[llvm] 1076082 - [Support]: Introduce the `HashBuilder` interface.
Alexandre Rames via llvm-commits
llvm-commits at lists.llvm.org
Thu Aug 26 09:20:55 PDT 2021
Author: Alexandre Rames
Date: 2021-08-26T09:20:50-07:00
New Revision: 1076082a0d97bd5c16a25ee7cf3dbb6ee4b5a9fe
URL: https://github.com/llvm/llvm-project/commit/1076082a0d97bd5c16a25ee7cf3dbb6ee4b5a9fe
DIFF: https://github.com/llvm/llvm-project/commit/1076082a0d97bd5c16a25ee7cf3dbb6ee4b5a9fe.diff
LOG: [Support]: Introduce the `HashBuilder` interface.
The `HashBuilder` interface allows conveniently building hashes of various data
types, without relying on the underlying hasher type to know about hashed data
types.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D106910
Added:
llvm/include/llvm/Support/HashBuilder.h
llvm/unittests/Support/HashBuilderTest.cpp
Modified:
llvm/unittests/Support/CMakeLists.txt
Removed:
################################################################################
diff --git a/llvm/include/llvm/Support/HashBuilder.h b/llvm/include/llvm/Support/HashBuilder.h
new file mode 100644
index 0000000000000..23f73f1b659dc
--- /dev/null
+++ b/llvm/include/llvm/Support/HashBuilder.h
@@ -0,0 +1,404 @@
+//===- llvm/Support/HashBuilder.h - Convenient hashing interface-*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements an interface allowing to conveniently build hashes of
+// various data types, without relying on the underlying hasher type to know
+// about hashed data types.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_SUPPORT_HASHBUILDER_H
+#define LLVM_SUPPORT_HASHBUILDER_H
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/Endian.h"
+#include "llvm/Support/type_traits.h"
+
+#include <iterator>
+#include <utility>
+
+namespace llvm {
+
+/// Declares the hasher member, and functions forwarding directly to the hasher.
+template <typename HasherT> class HashBuilderBase {
+public:
+ HasherT &getHasher() { return Hasher; }
+
+ /// Forward to `HasherT::update(ArrayRef<uint8_t>)`.
+ ///
+ /// This may not take the size of `Data` into account.
+ /// Users of this function should pay attention to respect endianness
+ /// contraints.
+ void update(ArrayRef<uint8_t> Data) { this->getHasher().update(Data); }
+
+ /// Forward to `HasherT::update(ArrayRef<uint8_t>)`.
+ ///
+ /// This may not take the size of `Data` into account.
+ /// Users of this function should pay attention to respect endianness
+ /// contraints.
+ void update(StringRef Data) {
+ update(makeArrayRef(reinterpret_cast<const uint8_t *>(Data.data()),
+ Data.size()));
+ }
+
+ /// Forward to `HasherT::final()` if available.
+ template <typename HasherT_ = HasherT> StringRef final() {
+ return this->getHasher().final();
+ }
+
+ /// Forward to `HasherT::result()` if available.
+ template <typename HasherT_ = HasherT> StringRef result() {
+ return this->getHasher().result();
+ }
+
+protected:
+ explicit HashBuilderBase(HasherT &Hasher) : Hasher(Hasher) {}
+
+ template <typename... ArgTypes>
+ explicit HashBuilderBase(ArgTypes &&...Args)
+ : OptionalHasher(in_place, std::forward<ArgTypes>(Args)...),
+ Hasher(*OptionalHasher) {}
+
+private:
+ Optional<HasherT> OptionalHasher;
+ HasherT &Hasher;
+};
+
+/// Implementation of the `HashBuilder` interface.
+///
+/// `support::endianness::native` is not supported. `HashBuilder` is
+/// expected to canonicalize `support::endianness::native` to one of
+/// `support::endianness::big` or `support::endianness::little`.
+template <typename HasherT, support::endianness Endianness>
+class HashBuilderImpl : public HashBuilderBase<HasherT> {
+ static_assert(Endianness != support::endianness::native,
+ "HashBuilder should canonicalize endianness");
+ /// Trait to indicate whether a type's bits can be hashed directly (after
+ /// endianness correction).
+ template <typename U>
+ struct IsHashableData
+ : std::integral_constant<bool, is_integral_or_enum<U>::value> {};
+
+public:
+ explicit HashBuilderImpl(HasherT &Hasher)
+ : HashBuilderBase<HasherT>(Hasher) {}
+ template <typename... ArgTypes>
+ explicit HashBuilderImpl(ArgTypes &&...Args)
+ : HashBuilderBase<HasherT>(Args...) {}
+
+ /// Implement hashing for hashable data types, e.g. integral or enum values.
+ template <typename T>
+ std::enable_if_t<IsHashableData<T>::value, HashBuilderImpl &> add(T Value) {
+ return adjustForEndiannessAndAdd(Value);
+ }
+
+ /// Support hashing `ArrayRef`.
+ ///
+ /// `Value.size()` is taken into account to ensure cases like
+ /// ```
+ /// builder.add({1});
+ /// builder.add({2, 3});
+ /// ```
+ /// and
+ /// ```
+ /// builder.add({1, 2});
+ /// builder.add({3});
+ /// ```
+ /// do not collide.
+ template <typename T> HashBuilderImpl &add(ArrayRef<T> Value) {
+ // As of implementation time, simply calling `addRange(Value)` would also go
+ // through the `update` fast path. But that would rely on the implementation
+ // details of `ArrayRef::begin()` and `ArrayRef::end()`. Explicitly call
+ // `update` to guarantee the fast path.
+ add(Value.size());
+ if (IsHashableData<T>::value &&
+ Endianness == support::endian::system_endianness()) {
+ this->update(
+ makeArrayRef(reinterpret_cast<const uint8_t *>(Value.begin()),
+ Value.size() * sizeof(T)));
+ } else {
+ for (auto &V : Value)
+ add(V);
+ }
+ return *this;
+ }
+
+ /// Support hashing `StringRef`.
+ ///
+ /// `Value.size()` is taken into account to ensure cases like
+ /// ```
+ /// builder.add("a");
+ /// builder.add("bc");
+ /// ```
+ /// and
+ /// ```
+ /// builder.add("ab");
+ /// builder.add("c");
+ /// ```
+ /// do not collide.
+ HashBuilderImpl &add(StringRef Value) {
+ // As of implementation time, simply calling `addRange(Value)` would also go
+ // through `update`. But that would rely on the implementation of
+ // `StringRef::begin()` and `StringRef::end()`. Explicitly call `update` to
+ // guarantee the fast path.
+ add(Value.size());
+ this->update(makeArrayRef(reinterpret_cast<const uint8_t *>(Value.begin()),
+ Value.size()));
+ return *this;
+ }
+
+ template <typename T>
+ using HasAddHashT =
+ decltype(addHash(std::declval<HashBuilderImpl &>(), std::declval<T &>()));
+ /// Implement hashing for user-defined `struct`s.
+ ///
+ /// Any user-define `struct` can participate in hashing via `HashBuilder` by
+ /// providing a `addHash` templated function.
+ ///
+ /// ```
+ /// template <typename HasherT, support::endianness Endianness>
+ /// void addHash(HashBuilder<HasherT, Endianness> &HBuilder,
+ /// const UserDefinedStruct &Value);
+ /// ```
+ ///
+ /// For example:
+ /// ```
+ /// struct SimpleStruct {
+ /// char c;
+ /// int i;
+ /// };
+ ///
+ /// template <typename HasherT, support::endianness Endianness>
+ /// void addHash(HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ /// const SimpleStruct &Value) {
+ /// HBuilder.add(Value.c);
+ /// HBuilder.add(Value.i);
+ /// }
+ /// ```
+ ///
+ /// To avoid endianness issues, specializations of `addHash` should
+ /// generally rely on exising `add`, `addRange`, and `addRangeElements`
+ /// functions. If directly using `update`, an implementation must correctly
+ /// handle endianness.
+ ///
+ /// ```
+ /// struct __attribute__ ((packed)) StructWithFastHash {
+ /// int I;
+ /// char C;
+ ///
+ /// // If possible, we want to hash both `I` and `C` in a single
+ /// // `update` call for performance concerns.
+ /// template <typename HasherT, support::endianness Endianness>
+ /// friend void addHash(HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ /// const StructWithFastHash &Value) {
+ /// if (Endianness == support::endian::system_endianness()) {
+ /// HBuilder.update(makeArrayRef(
+ /// reinterpret_cast<const uint8_t *>(&Value), sizeof(Value)));
+ /// } else {
+ /// // Rely on existing `add` methods to handle endianness.
+ /// HBuilder.add(Value.I);
+ /// HBuilder.add(Value.C);
+ /// }
+ /// }
+ /// };
+ /// ```
+ ///
+ /// To avoid collisions, specialization of `addHash` for variable-size
+ /// types must take the size into account.
+ ///
+ /// For example:
+ /// ```
+ /// struct CustomContainer {
+ /// private:
+ /// size_t Size;
+ /// int Elements[100];
+ ///
+ /// public:
+ /// CustomContainer(size_t Size) : Size(Size) {
+ /// for (size_t I = 0; I != Size; ++I)
+ /// Elements[I] = I;
+ /// }
+ /// template <typename HasherT, support::endianness Endianness>
+ /// friend void addHash(HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ /// const CustomContainer &Value) {
+ /// if (Endianness == support::endian::system_endianness()) {
+ /// HBuilder.update(makeArrayRef(
+ /// reinterpret_cast<const uint8_t *>(&Value.Size),
+ /// sizeof(Value.Size) + Value.Size * sizeof(Value.Elements[0])));
+ /// } else {
+ /// // `addRange` will take care of encoding the size.
+ /// HBuilder.addRange(&Value.Elements[0], &Value.Elements[0] +
+ /// Value.Size);
+ /// }
+ /// }
+ /// };
+ /// ```
+ template <typename T>
+ std::enable_if_t<is_detected<HasAddHashT, T>::value &&
+ !IsHashableData<T>::value,
+ HashBuilderImpl &>
+ add(const T &Value) {
+ addHash(*this, Value);
+ return *this;
+ }
+
+ template <typename T1, typename T2>
+ HashBuilderImpl &add(const std::pair<T1, T2> &Value) {
+ add(Value.first);
+ add(Value.second);
+ return *this;
+ }
+
+ template <typename... Ts> HashBuilderImpl &add(const std::tuple<Ts...> &Arg) {
+ return addTupleHelper(Arg, typename std::index_sequence_for<Ts...>());
+ }
+
+ /// A convenenience variadic helper.
+ /// It simply iterates over its arguments, in order.
+ /// ```
+ /// add(Arg1, Arg2);
+ /// ```
+ /// is equivalent to
+ /// ```
+ /// add(Arg1)
+ /// add(Arg2)
+ /// ```
+ template <typename T, typename... Ts>
+ typename std::enable_if<(sizeof...(Ts) >= 1), HashBuilderImpl &>::type
+ add(const T &FirstArg, const Ts &...Args) {
+ add(FirstArg);
+ add(Args...);
+ return *this;
+ }
+
+ template <typename ForwardIteratorT>
+ HashBuilderImpl &addRange(ForwardIteratorT First, ForwardIteratorT Last) {
+ add(std::distance(First, Last));
+ return addRangeElements(First, Last);
+ }
+
+ template <typename RangeT> HashBuilderImpl &addRange(const RangeT &Range) {
+ return addRange(adl_begin(Range), adl_end(Range));
+ }
+
+ template <typename ForwardIteratorT>
+ HashBuilderImpl &addRangeElements(ForwardIteratorT First,
+ ForwardIteratorT Last) {
+ return addRangeElementsImpl(
+ First, Last,
+ typename std::iterator_traits<ForwardIteratorT>::iterator_category());
+ }
+
+ template <typename RangeT>
+ HashBuilderImpl &addRangeElements(const RangeT &Range) {
+ return addRangeElements(adl_begin(Range), adl_end(Range));
+ }
+
+ template <typename T>
+ using HasByteSwapT = decltype(support::endian::byte_swap(
+ std::declval<T &>(), support::endianness::little));
+ /// Adjust `Value` for the target endianness and add it to the hash.
+ template <typename T>
+ std::enable_if_t<is_detected<HasByteSwapT, T>::value, HashBuilderImpl &>
+ adjustForEndiannessAndAdd(const T &Value) {
+ T SwappedValue = support::endian::byte_swap(Value, Endianness);
+ this->update(makeArrayRef(reinterpret_cast<const uint8_t *>(&SwappedValue),
+ sizeof(SwappedValue)));
+ return *this;
+ }
+
+private:
+ template <typename... Ts, std::size_t... Indices>
+ HashBuilderImpl &addTupleHelper(const std::tuple<Ts...> &Arg,
+ std::index_sequence<Indices...>) {
+ add(std::get<Indices>(Arg)...);
+ return *this;
+ }
+
+ // FIXME: Once available, specialize this function for `contiguous_iterator`s,
+ // and use it for `ArrayRef` and `StringRef`.
+ template <typename ForwardIteratorT>
+ HashBuilderImpl &addRangeElementsImpl(ForwardIteratorT First,
+ ForwardIteratorT Last,
+ std::forward_iterator_tag) {
+ for (auto It = First; It != Last; ++It)
+ add(*It);
+ return *this;
+ }
+
+ template <typename T>
+ std::enable_if_t<IsHashableData<T>::value &&
+ Endianness == support::endian::system_endianness(),
+ HashBuilderImpl &>
+ addRangeElementsImpl(T *First, T *Last, std::forward_iterator_tag) {
+ this->update(makeArrayRef(reinterpret_cast<const uint8_t *>(First),
+ (Last - First) * sizeof(T)));
+ return *this;
+ }
+};
+
+/// Interface to help hash various types through a hasher type.
+///
+/// Via provided specializations of `add`, `addRange`, and `addRangeElements`
+/// functions, various types (e.g. `ArrayRef`, `StringRef`, etc.) can be hashed
+/// without requiring any knowledge of hashed types from the hasher type.
+///
+/// The only method expected from the templated hasher type `HasherT` is:
+/// * void update(ArrayRef<uint8_t> Data)
+///
+/// Additionally, the following methods will be forwarded to the hasher type:
+/// * decltype(std::declval<HasherT &>().final()) final()
+/// * decltype(std::declval<HasherT &>().result()) result()
+///
+/// From a user point of view, the interface provides the following:
+/// * `template<typename T> add(const T &Value)`
+/// The `add` function implements hashing of various types.
+/// * `template <typename ItT> void addRange(ItT First, ItT Last)`
+/// The `addRange` function is designed to aid hashing a range of values.
+/// It explicitly adds the size of the range in the hash.
+/// * `template <typename ItT> void addRangeElements(ItT First, ItT Last)`
+/// The `addRangeElements` function is also designed to aid hashing a range of
+/// values. In contrast to `addRange`, it **ignores** the size of the range,
+/// behaving as if elements were added one at a time with `add`.
+///
+/// User-defined `struct` types can participate in this interface by providing
+/// an `addHash` templated function. See the associated template specialization
+/// for details.
+///
+/// This interface does not impose requirements on the hasher
+/// `update(ArrayRef<uint8_t> Data)` method. We want to avoid collisions for
+/// variable-size types; for example for
+/// ```
+/// builder.add({1});
+/// builder.add({2, 3});
+/// ```
+/// and
+/// ```
+/// builder.add({1, 2});
+/// builder.add({3});
+/// ```
+/// . Thus, specializations of `add` and `addHash` for variable-size types must
+/// not assume that the hasher type considers the size as part of the hash; they
+/// must explicitly add the size to the hash. See for example specializations
+/// for `ArrayRef` and `StringRef`.
+///
+/// Additionally, since types are eventually forwarded to the hasher's
+/// `void update(ArrayRef<uint8_t>)` method, endianness plays a role in the hash
+/// computation (for example when computing `add((int)123)`).
+/// Specifiying a non-`native` `Endianness` template parameter allows to compute
+/// stable hash across platforms with
diff erent endianness.
+template <class HasherT, support::endianness Endianness>
+using HashBuilder =
+ HashBuilderImpl<HasherT, (Endianness == support::endianness::native
+ ? support::endian::system_endianness()
+ : Endianness)>;
+} // end namespace llvm
+
+#endif // LLVM_SUPPORT_HASHBUILDER_H
diff --git a/llvm/unittests/Support/CMakeLists.txt b/llvm/unittests/Support/CMakeLists.txt
index f1145681389e9..b44d9ee27188d 100644
--- a/llvm/unittests/Support/CMakeLists.txt
+++ b/llvm/unittests/Support/CMakeLists.txt
@@ -39,6 +39,7 @@ add_llvm_unittest(SupportTests
FormatVariadicTest.cpp
FSUniqueIDTest.cpp
GlobPatternTest.cpp
+ HashBuilderTest.cpp
Host.cpp
IndexedAccessorTest.cpp
InstructionCostTest.cpp
diff --git a/llvm/unittests/Support/HashBuilderTest.cpp b/llvm/unittests/Support/HashBuilderTest.cpp
new file mode 100644
index 0000000000000..48e25558f0781
--- /dev/null
+++ b/llvm/unittests/Support/HashBuilderTest.cpp
@@ -0,0 +1,336 @@
+//===- llvm/unittest/Support/HashBuilderTest.cpp - HashBuilder unit tests -===//
+//
+// 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 "llvm/Support/HashBuilder.h"
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/Support/MD5.h"
+#include "llvm/Support/SHA1.h"
+#include "llvm/Support/SHA256.h"
+#include "gtest/gtest.h"
+
+#include <list>
+#include <string>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+// gtest utilities and macros rely on using a single type. So wrap both the
+// hasher type and endianness.
+template <typename _HasherT, llvm::support::endianness _Endianness>
+struct HasherTAndEndianness {
+ using HasherT = _HasherT;
+ static constexpr llvm::support::endianness Endianness = _Endianness;
+};
+using HasherTAndEndiannessToTest =
+ ::testing::Types<HasherTAndEndianness<llvm::MD5, llvm::support::big>,
+ HasherTAndEndianness<llvm::MD5, llvm::support::little>,
+ HasherTAndEndianness<llvm::MD5, llvm::support::native>,
+ HasherTAndEndianness<llvm::SHA1, llvm::support::big>,
+ HasherTAndEndianness<llvm::SHA1, llvm::support::little>,
+ HasherTAndEndianness<llvm::SHA1, llvm::support::native>,
+ HasherTAndEndianness<llvm::SHA256, llvm::support::big>,
+ HasherTAndEndianness<llvm::SHA256, llvm::support::little>,
+ HasherTAndEndianness<llvm::SHA256, llvm::support::native>>;
+template <typename HasherT> class HashBuilderTest : public testing::Test {};
+TYPED_TEST_SUITE(HashBuilderTest, HasherTAndEndiannessToTest);
+
+template <typename HasherTAndEndianness>
+using HashBuilder = llvm::HashBuilder<typename HasherTAndEndianness::HasherT,
+ HasherTAndEndianness::Endianness>;
+
+template <typename HasherTAndEndianness, typename... Ts>
+static std::string hashWithBuilder(const Ts &...Args) {
+ return HashBuilder<HasherTAndEndianness>().add(Args...).final().str();
+}
+
+template <typename HasherTAndEndianness, typename... Ts>
+static std::string hashRangeWithBuilder(const Ts &...Args) {
+ return HashBuilder<HasherTAndEndianness>().addRange(Args...).final().str();
+}
+
+// All the test infrastructure relies on the variadic helpers. Test them first.
+TYPED_TEST(HashBuilderTest, VariadicHelpers) {
+ {
+ HashBuilder<TypeParam> HBuilder;
+
+ HBuilder.add(100);
+ HBuilder.add('c');
+ HBuilder.add("string");
+
+ EXPECT_EQ(HBuilder.final(), hashWithBuilder<TypeParam>(100, 'c', "string"));
+ }
+
+ {
+ HashBuilder<TypeParam> HBuilder;
+
+ std::vector<int> Vec{100, 101, 102};
+ HBuilder.addRange(Vec);
+
+ EXPECT_EQ(HBuilder.final(), hashRangeWithBuilder<TypeParam>(Vec));
+ }
+
+ {
+ HashBuilder<TypeParam> HBuilder;
+
+ std::vector<int> Vec{200, 201, 202};
+ HBuilder.addRange(Vec.begin(), Vec.end());
+
+ EXPECT_EQ(HBuilder.final(),
+ hashRangeWithBuilder<TypeParam>(Vec.begin(), Vec.end()));
+ }
+}
+
+TYPED_TEST(HashBuilderTest, AddRangeElements) {
+ HashBuilder<TypeParam> HBuilder;
+ int Values[] = {1, 2, 3};
+ HBuilder.addRangeElements(llvm::ArrayRef<int>(Values));
+ EXPECT_EQ(HBuilder.final(), hashWithBuilder<TypeParam>(1, 2, 3));
+}
+
+TYPED_TEST(HashBuilderTest, AddHashableData) {
+ using HE = TypeParam;
+
+ auto ByteSwapAndHashWithHasher = [](auto Data) {
+ using H = typename HE::HasherT;
+ constexpr auto E = HE::Endianness;
+ H Hasher;
+ auto SwappedData = llvm::support::endian::byte_swap(Data, E);
+ Hasher.update(llvm::makeArrayRef(
+ reinterpret_cast<const uint8_t *>(&SwappedData), sizeof(Data)));
+ return static_cast<std::string>(Hasher.final());
+ };
+
+ char C = 'c';
+ int32_t I = 0x12345678;
+ uint64_t UI64 = static_cast<uint64_t>(1) << 50;
+ enum TestEnumeration : uint16_t { TE_One = 1, TE_Two = 2 };
+ TestEnumeration Enum = TE_Two;
+
+ EXPECT_EQ(ByteSwapAndHashWithHasher(C), hashWithBuilder<HE>(C));
+ EXPECT_EQ(ByteSwapAndHashWithHasher(I), hashWithBuilder<HE>(I));
+ EXPECT_EQ(ByteSwapAndHashWithHasher(UI64), hashWithBuilder<HE>(UI64));
+ EXPECT_EQ(ByteSwapAndHashWithHasher(Enum), hashWithBuilder<HE>(Enum));
+}
+
+struct SimpleStruct {
+ char C;
+ int I;
+};
+
+template <typename HasherT, llvm::support::endianness Endianness>
+void addHash(llvm::HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ const SimpleStruct &Value) {
+ HBuilder.add(Value.C);
+ HBuilder.add(Value.I);
+}
+
+struct StructWithoutCopyOrMove {
+ int I;
+ StructWithoutCopyOrMove() = default;
+ StructWithoutCopyOrMove(const StructWithoutCopyOrMove &) = delete;
+ StructWithoutCopyOrMove &operator=(const StructWithoutCopyOrMove &) = delete;
+
+ template <typename HasherT, llvm::support::endianness Endianness>
+ friend void addHash(llvm::HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ const StructWithoutCopyOrMove &Value) {
+ HBuilder.add(Value.I);
+ }
+};
+
+// The struct and associated tests are simplified to avoid failures caused by
+//
diff erent alignments on
diff erent platforms.
+struct /* __attribute__((packed)) */ StructWithFastHash {
+ int I;
+ // char C;
+
+ // If possible, we want to hash both `I` and `C` in a single `update`
+ // call for performance concerns.
+ template <typename HasherT, llvm::support::endianness Endianness>
+ friend void addHash(llvm::HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ const StructWithFastHash &Value) {
+ if (Endianness == llvm::support::endian::system_endianness()) {
+ HBuilder.update(llvm::makeArrayRef(
+ reinterpret_cast<const uint8_t *>(&Value), sizeof(Value)));
+ } else {
+ // Rely on existing `add` methods to handle endianness.
+ HBuilder.add(Value.I);
+ // HBuilder.add(Value.C);
+ }
+ }
+};
+
+struct CustomContainer {
+private:
+ size_t Size;
+ int Elements[100];
+
+public:
+ CustomContainer(size_t Size) : Size(Size) {
+ for (size_t I = 0; I != Size; ++I)
+ Elements[I] = I;
+ }
+ template <typename HasherT, llvm::support::endianness Endianness>
+ friend void addHash(llvm::HashBuilderImpl<HasherT, Endianness> &HBuilder,
+ const CustomContainer &Value) {
+ if (Endianness == llvm::support::endian::system_endianness()) {
+ HBuilder.update(llvm::makeArrayRef(
+ reinterpret_cast<const uint8_t *>(&Value.Size),
+ sizeof(Value.Size) + Value.Size * sizeof(Value.Elements[0])));
+ } else {
+ HBuilder.addRange(&Value.Elements[0], &Value.Elements[0] + Value.Size);
+ }
+ }
+};
+
+TYPED_TEST(HashBuilderTest, HashUserDefinedStruct) {
+ using HE = TypeParam;
+ EXPECT_EQ(hashWithBuilder<HE>(SimpleStruct{'c', 123}),
+ hashWithBuilder<HE>('c', 123));
+ EXPECT_EQ(hashWithBuilder<HE>(StructWithoutCopyOrMove{1}),
+ hashWithBuilder<HE>(1));
+ EXPECT_EQ(hashWithBuilder<HE>(StructWithFastHash{123}),
+ hashWithBuilder<HE>(123));
+ EXPECT_EQ(hashWithBuilder<HE>(CustomContainer(3)),
+ hashWithBuilder<HE>(static_cast<size_t>(3), 0, 1, 2));
+}
+
+TYPED_TEST(HashBuilderTest, HashArrayRefHashableDataTypes) {
+ using HE = TypeParam;
+ int Values[] = {1, 20, 0x12345678};
+ llvm::ArrayRef<int> Array(Values);
+ EXPECT_NE(hashWithBuilder<HE>(Array), hashWithBuilder<HE>(1, 20, 0x12345678));
+ EXPECT_EQ(hashWithBuilder<HE>(Array),
+ hashRangeWithBuilder<HE>(Array.begin(), Array.end()));
+ EXPECT_EQ(
+ hashWithBuilder<HE>(Array),
+ hashRangeWithBuilder<HE>(Array.data(), Array.data() + Array.size()));
+}
+
+TYPED_TEST(HashBuilderTest, HashArrayRef) {
+ using HE = TypeParam;
+ int Values[] = {1, 2, 3};
+ llvm::ArrayRef<int> Array123(&Values[0], 3);
+ llvm::ArrayRef<int> Array12(&Values[0], 2);
+ llvm::ArrayRef<int> Array1(&Values[0], 1);
+ llvm::ArrayRef<int> Array23(&Values[1], 2);
+ llvm::ArrayRef<int> Array3(&Values[2], 1);
+ llvm::ArrayRef<int> ArrayEmpty(&Values[0], static_cast<size_t>(0));
+
+ auto Hash123andEmpty = hashWithBuilder<HE>(Array123, ArrayEmpty);
+ auto Hash12And3 = hashWithBuilder<HE>(Array12, Array3);
+ auto Hash1And23 = hashWithBuilder<HE>(Array1, Array23);
+ auto HashEmptyAnd123 = hashWithBuilder<HE>(ArrayEmpty, Array123);
+
+ EXPECT_NE(Hash123andEmpty, Hash12And3);
+ EXPECT_NE(Hash123andEmpty, Hash1And23);
+ EXPECT_NE(Hash123andEmpty, HashEmptyAnd123);
+ EXPECT_NE(Hash12And3, Hash1And23);
+ EXPECT_NE(Hash12And3, HashEmptyAnd123);
+ EXPECT_NE(Hash1And23, HashEmptyAnd123);
+}
+
+TYPED_TEST(HashBuilderTest, HashArrayRefNonHashableDataTypes) {
+ using HE = TypeParam;
+ SimpleStruct Values[] = {{'a', 100}, {'b', 200}};
+ llvm::ArrayRef<SimpleStruct> Array(Values);
+ EXPECT_NE(
+ hashWithBuilder<HE>(Array),
+ hashWithBuilder<HE>(SimpleStruct{'a', 100}, SimpleStruct{'b', 200}));
+}
+
+TYPED_TEST(HashBuilderTest, HashStringRef) {
+ using HE = TypeParam;
+ llvm::StringRef SEmpty("");
+ llvm::StringRef S1("1");
+ llvm::StringRef S12("12");
+ llvm::StringRef S123("123");
+ llvm::StringRef S23("23");
+ llvm::StringRef S3("3");
+
+ auto Hash123andEmpty = hashWithBuilder<HE>(S123, SEmpty);
+ auto Hash12And3 = hashWithBuilder<HE>(S12, S3);
+ auto Hash1And23 = hashWithBuilder<HE>(S1, S23);
+ auto HashEmptyAnd123 = hashWithBuilder<HE>(SEmpty, S123);
+
+ EXPECT_NE(Hash123andEmpty, Hash12And3);
+ EXPECT_NE(Hash123andEmpty, Hash1And23);
+ EXPECT_NE(Hash123andEmpty, HashEmptyAnd123);
+ EXPECT_NE(Hash12And3, Hash1And23);
+ EXPECT_NE(Hash12And3, HashEmptyAnd123);
+ EXPECT_NE(Hash1And23, HashEmptyAnd123);
+}
+
+TYPED_TEST(HashBuilderTest, HashStdString) {
+ using HE = TypeParam;
+ EXPECT_EQ(hashWithBuilder<HE>(std::string("123")),
+ hashWithBuilder<HE>(llvm::StringRef("123")));
+}
+
+TYPED_TEST(HashBuilderTest, HashStdPair) {
+ using HE = TypeParam;
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_pair(1, "string")),
+ hashWithBuilder<HE>(1, "string"));
+
+ std::pair<StructWithoutCopyOrMove, std::string> Pair;
+ Pair.first.I = 1;
+ Pair.second = "string";
+ EXPECT_EQ(hashWithBuilder<HE>(Pair), hashWithBuilder<HE>(1, "string"));
+}
+
+TYPED_TEST(HashBuilderTest, HashStdTuple) {
+ using HE = TypeParam;
+
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_tuple(1)), hashWithBuilder<HE>(1));
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_tuple(2ULL)),
+ hashWithBuilder<HE>(2ULL));
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_tuple("three")),
+ hashWithBuilder<HE>("three"));
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_tuple(1, 2ULL)),
+ hashWithBuilder<HE>(1, 2ULL));
+ EXPECT_EQ(hashWithBuilder<HE>(std::make_tuple(1, 2ULL, "three")),
+ hashWithBuilder<HE>(1, 2ULL, "three"));
+
+ std::tuple<StructWithoutCopyOrMove, std::string> Tuple;
+ std::get<0>(Tuple).I = 1;
+ std::get<1>(Tuple) = "two";
+
+ EXPECT_EQ(hashWithBuilder<HE>(Tuple), hashWithBuilder<HE>(1, "two"));
+}
+
+TYPED_TEST(HashBuilderTest, HashRangeWithForwardIterator) {
+ using HE = TypeParam;
+ std::list<int> List;
+ List.push_back(1);
+ List.push_back(2);
+ List.push_back(3);
+ EXPECT_NE(hashRangeWithBuilder<HE>(List), hashWithBuilder<HE>(1, 2, 3));
+}
+
+TEST(CustomHasher, CustomHasher) {
+ struct SumHash {
+ explicit SumHash(uint8_t Seed1, uint8_t Seed2) : Hash(Seed1 + Seed2) {}
+ void update(llvm::ArrayRef<uint8_t> Data) {
+ for (uint8_t C : Data)
+ Hash += C;
+ }
+ uint8_t Hash;
+ };
+
+ {
+ llvm::HashBuilder<SumHash, llvm::support::endianness::little> HBuilder(0,
+ 1);
+ EXPECT_EQ(HBuilder.add(0x02, 0x03, 0x400).getHasher().Hash, 0xa);
+ }
+ {
+ llvm::HashBuilder<SumHash, llvm::support::endianness::little> HBuilder(2,
+ 3);
+ EXPECT_EQ(HBuilder.add("ab", 'c').getHasher().Hash,
+ static_cast<uint8_t>(/*seeds*/ 2 + 3 + /*range size*/ 2 +
+ /*characters*/ 'a' + 'b' + 'c'));
+ }
+}
More information about the llvm-commits
mailing list