AMDGPU/SI: Fix more cases of losing exec operands

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@247230 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/AMDGPU/SIInstrFormats.td

@@ -611,15 +611,13 @@ class VINTRPCommon <dag outs, dag ins, string asm, list<dag> pattern> :
 // Vector I/O operations
 //===----------------------------------------------------------------------===//
 
-let Uses = [EXEC] in {
-
 class DS <dag outs, dag ins, string asm, list<dag> pattern> :
     InstSI <outs, ins, asm, pattern> {
 
   let LGKM_CNT = 1;
   let DS = 1;
   let UseNamedOperandTable = 1;
-  let Uses = [M0];
+  let Uses = [M0, EXEC];
 
   // Most instruction load and store data, so set this as the default.
   let mayLoad = 1;
@@ -636,6 +634,7 @@ class MUBUF <dag outs, dag ins, string asm, list<dag> pattern> :
   let VM_CNT = 1;
   let EXP_CNT = 1;
   let MUBUF = 1;
+  let Uses = [EXEC];
 
   let hasSideEffects = 0;
   let UseNamedOperandTable = 1;
@@ -649,6 +648,7 @@ class MTBUF <dag outs, dag ins, string asm, list<dag> pattern> :
   let VM_CNT = 1;
   let EXP_CNT = 1;
   let MTBUF = 1;
+  let Uses = [EXEC];
 
   let hasSideEffects = 0;
   let UseNamedOperandTable = 1;
@@ -678,9 +678,7 @@ class MIMG <bits<7> op, dag outs, dag ins, string asm, list<dag> pattern> :
   let VM_CNT = 1;
   let EXP_CNT = 1;
   let MIMG = 1;
+  let Uses = [EXEC];
 
   let hasSideEffects = 0; // XXX ????
 }
-
-
-} // End Uses = [EXEC]

lib/Target/AMDGPU/SIInstrInfo.cpp

@@ -82,6 +82,7 @@ bool SIInstrInfo::isReallyTriviallyReMaterializable(const MachineInstr *MI,
   switch (MI->getOpcode()) {
   case AMDGPU::V_MOV_B32_e32:
   case AMDGPU::V_MOV_B32_e64:
+  case AMDGPU::V_MOV_B64_PSEUDO:
     return true;
   default:
     return false;
@@ -996,9 +997,6 @@ bool SIInstrInfo::FoldImmediate(MachineInstr *UseMI, MachineInstr *DefMI,
           AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src2));
       }
 
-      UseMI->RemoveOperand(AMDGPU::getNamedOperandIdx(Opc,
-                                                      AMDGPU::OpName::src2));
-      // ChangingToImmediate adds Src2 back to the instruction.
       Src2->ChangeToImmediate(Imm);
 
       removeModOperands(*UseMI);

lib/Target/AMDGPU/SIInstructions.td

@@ -1367,7 +1367,7 @@ defm V_RSQ_CLAMP_F64 : VOP1InstSI <vop1<0x32>, "v_rsq_clamp_f64",
 // VINTRP Instructions
 //===----------------------------------------------------------------------===//
 
-let Uses = [M0] in {
+let Uses = [M0, EXEC] in {
 
 // FIXME: Specify SchedRW for VINTRP insturctions.
 
@@ -1412,7 +1412,7 @@ defm V_INTERP_MOV_F32 : VINTRP_m <
   [(set f32:$dst, (AMDGPUinterp_mov (i32 imm:$src0), (i32 imm:$attr_chan),
                                     (i32 imm:$attr)))]>;
 
-} // End Uses = [M0]
+} // End Uses = [M0, EXEC]
 
 //===----------------------------------------------------------------------===//
 // VOP2 Instructions
@@ -1769,7 +1769,7 @@ let SchedRW = [WriteDouble, WriteSALU] in {
 defm V_DIV_SCALE_F64 : VOP3b_64 <vop3<0x16e, 0x1e1>, "v_div_scale_f64", []>;
 } // let SchedRW = [WriteDouble]
 
-let isCommutable = 1, Uses = [VCC] in {
+let isCommutable = 1, Uses = [VCC, EXEC] in {
 
 let SchedRW = [WriteFloatFMA] in {
 // v_div_fmas_f32:
@@ -1793,7 +1793,7 @@ defm V_DIV_FMAS_F64 : VOP3_VCC_Inst <vop3<0x170, 0x1e3>, "v_div_fmas_f64",
 >;
 
 } // End SchedRW = [WriteDouble]
-} // End isCommutable = 1
+} // End isCommutable = 1, Uses = [VCC, EXEC]
 
 //def V_MSAD_U8 : VOP3_U8 <0x00000171, "v_msad_u8", []>;
 //def V_QSAD_U8 : VOP3_U8 <0x00000172, "v_qsad_u8", []>;
@@ -1842,7 +1842,7 @@ def V_CNDMASK_B64_PSEUDO : VOP3Common <(outs VReg_64:$dst),
   (ins VSrc_64:$src0, VSrc_64:$src1, SSrc_64:$src2), "", []
 >;
 
-let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in {
+let hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC] in {
 // 64-bit vector move instruction.  This is mainly used by the SIFoldOperands
 // pass to enable folding of inline immediates.
 def V_MOV_B64_PSEUDO : InstSI <(outs VReg_64:$dst), (ins VSrc_64:$src0), "", []>;
@@ -1984,7 +1984,7 @@ def SI_INDIRECT_DST_V16 : SI_INDIRECT_DST<VReg_512>;
 
 multiclass SI_SPILL_SGPR <RegisterClass sgpr_class> {
 
-  let UseNamedOperandTable = 1 in {
+  let UseNamedOperandTable = 1, Uses = [EXEC] in {
     def _SAVE : InstSI <
       (outs),
       (ins sgpr_class:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,
@@ -2016,7 +2016,7 @@ defm SI_SPILL_S256 : SI_SPILL_SGPR <SReg_256>;
 defm SI_SPILL_S512 : SI_SPILL_SGPR <SReg_512>;
 
 multiclass SI_SPILL_VGPR <RegisterClass vgpr_class> {
-  let UseNamedOperandTable = 1, VGPRSpill = 1 in {
+  let UseNamedOperandTable = 1, VGPRSpill = 1, Uses = [EXEC] in {
     def _SAVE : InstSI <
       (outs),
       (ins vgpr_class:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,