[AMDGPU] Change encoding of gfx1250 ld_scale (#167777)

The unused field src2 is now encoded to vgpr0 accoring to the
latest guidelines.

Fixes: SWDEV-565846

Author: rampitec

llvm/lib/Target/AMDGPU/VOP3PInstructions.td

@@ -2218,7 +2218,7 @@ class VOP3PX2e <bits<8> op, bits<8> LdScaleOp, VOP3PWMMA_Profile P> : Enc128, VO
   let Inst{23-16} = LdScaleOp;
   let Inst{40-32} = scale_src0;
   let Inst{49-41} = scale_src1;
-  let Inst{58-50} = 0; // scale src2
+  let Inst{58-50} = 0x100; // scale src2 = vgpr0 (dummy)
   let Inst{59}    = matrix_b_scale{0}; // scale_op_sel_hi(0)
   let Inst{60}    = 0;                 // scale_op_sel_hi(1)
   let Inst{63-61} = {0, matrix_a_scale_fmt{1-0}}; // neg (lo)
@@ -2433,6 +2433,15 @@ multiclass VOP3P_Real_with_name_gfx12<bits<8> op,
                           string asmName = !cast<VOP3P_Pseudo>(NAME).Mnemonic> :
   VOP3P_Real_with_name<GFX12Gen, op, backing_ps_name, asmName>;
 
+multiclass VOP3P_Real_LD_SCALE_gfx1250<bits<8> op> {
+  defvar ps = !cast<VOP3P_Pseudo>(NAME);
+  def _gfx1250 :
+    VOP3P_Real_Gen<ps, GFX1250Gen, ps.Mnemonic>,
+    VOP3Pe_gfx11_gfx12<op, ps.Pfl> {
+      let Inst{58-50} = 0x100; // scale src2 = vgpr0 (dummy)
+    }
+}
+
 defm V_PK_MIN_NUM_F16 : VOP3P_Real_with_name_gfx12<0x1b, "V_PK_MIN_F16", "v_pk_min_num_f16">;
 defm V_PK_MAX_NUM_F16 : VOP3P_Real_with_name_gfx12<0x1c, "V_PK_MAX_F16", "v_pk_max_num_f16">;
 
@@ -2462,8 +2471,8 @@ defm V_FMA_MIX_F32_BF16 : VOP3P_Realtriple<GFX1250Gen, 0x3d>;
 defm V_FMA_MIXLO_BF16   : VOP3P_Realtriple<GFX1250Gen, 0x3e>;
 defm V_FMA_MIXHI_BF16   : VOP3P_Realtriple<GFX1250Gen, 0x3f>;
 
-defm V_WMMA_LD_SCALE_PAIRED_B32   : VOP3P_Real_gfx1250<0x35>;
-defm V_WMMA_LD_SCALE16_PAIRED_B64 : VOP3P_Real_gfx1250<0x3a>;
+defm V_WMMA_LD_SCALE_PAIRED_B32   : VOP3P_Real_LD_SCALE_gfx1250<0x35>;
+defm V_WMMA_LD_SCALE16_PAIRED_B64 : VOP3P_Real_LD_SCALE_gfx1250<0x3a>;
 
 let AssemblerPredicate = isGFX1250Plus in
 def : AMDGPUMnemonicAlias<"v_fma_mix_f32_f16",  "v_fma_mix_f32">;