diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp
index 9732933cd84..74244b0907e 100644
--- a/src/amd/compiler/aco_lower_to_hw_instr.cpp
+++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp
@@ -2224,6 +2224,8 @@ lower_to_hw_instr(Program* program)
                   } else if (offset == 0 && signext && (bits == 8 || bits == 16)) {
                      bld.sop1(bits == 8 ? aco_opcode::s_sext_i32_i8 : aco_opcode::s_sext_i32_i16,
                               dst, op);
+                  } else if (ctx.program->gfx_level >= GFX9 && offset == 0 && bits == 16) {
+                     bld.sop2(aco_opcode::s_pack_ll_b32_b16, dst, op, Operand::zero());
                   } else {
                      bld.sop2(signext ? aco_opcode::s_bfe_i32 : aco_opcode::s_bfe_u32, dst,
                               bld.def(s1, scc), op, Operand::c32((bits << 16) | offset));
diff --git a/src/amd/compiler/tests/test_to_hw_instr.cpp b/src/amd/compiler/tests/test_to_hw_instr.cpp
index 5e71d294a88..4f6aa44623f 100644
--- a/src/amd/compiler/tests/test_to_hw_instr.cpp
+++ b/src/amd/compiler/tests/test_to_hw_instr.cpp
@@ -648,7 +648,8 @@ BEGIN_TEST(to_hw_instr.extract)
       EXT(2, 8)
       //! s1: %_:s[0],  s1: %_:scc = @s_shr %_:s[1], 24
       EXT(3, 8)
-      //~gfx.*_unsigned! s1: %_:s[0],  s1: %_:scc = @s_bfe %_:s[1], 0x100000
+      //~gfx(7|8)_unsigned! s1: %_:s[0],  s1: %_:scc = @s_bfe %_:s[1], 0x100000
+      //~gfx(9|11)_unsigned! s1: %_:s[0] = s_pack_ll_b32_b16 %_:s[1], 0
       //~gfx.*_signed! s1: %_:s[0] = s_sext_i32_i16 %_:s[1]
       EXT(0, 16)
       //! s1: %_:s[0],  s1: %_:scc = @s_shr %_:s[1], 16