merge main into amd-staging

Change-Id: Ibcd11a4bcf026ae9244631b875963ea820f2fff5

Author: Jenkins

clang/lib/AST/ByteCode/Program.cpp

@@ -284,15 +284,13 @@ Record *Program::getOrCreateRecord(const RecordDecl *RD) {
   if (!RD->isCompleteDefinition())
     return nullptr;
 
-  // Deduplicate records.
-  if (auto It = Records.find(RD); It != Records.end())
+  // Return an existing record if available.  Otherwise, we insert nullptr now
+  // and replace that later, so recursive calls to this function with the same
+  // RecordDecl don't run into infinite recursion.
+  auto [It, Inserted] = Records.try_emplace(RD);
+  if (!Inserted)
     return It->second;
 
-  // We insert nullptr now and replace that later, so recursive calls
-  // to this function with the same RecordDecl don't run into
-  // infinite recursion.
-  Records.insert({RD, nullptr});
-
   // Number of bytes required by fields and base classes.
   unsigned BaseSize = 0;
   // Number of bytes required by virtual base.

clang/lib/StaticAnalyzer/Core/CheckerManager.cpp

@@ -66,13 +66,10 @@ void CheckerManager::runCheckersOnASTDecl(const Decl *D, AnalysisManager& mgr,
   assert(D);
 
   unsigned DeclKind = D->getKind();
-  CachedDeclCheckers *checkers = nullptr;
-  CachedDeclCheckersMapTy::iterator CCI = CachedDeclCheckersMap.find(DeclKind);
-  if (CCI != CachedDeclCheckersMap.end()) {
-    checkers = &(CCI->second);
-  } else {
+  auto [CCI, Inserted] = CachedDeclCheckersMap.try_emplace(DeclKind);
+  CachedDeclCheckers *checkers = &(CCI->second);
+  if (Inserted) {
     // Find the checkers that should run for this Decl and cache them.
-    checkers = &CachedDeclCheckersMap[DeclKind];
     for (const auto &info : DeclCheckers)
       if (info.IsForDeclFn(D))
         checkers->push_back(info.CheckFn);
@@ -896,14 +893,13 @@ CheckerManager::getCachedStmtCheckersFor(const Stmt *S, bool isPreVisit) {
   assert(S);
 
   unsigned Key = (S->getStmtClass() << 1) | unsigned(isPreVisit);
-  CachedStmtCheckersMapTy::iterator CCI = CachedStmtCheckersMap.find(Key);
-  if (CCI != CachedStmtCheckersMap.end())
-    return CCI->second;
-
-  // Find the checkers that should run for this Stmt and cache them.
-  CachedStmtCheckers &Checkers = CachedStmtCheckersMap[Key];
-  for (const auto &Info : StmtCheckers)
-    if (Info.IsPreVisit == isPreVisit && Info.IsForStmtFn(S))
-      Checkers.push_back(Info.CheckFn);
+  auto [CCI, Inserted] = CachedStmtCheckersMap.try_emplace(Key);
+  CachedStmtCheckers &Checkers = CCI->second;
+  if (Inserted) {
+    // Find the checkers that should run for this Stmt and cache them.
+    for (const auto &Info : StmtCheckers)
+      if (Info.IsPreVisit == isPreVisit && Info.IsForStmtFn(S))
+        Checkers.push_back(Info.CheckFn);
+  }
   return Checkers;
 }

libc/newhdrgen/yaml/gpu/rpc.yaml

@@ -16,7 +16,7 @@ functions:
   - name: rpc_host_call
     standards:
       - GPUExtensions
-    return_type: void
+    return_type: unsigned long long
     arguments:
       - type: void *
       - type: void *

libc/spec/gpu_ext.td

@@ -7,7 +7,7 @@ def GPUExtensions : StandardSpec<"GPUExtensions"> {
     [
         FunctionSpec<
             "rpc_host_call",
-            RetValSpec<VoidType>,
+            RetValSpec<UnsignedLongLongType>,
             [ArgSpec<VoidPtr>, ArgSpec<VoidPtr>, ArgSpec<SizeTType>]
         >,
     ]

libc/src/gpu/rpc_host_call.cpp

@@ -17,14 +17,19 @@ namespace LIBC_NAMESPACE_DECL {
 
 // This calls the associated function pointer on the RPC server with the given
 // arguments. We expect that the pointer here is a valid pointer on the server.
-LLVM_LIBC_FUNCTION(void, rpc_host_call, (void *fn, void *data, size_t size)) {
+LLVM_LIBC_FUNCTION(unsigned long long, rpc_host_call,
+                   (void *fn, void *data, size_t size)) {
   rpc::Client::Port port = rpc::client.open<RPC_HOST_CALL>();
   port.send_n(data, size);
   port.send([=](rpc::Buffer *buffer) {
     buffer->data[0] = reinterpret_cast<uintptr_t>(fn);
   });
-  port.recv([](rpc::Buffer *) {});
+  unsigned long long ret;
+  port.recv([&](rpc::Buffer *buffer) {
+    ret = static_cast<unsigned long long>(buffer->data[0]);
+  });
   port.close();
+  return ret;
 }
 
 } // namespace LIBC_NAMESPACE_DECL

libc/src/gpu/rpc_host_call.h

@@ -14,7 +14,7 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-void rpc_host_call(void *fn, void *buffer, size_t size);
+unsigned long long rpc_host_call(void *fn, void *buffer, size_t size);
 
 } // namespace LIBC_NAMESPACE_DECL
 

libc/src/string/memory_utils/aarch64/inline_memcmp.h

@@ -8,7 +8,7 @@
 #ifndef LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMCMP_H
 #define LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMCMP_H
 
-#include "src/__support/macros/config.h"       // LIBC_INLINE
+#include "src/__support/macros/attributes.h"   // LIBC_INLINE
 #include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
 #include "src/string/memory_utils/op_aarch64.h"
 #include "src/string/memory_utils/op_generic.h"

libc/src/string/memory_utils/aarch64/inline_memcpy.h

@@ -8,7 +8,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_AARCH64_INLINE_MEMCPY_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_AARCH64_INLINE_MEMCPY_H
 
-#include "src/__support/macros/config.h" // LIBC_INLINE
+#include "src/__support/macros/attributes.h" // LIBC_INLINE
 #include "src/string/memory_utils/op_builtin.h"
 #include "src/string/memory_utils/utils.h"
 

libc/src/string/memory_utils/aarch64/inline_memmove.h

@@ -8,7 +8,7 @@
 #ifndef LIBC_SRC_STRING_MEMORY_UTILS_AARCH64_INLINE_MEMMOVE_H
 #define LIBC_SRC_STRING_MEMORY_UTILS_AARCH64_INLINE_MEMMOVE_H
 
-#include "src/__support/macros/config.h"        // LIBC_INLINE
+#include "src/__support/macros/attributes.h"    // LIBC_INLINE
 #include "src/string/memory_utils/op_aarch64.h" // aarch64::kNeon
 #include "src/string/memory_utils/op_builtin.h"
 #include "src/string/memory_utils/op_generic.h"

libc/src/string/memory_utils/generic/aligned_access.h

@@ -13,7 +13,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_GENERIC_ALIGNED_ACCESS_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_GENERIC_ALIGNED_ACCESS_H
 
-#include "src/__support/macros/config.h" // LIBC_INLINE
+#include "src/__support/macros/attributes.h" // LIBC_INLINE
 #include "src/string/memory_utils/generic/byte_per_byte.h"
 #include "src/string/memory_utils/op_generic.h" // generic::splat
 #include "src/string/memory_utils/utils.h"      // Ptr, CPtr

libc/src/string/memory_utils/generic/byte_per_byte.h

@@ -12,7 +12,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_GENERIC_BYTE_PER_BYTE_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_GENERIC_BYTE_PER_BYTE_H
 
-#include "src/__support/macros/config.h"       // LIBC_INLINE
+#include "src/__support/macros/attributes.h"   // LIBC_INLINE
 #include "src/__support/macros/optimization.h" // LIBC_LOOP_NOUNROLL
 #include "src/string/memory_utils/utils.h"     // Ptr, CPtr
 

libc/src/string/memory_utils/inline_memcmp.h

@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMCMP_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMCMP_H
 
-#include "src/__support/macros/config.h"                   // LIBC_INLINE
+#include "src/__support/macros/attributes.h"               // LIBC_INLINE
 #include "src/__support/macros/properties/architectures.h" // LIBC_TARGET_ARCH_IS_
 #include "src/string/memory_utils/utils.h"                 // Ptr, CPtr
 

libc/src/string/memory_utils/inline_memcpy.h

@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMCPY_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMCPY_H
 
-#include "src/__support/macros/config.h"                   // LIBC_INLINE
+#include "src/__support/macros/attributes.h"               // LIBC_INLINE
 #include "src/__support/macros/properties/architectures.h" // LIBC_TARGET_ARCH_IS_
 #include "src/string/memory_utils/utils.h"                 // Ptr, CPtr
 

libc/src/string/memory_utils/inline_memset.h

@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMSET_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_INLINE_MEMSET_H
 
-#include "src/__support/macros/config.h"                   // LIBC_INLINE
+#include "src/__support/macros/attributes.h"               // LIBC_INLINE
 #include "src/__support/macros/properties/architectures.h" // LIBC_TARGET_ARCH_IS_
 #include "src/string/memory_utils/utils.h"                 // Ptr, CPtr
 

libc/src/string/memory_utils/x86_64/inline_memcmp.h

@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMCMP_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMCMP_H
 
-#include "src/__support/macros/config.h"       // LIBC_INLINE
+#include "src/__support/macros/attributes.h"   // LIBC_INLINE
 #include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
 #include "src/string/memory_utils/op_generic.h"
 #include "src/string/memory_utils/op_x86.h"

libc/src/string/memory_utils/x86_64/inline_memcpy.h

@@ -9,7 +9,6 @@
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMCPY_H
 
 #include "src/__support/macros/attributes.h"   // LIBC_INLINE_VAR
-#include "src/__support/macros/config.h"       // LIBC_INLINE
 #include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
 #include "src/string/memory_utils/op_builtin.h"
 #include "src/string/memory_utils/op_x86.h"

libc/src/string/memory_utils/x86_64/inline_memmove.h

@@ -8,7 +8,7 @@
 #ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMMOVE_H
 #define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_X86_64_INLINE_MEMMOVE_H
 
-#include "src/__support/macros/config.h" // LIBC_INLINE
+#include "src/__support/macros/attributes.h" // LIBC_INLINE
 #include "src/string/memory_utils/op_builtin.h"
 #include "src/string/memory_utils/op_generic.h"
 #include "src/string/memory_utils/op_x86.h"

libc/utils/gpu/server/rpc_server.cpp

@@ -319,13 +319,18 @@ rpc_status_t handle_server_impl(
   }
   case RPC_HOST_CALL: {
     uint64_t sizes[lane_size] = {0};
+    unsigned long long results[lane_size] = {0};
     void *args[lane_size] = {nullptr};
     port->recv_n(args, sizes,
                  [&](uint64_t size) { return temp_storage.alloc(size); });
     port->recv([&](rpc::Buffer *buffer, uint32_t id) {
-      reinterpret_cast<void (*)(void *)>(buffer->data[0])(args[id]);
+      using func_ptr_t = unsigned long long (*)(void *);
+      auto func = reinterpret_cast<func_ptr_t>(buffer->data[0]);
+      results[id] = func(args[id]);
+    });
+    port->send([&](rpc::Buffer *buffer, uint32_t id) {
+      buffer->data[0] = static_cast<uint64_t>(results[id]);
     });
-    port->send([&](rpc::Buffer *, uint32_t id) {});
     break;
   }
   case RPC_FEOF: {

libcxx/test/benchmarks/CMakeLists.txt

@@ -147,6 +147,9 @@ set(BENCHMARK_TESTS
     deque_iterator.bench.cpp
     exception_ptr.bench.cpp
     filesystem.bench.cpp
+    format/write_double_comparison.bench.cpp
+    format/write_int_comparison.bench.cpp
+    format/write_string_comparison.bench.cpp
     format_to_n.bench.cpp
     format_to.bench.cpp
     format.bench.cpp

libcxx/test/benchmarks/format.bench.cpp

@@ -20,8 +20,10 @@ static void BM_format_string(benchmark::State& state) {
   size_t size = state.range(0);
   std::basic_string<CharT> str(size, CharT('*'));
 
-  while (state.KeepRunningBatch(str.size()))
-    benchmark::DoNotOptimize(std::format(CSTR("{}"), str));
+  while (state.KeepRunningBatch(str.size())) {
+    std::basic_string<CharT> s = std::format(CSTR("{}"), str);
+    benchmark::DoNotOptimize(s);
+  }
 
   state.SetBytesProcessed(state.iterations() * size * sizeof(CharT));
 }

libcxx/test/benchmarks/format/write_double_comparison.bench.cpp

@@ -0,0 +1,117 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <array>
+#include <charconv>
+#include <cstdio>
+#include <format>
+#include <iterator>
+#include <list>
+#include <random>
+#include <vector>
+
+#include "benchmark/benchmark.h"
+
+std::array data = [] {
+  std::uniform_real_distribution<double> distribution;
+  std::mt19937 generator;
+  std::array<double, 1000> result;
+  std::generate_n(result.begin(), result.size(), [&] { return distribution(generator); });
+  return result;
+}();
+
+static void BM_sprintf(benchmark::State& state) {
+  std::array<char, 100> output;
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      sprintf(output.data(), "%f", value);
+      benchmark::DoNotOptimize(output.data());
+    }
+}
+
+static void BM_to_string(benchmark::State& state) {
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      std::string s = std::to_string(value);
+      benchmark::DoNotOptimize(s);
+    }
+}
+
+static void BM_to_chars(benchmark::State& state) {
+  std::array<char, 100> output;
+
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      std::to_chars(output.data(), output.data() + output.size(), value);
+      benchmark::DoNotOptimize(output.data());
+    }
+}
+
+static void BM_to_chars_as_string(benchmark::State& state) {
+  std::array<char, 100> output;
+
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      char* end = std::to_chars(output.data(), output.data() + output.size(), value).ptr;
+      std::string s{output.data(), end};
+      benchmark::DoNotOptimize(s);
+    }
+}
+
+static void BM_format(benchmark::State& state) {
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      std::string s = std::format("{}", value);
+      benchmark::DoNotOptimize(s);
+    }
+}
+
+template <class C>
+static void BM_format_to_back_inserter(benchmark::State& state) {
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      C c;
+      std::format_to(std::back_inserter(c), "{}", value);
+      benchmark::DoNotOptimize(c);
+    }
+}
+
+template <class F>
+static void BM_format_to_iterator(benchmark::State& state, F&& f) {
+  auto output = f();
+  while (state.KeepRunningBatch(data.size()))
+    for (auto value : data) {
+      std::format_to(std::begin(output), "{}", value);
+      benchmark::DoNotOptimize(std::begin(output));
+    }
+}
+
+BENCHMARK(BM_sprintf);
+BENCHMARK(BM_to_string);
+BENCHMARK(BM_to_chars);
+BENCHMARK(BM_to_chars_as_string);
+BENCHMARK(BM_format);
+BENCHMARK_TEMPLATE(BM_format_to_back_inserter, std::string);
+BENCHMARK_TEMPLATE(BM_format_to_back_inserter, std::vector<char>);
+BENCHMARK_TEMPLATE(BM_format_to_back_inserter, std::list<char>);
+BENCHMARK_CAPTURE(BM_format_to_iterator, <std::array>, ([] {
+                    std::array<char, 100> a;
+                    return a;
+                  }));
+BENCHMARK_CAPTURE(BM_format_to_iterator, <std::string>, ([] {
+                    std::string s;
+                    s.resize(100);
+                    return s;
+                  }));
+BENCHMARK_CAPTURE(BM_format_to_iterator, <std::vector>, ([] {
+                    std::vector<char> v;
+                    v.resize(100);
+                    return v;
+                  }));
+
+BENCHMARK_MAIN();