merge main into amd-staging

Change-Id: I9b7530b0a747ecc4217933773520b2d5eb162e69

clang/lib/Format/TokenAnnotator.cpp

@@ -250,7 +250,7 @@ class AnnotatingParser {
         if (Precedence > prec::Conditional && Precedence < prec::Relational)
           return false;
       }
-      if (Prev.is(TT_ConditionalExpr))
+      if (Prev.isOneOf(tok::question, tok::colon) && !Style.isProto())
         SeenTernaryOperator = true;
       updateParameterCount(Left, CurrentToken);
       if (Style.Language == FormatStyle::LK_Proto) {
@@ -2875,6 +2875,8 @@ class AnnotatingParser {
     // Search for unexpected tokens.
     for (auto *Prev = BeforeRParen; Prev != LParen; Prev = Prev->Previous) {
       if (Prev->is(tok::r_paren)) {
+        if (Prev->is(TT_CastRParen))
+          return false;
         Prev = Prev->MatchingParen;
         if (!Prev)
           return false;

clang/unittests/Format/TokenAnnotatorTest.cpp

@@ -620,6 +620,11 @@ TEST_F(TokenAnnotatorTest, UnderstandsNonTemplateAngleBrackets) {
   EXPECT_TOKEN(Tokens[2], tok::less, TT_BinaryOperator);
   EXPECT_TOKEN(Tokens[8], tok::greater, TT_BinaryOperator);
 
+  Tokens = annotate("return checklower ? a < b : a > b;");
+  ASSERT_EQ(Tokens.size(), 12u) << Tokens;
+  EXPECT_TOKEN(Tokens[4], tok::less, TT_BinaryOperator);
+  EXPECT_TOKEN(Tokens[8], tok::greater, TT_BinaryOperator);
+
   Tokens = annotate("return A < B ^ A > B;");
   ASSERT_EQ(Tokens.size(), 10u) << Tokens;
   EXPECT_TOKEN(Tokens[2], tok::less, TT_BinaryOperator);
@@ -747,6 +752,12 @@ TEST_F(TokenAnnotatorTest, UnderstandsCasts) {
   EXPECT_TOKEN(Tokens[9], tok::r_paren, TT_CastRParen);
   EXPECT_TOKEN(Tokens[10], tok::amp, TT_UnaryOperator);
 
+  Tokens = annotate("int result = ((int)a) - b;");
+  ASSERT_EQ(Tokens.size(), 13u) << Tokens;
+  EXPECT_TOKEN(Tokens[6], tok::r_paren, TT_CastRParen);
+  EXPECT_TOKEN(Tokens[8], tok::r_paren, TT_Unknown);
+  EXPECT_TOKEN(Tokens[9], tok::minus, TT_BinaryOperator);
+
   auto Style = getLLVMStyle();
   Style.TypeNames.push_back("Foo");
   Tokens = annotate("#define FOO(bar) foo((Foo)&bar)", Style);

clang/utils/TableGen/ClangDiagnosticsEmitter.cpp

@@ -1634,7 +1634,7 @@ static void emitDiagArrays(std::map<std::string, GroupInfo> &DiagsInGroup,
 ///     \000\020#pragma-messages\t#warnings\020CFString-literal"
 ///   };
 /// \endcode
-static void emitDiagGroupNames(StringToOffsetTable &GroupNames,
+static void emitDiagGroupNames(const StringToOffsetTable &GroupNames,
                                raw_ostream &OS) {
   OS << "static const char DiagGroupNames[] = {\n";
   GroupNames.EmitString(OS);
@@ -1656,7 +1656,7 @@ static void emitDiagGroupNames(StringToOffsetTable &GroupNames,
 static void emitAllDiagArrays(std::map<std::string, GroupInfo> &DiagsInGroup,
                               RecordVec &DiagsInPedantic,
                               RecordVec &GroupsInPedantic,
-                              StringToOffsetTable &GroupNames,
+                              const StringToOffsetTable &GroupNames,
                               raw_ostream &OS) {
   OS << "\n#ifdef GET_DIAG_ARRAYS\n";
   emitDiagArrays(DiagsInGroup, DiagsInPedantic, OS);
@@ -1683,7 +1683,8 @@ static void emitAllDiagArrays(std::map<std::string, GroupInfo> &DiagsInGroup,
 static void emitDiagTable(std::map<std::string, GroupInfo> &DiagsInGroup,
                           RecordVec &DiagsInPedantic,
                           RecordVec &GroupsInPedantic,
-                          StringToOffsetTable &GroupNames, raw_ostream &OS) {
+                          const StringToOffsetTable &GroupNames,
+                          raw_ostream &OS) {
   unsigned MaxLen = 0;
 
   for (auto const &I: DiagsInGroup)
@@ -1705,7 +1706,7 @@ static void emitDiagTable(std::map<std::string, GroupInfo> &DiagsInGroup,
     OS << I.first << " */, ";
     // Store a pascal-style length byte at the beginning of the string.
     std::string Name = char(I.first.size()) + I.first;
-    OS << GroupNames.GetOrAddStringOffset(Name, false) << ", ";
+    OS << *GroupNames.GetStringOffset(Name) << ", ";
 
     // Special handling for 'pedantic'.
     const bool IsPedantic = I.first == "pedantic";

compiler-rt/lib/nsan/nsan.cpp

@@ -409,21 +409,21 @@ __nsan_dump_shadow_mem(const u8 *addr, size_t size_bytes, size_t bytes_per_line,
   }
 }
 
-alignas(16) SANITIZER_INTERFACE_ATTRIBUTE
+alignas(64) SANITIZER_INTERFACE_ATTRIBUTE
     thread_local uptr __nsan_shadow_ret_tag = 0;
 
-alignas(16) SANITIZER_INTERFACE_ATTRIBUTE
+alignas(64) SANITIZER_INTERFACE_ATTRIBUTE
     thread_local char __nsan_shadow_ret_ptr[kMaxVectorWidth *
                                             sizeof(__float128)];
 
-alignas(16) SANITIZER_INTERFACE_ATTRIBUTE
+alignas(64) SANITIZER_INTERFACE_ATTRIBUTE
     thread_local uptr __nsan_shadow_args_tag = 0;
 
 // Maximum number of args. This should be enough for anyone (tm). An alternate
 // scheme is to have the generated code create an alloca and make
 // __nsan_shadow_args_ptr point ot the alloca.
 constexpr const int kMaxNumArgs = 128;
-alignas(16) SANITIZER_INTERFACE_ATTRIBUTE
+alignas(64) SANITIZER_INTERFACE_ATTRIBUTE
     thread_local char __nsan_shadow_args_ptr[kMaxVectorWidth * kMaxNumArgs *
                                              sizeof(__float128)];
 
@@ -445,6 +445,32 @@ int32_t checkFT(const FT value, ShadowFT Shadow, CheckTypeT CheckType,
   const InternalFT check_value = value;
   const InternalFT check_shadow = Shadow;
 
+  // We only check for NaNs in the value, not the shadow.
+  if (flags().check_nan && isnan(check_value)) {
+    GET_CALLER_PC_BP;
+    BufferedStackTrace stack;
+    stack.Unwind(pc, bp, nullptr, false);
+    if (GetSuppressionForStack(&stack, CheckKind::Consistency)) {
+      // FIXME: optionally print.
+      return flags().resume_after_suppression ? kResumeFromValue
+                                              : kContinueWithShadow;
+    }
+    Decorator D;
+    Printf("%s", D.Warning());
+    Printf("WARNING: NumericalStabilitySanitizer: NaN detected\n");
+    Printf("%s", D.Default());
+    stack.Print();
+    if (flags().halt_on_error) {
+      if (common_flags()->abort_on_error)
+        Printf("ABORTING\n");
+      else
+        Printf("Exiting\n");
+      Die();
+    }
+    // Performing other tests for NaN values is meaningless when dealing with numbers.
+    return kResumeFromValue;
+  }
+
   // See this article for an interesting discussion of how to compare floats:
   // https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
   static constexpr const FT Eps = FTInfo<FT>::kEpsilon;

compiler-rt/lib/nsan/nsan_flags.inc

@@ -48,3 +48,5 @@ NSAN_FLAG(bool, enable_loadtracking_stats, false,
           "due to invalid or unknown types.")
 NSAN_FLAG(bool, poison_in_free, true, "")
 NSAN_FLAG(bool, print_stats_on_exit, false, "If true, print stats on exit.")
+NSAN_FLAG(bool, check_nan, false,
+          "If true, check the floating-point number is nan")

compiler-rt/test/nsan/nan.cpp

@@ -0,0 +1,25 @@
+// RUN: %clangxx_nsan -O0 -g %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+
+// RUN: %clangxx_nsan -O3 -g %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+
+// RUN: %clangxx_nsan -O0 -g %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=1 not %run %t
+
+#include <cmath>
+#include <cstdio>
+
+// This function returns a NaN value for triggering the NaN detection.
+__attribute__((noinline)) float ReturnNaN(float p, float q) {
+  float ret = p / q;
+  return ret;
+  // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+}
+
+int main() {
+  float val = ReturnNaN(0., 0.);
+  printf("%f\n", val);
+  // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+  return 0;
+}

compiler-rt/test/nsan/softmax.cpp

@@ -0,0 +1,54 @@
+// RUN: %clangxx_nsan -O0 -g -DSOFTMAX=softmax %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0,log2_max_relative_error=19 %run %t 2>&1 | FileCheck %s
+
+// RUN: %clangxx_nsan -O3 -g -DSOFTMAX=softmax %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0,log2_max_relative_error=19 %run %t 2>&1 | FileCheck %s
+
+// RUN: %clangxx_nsan -O0 -g -DSOFTMAX=stable_softmax %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=1,log2_max_relative_error=19 %run %t 
+
+// RUN: %clangxx_nsan -O3 -g -DSOFTMAX=stable_softmax %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=1,log2_max_relative_error=19 %run %t
+
+#include<iostream>
+#include<vector>
+#include<algorithm>
+#include<cmath>
+
+// unstable softmax
+template <typename T>
+__attribute__((noinline)) void softmax(std::vector<T> &values) {
+    T sum_exp = 0.0;
+    for (auto &i: values) {
+      i = std::exp(i);
+      sum_exp += i;
+    }
+    for (auto &i: values) {
+      i /= sum_exp;
+    }
+}
+
+// use max value to avoid overflow
+// \sigma_i exp(x_i) / \sum_j exp(x_j) = \sigma_i exp(x_i - max(x)) / \sum_j exp(x_j - max(x))
+template <typename T>
+__attribute__((noinline)) void stable_softmax(std::vector<T> &values) {
+  T sum_exp = 0.0;
+  T max_values = *std::max_element(values.begin(), values.end());
+  for (auto &i: values) {
+    i = std::exp(i - max_values);
+    sum_exp += i;
+  }
+  for (auto &i:values) {
+    i /= sum_exp;
+  }
+}
+
+int main() {
+  std::vector<double> data = {1000, 1001, 1002};
+  SOFTMAX(data);
+  for (auto i: data) {
+    printf("%f", i);
+    // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+  }
+  return 0;
+}

compiler-rt/test/nsan/vec_sqrt.cpp

@@ -0,0 +1,34 @@
+// RUN: %clangxx_nsan -O0 -g -mavx %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+// RUN: %clangxx_nsan -O3 -g -mavx %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+
+#include <cmath>
+#include <immintrin.h>
+#include <iostream>
+
+void simd_sqrt(const float *input, float *output, size_t size) {
+  size_t i = 0;
+  for (; i + 7 < size; i += 8) {
+    __m256 vec = _mm256_loadu_ps(&input[i]);
+    __m256 result = _mm256_sqrt_ps(vec);
+    _mm256_storeu_ps(&output[i], result);
+  }
+  for (; i < size; ++i) {
+    output[i] = std::sqrt(input[i]);
+    // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+  }
+}
+
+int main() {
+  float input[] = {1.0,  2.0,   -3.0,  4.0,   5.0,   6.0,  7.0,
+                   8.0,  9.0,   -10.0, 11.0,  12.0,  13.0, 14.0,
+                   15.0, -16.0, 17.0,  -18.0, -19.0, -20.0};
+  float output[20];
+  simd_sqrt(input, output, 20);
+  for (int i = 0; i < 20; ++i) {
+    std::cout << output[i] << std::endl;
+    // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+  }
+  return 0;
+}

compiler-rt/test/nsan/vec_sqrt_ext.cpp

@@ -0,0 +1,25 @@
+// RUN: %clangxx_nsan -O0 -g -mavx %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+// RUN: %clangxx_nsan -O3 -g -mavx %s -o %t
+// RUN: NSAN_OPTIONS=check_nan=true,halt_on_error=0 %run %t 2>&1 | FileCheck %s
+#include <iostream>
+#include <cmath>
+
+typedef float v8sf __attribute__ ((vector_size(32)));
+
+v8sf simd_sqrt(v8sf a) {
+  return __builtin_elementwise_sqrt(a);
+  // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+}
+
+int main() {
+  v8sf a = {-1.0, -2.0, -3.0, 4.0, 5.0, 6.0, 7.0, 8.0};
+  a = simd_sqrt(a);
+
+  // This prevents DCE.
+  for (size_t i = 0; i < 8; ++i) {
+    std::cout << a[i] << std::endl;
+    // CHECK: WARNING: NumericalStabilitySanitizer: NaN detected
+  }
+  return 0;
+}

lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp

@@ -697,6 +697,10 @@ void TypeSystemClang::CreateASTContext() {
   TargetInfo *target_info = getTargetInfo();
   if (target_info)
     m_ast_up->InitBuiltinTypes(*target_info);
+  else if (auto *log = GetLog(LLDBLog::Expressions))
+    LLDB_LOG(log,
+             "Failed to initialize builtin ASTContext types for target '{0}'",
+             m_target_triple);
 
   GetASTMap().Insert(m_ast_up.get(), this);
 

llvm/CMakeLists.txt

@@ -576,6 +576,8 @@ set(LLVM_USE_STATIC_ZSTD FALSE CACHE BOOL "Use static version of zstd. Can be TR
 
 set(LLVM_ENABLE_CURL "OFF" CACHE STRING "Use libcurl for the HTTP client if available. Can be ON, OFF, or FORCE_ON")
 
+set(LLVM_HAS_LOGF128 "OFF" CACHE STRING "Use logf128 to constant fold fp128 logarithm calls. Can be ON, OFF, or FORCE_ON")
+
 set(LLVM_ENABLE_HTTPLIB "OFF" CACHE STRING "Use cpp-httplib HTTP server library if available. Can be ON, OFF, or FORCE_ON")
 
 set(LLVM_Z3_INSTALL_DIR "" CACHE STRING "Install directory of the Z3 solver.")

llvm/cmake/config-ix.cmake

@@ -246,6 +246,17 @@ else()
   set(HAVE_LIBEDIT 0)
 endif()
 
+if(LLVM_HAS_LOGF128)
+  include(CheckCXXSymbolExists)
+  check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
+
+  if(LLVM_HAS_LOGF128 STREQUAL FORCE_ON AND NOT HAS_LOGF128)
+    message(FATAL_ERROR "Failed to configure logf128")
+  endif()
+
+  set(LLVM_HAS_LOGF128 "${HAS_LOGF128}")
+endif()
+
 # function checks
 check_symbol_exists(arc4random "stdlib.h" HAVE_DECL_ARC4RANDOM)
 find_package(Backtrace)
@@ -259,13 +270,6 @@ if(C_SUPPORTS_WERROR_UNGUARDED_AVAILABILITY_NEW)
   set(CMAKE_REQUIRED_FLAGS "${CMAKE_REQUIRED_FLAGS} -Werror=unguarded-availability-new")
 endif()
 
-check_cxx_symbol_exists(logf128 cmath HAS_LOGF128)
-check_symbol_exists(__powerpc__ "" __PPC64LE)
-if(HAS_LOGF128 AND NOT __PPC64LE)
-    set(LLVM_HAS_LOGF128 On)
-    add_compile_definitions(HAS_LOGF128)
-endif()
-
 # Determine whether we can register EH tables.
 check_symbol_exists(__register_frame "${CMAKE_CURRENT_LIST_DIR}/unwind.h" HAVE_REGISTER_FRAME)
 check_symbol_exists(__deregister_frame "${CMAKE_CURRENT_LIST_DIR}/unwind.h" HAVE_DEREGISTER_FRAME)

llvm/include/llvm/ADT/APFloat.h

@@ -19,6 +19,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/FloatingPointMode.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/float128.h"
 #include <memory>
 
 #define APFLOAT_DISPATCH_ON_SEMANTICS(METHOD_CALL)                             \
@@ -377,6 +378,9 @@ class IEEEFloat final : public APFloatBase {
   Expected<opStatus> convertFromString(StringRef, roundingMode);
   APInt bitcastToAPInt() const;
   double convertToDouble() const;
+#ifdef HAS_IEE754_FLOAT128
+  float128 convertToQuad() const;
+#endif
   float convertToFloat() const;
 
   /// @}
@@ -1270,9 +1274,14 @@ class APFloat : public APFloatBase {
   /// shorter semantics, like IEEEsingle and others.
   double convertToDouble() const;
 
-  /// Return true if this APFloat has quadruple precision floating point
-  /// semantics
-  bool isValidIEEEQuad() const;
+  /// Converts this APFloat to host float value.
+  ///
+  /// \pre The APFloat must be built using semantics, that can be represented by
+  /// the host float type without loss of precision. It can be IEEEquad and
+  /// shorter semantics, like IEEEdouble and others.
+#ifdef HAS_IEE754_FLOAT128
+  float128 convertToQuad() const;
+#endif
 
   /// Converts this APFloat to host float value.
   ///