diff --git a/src/amd/compiler/aco_insert_waitcnt.cpp b/src/amd/compiler/aco_insert_waitcnt.cpp
index 97cf9229255..d1123ea893c 100644
--- a/src/amd/compiler/aco_insert_waitcnt.cpp
+++ b/src/amd/compiler/aco_insert_waitcnt.cpp
@@ -575,16 +575,6 @@ kill(wait_imm& imm, depctr_wait& depctr, Instruction* instr, wait_ctx& ctx,
       force_waitcnt(ctx, imm);
    }
 
-   /* sendmsg(dealloc_vgprs) releases scratch, so this isn't safe if there is a in-progress
-    * scratch store.
-    */
-   if (ctx.gfx_level >= GFX11 && instr->opcode == aco_opcode::s_sendmsg &&
-       instr->salu().imm == sendmsg_dealloc_vgprs) {
-      barrier_info& bar = ctx.bar[barrier_info_release_dep];
-      imm.combine(bar.imm[ffs(storage_scratch) - 1]);
-      imm.combine(bar.imm[ffs(storage_vgpr_spill) - 1]);
-   }
-
    check_instr(ctx, imm, instr);
 
    if (instr->opcode == aco_opcode::ds_ordered_count &&
diff --git a/src/amd/compiler/aco_ir.cpp b/src/amd/compiler/aco_ir.cpp
index 0a53afac5c1..d2a208c2fec 100644
--- a/src/amd/compiler/aco_ir.cpp
+++ b/src/amd/compiler/aco_ir.cpp
@@ -1657,32 +1657,6 @@ depctr_wait::pack() const
    return imm;
 }
 
-bool
-dealloc_vgprs(Program* program)
-{
-   if (program->gfx_level < GFX11)
-      return false;
-
-   /* If we insert the sendmsg on GFX11.5, the export priority workaround will require us to insert
-    * a wait after exports. There might still be pending VMEM stores for PS parameter exports,
-    * except NGG lowering usually inserts a memory barrier. This means there is unlikely to be any
-    * pending VMEM stores or exports if we insert the sendmsg for these stages. */
-   if (program->gfx_level == GFX11_5 && (program->stage.hw == AC_HW_NEXT_GEN_GEOMETRY_SHADER ||
-                                         program->stage.hw == AC_HW_PIXEL_SHADER))
-      return false;
-
-   Block& block = program->blocks.back();
-
-   /* don't bother checking if there is a pending VMEM store or export: there almost always is */
-   Builder bld(program);
-   if (!block.instructions.empty() && block.instructions.back()->opcode == aco_opcode::s_endpgm) {
-      bld.reset(&block.instructions, block.instructions.begin() + (block.instructions.size() - 1));
-      bld.sopp(aco_opcode::s_sendmsg, sendmsg_dealloc_vgprs);
-   }
-
-   return true;
-}
-
 bool
 Instruction::isTrans() const noexcept
 {
diff --git a/src/amd/compiler/aco_ir.h b/src/amd/compiler/aco_ir.h
index 92121b2aa42..d970224e599 100644
--- a/src/amd/compiler/aco_ir.h
+++ b/src/amd/compiler/aco_ir.h
@@ -2440,7 +2440,6 @@ void spill(Program* program);
 void insert_waitcnt(Program* program);
 void insert_delay_alu(Program* program);
 void combine_delay_alu(Program* program);
-bool dealloc_vgprs(Program* program);
 void insert_NOPs(Program* program);
 void form_hard_clauses(Program* program);
 unsigned emit_program(Program* program, std::vector<uint32_t>& code,
diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp
index 1f16271fa3c..6d472ee9495 100644
--- a/src/amd/compiler/aco_lower_to_hw_instr.cpp
+++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp
@@ -2273,8 +2273,6 @@ lower_to_hw_instr(Program* program)
 
    int end_with_regs_block_index = -1;
 
-   bool should_dealloc_vgprs = dealloc_vgprs(program);
-
    for (int block_idx = program->blocks.size() - 1; block_idx >= 0; block_idx--) {
       Block* block = &program->blocks[block_idx];
       lower_context ctx;
@@ -2404,9 +2402,6 @@ lower_to_hw_instr(Program* program)
                               instr2->definitions[0].isFixed() &&
                               instr2->definitions[0].physReg() == exec)
                         continue;
-                     else if (instr2->opcode == aco_opcode::s_sendmsg &&
-                              instr2->salu().imm == sendmsg_dealloc_vgprs)
-                        continue;
 
                      ignore_early_exit = false;
                   }
@@ -2431,12 +2426,6 @@ lower_to_hw_instr(Program* program)
                   block = &program->blocks[block_idx];
                   ctx.block = block;
 
-                  /* sendmsg(dealloc_vgprs) releases scratch, so it isn't safe if there is an
-                   * in-progress scratch store. */
-                  wait_imm wait;
-                  if (should_dealloc_vgprs && uses_scratch(program))
-                     wait.vs = 0;
-
                   bld.reset(discard_block);
                   if (program->has_pops_overlapped_waves_wait &&
                       (program->gfx_level >= GFX11 || discard_sends_pops_done)) {
@@ -2444,6 +2433,7 @@ 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.
                       */
+                     wait_imm wait;
                      if (program->gfx_level >= GFX10)
                         wait.vs = 0;
                      wait.vm = 0;
@@ -2462,10 +2452,6 @@ lower_to_hw_instr(Program* program)
                      bld.exp(aco_opcode::exp, Operand(v1), Operand(v1), Operand(v1), Operand(v1), 0,
                              target, false, true, true);
 
-                  wait.build_waitcnt(bld);
-                  if (should_dealloc_vgprs)
-                     bld.sopp(aco_opcode::s_sendmsg, sendmsg_dealloc_vgprs);
-
                   bld.sopp(aco_opcode::s_endpgm);
 
                   bld.reset(&ctx.instructions);
diff --git a/src/amd/compiler/tests/test_isel.cpp b/src/amd/compiler/tests/test_isel.cpp
index ab6676875b4..4a1bf9f9cf8 100644
--- a/src/amd/compiler/tests/test_isel.cpp
+++ b/src/amd/compiler/tests/test_isel.cpp
@@ -174,13 +174,9 @@ BEGIN_TEST(isel.discard_early_exit.mrtz)
 
    /* On GFX11, the discard early exit must use mrtz if the shader exports only depth. */
    //>> exp mrtz v#_, off, off, off done    ; $_ $_
-   //! s_nop 0                              ; $_
-   //! s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; $_
    //! s_endpgm                             ; $_
    //! BB1:
    //! exp mrtz off, off, off, off done     ; $_ $_
-   //! s_nop 0                              ; $_
-   //! s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; $_
    //! s_endpgm                             ; $_
 
    PipelineBuilder pbld(get_vk_device(GFX11));
@@ -203,13 +199,9 @@ BEGIN_TEST(isel.discard_early_exit.mrt0)
 
    /* On GFX11, the discard early exit must use mrt0 if the shader exports color. */
    //>> exp mrt0 v#x, v#x, v#x, v#x done    ; $_ $_
-   //! s_nop 0                              ; $_
-   //! s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; $_
    //! s_endpgm                             ; $_
    //! BB1:
    //! exp mrt0 off, off, off, off done     ; $_ $_
-   //! s_nop 0                              ; $_
-   //! s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; $_
    //! s_endpgm                             ; $_
 
    PipelineBuilder pbld(get_vk_device(GFX11));
diff --git a/src/amd/compiler/tests/test_to_hw_instr.cpp b/src/amd/compiler/tests/test_to_hw_instr.cpp
index c638016198b..17e83fe2295 100644
--- a/src/amd/compiler/tests/test_to_hw_instr.cpp
+++ b/src/amd/compiler/tests/test_to_hw_instr.cpp
@@ -218,7 +218,6 @@ BEGIN_TEST(to_hw_instr.swap_subdword)
       bld.pseudo(aco_opcode::p_parallelcopy, Definition(v0_lo, v1b), Definition(v0_hi, v1b),
                  Operand(v0_hi, v1b), Operand(v0_lo, v1b));
 
-      //~gfx11! s_sendmsg sendmsg(dealloc_vgprs)
       //! s_endpgm
 
       finish_to_hw_instr_test();
@@ -369,7 +368,6 @@ BEGIN_TEST(to_hw_instr.subdword_constant)
       bld.pseudo(aco_opcode::p_unit_test, Operand::c32(17u));
       bld.pseudo(aco_opcode::p_parallelcopy, Definition(v0_lo, v1b), Operand::zero(1));
 
-      //~gfx11! s_sendmsg sendmsg(dealloc_vgprs)
       //! s_endpgm
 
       finish_to_hw_instr_test();
@@ -500,7 +498,6 @@ BEGIN_TEST(to_hw_instr.extract)
 
          finish_to_hw_instr_test();
 
-         //~gfx11_.*! s_sendmsg sendmsg(dealloc_vgprs)
          //! s_endpgm
       }
    }
@@ -618,7 +615,6 @@ BEGIN_TEST(to_hw_instr.insert)
 
       finish_to_hw_instr_test();
 
-      //~gfx11! s_sendmsg sendmsg(dealloc_vgprs)
       //! s_endpgm
    }
 END_TEST
@@ -838,7 +834,6 @@ BEGIN_TEST(to_hw_instr.pack2x16_constant)
       bld.pseudo(aco_opcode::p_parallelcopy, Definition(v0_lo, v2b), Definition(v0_hi, v2b),
                  Operand(v1_lo, v2b), Operand::c16(0x00f0));
 
-      //~gfx11! s_sendmsg sendmsg(dealloc_vgprs)
       //! s_endpgm
 
       finish_to_hw_instr_test();
@@ -855,7 +850,6 @@ BEGIN_TEST(to_hw_instr.mov_b16_sgpr_src)
    bld.pseudo(aco_opcode::p_extract_vector, Definition(PhysReg(256), v2b), Operand(PhysReg(0), s1),
               Operand::c32(1));
 
-   //! s_sendmsg sendmsg(dealloc_vgprs)
    //! s_endpgm
 
    finish_to_hw_instr_test();