[PowerPC] Add support for intrinsics dcbfps and dcbstps in P10.

Summary: This patch added support for the intrinsics llvm.ppc.dcbfps and llvm.ppc.dcbstps.
dcbfps and dcbstps are actually extended mnemonics of dcbf.
dcbfps RA,RB ---> dcbf RA,RB,4
dcbstps RA,RB ---> dcbf RA,RB,6

Reviewed By: amyk, steven.zhang

Differential Revision: https://reviews.llvm.org/D91323

llvm/include/llvm/IR/IntrinsicsPowerPC.td

@@ -18,10 +18,12 @@
 let TargetPrefix = "ppc" in {  // All intrinsics start with "llvm.ppc.".
   // dcba/dcbf/dcbi/dcbst/dcbt/dcbz/dcbzl(PPC970) instructions.
   def int_ppc_dcba  : Intrinsic<[], [llvm_ptr_ty], []>;
-  def int_ppc_dcbf  : GCCBuiltin<"__builtin_dcbf">,
-                      Intrinsic<[], [llvm_ptr_ty], []>;
-  def int_ppc_dcbfl : Intrinsic<[], [llvm_ptr_ty], []>;
-  def int_ppc_dcbflp: Intrinsic<[], [llvm_ptr_ty], []>;
+  def int_ppc_dcbf : GCCBuiltin<"__builtin_dcbf">,
+                      Intrinsic<[], [llvm_ptr_ty], [IntrArgMemOnly]>;
+  def int_ppc_dcbfl : Intrinsic<[], [llvm_ptr_ty], [IntrArgMemOnly]>;
+  def int_ppc_dcbflp : Intrinsic<[], [llvm_ptr_ty], [IntrArgMemOnly]>;
+  def int_ppc_dcbfps : Intrinsic<[], [llvm_ptr_ty], [IntrArgMemOnly]>;
+  def int_ppc_dcbstps : Intrinsic<[], [llvm_ptr_ty], [IntrArgMemOnly]>;
   def int_ppc_dcbi  : Intrinsic<[], [llvm_ptr_ty], []>;
   def int_ppc_dcbst : Intrinsic<[], [llvm_ptr_ty], []>;
   def int_ppc_dcbt  : Intrinsic<[], [llvm_ptr_ty],

llvm/lib/Target/PowerPC/AsmParser/PPCAsmParser.cpp

@@ -774,12 +774,18 @@ void PPCAsmParser::ProcessInstruction(MCInst &Inst,
   }
   case PPC::DCBFx:
   case PPC::DCBFL:
-  case PPC::DCBFLP: {
+  case PPC::DCBFLP:
+  case PPC::DCBFPS:
+  case PPC::DCBSTPS: {
     int L = 0;
     if (Opcode == PPC::DCBFL)
       L = 1;
     else if (Opcode == PPC::DCBFLP)
       L = 3;
+    else if (Opcode == PPC::DCBFPS)
+      L = 4;
+    else if (Opcode == PPC::DCBSTPS)
+      L = 6;
 
     MCInst TmpInst;
     TmpInst.setOpcode(PPC::DCBF);

llvm/lib/Target/PowerPC/MCTargetDesc/PPCInstPrinter.cpp

@@ -184,12 +184,18 @@ void PPCInstPrinter::printInst(const MCInst *MI, uint64_t Address,
 
   if (MI->getOpcode() == PPC::DCBF) {
     unsigned char L = MI->getOperand(0).getImm();
-    if (!L || L == 1 || L == 3) {
-      O << "\tdcbf";
-      if (L == 1 || L == 3)
+    if (!L || L == 1 || L == 3 || L == 4 || L == 6) {
+      O << "\tdcb";
+      if (L != 6)
+        O << "f";
+      if (L == 1)
         O << "l";
       if (L == 3)
-        O << "p";
+        O << "lp";
+      if (L == 4)
+        O << "ps";
+      if (L == 6)
+        O << "stps";
       O << " ";
 
       printOperand(MI, 1, O);

llvm/lib/Target/PowerPC/PPCInstrInfo.td

@@ -4609,6 +4609,16 @@ def : Pat<(int_ppc_dcbfl xoaddr:$dst),
 def : Pat<(int_ppc_dcbflp xoaddr:$dst),
           (DCBF 3, xoaddr:$dst)>;
 
+let Predicates = [IsISA3_1] in {
+  def DCBFPS  : PPCAsmPseudo<"dcbfps $dst", (ins memrr:$dst)>;
+  def DCBSTPS : PPCAsmPseudo<"dcbstps $dst", (ins memrr:$dst)>;
+
+  def : Pat<(int_ppc_dcbfps xoaddr:$dst),
+            (DCBF 4, xoaddr:$dst)>;
+  def : Pat<(int_ppc_dcbstps xoaddr:$dst),
+            (DCBF 6, xoaddr:$dst)>;
+}
+
 def : InstAlias<"crset $bx", (CREQV crbitrc:$bx, crbitrc:$bx, crbitrc:$bx)>;
 def : InstAlias<"crclr $bx", (CRXOR crbitrc:$bx, crbitrc:$bx, crbitrc:$bx)>;
 def : InstAlias<"crmove $bx, $by", (CROR crbitrc:$bx, crbitrc:$by, crbitrc:$by)>;

llvm/test/CodeGen/PowerPC/dcbf-p10.ll

@@ -0,0 +1,37 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu < %s \
+; RUN:     -verify-machineinstrs -ppc-asm-full-reg-names \
+; RUN:     -ppc-vsr-nums-as-vr -mcpu=pwr10 | FileCheck %s
+; RUN: llc -mtriple=powerpc64-unknown-linux-gnu < %s \
+; RUN:     -verify-machineinstrs -ppc-asm-full-reg-names \
+; RUN:     -ppc-vsr-nums-as-vr -mcpu=pwr10 | FileCheck %s
+
+; Function Attrs: nounwind
+define void @dcbfps_test(i8* %a) {
+; CHECK-LABEL: dcbfps_test:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi r3, r3, 3
+; CHECK-NEXT:    dcbfps 0, r3
+; CHECK-NEXT:    blr
+entry:
+  %add.a = getelementptr inbounds i8, i8* %a, i64 3
+  tail call void @llvm.ppc.dcbfps(i8* %add.a)
+ret void
+}
+
+declare void @llvm.ppc.dcbfps(i8*)
+
+; Function Attrs: nounwind
+define void @dcbstps_test(i8* %a) {
+; CHECK-LABEL: dcbstps_test:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi r3, r3, 3
+; CHECK-NEXT:    dcbstps 0, r3
+; CHECK-NEXT:    blr
+entry:
+  %add.a = getelementptr inbounds i8, i8* %a, i64 3
+  tail call void @llvm.ppc.dcbstps(i8* %add.a)
+ret void
+}
+
+declare void @llvm.ppc.dcbstps(i8*)