[ARM][AArch64] Add ARM specific builtin for clz that is not undefined…

… for 0 in ubsan.

D152023 made ubsan consider __builtin_clz of 0 undefined regardless of
the target. This ensures portability and matches gcc.

This causes the ACLE intrinsics to also be considered to also be
considered to be undefined for 0 since they used the generic builtins
as their implementation.

This patch adds builtins for ARM that ubsan doesn't know about to make
the behavior defined for 0. Alternatively, I could have added a zero
check to the intrinsics, but the dedicated builtin will give better -O0
codegen.

Fixes #63113.

Reviewed By: tmatheson

Differential Revision: https://reviews.llvm.org/D154915

clang/include/clang/Basic/BuiltinsAArch64.def

@@ -39,6 +39,8 @@ BUILTIN(__builtin_arm_rbit, "UiUi", "nc")
 BUILTIN(__builtin_arm_rbit64, "WUiWUi", "nc")
 BUILTIN(__builtin_arm_cls, "UiZUi", "nc")
 BUILTIN(__builtin_arm_cls64, "UiWUi", "nc")
+BUILTIN(__builtin_arm_clz, "UiZUi", "nc")
+BUILTIN(__builtin_arm_clz64, "UiWUi", "nc")
 
 // HINT
 BUILTIN(__builtin_arm_nop, "v", "")

clang/include/clang/Basic/BuiltinsARM.def

@@ -119,6 +119,8 @@ BUILTIN(__builtin_arm_smusdx, "iii", "nc")
 
 // Bit manipulation
 BUILTIN(__builtin_arm_rbit, "UiUi", "nc")
+BUILTIN(__builtin_arm_clz, "UiZUi", "nc")
+BUILTIN(__builtin_arm_clz64, "UiWUi", "nc")
 BUILTIN(__builtin_arm_cls, "UiZUi", "nc")
 BUILTIN(__builtin_arm_cls64, "UiWUi", "nc")
 

clang/lib/CodeGen/CGBuiltin.cpp

@@ -7906,6 +7906,17 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
         CGM.getIntrinsic(Intrinsic::bitreverse, Arg->getType()), Arg, "rbit");
   }
 
+  if (BuiltinID == clang::ARM::BI__builtin_arm_clz ||
+      BuiltinID == clang::ARM::BI__builtin_arm_clz64) {
+    llvm::Value *Arg = EmitScalarExpr(E->getArg(0));
+    Function *F = CGM.getIntrinsic(Intrinsic::ctlz, Arg->getType());
+    Value *Res = Builder.CreateCall(F, {Arg, Builder.getInt1(false)});
+    if (BuiltinID == clang::ARM::BI__builtin_arm_clz64)
+      Res = Builder.CreateTrunc(Res, Builder.getInt32Ty());
+    return Res;
+  }
+
+
   if (BuiltinID == clang::ARM::BI__builtin_arm_cls) {
     llvm::Value *Arg = EmitScalarExpr(E->getArg(0));
     return Builder.CreateCall(CGM.getIntrinsic(Intrinsic::arm_cls), Arg, "cls");
@@ -9988,6 +9999,16 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
         CGM.getIntrinsic(Intrinsic::bitreverse, Arg->getType()), Arg, "rbit");
   }
 
+  if (BuiltinID == clang::AArch64::BI__builtin_arm_clz ||
+      BuiltinID == clang::AArch64::BI__builtin_arm_clz64) {
+    llvm::Value *Arg = EmitScalarExpr(E->getArg(0));
+    Function *F = CGM.getIntrinsic(Intrinsic::ctlz, Arg->getType());
+    Value *Res = Builder.CreateCall(F, {Arg, Builder.getInt1(false)});
+    if (BuiltinID == clang::AArch64::BI__builtin_arm_clz64)
+      Res = Builder.CreateTrunc(Res, Builder.getInt32Ty());
+    return Res;
+  }
+
   if (BuiltinID == clang::AArch64::BI__builtin_arm_cls) {
     llvm::Value *Arg = EmitScalarExpr(E->getArg(0));
     return Builder.CreateCall(CGM.getIntrinsic(Intrinsic::aarch64_cls), Arg,

clang/lib/Headers/arm_acle.h

@@ -140,17 +140,21 @@ __rorl(unsigned long __x, uint32_t __y) {
 /* CLZ */
 static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
 __clz(uint32_t __t) {
-  return (unsigned int)__builtin_clz(__t);
+  return __builtin_arm_clz(__t);
 }
 
 static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
 __clzl(unsigned long __t) {
-  return (unsigned int)__builtin_clzl(__t);
+#if __SIZEOF_LONG__ == 4
+  return __builtin_arm_clz(__t);
+#else
+  return __builtin_arm_clz64(__t);
+#endif
 }
 
 static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
 __clzll(uint64_t __t) {
-  return (unsigned int)__builtin_clzll(__t);
+  return __builtin_arm_clz64(__t);
 }
 
 /* CLS */