[AMDGPU][MC][GFX10] Added sdwa/dpp versions of v_cndmask_b32

[dp]

Summary

See https://bugs.llvm.org/show_bug.cgi?id=43608

[rampitec]

lib/Target/AMDGPU/VOP2Instructions.td

978

Please add foreach like in D69010. Here and below.

1014

Also needs foreach,

[dp]

Updated by integrating changes from D69010.

[rampitec]

Still missing foreach around dpp.

[dp]

In D69096#1713435, @rampitec wrote:

Still missing foreach around dpp.

I added them but found no differences in *.inc files.
Are they necessary as a reserve for the future?

[rampitec]

In D69096#1713450, @dp wrote:

In D69096#1713435, @rampitec wrote:

Still missing foreach around dpp.

I added them but found no differences in *.inc files.
Are they necessary as a reserve for the future?

Yes, it is better to have them so we do not have to check every time we are not producing garbage.

[dp]

Added more guards to avoid generating unused instructions.

[rampitec]

LGTM

lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp

void AMDGPUInstPrinter::printOperand(const MCInst *MI, unsigned OpNo,
   // Print default vcc/vcc_lo operand of v_cndmask_b32_e32.
   switch (MI->getOpcode()) {
   default: break;
 
   case AMDGPU::V_CNDMASK_B32_e32_gfx10:
   case AMDGPU::V_ADD_CO_CI_U32_e32_gfx10:
   case AMDGPU::V_SUB_CO_CI_U32_e32_gfx10:
   case AMDGPU::V_SUBREV_CO_CI_U32_e32_gfx10:
+  case AMDGPU::V_CNDMASK_B32_dpp_gfx10:
   case AMDGPU::V_ADD_CO_CI_U32_dpp_gfx10:
   case AMDGPU::V_SUB_CO_CI_U32_dpp_gfx10:
   case AMDGPU::V_SUBREV_CO_CI_U32_dpp_gfx10:
+  case AMDGPU::V_CNDMASK_B32_dpp8_gfx10:
   case AMDGPU::V_ADD_CO_CI_U32_dpp8_gfx10:
   case AMDGPU::V_SUB_CO_CI_U32_dpp8_gfx10:
   case AMDGPU::V_SUBREV_CO_CI_U32_dpp8_gfx10:
 
   case AMDGPU::V_CNDMASK_B32_e32_gfx6_gfx7:
   case AMDGPU::V_CNDMASK_B32_e32_vi:
     if ((int)OpNo == AMDGPU::getNamedOperandIdx(MI->getOpcode(),
                                                 AMDGPU::OpName::src1))
void AMDGPUInstPrinter::printOperandAndIntInputMods(const MCInst *MI,
   printOperand(MI, OpNo + 1, STI, O);
   if (InputModifiers & SISrcMods::SEXT)
     O << ')';
 
   // Print default vcc/vcc_lo operand of VOP2b.
   switch (MI->getOpcode()) {
   default: break;
 
+  case AMDGPU::V_CNDMASK_B32_sdwa_gfx10:
   case AMDGPU::V_ADD_CO_CI_U32_sdwa_gfx10:
   case AMDGPU::V_SUB_CO_CI_U32_sdwa_gfx10:
   case AMDGPU::V_SUBREV_CO_CI_U32_sdwa_gfx10:
     if ((int)OpNo + 1 == AMDGPU::getNamedOperandIdx(MI->getOpcode(),
                                                     AMDGPU::OpName::src1))
       printDefaultVccOperand(OpNo, STI, O);
     break;
   }

lib/Target/AMDGPU/VOP2Instructions.td

multiclass VOP2_Real_dpp8_gfx10_with_name<bits<6> op, string opName,
         VOP2_Pseudo ps = !cast<VOP2_Pseudo>(opName#"_e32");
         let AsmString = asmName # ps.Pfl.AsmDPP8;
         let DecoderNamespace = "DPP8";
       }
     }
   } // End DecoderNamespace = "SDWA10"
 
   //===------------------------------ VOP2be ------------------------------===//
-  multiclass VOP2be_Real_gfx10<bits<6> op, string opName, string asmName> {
+  multiclass VOP2be_Real_e32_gfx10<bits<6> op, string opName, string asmName> {
     def _e32_gfx10 :
       VOP2_Real<!cast<VOP2_Pseudo>(opName#"_e32"), SIEncodingFamily.GFX10>,
       VOP2e<op{5-0}, !cast<VOP2_Pseudo>(opName#"_e32").Pfl> {
         VOP2_Pseudo Ps = !cast<VOP2_Pseudo>(opName#"_e32");
         let AsmString = asmName # !subst(", vcc", "", Ps.AsmOperands);
       }
+  }
+  multiclass VOP2be_Real_e64_gfx10<bits<6> op, string opName, string asmName> {
     def _e64_gfx10 :
       VOP3_Real<!cast<VOP3_Pseudo>(opName#"_e64"), SIEncodingFamily.GFX10>,
       VOP3be_gfx10<{0, 1, 0, 0, op{5-0}},
                    !cast<VOP3_Pseudo>(opName#"_e64").Pfl> {
         VOP3_Pseudo Ps = !cast<VOP3_Pseudo>(opName#"_e64");
         let AsmString = asmName # Ps.AsmOperands;
       }
+  }
+  multiclass VOP2be_Real_sdwa_gfx10<bits<6> op, string opName, string asmName> {
     foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtSDWA9>.ret in
     def _sdwa_gfx10 :
       VOP_SDWA10_Real<!cast<VOP2_SDWA_Pseudo>(opName#"_sdwa")>,
       VOP2_SDWA9Ae<op{5-0}, !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa").Pfl> {
         VOP2_SDWA_Pseudo Ps = !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa");
         let AsmString = asmName # !subst(", vcc", "", Ps.AsmOperands);
         let DecoderNamespace = "SDWA10";
       }
-    foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtDPP>.ret in
-    def _dpp_gfx10 :
-      VOP2_DPP16<op, !cast<VOP2_DPP_Pseudo>(opName#"_dpp"), asmName> {
-        string AsmDPP = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP16;
-        let AsmString = asmName # !subst(", vcc", "", AsmDPP);
-        let DecoderNamespace = "SDWA10";
-      }
-    foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtDPP>.ret in
-    def _dpp8_gfx10 :
-      VOP2_DPP8<op, !cast<VOP2_Pseudo>(opName#"_e32"), asmName> {
-        string AsmDPP8 = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP8;
-        let AsmString = asmName # !subst(", vcc", "", AsmDPP8);
-        let DecoderNamespace = "DPP8";
-      }
-
-    let WaveSizePredicate = isWave32 in {
-      foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtSDWA9>.ret in
-      def _sdwa_w32_gfx10 :
-        Base_VOP_SDWA10_Real<!cast<VOP2_SDWA_Pseudo>(opName#"_sdwa")>,
-        VOP2_SDWA9Ae<op{5-0}, !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa").Pfl> {
-          VOP2_SDWA_Pseudo Ps = !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa");
-          let AsmString = asmName # !subst("vcc", "vcc_lo", Ps.AsmOperands);
-          let isAsmParserOnly = 1;
-          let DecoderNamespace = "SDWA10";
-        }
-      def _dpp_w32_gfx10 :
-        Base_VOP2_DPP16<op, !cast<VOP2_DPP_Pseudo>(opName#"_dpp"), asmName> {
-          string AsmDPP = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP16;
-          let AsmString = asmName # !subst("vcc", "vcc_lo", AsmDPP);
-          let isAsmParserOnly = 1;
-        }
-      def _dpp8_w32_gfx10 :
-        VOP2_DPP8<op, !cast<VOP2_Pseudo>(opName#"_e32"), asmName> {
-          string AsmDPP8 = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP8;
-          let AsmString = asmName # !subst("vcc", "vcc_lo", AsmDPP8);
-          let isAsmParserOnly = 1;
-        }
-    } // End WaveSizePredicate = isWave32
-
-    let WaveSizePredicate = isWave64 in {
-      foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtSDWA9>.ret in
-      def _sdwa_w64_gfx10 :
-        Base_VOP_SDWA10_Real<!cast<VOP2_SDWA_Pseudo>(opName#"_sdwa")>,
-        VOP2_SDWA9Ae<op{5-0}, !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa").Pfl> {
-          VOP2_SDWA_Pseudo Ps = !cast<VOP2_SDWA_Pseudo>(opName#"_sdwa");
-          let AsmString = asmName # Ps.AsmOperands;
-          let isAsmParserOnly = 1;
-          let DecoderNamespace = "SDWA10";
-        }
-      def _dpp_w64_gfx10 :
-        Base_VOP2_DPP16<op, !cast<VOP2_DPP_Pseudo>(opName#"_dpp"), asmName> {
-          string AsmDPP = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP16;
-          let AsmString = asmName # AsmDPP;
-          let isAsmParserOnly = 1;
-        }
-      def _dpp8_w64_gfx10 :
-        VOP2_DPP8<op, !cast<VOP2_Pseudo>(opName#"_e32"), asmName> {
-          string AsmDPP8 = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP8;
-          let AsmString = asmName # AsmDPP8;
-          let isAsmParserOnly = 1;
-        }
-    } // End WaveSizePredicate = isWave64
+        let WaveSizePredicate = isWave32;
+        let WaveSizePredicate = isWave64;
+      }
+  }
+  multiclass VOP2be_Real_dpp_gfx10<bits<6> op, string opName, string asmName> {
+    foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtDPP>.ret in
+    def _dpp_gfx10 :
+      VOP2_DPP16<op, !cast<VOP2_DPP_Pseudo>(opName#"_dpp"), asmName> {
+        string AsmDPP = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP16;
+        let AsmString = asmName # !subst(", vcc", "", AsmDPP);
+        let DecoderNamespace = "SDWA10";
+      }
+    def _dpp_w32_gfx10 :
+      Base_VOP2_DPP16<op, !cast<VOP2_DPP_Pseudo>(opName#"_dpp"), asmName> {
+        string AsmDPP = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP16;
+        let AsmString = asmName # !subst("vcc", "vcc_lo", AsmDPP);
+        let isAsmParserOnly = 1;
+        let WaveSizePredicate = isWave32;
+        let WaveSizePredicate = isWave64;
+      }
+  }
+  multiclass VOP2be_Real_dpp8_gfx10<bits<6> op, string opName, string asmName> {
+    foreach _ = BoolToList<!cast<VOP2_Pseudo>(opName#"_e32").Pfl.HasExtDPP>.ret in
+    def _dpp8_gfx10 :
+      VOP2_DPP8<op, !cast<VOP2_Pseudo>(opName#"_e32"), asmName> {
+        string AsmDPP8 = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP8;
+        let AsmString = asmName # !subst(", vcc", "", AsmDPP8);
+        let DecoderNamespace = "DPP8";
+      }
+    def _dpp8_w32_gfx10 :
+      VOP2_DPP8<op, !cast<VOP2_Pseudo>(opName#"_e32"), asmName> {
+        string AsmDPP8 = !cast<VOP2_Pseudo>(opName#"_e32").Pfl.AsmDPP8;
+        let AsmString = asmName # !subst("vcc", "vcc_lo", AsmDPP8);
+        let isAsmParserOnly = 1;
+        let WaveSizePredicate = isWave32;
+        let WaveSizePredicate = isWave64;
   }
 
   //===----------------------------- VOP3Only -----------------------------===//
   multiclass VOP3Only_Real_gfx10<bits<10> op> {
     def _e64_gfx10 :
       VOP3_Real<!cast<VOP3_Pseudo>(NAME#"_e64"), SIEncodingFamily.GFX10>,
       VOP3e_gfx10<op, !cast<VOP3_Pseudo>(NAME#"_e64").Pfl>;
   }
 
   //===---------------------------- VOP3beOnly ----------------------------===//
   multiclass VOP3beOnly_Real_gfx10<bits<10> op, string opName, string asmName> {
     def _e64_gfx10 :
       VOP3_Real<!cast<VOP3_Pseudo>(opName#"_e64"), SIEncodingFamily.GFX10>,
       VOP3be_gfx10<op, !cast<VOP3_Pseudo>(opName#"_e64").Pfl> {
         VOP3_Pseudo Ps = !cast<VOP3_Pseudo>(opName#"_e64");
         let AsmString = asmName # Ps.AsmOperands;
       }
   }
 } // End AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10"
 
-multiclass Base_VOP2_Real_gfx10<bits<6> op> :
-  VOP2_Real_e32_gfx10<op>, VOP2_Real_e64_gfx10<op>;
+multiclass VOP2be_Real_gfx10<bits<6> op, string opName, string asmName> :
+  VOP2be_Real_e32_gfx10<op, opName, asmName>,
+  VOP2be_Real_e64_gfx10<op, opName, asmName>,
+  VOP2be_Real_sdwa_gfx10<op, opName, asmName>,
+  VOP2be_Real_dpp_gfx10<op, opName, asmName>,
+  VOP2be_Real_dpp8_gfx10<op, opName, asmName>;
+
+multiclass VOP2e_Real_gfx10<bits<6> op, string opName, string asmName> :
+  VOP2_Real_e32_gfx10<op>,
+  VOP2_Real_e64_gfx10<op>,
+  VOP2be_Real_sdwa_gfx10<op, opName, asmName>,
+  VOP2be_Real_dpp_gfx10<op, opName, asmName>,
+  VOP2be_Real_dpp8_gfx10<op, opName, asmName>;
 
 multiclass VOP2_Real_gfx10<bits<6> op> :
   VOP2_Real_e32_gfx10<op>, VOP2_Real_e64_gfx10<op>,
   VOP2_Real_sdwa_gfx10<op>, VOP2_Real_dpp_gfx10<op>, VOP2_Real_dpp8_gfx10<op>;
 
 multiclass VOP2_Real_gfx10_with_name<bits<6> op, string opName,
                                      string asmName> :
   VOP2_Real_e32_gfx10_with_name<op, opName, asmName>,
   VOP2_Real_e64_gfx10_with_name<op, opName, asmName>,
   VOP2_Real_sdwa_gfx10_with_name<op, opName, asmName>,
   VOP2_Real_dpp_gfx10_with_name<op, opName, asmName>,
   VOP2_Real_dpp8_gfx10_with_name<op, opName, asmName>;
 
-defm V_CNDMASK_B32   : Base_VOP2_Real_gfx10<0x001>;
 defm V_XNOR_B32      : VOP2_Real_gfx10<0x01e>;
 defm V_FMAC_F32      : VOP2_Real_gfx10<0x02b>;
 defm V_FMAMK_F32     : VOP2Only_Real_MADK_gfx10<0x02c>;
 defm V_FMAAK_F32     : VOP2Only_Real_MADK_gfx10<0x02d>;
 defm V_ADD_F16       : VOP2_Real_gfx10<0x032>;
 defm V_SUB_F16       : VOP2_Real_gfx10<0x033>;
 defm V_SUBREV_F16    : VOP2_Real_gfx10<0x034>;
 defm V_MUL_F16       : VOP2_Real_gfx10<0x035>;

 // VOP2 carry-in, carry-out.
 defm V_ADD_CO_CI_U32 :
   VOP2be_Real_gfx10<0x028, "V_ADDC_U32", "v_add_co_ci_u32">;
 defm V_SUB_CO_CI_U32 :
   VOP2be_Real_gfx10<0x029, "V_SUBB_U32", "v_sub_co_ci_u32">;
 defm V_SUBREV_CO_CI_U32 :
   VOP2be_Real_gfx10<0x02a, "V_SUBBREV_U32", "v_subrev_co_ci_u32">;
 
+defm V_CNDMASK_B32 :
+  VOP2e_Real_gfx10<0x001, "V_CNDMASK_B32", "v_cndmask_b32">;
+
 // VOP3 only.
 defm V_BFM_B32            : VOP3Only_Real_gfx10<0x363>;
 defm V_BCNT_U32_B32       : VOP3Only_Real_gfx10<0x364>;
 defm V_MBCNT_LO_U32_B32   : VOP3Only_Real_gfx10<0x365>;
 defm V_MBCNT_HI_U32_B32   : VOP3Only_Real_gfx10<0x366>;
 defm V_LDEXP_F32          : VOP3Only_Real_gfx10<0x362>;
 defm V_CVT_PKNORM_I16_F32 : VOP3Only_Real_gfx10<0x368>;
 defm V_CVT_PKNORM_U16_F32 : VOP3Only_Real_gfx10<0x369>;

test/MC/AMDGPU/gfx10_asm_dpp8.s

 // GFX10: encoding: [0xea,0x04,0x0a,0x72,0x01,0x88,0xc6,0xfa]
 
 v_min_f16_dpp v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
 // GFX10: encoding: [0xea,0x04,0x0a,0x74,0x01,0x88,0xc6,0xfa]
 
 v_ldexp_f16_dpp v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
 // GFX10: encoding: [0xea,0x04,0x0a,0x76,0x01,0x88,0xc6,0xfa]
 
+v_cndmask_b32_dpp v0, v1, v2, vcc_lo dpp8:[7,6,5,4,3,2,1,0]
+// W32: v_cndmask_b32_dpp v0, v1, v2, vcc_lo  dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0xe9,0x04,0x00,0x02,0x01,0x77,0x39,0x05]
+// W64-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_dpp v0, v1, v2, vcc_lo dpp8:[0,1,2,3,4,5,6,7] fi:1
+// W32: v_cndmask_b32_dpp v0, v1, v2, vcc_lo  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+// W64-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_dpp v0, v1, v2, vcc dpp8:[7,6,5,4,3,2,1,0]
+// W64: v_cndmask_b32_dpp v0, v1, v2, vcc  dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0xe9,0x04,0x00,0x02,0x01,0x77,0x39,0x05]
+// W32-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_dpp v0, v1, v2, vcc dpp8:[0,1,2,3,4,5,6,7] fi:1
+// W64: v_cndmask_b32_dpp v0, v1, v2, vcc  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+// W32-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_dpp v0, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
+// W32: v_cndmask_b32_dpp v0, v1, v2, vcc_lo  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+// W64: v_cndmask_b32_dpp v0, v1, v2, vcc  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+
 v_add_co_ci_u32_dpp v0, vcc_lo, v0, v0, vcc_lo dpp8:[7,6,5,4,3,2,1,0]
 // W32: [0xe9,0x00,0x00,0x50,0x00,0x77,0x39,0x05]
 // W64-ERR: error: instruction not supported on this GPU
 
 v_sub_co_ci_u32_dpp v0, vcc_lo, v0, v0, vcc_lo dpp8:[7,6,5,4,3,2,1,0] fi:0
 // W32: [0xe9,0x00,0x00,0x52,0x00,0x77,0x39,0x05]
 // W64-ERR: error: instruction not supported on this GPU
 

test/MC/AMDGPU/wave32.s

 v_cndmask_b32_e32 v1, v2, v3, vcc_lo
 // GFX1032: v_cndmask_b32_e32 v1, v2, v3, vcc_lo ; encoding: [0x02,0x07,0x02,0x02]
 // GFX1064-ERR: error: instruction not supported on this GPU
 
 v_cndmask_b32_e32 v1, v2, v3, vcc
 // GFX1032-ERR: error: instruction not supported on this GPU
 // GFX1064: v_cndmask_b32_e32 v1, v2, v3, vcc ; encoding: [0x02,0x07,0x02,0x02]
 
+v_cndmask_b32_sdwa v5, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD
+// GFX1032: v_cndmask_b32_sdwa v5, v1, v2, vcc_lo  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x0a,0x02,0x01,0x16,0x06,0x06]
+// GFX1064: v_cndmask_b32_sdwa v5, v1, v2, vcc  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x0a,0x02,0x01,0x16,0x06,0x06]
+
+v_cndmask_b32_sdwa v5, v1, v2, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD
+// GFX1032: v_cndmask_b32_sdwa v5, v1, v2, vcc_lo  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x0a,0x02,0x01,0x16,0x06,0x06]
+// GFX1064-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_sdwa v5, v1, v2, vcc dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD
+// GFX1032-ERR: error: instruction not supported on this GPU
+// GFX1064: v_cndmask_b32_sdwa v5, v1, v2, vcc  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x0a,0x02,0x01,0x16,0x06,0x06]
+
+v_cndmask_b32_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+// GFX1032: v_cndmask_b32_dpp v5, v1, v2, vcc_lo  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x02,0x01,0xe4,0x00,0x00]
+// GFX1064: v_cndmask_b32_dpp v5, v1, v2, vcc  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x02,0x01,0xe4,0x00,0x00]
+
+v_cndmask_b32_dpp v5, v1, v2, vcc_lo quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+// GFX1032: v_cndmask_b32_dpp v5, v1, v2, vcc_lo  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x02,0x01,0xe4,0x00,0x00]
+// GFX1064-ERR: error: instruction not supported on this GPU
+
+v_cndmask_b32_dpp v5, v1, v2, vcc quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+// GFX1032-ERR: error: instruction not supported on this GPU
+// GFX1064: v_cndmask_b32_dpp v5, v1, v2, vcc  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x02,0x01,0xe4,0x00,0x00]
+
 v_add_co_u32_e32 v2, vcc_lo, s0, v2
 // GFX1032-ERR: error: instruction not supported on this GPU
 // GFX1064-ERR: error: instruction not supported on this GPU
 
 v_add_co_u32_e32 v2, vcc, s0, v2
 // GFX1032-ERR: error: instruction not supported on this GPU
 // GFX1064-ERR: error: instruction not supported on this GPU
 

test/MC/Disassembler/AMDGPU/gfx10_dasm_dpp16.txt

 0xfa,0x04,0x0a,0x3a,0x01,0x1b,0x00,0x00
 
 # GFX10: v_xnor_b32_dpp v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x00,0x00]
 0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x00,0x00
 
 # GFX10: v_mac_f32_dpp v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x3e,0x01,0x1b,0x00,0x00]
 0xfa,0x04,0x0a,0x3e,0x01,0x1b,0x00,0x00
 
+# W32: v_cndmask_b32_dpp v0, v1, v2, vcc_lo  dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0xe9,0x04,0x00,0x02,0x01,0x77,0x39,0x05]
+# W64: v_cndmask_b32_dpp v0, v1, v2, vcc  dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0xe9,0x04,0x00,0x02,0x01,0x77,0x39,0x05]
+0xe9,0x04,0x00,0x02,0x01,0x77,0x39,0x05
+
+# W32: v_cndmask_b32_dpp v0, v1, v2, vcc_lo  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+# W64: v_cndmask_b32_dpp v0, v1, v2, vcc  dpp8:[0,1,2,3,4,5,6,7] fi:1 ; encoding: [0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa]
+0xea,0x04,0x00,0x02,0x01,0x88,0xc6,0xfa
+
 # W32: v_add_co_ci_u32_dpp v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x50,0x01,0x1b,0x00,0x00]
 # W64: v_add_co_ci_u32_dpp v5, vcc, v1, v2, vcc quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x50,0x01,0x1b,0x00,0x00]
 0xfa,0x04,0x0a,0x50,0x01,0x1b,0x00,0x00
 
 # W32: v_sub_co_ci_u32_dpp v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x52,0x01,0x1b,0x00,0x00]
 # W64: v_sub_co_ci_u32_dpp v5, vcc, v1, v2, vcc quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x52,0x01,0x1b,0x00,0x00]
 0xfa,0x04,0x0a,0x52,0x01,0x1b,0x00,0x00
 

test/MC/Disassembler/AMDGPU/wave32.txt

 # GFX1032: v_cndmask_b32_e32 v5, 0, v2, vcc_lo
 # GFX1064: v_cndmask_b32_e32 v5, 0, v2, vcc ;
 0x80,0x04,0x0a,0x02
 
 # GFX1032: v_cndmask_b32_e32 v1, v2, v3, vcc_lo
 # GFX1064: v_cndmask_b32_e32 v1, v2, v3, vcc ;
 0x02,0x07,0x02,0x02
 
+# GFX1032: v_cndmask_b32_sdwa v5, v1, v2, vcc_lo  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD
+# GFX1064: v_cndmask_b32_sdwa v5, v1, v2, vcc  dst_sel:DWORD dst_unused:UNUSED_PRESERVE src0_sel:DWORD src1_sel:DWORD
+0xf9,0x04,0x0a,0x02,0x01,0x16,0x06,0x06
+
+# GFX1032: v_cndmask_b32_dpp v5, v1, v2, vcc_lo  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+# GFX1064: v_cndmask_b32_dpp v5, v1, v2, vcc  quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+0xfa,0x04,0x0a,0x02,0x01,0xe4,0x00,0x00
+
 # GFX1032: v_add_co_u32_e64 v2, vcc_lo, s0, v2
 # GFX1064: v_add_co_u32_e64 v2, vcc, s0, v2
 0x02,0x6a,0x0f,0xd7,0x00,0x04,0x02,0x00
 
 # GFX1032: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo
 # GFX1064: v_add_co_ci_u32_e32 v3, vcc, v3, v4, vcc ;
 0x03,0x09,0x06,0x50