From: Georg Lehmann <dadschoorse@gmail.com>
Date: Tue, 21 Mar 2023 11:27:28 +0000 (+0100)
Subject: aco/to_hw_instr: use VOP1 opsel for v_mov_b16
X-Git-Tag: upstream/23.3.3~10837
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=8ee1519ceecd9105855b3c9aad85e729565b6c6c;p=platform%2Fupstream%2Fmesa.git

aco/to_hw_instr: use VOP1 opsel for v_mov_b16

Foz-DB GFX1100:
Totals from 4661 (3.46% of 134864) affected shaders:
CodeSize: 36500568 -> 36391704 (-0.30%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22069>
---

diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp
index 2d92b63..dc8dba5 100644
--- a/src/amd/compiler/aco_lower_to_hw_instr.cpp
+++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp
@@ -1109,23 +1109,15 @@ emit_v_mov_b16(Builder& bld, Definition dst, Operand op)
       if (!op.isLiteral() && op.physReg() >= 240) {
          /* v_add_f16 is smaller because it can use 16bit fp inline constants. */
          Instruction* instr = bld.vop2_e64(aco_opcode::v_add_f16, dst, op, Operand::zero());
-         if (dst.physReg().byte() == 2)
-            instr->valu().opsel = 0x8;
+         instr->valu().opsel[3] = dst.physReg().byte() == 2;
          return;
       }
       op = Operand::c32((int32_t)(int16_t)op.constantValue());
    }
 
-   if (!dst.physReg().byte() && !op.physReg().byte()) {
-      bld.vop1(aco_opcode::v_mov_b16, dst, op);
-   } else {
-      // TODO: this can use VOP1 for vgpr0-127 with assembler support
-      Instruction* instr = bld.vop1_e64(aco_opcode::v_mov_b16, dst, op);
-      if (op.physReg().byte() == 2)
-         instr->valu().opsel |= 0x1;
-      if (dst.physReg().byte() == 2)
-         instr->valu().opsel |= 0x8;
-   }
+   Instruction* instr = bld.vop1(aco_opcode::v_mov_b16, dst, op);
+   instr->valu().opsel[0] = op.physReg().byte() == 2;
+   instr->valu().opsel[3] = dst.physReg().byte() == 2;
 }
 
 void