aco: fix single-alignbyte do_pack_2x16() path with fp inline constants

We were using a 16-bit inline constant with a 32-bit instruction and the
test would have created
"v1: %_:v[0] = v_alignbyte_b32 0.5, %_:v[1][16:32], 2" instead.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16296>

diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp
index 40798da..6e9606c 100644 (file)
--- a/src/amd/compiler/aco_lower_to_hw_instr.cpp
+++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp
@@ -1452,7 +1452,11 @@ do_pack_2x16(lower_context* ctx, Builder& bld, Definition def, Operand lo, Opera
    if (lo.physReg().byte() == 2 && hi.physReg().byte() == 0 &&
        (!hi.isConstant() || !Operand::c32(hi.constantValue()).isLiteral() ||
         ctx->program->gfx_level >= GFX10)) {
-      bld.vop3(aco_opcode::v_alignbyte_b32, def, hi, lo, Operand::c32(2u));
+      if (hi.isConstant())
+         bld.vop3(aco_opcode::v_alignbyte_b32, def, Operand::c32(hi.constantValue()), lo,
+                  Operand::c32(2u));
+      else
+         bld.vop3(aco_opcode::v_alignbyte_b32, def, hi, lo, Operand::c32(2u));
       return;
    }
 

diff --git a/src/amd/compiler/tests/test_to_hw_instr.cpp b/src/amd/compiler/tests/test_to_hw_instr.cpp
index d4799d6..38a95ea 100644 (file)
--- a/src/amd/compiler/tests/test_to_hw_instr.cpp
+++ b/src/amd/compiler/tests/test_to_hw_instr.cpp
@@ -838,3 +838,27 @@ BEGIN_TEST(to_hw_instr.swap_linear_vgpr)
 
    finish_to_hw_instr_test();
 END_TEST
+
+BEGIN_TEST(to_hw_instr.pack2x16_alignbyte_constant)
+   PhysReg v0_lo{256};
+   PhysReg v0_hi{256};
+   PhysReg v1_hi{257};
+   v0_hi.reg_b += 2;
+   v1_hi.reg_b += 2;
+
+   if (!setup_cs(NULL, GFX10))
+      return;
+
+   /* prevent usage of v_pack_b32_f16 */
+   program->blocks[0].fp_mode.denorm16_64 = fp_denorm_flush;
+
+   //>> p_unit_test 0
+   //! v1: %_:v[0] = v_alignbyte_b32 0x3800, %_:v[1][16:32], 2
+   bld.pseudo(aco_opcode::p_unit_test, Operand::zero());
+   bld.pseudo(aco_opcode::p_parallelcopy, Definition(v0_lo, v2b), Definition(v0_hi, v2b),
+              Operand(v1_hi, v2b), Operand::c16(0x3800));
+
+   //! s_endpgm
+
+   finish_to_hw_instr_test();
+END_TEST