merge main into amd-staging

Change-Id: Ief047b8931f227872f3af719811c6baaea45d94e

Author: animeshk-amd

compiler-rt/test/profile/ContinuousSyncMode/online-merging.c

@@ -8,9 +8,9 @@
 // Create two DSOs and a driver program that uses them.
 // RUN: echo "void dso1(void) {}" > dso1.c
 // RUN: echo "void dso2(void) {}" > dso2.c
-// RUN: %clang_pgogen_cont %shared_lib_flag -o %t.dir/dso1.dylib dso1.c -mllvm -instrprof-atomic-counter-update-all=1
-// RUN: %clang_pgogen_cont %shared_lib_flag -o %t.dir/dso2.dylib dso2.c -mllvm -instrprof-atomic-counter-update-all=1
-// RUN: %clang_pgogen_cont -o main.exe %s %t.dir/dso1.dylib %t.dir/dso2.dylib -mllvm -instrprof-atomic-counter-update-all=1
+// RUN: %clang_pgogen_cont %shared_lib_flag -o %t.dir/dso1.dylib dso1.c -fprofile-update=atomic
+// RUN: %clang_pgogen_cont %shared_lib_flag -o %t.dir/dso2.dylib dso2.c -fprofile-update=atomic
+// RUN: %clang_pgogen_cont -o main.exe %s %t.dir/dso1.dylib %t.dir/dso2.dylib -fprofile-update=atomic
 //
 // === Round 1 ===
 // Test merging+continuous mode without any file contention.

libc/fuzzing/__support/hashtable_fuzz.cpp

@@ -10,6 +10,7 @@
 ///
 //===----------------------------------------------------------------------===//
 #include "include/llvm-libc-types/ENTRY.h"
+#include "src/__support/CPP/bit.h"
 #include "src/__support/CPP/string_view.h"
 #include "src/__support/HashTable/table.h"
 #include "src/__support/macros/config.h"
@@ -81,15 +82,14 @@ static struct {
 
   template <typename T> T next() {
     static_assert(cpp::is_integral<T>::value, "T must be an integral type");
-    union {
-      T result;
-      char data[sizeof(T)];
-    };
-    for (size_t i = 0; i < sizeof(result); i++)
+
+    char data[sizeof(T)];
+
+    for (size_t i = 0; i < sizeof(T); i++)
       data[i] = buffer[i];
-    buffer += sizeof(result);
-    remaining -= sizeof(result);
-    return result;
+    buffer += sizeof(T);
+    remaining -= sizeof(T);
+    return cpp::bit_cast<T>(data);
   }
 
   cpp::string_view next_string() {

libc/src/__support/HashTable/generic/bitmask_impl.inc

@@ -6,6 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "src/__support/CPP/bit.h"
 #include "src/__support/common.h"
 #include "src/__support/endian_internal.h"
 #include "src/__support/macros/config.h"
@@ -44,13 +45,11 @@ struct Group {
 
   // Load a group of control words from an arbitary address.
   LIBC_INLINE static Group load(const void *addr) {
-    union {
-      bitmask_t value;
-      char bytes[sizeof(bitmask_t)];
-    } data;
+    char bytes[sizeof(bitmask_t)];
+
     for (size_t i = 0; i < sizeof(bitmask_t); ++i)
-      data.bytes[i] = static_cast<const char *>(addr)[i];
-    return {data.value};
+      bytes[i] = static_cast<const char *>(addr)[i];
+    return Group{cpp::bit_cast<bitmask_t>(bytes)};
   }
 
   // Load a group of control words from an aligned address.

libc/src/__support/hash.h

@@ -13,8 +13,8 @@
 #include "src/__support/CPP/limits.h"        // numeric_limits
 #include "src/__support/macros/attributes.h" // LIBC_INLINE
 #include "src/__support/macros/config.h"
-#include "src/__support/uint128.h"           // UInt128
-#include <stdint.h>                          // For uint64_t
+#include "src/__support/uint128.h" // UInt128
+#include <stdint.h>                // For uint64_t
 
 namespace LIBC_NAMESPACE_DECL {
 namespace internal {
@@ -34,25 +34,23 @@ LIBC_INLINE uint64_t folded_multiply(uint64_t x, uint64_t y) {
 // Therefore, we use a union to read the value.
 template <typename T> LIBC_INLINE T read_little_endian(const void *ptr) {
   const uint8_t *bytes = static_cast<const uint8_t *>(ptr);
-  union {
-    T value;
-    uint8_t buffer[sizeof(T)];
-  } data;
+  uint8_t buffer[sizeof(T)];
 #if __BYTE_ORDER__ != __ORDER_LITTLE_ENDIAN__
-  // Compiler should able to optimize this as a load followed by a byte swap.
-  // On aarch64 (-mbig-endian), this compiles to the following for int:
+  // Compiler should able to optimize this as a load followed by a byte
+  // swap. On aarch64 (-mbig-endian), this compiles to the following for
+  // int:
   //      ldr     w0, [x0]
   //      rev     w0, w0
   //      ret
   for (size_t i = 0; i < sizeof(T); ++i) {
-    data.buffer[i] = bytes[sizeof(T) - i - 1];
+    buffer[i] = bytes[sizeof(T) - i - 1];
   }
 #else
   for (size_t i = 0; i < sizeof(T); ++i) {
-    data.buffer[i] = bytes[i];
+    buffer[i] = bytes[i];
   }
 #endif
-  return data.value;
+  return cpp::bit_cast<T>(buffer);
 }
 
 // Specialized read functions for small values. size must be <= 8.

libc/src/math/generic/exp10m1f16.cpp

@@ -119,6 +119,9 @@ LLVM_LIBC_FUNCTION(float16, exp10m1f16, (float16 x)) {
 
     // When |x| <= 2^(-3).
     if (x_abs <= 0x3000U) {
+      if (LIBC_UNLIKELY(x_abs == 0))
+        return x;
+
       if (auto r = EXP10M1F16_EXCEPTS_LO.lookup(x_u);
           LIBC_UNLIKELY(r.has_value()))
         return r.value();

libc/src/math/generic/tanhf16.cpp

@@ -64,6 +64,9 @@ LLVM_LIBC_FUNCTION(float16, tanhf16, (float16 x)) {
 
     // When |x| <= 0x1.d2p-4.
     if (x_abs <= 0x2f48U) {
+      if (LIBC_UNLIKELY(x_abs == 0))
+        return x;
+
       float xf = x;
       float xf_sq = xf * xf;
       // Degree-7 Taylor expansion generated by Sollya with the following

libc/test/src/__support/HashTable/group_test.cpp

@@ -8,6 +8,7 @@
 
 #include "src/__support/HashTable/bitmask.h"
 
+#include "src/__support/CPP/bit.h"
 #include "src/__support/macros/config.h"
 #include "src/stdlib/rand.h"
 #include "test/UnitTest/Test.h"
@@ -28,14 +29,13 @@ TEST(LlvmLibcHashTableBitMaskTest, Match) {
   size_t appearance[4][sizeof(Group)];
   ByteArray array{};
 
-  union {
-    uintptr_t random;
-    int data[sizeof(uintptr_t) / sizeof(int)];
-  };
+  int data[sizeof(uintptr_t) / sizeof(int)];
 
   for (int &i : data)
     i = rand();
 
+  uintptr_t random = cpp::bit_cast<uintptr_t>(data);
+
   for (size_t i = 0; i < sizeof(Group); ++i) {
     size_t choice = random % 4;
     random /= 4;
@@ -62,14 +62,13 @@ TEST(LlvmLibcHashTableBitMaskTest, MaskAvailable) {
   for (size_t i = 0; i < sizeof(Group); ++i) {
     ByteArray array{};
 
-    union {
-      uintptr_t random;
-      int data[sizeof(uintptr_t) / sizeof(int)];
-    };
+    int data[sizeof(uintptr_t) / sizeof(int)];
 
     for (int &j : data)
       j = rand();
 
+    uintptr_t random = cpp::bit_cast<uintptr_t>(data);
+
     ASSERT_FALSE(Group::load(array.data).mask_available().any_bit_set());
 
     array.data[i] = 0x80;

libc/test/src/__support/HashTable/table_test.cpp

@@ -82,7 +82,7 @@ TEST(LlvmLibcTableTest, GrowthSequence) {
 }
 
 TEST(LlvmLibcTableTest, Insertion) {
-  union key {
+  struct key {
     char bytes[2];
   } keys[256];
   for (size_t k = 0; k < 256; ++k) {

lldb/test/Shell/Unwind/windows-unaligned-x86_64.test

@@ -17,7 +17,7 @@ breakpoint set -n func
 # CHECK: Breakpoint 1: where = {{.*}}`{{(::)?}}func
 
 process launch
-# CHECK: stop reason = breakpoint 1.1
+# CHECK: stop reason = breakpoint 1
 
 thread backtrace
 # CHECK: frame #0: {{.*}}`{{(::)?}}func

llvm/lib/Linker/IRMover.cpp

@@ -82,8 +82,6 @@ class TypeMapTy : public ValueMapTypeRemapper {
   Type *get(Type *SrcTy);
   Type *get(Type *SrcTy, SmallPtrSet<StructType *, 8> &Visited);
 
-  void finishType(StructType *DTy, StructType *STy, ArrayRef<Type *> ETypes);
-
   FunctionType *get(FunctionType *T) {
     return cast<FunctionType>(get((Type *)T));
   }
@@ -233,20 +231,6 @@ Error TypeMapTy::linkDefinedTypeBodies() {
   return Error::success();
 }
 
-void TypeMapTy::finishType(StructType *DTy, StructType *STy,
-                           ArrayRef<Type *> ETypes) {
-  DTy->setBody(ETypes, STy->isPacked());
-
-  // Steal STy's name.
-  if (STy->hasName()) {
-    SmallString<16> TmpName = STy->getName();
-    STy->setName("");
-    DTy->setName(TmpName);
-  }
-
-  DstStructTypesSet.addNonOpaque(DTy);
-}
-
 Type *TypeMapTy::get(Type *Ty) {
   SmallPtrSet<StructType *, 8> Visited;
   return get(Ty, Visited);
@@ -292,17 +276,9 @@ Type *TypeMapTy::get(Type *Ty, SmallPtrSet<StructType *, 8> &Visited) {
     AnyChange |= ElementTypes[I] != Ty->getContainedType(I);
   }
 
-  // If we found our type while recursively processing stuff, just use it.
+  // Refresh Entry after recursively processing stuff.
   Entry = &MappedTypes[Ty];
-  if (*Entry) {
-    if (auto *DTy = dyn_cast<StructType>(*Entry)) {
-      if (DTy->isOpaque()) {
-        auto *STy = cast<StructType>(Ty);
-        finishType(DTy, STy, ElementTypes);
-      }
-    }
-    return *Entry;
-  }
+  assert(!*Entry && "Recursive type!");
 
   // If all of the element types mapped directly over and the type is not
   // a named struct, then the type is usable as-is.
@@ -350,8 +326,17 @@ Type *TypeMapTy::get(Type *Ty, SmallPtrSet<StructType *, 8> &Visited) {
       return *Entry = Ty;
     }
 
-    StructType *DTy = StructType::create(Ty->getContext());
-    finishType(DTy, STy, ElementTypes);
+    StructType *DTy =
+        StructType::create(Ty->getContext(), ElementTypes, "", STy->isPacked());
+
+    // Steal STy's name.
+    if (STy->hasName()) {
+      SmallString<16> TmpName = STy->getName();
+      STy->setName("");
+      DTy->setName(TmpName);
+    }
+
+    DstStructTypesSet.addNonOpaque(DTy);
     return *Entry = DTy;
   }
   }

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

@@ -920,8 +920,7 @@ bool AArch64DAGToDAGISel::SelectRDVLImm(SDValue N, SDValue &Imm) {
   if ((MulImm % std::abs(Scale)) == 0) {
     int64_t RDVLImm = MulImm / Scale;
     if ((RDVLImm >= Low) && (RDVLImm <= High)) {
-      Imm = CurDAG->getSignedConstant(RDVLImm, SDLoc(N), MVT::i32,
-                                      /*isTarget=*/true);
+      Imm = CurDAG->getSignedTargetConstant(RDVLImm, SDLoc(N), MVT::i32);
       return true;
     }
   }
@@ -4283,7 +4282,7 @@ bool AArch64DAGToDAGISel::SelectSVESignedArithImm(SDValue N, SDValue &Imm) {
     int64_t ImmVal = CNode->getSExtValue();
     SDLoc DL(N);
     if (ImmVal >= -128 && ImmVal < 128) {
-      Imm = CurDAG->getSignedConstant(ImmVal, DL, MVT::i32, /*isTarget=*/true);
+      Imm = CurDAG->getSignedTargetConstant(ImmVal, DL, MVT::i32);
       return true;
     }
   }

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -9299,8 +9299,7 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
     // Each tail call may have to adjust the stack by a different amount, so
     // this information must travel along with the operation for eventual
     // consumption by emitEpilogue.
-    Ops.push_back(
-        DAG.getSignedConstant(FPDiff, DL, MVT::i32, /*isTarget=*/true));
+    Ops.push_back(DAG.getSignedTargetConstant(FPDiff, DL, MVT::i32));
   }
 
   if (CLI.PAI) {