[PowerPC] Add PowerPC population count, reversed load and store related builtins and instrinsics for XL compatibility

[NeHuang]

Summary

This patch is in a series of patches to provide builtins for compatibility
with the XL compiler. This patch adds the builtins and instrisics for population
count, reversed load and store related operations.

[nemanjai]

LGTM as long as the redundant clear is removed.

clang/lib/CodeGen/CGBuiltin.cpp

15202

Isn't this just Ops[0]?

llvm/lib/Target/PowerPC/PPCInstrInfo.td

5282

The clear is redundant. Why do we need to clear the bits that we won't store to begin with?

clang/include/clang/Basic/BuiltinsPPC.def

 #   define BUILTIN(ID, TYPES, ATTRS)
 #endif
 
 #define UNALIASED_CUSTOM_BUILTIN(ID, TYPES, ACCUMULATE) \
    CUSTOM_BUILTIN(ID, ID, TYPES, ACCUMULATE)
 
 // builtins for compatibility with the XL compiler
 BUILTIN(__builtin_ppc_popcntb, "ULiULi", "")
+BUILTIN(__builtin_ppc_poppar4, "iUi", "")
+BUILTIN(__builtin_ppc_poppar8, "iULLi", "")
 BUILTIN(__builtin_ppc_eieio, "v", "")
 BUILTIN(__builtin_ppc_iospace_eieio, "v", "")
 BUILTIN(__builtin_ppc_isync, "v", "")
 BUILTIN(__builtin_ppc_lwsync, "v", "")
 BUILTIN(__builtin_ppc_iospace_lwsync, "v", "")
 BUILTIN(__builtin_ppc_sync, "v", "")
 BUILTIN(__builtin_ppc_iospace_sync, "v", "")
 BUILTIN(__builtin_ppc_dcbfl, "vvC*", "")

 BUILTIN(__builtin_ppc_mulhwu, "UiUiUi", "")
 BUILTIN(__builtin_ppc_maddhd, "LLiLLiLLiLLi", "")
 BUILTIN(__builtin_ppc_maddhdu, "ULLiULLiULLiULLi", "")
 BUILTIN(__builtin_ppc_maddld, "LLiLLiLLiLLi", "")
 // Rotate
 BUILTIN(__builtin_ppc_rlwnm, "UiUiIUiIUi", "")
 BUILTIN(__builtin_ppc_rlwimi, "UiUiUiIUiIUi", "")
 BUILTIN(__builtin_ppc_rldimi, "ULLiULLiULLiIUiIULLi", "")
+// load
+BUILTIN(__builtin_ppc_load2r, "UiUs*", "")
+BUILTIN(__builtin_ppc_load4r, "UiUi*", "")
+BUILTIN(__builtin_ppc_load8r, "ULLiULLi*", "")
+// store
+BUILTIN(__builtin_ppc_store2r, "vUiUs*", "")
+BUILTIN(__builtin_ppc_store4r, "vUiUi*", "")
+BUILTIN(__builtin_ppc_store8r, "vULLiULLi*", "")
 
 BUILTIN(__builtin_ppc_get_timebase, "ULLi", "n")
 
 // This is just a placeholder, the types and attributes are wrong.
 BUILTIN(__builtin_altivec_vaddcuw, "V4UiV4UiV4Ui", "")
 
 BUILTIN(__builtin_altivec_vaddsbs, "V16ScV16ScV16Sc", "")
 BUILTIN(__builtin_altivec_vaddubs, "V16UcV16UcV16Uc", "")

clang/lib/Basic/Targets/PPC.cpp

for (const auto &Feature : Features) {
     // all.
   }
 
   return true;
 }
 
 static void defineXLCompatMacros(MacroBuilder &Builder) {
   Builder.defineMacro("__popcntb", "__builtin_ppc_popcntb");
+  Builder.defineMacro("__poppar4", "__builtin_ppc_poppar4");
+  Builder.defineMacro("__poppar8", "__builtin_ppc_poppar8");
   Builder.defineMacro("__eieio", "__builtin_ppc_eieio");
   Builder.defineMacro("__iospace_eieio", "__builtin_ppc_iospace_eieio");
   Builder.defineMacro("__isync", "__builtin_ppc_isync");
   Builder.defineMacro("__lwsync", "__builtin_ppc_lwsync");
   Builder.defineMacro("__iospace_lwsync", "__builtin_ppc_iospace_lwsync");
   Builder.defineMacro("__sync", "__builtin_ppc_sync");
   Builder.defineMacro("__iospace_sync", "__builtin_ppc_iospace_sync");
   Builder.defineMacro("__dcbfl", "__builtin_ppc_dcbfl");
static void defineXLCompatMacros(MacroBuilder &Builder) {
   Builder.defineMacro("__mulhw", "__builtin_ppc_mulhw");
   Builder.defineMacro("__mulhwu", "__builtin_ppc_mulhwu");
   Builder.defineMacro("__maddhd", "__builtin_ppc_maddhd");
   Builder.defineMacro("__maddhdu", "__builtin_ppc_maddhdu");
   Builder.defineMacro("__maddld", "__builtin_ppc_maddld");
   Builder.defineMacro("__rlwnm", "__builtin_ppc_rlwnm");
   Builder.defineMacro("__rlwimi", "__builtin_ppc_rlwimi");
   Builder.defineMacro("__rldimi", "__builtin_ppc_rldimi");
+  Builder.defineMacro("__load2r", "__builtin_ppc_load2r");
+  Builder.defineMacro("__load4r", "__builtin_ppc_load4r");
+  Builder.defineMacro("__load8r", "__builtin_ppc_load8r");
+  Builder.defineMacro("__store2r", "__builtin_ppc_store2r");
+  Builder.defineMacro("__store4r", "__builtin_ppc_store4r");
+  Builder.defineMacro("__store8r", "__builtin_ppc_store8r");
 }
 
 /// PPCTargetInfo::getTargetDefines - Return a set of the PowerPC-specific
 /// #defines that are not tied to a specific subtarget.
 void PPCTargetInfo::getTargetDefines(const LangOptions &Opts,
                                      MacroBuilder &Builder) const {
 
   defineXLCompatMacros(Builder);

clang/lib/CodeGen/CGBuiltin.cpp

Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
   // __rlwnm(rs, shift, mask)
   // rotl(rs, shift) & mask
   case PPC::BI__builtin_ppc_rlwnm: {
     llvm::Type *Ty = Ops[0]->getType();
     Function *F = CGM.getIntrinsic(Intrinsic::fshl, Ty);
     Value *Shift = Builder.CreateCall(F, {Ops[0], Ops[0], Ops[1]});
     return Builder.CreateAnd(Shift, Ops[2]);
   }
+  case PPC::BI__builtin_ppc_poppar4:
+  case PPC::BI__builtin_ppc_poppar8: {
+    llvm::Type *ArgType = Ops[0]->getType();
+    Function *F = CGM.getIntrinsic(Intrinsic::ctpop, ArgType);
+    Value *Tmp = Builder.CreateCall(F, Ops[0]);
+
+    llvm::Type *ResultType = ConvertType(E->getType());
+    Value *Result = Builder.CreateAnd(Tmp, llvm::ConstantInt::get(ArgType, 1));
+    if (Result->getType() != ResultType)
+      Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
+                                     "cast");
+    return Result;
+  }
+
   // Copy sign
   case PPC::BI__builtin_vsx_xvcpsgnsp:
   case PPC::BI__builtin_vsx_xvcpsgndp: {
     llvm::Type *ResultType = ConvertType(E->getType());
     Value *X = EmitScalarExpr(E->getArg(0));
     Value *Y = EmitScalarExpr(E->getArg(1));
     ID = Intrinsic::copysign;
     llvm::Function *F = CGM.getIntrinsic(ID, ResultType);

clang/lib/Sema/SemaChecking.cpp

static bool isPPC_64Builtin(unsigned BuiltinID) {
   case PPC::BI__builtin_ppc_trapd:
   case PPC::BI__builtin_ppc_cmpeqb:
   case PPC::BI__builtin_ppc_setb:
   case PPC::BI__builtin_ppc_mulhd:
   case PPC::BI__builtin_ppc_mulhdu:
   case PPC::BI__builtin_ppc_maddhd:
   case PPC::BI__builtin_ppc_maddhdu:
   case PPC::BI__builtin_ppc_maddld:
+  case PPC::BI__builtin_ppc_load8r:
+  case PPC::BI__builtin_ppc_store8r:
     return true;
   }
   return false;
 }
 
 static bool SemaFeatureCheck(Sema &S, CallExpr *TheCall,
                              StringRef FeatureToCheck, unsigned DiagID,
                              StringRef DiagArg = "") {

clang/test/CodeGen/builtins-ppc-xlcompat-load-store-reversed-64bit-only.c

+// REQUIRES: powerpc-registered-target.
+// RUN: %clang_cc1 -triple powerpc64-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc64le-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr8 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc64-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+// RUN: not %clang_cc1 -triple powerpc-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 2>&1 | FileCheck %s --check-prefix=CHECK-32-ERROR
+
+extern unsigned long long ull;
+extern unsigned long long *ull_addr;
+
+// CHECK-LABEL: @test_builtin_ppc_store8r(
+// CHECK:         [[TMP0:%.*]] = load i64, i64* @ull, align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load i64*, i64** @ull_addr, align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[TMP1]] to i8*
+// CHECK-NEXT:    call void @llvm.ppc.store8r(i64 [[TMP0]], i8* [[TMP2]])
+// CHECK-NEXT:    ret void
+
+// CHECK-32-ERROR: error: this builtin is only available on 64-bit targets
+void test_builtin_ppc_store8r() {
+  __builtin_ppc_store8r(ull, ull_addr);
+}
+
+// CHECK-LABEL: @test_builtin_ppc_load8r(
+// CHECK:         [[TMP0:%.*]] = load i64*, i64** @ull_addr, align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[TMP0]] to i8*
+// CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.ppc.load8r(i8* [[TMP1]])
+// CHECK-NEXT:    ret i64 [[TMP2]]
+
+// CHECK-32-ERROR: error: this builtin is only available on 64-bit targets
+unsigned long long test_builtin_ppc_load8r() {
+  return __builtin_ppc_load8r(ull_addr);
+}

clang/test/CodeGen/builtins-ppc-xlcompat-load-store-reversed.c

+// REQUIRES: powerpc-registered-target.
+// RUN: %clang_cc1 -triple powerpc64-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc64le-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr8 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s --check-prefixes=CHECK-32B
+// RUN: %clang_cc1 -triple powerpc64-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+
+extern unsigned short us;
+extern unsigned int ui;
+extern unsigned short *us_addr;
+extern unsigned int *ui_addr;
+
+// CHECK-LABEL: @test_builtin_ppc_store2r(
+// CHECK:         [[TMP0:%.*]] = load i16, i16* @us, align 2
+// CHECK-NEXT:    [[CONV:%.*]] = zext i16 [[TMP0]] to i32
+// CHECK-NEXT:    [[TMP1:%.*]] = load i16*, i16** @us_addr, align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-NEXT:    call void @llvm.ppc.store2r(i32 [[CONV]], i8* [[TMP2]])
+// CHECK-NEXT:    ret void
+//
+// CHECK-32B-LABEL: @test_builtin_ppc_store2r(
+// CHECK-32B:         [[TMP0:%.*]] = load i16, i16* @us, align 2
+// CHECK-32B-NEXT:    [[CONV:%.*]] = zext i16 [[TMP0]] to i32
+// CHECK-32B-NEXT:    [[TMP1:%.*]] = load i16*, i16** @us_addr, align 4
+// CHECK-32B-NEXT:    [[TMP2:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-32B-NEXT:    call void @llvm.ppc.store2r(i32 [[CONV]], i8* [[TMP2]])
+// CHECK-32B-NEXT:    ret void
+//
+void test_builtin_ppc_store2r() {
+  __builtin_ppc_store2r(us, us_addr);
+}
+
+// CHECK-LABEL: @test_builtin_ppc_store4r(
+// CHECK:         [[TMP0:%.*]] = load i32, i32* @ui, align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = load i32*, i32** @ui_addr, align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to i8*
+// CHECK-NEXT:    call void @llvm.ppc.store4r(i32 [[TMP0]], i8* [[TMP2]])
+// CHECK-NEXT:    ret void
+//
+// CHECK-32B-LABEL: @test_builtin_ppc_store4r(
+// CHECK-32B:         [[TMP0:%.*]] = load i32, i32* @ui, align 4
+// CHECK-32B-NEXT:    [[TMP1:%.*]] = load i32*, i32** @ui_addr, align 4
+// CHECK-32B-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to i8*
+// CHECK-32B-NEXT:    call void @llvm.ppc.store4r(i32 [[TMP0]], i8* [[TMP2]])
+// CHECK-32B-NEXT:    ret void
+//
+void test_builtin_ppc_store4r() {
+  __builtin_ppc_store4r(ui, ui_addr);
+}
+
+// CHECK-LABEL: @test_builtin_ppc_load2r(
+// CHECK:         [[TMP0:%.*]] = load i16*, i16** @us_addr, align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[TMP0]] to i8*
+// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.load2r(i8* [[TMP1]])
+// CHECK-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP2]] to i16
+// CHECK-NEXT:    ret i16 [[CONV]]
+//
+// CHECK-32B-LABEL: @test_builtin_ppc_load2r(
+// CHECK-32B:         [[TMP0:%.*]] = load i16*, i16** @us_addr, align 4
+// CHECK-32B-NEXT:    [[TMP1:%.*]] = bitcast i16* [[TMP0]] to i8*
+// CHECK-32B-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.load2r(i8* [[TMP1]])
+// CHECK-32B-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP2]] to i16
+// CHECK-32B-NEXT:    ret i16 [[CONV]]
+//
+unsigned short test_builtin_ppc_load2r() {
+  return __builtin_ppc_load2r(us_addr);
+}
+
+// CHECK-LABEL: @test_builtin_ppc_load4r(
+// CHECK:         [[TMP0:%.*]] = load i32*, i32** @ui_addr, align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.load4r(i8* [[TMP1]])
+// CHECK-NEXT:    ret i32 [[TMP2]]
+//
+// CHECK-32B-LABEL: @test_builtin_ppc_load4r(
+// CHECK-32B:         [[TMP0:%.*]] = load i32*, i32** @ui_addr, align 4
+// CHECK-32B-NEXT:    [[TMP1:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-32B-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.load4r(i8* [[TMP1]])
+// CHECK-32B-NEXT:    ret i32 [[TMP2]]
+//
+unsigned int test_builtin_ppc_load4r() {
+  return __builtin_ppc_load4r(ui_addr);
+}

clang/test/CodeGen/builtins-ppc-xlcompat-popcnt.c

+// REQUIRES: powerpc-registered-target.
+// RUN: %clang_cc1 -triple powerpc64-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc64le-unknown-unknown \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr8 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+// RUN: %clang_cc1 -triple powerpc64-unknown-aix \
+// RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
+
+extern unsigned int ui;
+extern unsigned long long ull;
+
+// CHECK-LABEL: @test_builtin_ppc_poppar4(
+// CHECK:         [[TMP0:%.*]] = load i32, i32* @ui, align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* @ui, align 4
+// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ctpop.i32(i32 [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = and i32 [[TMP2]], 1
+// CHECK-NEXT:    ret i32 [[TMP3]]
+//
+int test_builtin_ppc_poppar4() {
+ return __builtin_ppc_poppar4(ui);
+}
+
+// CHECK-LABEL: @test_builtin_ppc_poppar8(
+// CHECK:         [[TMP0:%.*]] = load i64, i64* @ull, align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load i64, i64* @ull, align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = and i64 [[TMP2]], 1
+// CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[TMP3]] to i32
+// CHECK-NEXT:    ret i32 [[CAST]]
+//
+int test_builtin_ppc_poppar8() {
+ return __builtin_ppc_poppar8(ull);
+}

llvm/include/llvm/IR/IntrinsicsPowerPC.td

def int_ppc_maddhd
       : GCCBuiltin<"__builtin_ppc_maddhd">,
         Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty, llvm_i64_ty], [IntrNoMem]>;
   def int_ppc_maddhdu
       : GCCBuiltin<"__builtin_ppc_maddhdu">,
         Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty, llvm_i64_ty], [IntrNoMem]>;
   def int_ppc_maddld
       : GCCBuiltin<"__builtin_ppc_maddld">,
         Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty, llvm_i64_ty], [IntrNoMem]>;
+  // load
+  def int_ppc_load2r
+      : GCCBuiltin<"__builtin_ppc_load2r">,
+        Intrinsic<[llvm_i32_ty], [llvm_ptr_ty], [IntrReadMem, IntrArgMemOnly]>;
+  def int_ppc_load4r
+      : GCCBuiltin<"__builtin_ppc_load4r">,
+        Intrinsic<[llvm_i32_ty], [llvm_ptr_ty], [IntrReadMem, IntrArgMemOnly]>;
+  def int_ppc_load8r
+      : GCCBuiltin<"__builtin_ppc_load8r">,
+        Intrinsic<[llvm_i64_ty], [llvm_ptr_ty], [IntrReadMem, IntrArgMemOnly]>;
+  // store
+  def int_ppc_store2r
+      : GCCBuiltin<"__builtin_ppc_store2r">,
+        Intrinsic<[], [llvm_i32_ty, llvm_ptr_ty], [IntrWriteMem]>;
+  def int_ppc_store4r
+      : GCCBuiltin<"__builtin_ppc_store4r">,
+        Intrinsic<[], [llvm_i32_ty, llvm_ptr_ty], [IntrWriteMem]>;
+  def int_ppc_store8r
+      : GCCBuiltin<"__builtin_ppc_store8r">,
+        Intrinsic<[], [llvm_i64_ty, llvm_ptr_ty], [IntrWriteMem]>;
 }
 
 //===----------------------------------------------------------------------===//
 // PowerPC Atomic Intrinsic Definitions.
 let TargetPrefix = "ppc" in {
   class AtomicRMW128Intrinsic
     : Intrinsic<[llvm_i64_ty, llvm_i64_ty],
                 [llvm_ptr_ty, llvm_i64_ty, llvm_i64_ty],

llvm/lib/Target/PowerPC/PPCInstr64Bit.td

def : Pat<(i64 (int_ppc_maddld g8rc:$a, g8rc:$b, g8rc:$c)),
           (i64 (MADDLD8 $a, $b, $c))>;
 }
 
 let Predicates = [In64BitMode] in {
 def : Pat<(i64 (int_ppc_mulhd g8rc:$a, g8rc:$b)),
           (i64 (MULHD $a, $b))>;
 def : Pat<(i64 (int_ppc_mulhdu g8rc:$a, g8rc:$b)),
           (i64 (MULHDU $a, $b))>;
+def : Pat<(int_ppc_load8r ForceXForm:$ptr),
+          (LDBRX ForceXForm:$ptr)>;
+def : Pat<(int_ppc_store8r g8rc:$a, ForceXForm:$ptr),
+          (STDBRX g8rc:$a, ForceXForm:$ptr)>;
 }
 
 let Predicates = [IsISA3_0] in {
 // DARN (deliver random number)
 // L=0 for 32-bit, L=1 for conditioned random, L=2 for raw random
 def : Pat<(int_ppc_darn32), (EXTRACT_SUBREG (DARN 0), sub_32)>;
 def : Pat<(int_ppc_darn), (DARN 1)>;
 def : Pat<(int_ppc_darnraw), (DARN 2)>;

llvm/lib/Target/PowerPC/PPCInstrInfo.td

 def : Pat<(i32 (int_ppc_cmprb i32:$a, gprc:$b, gprc:$c)),
           (i32 (SETB (CMPRB u1imm:$a, $b, $c)))>;
 }
 def : Pat<(i32 (int_ppc_mulhw gprc:$a, gprc:$b)),
           (i32 (MULHW $a, $b))>;
 def : Pat<(i32 (int_ppc_mulhwu gprc:$a, gprc:$b)),
           (i32 (MULHWU $a, $b))>;
 
+def : Pat<(int_ppc_load2r ForceXForm:$ptr),
+          (LHBRX ForceXForm:$ptr)>;
+def : Pat<(int_ppc_load4r ForceXForm:$ptr),
+          (LWBRX ForceXForm:$ptr)>;
+def : Pat<(int_ppc_store2r gprc:$a, ForceXForm:$ptr),
+          (STHBRX gprc:$a, ForceXForm:$ptr)>;
+def : Pat<(int_ppc_store4r gprc:$a, ForceXForm:$ptr),
+          (STWBRX gprc:$a, ForceXForm:$ptr)>;
+
+
 // Fast 32-bit reverse bits algorithm:
 // Step 1: 1-bit swap (swap odd 1-bit and even 1-bit):
 // n = ((n >> 1) & 0x55555555) | ((n << 1) & 0xAAAAAAAA);
 // Step 2: 2-bit swap (swap odd 2-bit and even 2-bit):
 // n = ((n >> 2) & 0x33333333) | ((n << 2) & 0xCCCCCCCC);
 // Step 3: 4-bit swap (swap odd 4-bit and even 4-bit):
 // n = ((n >> 4) & 0x0F0F0F0F) | ((n << 4) & 0xF0F0F0F0);
 // Step 4: byte reverse (Suppose n = [B1,B2,B3,B4]):

llvm/test/CodeGen/PowerPC/builtins-ppc-xlcompat-load-store-reversed-64bit-only.ll

+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu \
+; RUN:   -mcpu=pwr8 < %s | FileCheck %s
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-linux-gnu \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-aix \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s
+
+@ull = external global i64, align 8
+@ull_addr = external global i64*, align 8
+
+define dso_local void @test_builtin_ppc_store8r() {
+; CHECK-LABEL: test_builtin_ppc_store8r:
+; CHECK:         stdbrx 3, 0, 4
+; CHECK-NEXT:    blr
+;
+entry:
+  %0 = load i64, i64* @ull, align 8
+  %1 = load i64*, i64** @ull_addr, align 8
+  %2 = bitcast i64* %1 to i8*
+  call void @llvm.ppc.store8r(i64 %0, i8* %2)
+  ret void
+}
+
+declare void @llvm.ppc.store8r(i64, i8*)
+
+define dso_local i64 @test_builtin_ppc_load8r() {
+; CHECK-LABEL: test_builtin_ppc_load8r:
+; CHECK:         ldbrx 3, 0, 3
+; CHECK-NEXT:    blr
+entry:
+  %0 = load i64*, i64** @ull_addr, align 8
+  %1 = bitcast i64* %0 to i8*
+  %2 = call i64 @llvm.ppc.load8r(i8* %1)
+  ret i64 %2
+}
+
+declare i64 @llvm.ppc.load8r(i8*)

llvm/test/CodeGen/PowerPC/builtins-ppc-xlcompat-load-store-reversed.ll

+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu \
+; RUN:   -mcpu=pwr8 < %s | FileCheck %s --check-prefix=CHECK-64B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-linux-gnu \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-64B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc-unknown-aix \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-32B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-aix \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-64B
+
+@us = external global i16, align 2
+@us_addr = external global i16*, align 8
+@ui = external global i32, align 4
+@ui_addr = external global i32*, align 8
+
+define dso_local void @test_builtin_ppc_store2r() {
+; CHECK-64B-LABEL: test_builtin_ppc_store2r:
+; CHECK-64B:         sthbrx 3, 0, 4
+; CHECK-64B-NEXT:    blr
+
+; CHECK-32B-LABEL: test_builtin_ppc_store2r:
+; CHECK-32B:         sthbrx 3, 0, 4
+; CHECK-32B-NEXT:    blr
+entry:
+  %0 = load i16, i16* @us, align 2
+  %conv = zext i16 %0 to i32
+  %1 = load i16*, i16** @us_addr, align 8
+  %2 = bitcast i16* %1 to i8*
+  call void @llvm.ppc.store2r(i32 %conv, i8* %2)
+  ret void
+}
+
+declare void @llvm.ppc.store2r(i32, i8*)
+
+define dso_local void @test_builtin_ppc_store4r() {
+; CHECK-64B-LABEL: test_builtin_ppc_store4r:
+; CHECK-64B:         stwbrx 3, 0, 4
+; CHECK-64B-NEXT:    blr
+
+; CHECK-32B-LABEL: test_builtin_ppc_store4r:
+; CHECK-32B:         stwbrx 3, 0, 4
+; CHECK-32B-NEXT:    blr
+entry:
+  %0 = load i32, i32* @ui, align 4
+  %1 = load i32*, i32** @ui_addr, align 8
+  %2 = bitcast i32* %1 to i8*
+  call void @llvm.ppc.store4r(i32 %0, i8* %2)
+  ret void
+}
+
+declare void @llvm.ppc.store4r(i32, i8*)
+
+define dso_local zeroext i16 @test_builtin_ppc_load2r() {
+; CHECK-64B-LABEL: test_builtin_ppc_load2r:
+; CHECK-64B:         lhbrx 3, 0, 3
+; CHECK-64B-NEXT:    clrldi 3, 3, 48
+; CHECK-64B-NEXT:    blr
+
+; CHECK-32B-LABEL: test_builtin_ppc_load2r:
+; CHECK-32B:         lhbrx 3, 0, 3
+; CHECK-32B-NEXT:    clrlwi 3, 3, 16
+; CHECK-32B-NEXT:    blr
+entry:
+  %0 = load i16*, i16** @us_addr, align 8
+  %1 = bitcast i16* %0 to i8*
+  %2 = call i32 @llvm.ppc.load2r(i8* %1)
+  %conv = trunc i32 %2 to i16
+  ret i16 %conv
+}
+
+declare i32 @llvm.ppc.load2r(i8*)
+
+define dso_local zeroext i32 @test_builtin_ppc_load4r() {
+; CHECK-64B-LABEL: test_builtin_ppc_load4r:
+; CHECK-64B:         lwbrx 3, 0, 3
+; CHECK-64B-NEXT:    blr
+
+; CHECK-32B-LABEL: test_builtin_ppc_load4r:
+; CHECK-32B:         lwbrx 3, 0, 3
+; CHECK-32B-NEXT:    blr
+entry:
+  %0 = load i32*, i32** @ui_addr, align 8
+  %1 = bitcast i32* %0 to i8*
+  %2 = call i32 @llvm.ppc.load4r(i8* %1)
+  ret i32 %2
+}
+
+declare i32 @llvm.ppc.load4r(i8*)

llvm/test/CodeGen/PowerPC/builtins-ppc-xlcompat-popcnt.ll

+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu \
+; RUN:   -mcpu=pwr8 < %s | FileCheck %s --check-prefix=CHECK-64B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-linux-gnu \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-64B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc-unknown-aix \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-32B
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-aix \
+; RUN:   -mcpu=pwr7 < %s | FileCheck %s --check-prefix=CHECK-64B
+
+@ui = external global i32, align 4
+@ull = external global i64, align 8
+
+define dso_local signext i32 @test_builtin_ppc_poppar4() {
+; CHECK-32B-LABEL: test_builtin_ppc_poppar4:
+; CHECK-32B:         popcntw 3, 3
+; CHECK-32B-NEXT:    clrlwi 3, 3, 31
+; CHECK-32B-NEXT:    blr
+; CHECK-64B-LABEL: test_builtin_ppc_poppar4:
+; CHECK-64B:         popcntw 3, 3
+; CHECK-64B-NEXT:    clrlwi 3, 3, 31
+; CHECK-64B-NEXT:    blr
+entry:
+  %0 = load i32, i32* @ui, align 4
+  %1 = load i32, i32* @ui, align 4
+  %2 = call i32 @llvm.ctpop.i32(i32 %1)
+  %3 = and i32 %2, 1
+  ret i32 %3
+}
+
+declare i32 @llvm.ctpop.i32(i32)
+
+define dso_local signext i32 @test_builtin_ppc_poppar8() {
+; CHECK-32B-LABEL: test_builtin_ppc_poppar8:
+; CHECK-32B:         xor 3, 3, 4
+; CHECK-32B-NEXT:    popcntw 3, 3
+; CHECK-32B-NEXT:    clrlwi 3, 3, 31
+; CHECK-32B-NEXT:    blr
+; CHECK-64B-LABEL: test_builtin_ppc_poppar8:
+; CHECK-64B:         popcntd 3, 3
+; CHECK-64B-NEXT:    clrldi 3, 3, 63
+; CHECK-64B-NEXT:    blr
+entry:
+  %0 = load i64, i64* @ull, align 8
+  %1 = load i64, i64* @ull, align 8
+  %2 = call i64 @llvm.ctpop.i64(i64 %1)
+  %3 = and i64 %2, 1
+  %cast = trunc i64 %3 to i32
+  ret i32 %cast
+}
+
+declare i64 @llvm.ctpop.i64(i64)