
+; CHECK-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; CHECK-NEXT:    v_mov_b32_e32 v3, v4
+; CHECK-NEXT:    v_mov_b32_e32 v2, v0
+; CHECK-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; CHECK-NEXT:    s_cbranch_execnz .LBB1_1
+; CHECK-NEXT:  ; %bb.2: ; %done
+; CHECK-NEXT:    s_or_b64 exec, exec, s[4:5]
+; CHECK-NEXT:    buffer_load_dword v0, v1, s[0:3], 0 offen
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dword v[6:7], v0, off
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+entry:
+  %alloca0 = alloca [16 x i32], addrspace(5)
+  br label %loop
+
+loop:
+  %inc = phi i32 [%a, %entry], [%b, %loop]
+  %divergent.alloca.phi = phi ptr addrspace(5) [ %alloca0, %entry ], [ %divergent.private, %loop ]
+  %ptr = getelementptr [16 x i32], ptr addrspace(5) %divergent.alloca.phi, i32 0, i32 %inc
+  store i32 %inc, ptr addrspace(5) %ptr
+  %inc.i = add i32 %inc, 1
+  %cnd = icmp uge i32 %inc.i, 16
+  br i1 %cnd, label %done, label %loop
+
+done:
+  %tmp1 = load i32, ptr addrspace(5) %divergent.alloca.phi
+  store i32 %tmp1, ptr addrspace(1) %out
+  ret void
+}