diff --git a/src/amd/compiler/aco_insert_waitcnt.cpp b/src/amd/compiler/aco_insert_waitcnt.cpp
index e6263d6f191..345ba6047bd 100644
--- a/src/amd/compiler/aco_insert_waitcnt.cpp
+++ b/src/amd/compiler/aco_insert_waitcnt.cpp
@@ -666,43 +666,7 @@ void
 emit_waitcnt(wait_ctx& ctx, std::vector<aco_ptr<Instruction>>& instructions, wait_imm& imm)
 {
    Builder bld(ctx.program, &instructions);
-
-   if (ctx.gfx_level >= GFX12) {
-      if (imm.vm != wait_imm::unset_counter && imm.lgkm != wait_imm::unset_counter) {
-         bld.sopp(aco_opcode::s_wait_loadcnt_dscnt, (imm.vm << 8) | imm.lgkm);
-         imm.vm = wait_imm::unset_counter;
-         imm.lgkm = wait_imm::unset_counter;
-      }
-
-      if (imm.vs != wait_imm::unset_counter && imm.lgkm != wait_imm::unset_counter) {
-         bld.sopp(aco_opcode::s_wait_storecnt_dscnt, (imm.vs << 8) | imm.lgkm);
-         imm.vs = wait_imm::unset_counter;
-         imm.lgkm = wait_imm::unset_counter;
-      }
-
-      aco_opcode op[wait_type_num];
-      op[wait_type_exp] = aco_opcode::s_wait_expcnt;
-      op[wait_type_lgkm] = aco_opcode::s_wait_dscnt;
-      op[wait_type_vm] = aco_opcode::s_wait_loadcnt;
-      op[wait_type_vs] = aco_opcode::s_wait_storecnt;
-      op[wait_type_sample] = aco_opcode::s_wait_samplecnt;
-      op[wait_type_bvh] = aco_opcode::s_wait_bvhcnt;
-      op[wait_type_km] = aco_opcode::s_wait_kmcnt;
-
-      for (unsigned i = 0; i < wait_type_num; i++) {
-         if (imm[i] != wait_imm::unset_counter)
-            bld.sopp(op[i], imm[i]);
-      }
-   } else {
-      if (imm.vs != wait_imm::unset_counter) {
-         assert(ctx.gfx_level >= GFX10);
-         bld.sopk(aco_opcode::s_waitcnt_vscnt, Operand(sgpr_null, s1), imm.vs);
-         imm.vs = wait_imm::unset_counter;
-      }
-      if (!imm.empty())
-         bld.sopp(aco_opcode::s_waitcnt, imm.pack(ctx.gfx_level));
-   }
-   imm = wait_imm();
+   imm.build_waitcnt(bld);
 }
 
 bool
diff --git a/src/amd/compiler/aco_ir.cpp b/src/amd/compiler/aco_ir.cpp
index 2c0b17a82ca..022dd60e87c 100644
--- a/src/amd/compiler/aco_ir.cpp
+++ b/src/amd/compiler/aco_ir.cpp
@@ -1328,6 +1328,50 @@ wait_imm::print(FILE* output) const
    }
 }
 
+void
+wait_imm::build_waitcnt(Builder& bld)
+{
+   enum amd_gfx_level gfx_level = bld.program->gfx_level;
+
+   if (gfx_level >= GFX12) {
+      if (vm != wait_imm::unset_counter && lgkm != wait_imm::unset_counter) {
+         bld.sopp(aco_opcode::s_wait_loadcnt_dscnt, (vm << 8) | lgkm);
+         vm = wait_imm::unset_counter;
+         lgkm = wait_imm::unset_counter;
+      }
+
+      if (vs != wait_imm::unset_counter && lgkm != wait_imm::unset_counter) {
+         bld.sopp(aco_opcode::s_wait_storecnt_dscnt, (vs << 8) | lgkm);
+         vs = wait_imm::unset_counter;
+         lgkm = wait_imm::unset_counter;
+      }
+
+      aco_opcode op[wait_type_num];
+      op[wait_type_exp] = aco_opcode::s_wait_expcnt;
+      op[wait_type_lgkm] = aco_opcode::s_wait_dscnt;
+      op[wait_type_vm] = aco_opcode::s_wait_loadcnt;
+      op[wait_type_vs] = aco_opcode::s_wait_storecnt;
+      op[wait_type_sample] = aco_opcode::s_wait_samplecnt;
+      op[wait_type_bvh] = aco_opcode::s_wait_bvhcnt;
+      op[wait_type_km] = aco_opcode::s_wait_kmcnt;
+
+      for (unsigned i = 0; i < wait_type_num; i++) {
+         if ((*this)[i] != wait_imm::unset_counter)
+            bld.sopp(op[i], (*this)[i]);
+      }
+   } else {
+      if (vs != wait_imm::unset_counter) {
+         assert(gfx_level >= GFX10);
+         bld.sopk(aco_opcode::s_waitcnt_vscnt, Operand(sgpr_null, s1), vs);
+         vs = wait_imm::unset_counter;
+      }
+      if (!empty())
+         bld.sopp(aco_opcode::s_waitcnt, pack(gfx_level));
+   }
+
+   *this = wait_imm();
+}
+
 bool
 should_form_clause(const Instruction* a, const Instruction* b)
 {
diff --git a/src/amd/compiler/aco_ir.h b/src/amd/compiler/aco_ir.h
index 2925a2fcdeb..6848f721eb4 100644
--- a/src/amd/compiler/aco_ir.h
+++ b/src/amd/compiler/aco_ir.h
@@ -182,6 +182,7 @@ enum wait_type {
 };
 
 struct Instruction;
+class Builder;
 
 struct wait_imm {
    static const uint8_t unset_counter = 0xff;
@@ -209,6 +210,8 @@ struct wait_imm {
 
    void print(FILE* output) const;
 
+   void build_waitcnt(Builder& bld);
+
    uint8_t& operator[](size_t i)
    {
       assert(i < wait_type_num);
diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp
index 0e18aa66069..9d7c895f195 100644
--- a/src/amd/compiler/aco_lower_to_hw_instr.cpp
+++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp
@@ -2380,13 +2380,13 @@ lower_to_hw_instr(Program* program)
                       * the waitcnt necessary before resuming overlapping waves as the normal
                       * waitcnt insertion doesn't work in a discard early exit block.
                       */
-                     if (program->gfx_level >= GFX10)
-                        bld.sopk(aco_opcode::s_waitcnt_vscnt, Operand(sgpr_null, s1), 0);
                      wait_imm pops_exit_wait_imm;
+                     if (program->gfx_level >= GFX10)
+                        pops_exit_wait_imm.vs = 0;
                      pops_exit_wait_imm.vm = 0;
                      if (program->has_smem_buffer_or_global_loads)
                         pops_exit_wait_imm.lgkm = 0;
-                     bld.sopp(aco_opcode::s_waitcnt, pops_exit_wait_imm.pack(program->gfx_level));
+                     pops_exit_wait_imm.build_waitcnt(bld);
                   }
                   if (discard_sends_pops_done)
                      bld.sopp(aco_opcode::s_sendmsg, sendmsg_ordered_ps_done);