Merge pull request #332 from fboemer/fboemer/seal-pr

fboemer/hexl-1.1.0

CMakeLists.txt

@@ -219,7 +219,7 @@ if(SEAL_USE_INTEL_HEXL)
         message(STATUS "Intel HEXL: download ...")
         seal_fetch_thirdparty_content(ExternalIntelHEXL)
     else()
-        find_package(HEXL 1.0.1)
+        find_package(HEXL 1.1.0)
         if (NOT TARGET HEXL::hexl)
             FATAL_ERROR("Intel HEXL: not found")
         endif()
@@ -461,8 +461,12 @@ else()
     endif()
 
     if(SEAL_USE_INTEL_HEXL)
-        target_include_directories(seal_shared PRIVATE $<BUILD_INTERFACE:${hexl_SOURCE_DIR}/hexl/include>)
-        target_link_libraries(seal_shared PRIVATE hexl)
+        get_target_property(
+            HEXL_INCLUDE_DIR
+            HEXL::hexl
+            INTERFACE_INCLUDE_DIRECTORIES)
+        target_include_directories(seal_shared PUBLIC ${HEXL_INCLUDE_DIR})
+        target_link_libraries(seal_shared PRIVATE HEXL::hexl)
     endif()
 endif()
 

README.md

@@ -116,7 +116,7 @@ The optional dependencies and their tested versions (other versions may work as
 
 | Optional dependency                                    | Tested version | Use                                              |
 | ------------------------------------------------------ | -------------- | ------------------------------------------------ |
-| [Intel HEXL](https://github.com/intel/hexl)            | 1.0.1          | Acceleration of low-level kernels                |
+| [Intel HEXL](https://github.com/intel/hexl)            | 1.1.0          | Acceleration of low-level kernels                |
 | [Microsoft GSL](https://github.com/microsoft/GSL)      | 3.1.0          | API extensions                                   |
 | [ZLIB](https://github.com/madler/zlib)                 | 1.2.11         | Compressed serialization                         |
 | [Zstandard](https://github.com/facebook/zstd)          | 1.4.5          | Compressed serialization (much faster than ZLIB) |

cmake/ExternalIntelHEXL.cmake

@@ -4,8 +4,8 @@
 FetchContent_Declare(
     hexl
     PREFIX hexl
-    GIT_REPOSITORY https://github.com/intel/hexl.git
-    GIT_TAG v1.0.1
+    GIT_REPOSITORY https://github.com/intel/hexl
+    GIT_TAG c28943d # v1.1.0
 )
 FetchContent_GetProperties(hexl)
 

native/src/seal/util/intel_seal_ext.h

@@ -4,12 +4,119 @@
 #pragma once
 
 #ifdef SEAL_USE_INTEL_HEXL
+#include "seal/memorymanager.h"
 #include "seal/util/locks.h"
 #include <unordered_map>
 #include "hexl/hexl.hpp"
 
 namespace intel
 {
+    namespace hexl
+    {
+        // Single threaded SEAL allocator adapter
+        template <>
+        struct NTT::AllocatorAdapter<seal::MemoryPoolHandle>
+            : public AllocatorInterface<NTT::AllocatorAdapter<seal::MemoryPoolHandle>>
+        {
+            AllocatorAdapter(seal::MemoryPoolHandle handle) : handle_(std::move(handle))
+            {}
+
+            ~AllocatorAdapter()
+            {}
+
+            // interface implementations
+            void *allocate_impl(std::size_t bytes_count)
+            {
+                cache_.push_back(static_cast<seal::util::MemoryPool &>(handle_).get_for_byte_count(bytes_count));
+                return cache_.back().get();
+            }
+
+            void deallocate_impl(void *p, std::size_t n)
+            {
+                (void)n;
+                auto it = std::remove_if(
+                    cache_.begin(), cache_.end(),
+                    [p](const seal::util::Pointer<seal::seal_byte> &seal_pointer) { return p == seal_pointer.get(); });
+
+#ifdef SEAL_DEBUG
+                if (it == cache_.end())
+                {
+                    throw std::logic_error("Inconsistent single-threaded allocator cache");
+                }
+#endif
+                cache_.erase(it, cache_.end());
+            }
+
+        private:
+            seal::MemoryPoolHandle handle_;
+            std::vector<seal::util::Pointer<seal::seal_byte>> cache_;
+        };
+
+        // Thread safe policy
+        struct SimpleThreadSafePolicy
+        {
+            SimpleThreadSafePolicy() : m_ptr(std::make_unique<std::mutex>())
+            {}
+
+            std::unique_lock<std::mutex> locker()
+            {
+                return std::unique_lock<std::mutex>{ *m_ptr };
+            };
+
+        private:
+            std::unique_ptr<std::mutex> m_ptr;
+        };
+
+        // Multithreaded SEAL allocator adapter
+        template <>
+        struct NTT::AllocatorAdapter<seal::MemoryPoolHandle, SimpleThreadSafePolicy>
+            : public AllocatorInterface<NTT::AllocatorAdapter<seal::MemoryPoolHandle, SimpleThreadSafePolicy>>
+        {
+            AllocatorAdapter(seal::MemoryPoolHandle handle, SimpleThreadSafePolicy &&policy)
+                : handle_(std::move(handle)), policy_(std::move(policy))
+            {}
+
+            ~AllocatorAdapter()
+            {}
+            // interface implementations
+            void *allocate_impl(std::size_t bytes_count)
+            {
+                {
+                    // to prevent inline optimization with deadlock
+                    auto accessor = policy_.locker();
+                    cache_.push_back(static_cast<seal::util::MemoryPool &>(handle_).get_for_byte_count(bytes_count));
+                    return cache_.back().get();
+                }
+            }
+
+            void deallocate_impl(void *p, std::size_t n)
+            {
+                (void)n;
+                {
+                    // to prevent inline optimization with deadlock
+                    auto accessor = policy_.locker();
+                    auto it = std::remove_if(
+                        cache_.begin(), cache_.end(), [p](const seal::util::Pointer<seal::seal_byte> &seal_pointer) {
+                            return p == seal_pointer.get();
+                        });
+
+#ifdef SEAL_DEBUG
+                    if (it == cache_.end())
+                    {
+                        throw std::logic_error("Inconsistent multi-threaded allocator cache");
+                    }
+#endif
+                    cache_.erase(it, cache_.end());
+                }
+            }
+
+        private:
+            seal::MemoryPoolHandle handle_;
+            SimpleThreadSafePolicy policy_;
+            std::vector<seal::util::Pointer<seal::seal_byte>> cache_;
+        };
+    } // namespace hexl
+
     namespace seal_ext
     {
         struct HashPair
@@ -61,7 +168,9 @@ namespace intel
             auto ntt_it = ntt_cache_.find(key);
             if (ntt_it == ntt_cache_.end())
             {
-                ntt_it = ntt_cache_.emplace(std::move(key), intel::hexl::NTT(N, modulus, root)).first;
+                intel::hexl::NTT ntt(
+                    N, modulus, root, seal::MemoryManager::GetPool(), intel::hexl::SimpleThreadSafePolicy{});
+                ntt_it = ntt_cache_.emplace(std::move(key), std::move(ntt)).first;
             }
             return ntt_it->second;
         }

native/src/seal/util/polyarithsmallmod.cpp

@@ -37,10 +37,14 @@ namespace seal
             }
 #endif
 
+#ifdef SEAL_USE_INTEL_HEXL
+            intel::hexl::EltwiseAddMod(result, poly, scalar, coeff_count, modulus.value());
+#else
             SEAL_ITERATE(iter(poly, result), coeff_count, [&](auto I) {
                 const uint64_t x = get<0>(I);
                 get<1>(I) = add_uint_mod(x, scalar, modulus);
             });
+#endif
         }
 
         void sub_poly_scalar_coeffmod(
@@ -65,10 +69,14 @@ namespace seal
             }
 #endif
 
+#ifdef SEAL_USE_INTEL_HEXL
+            intel::hexl::EltwiseSubMod(result, poly, scalar, coeff_count, modulus.value());
+#else
             SEAL_ITERATE(iter(poly, result), coeff_count, [&](auto I) {
                 const uint64_t x = get<0>(I);
                 get<1>(I) = sub_uint_mod(x, scalar, modulus);
             });
+#endif
         }
 
         void multiply_poly_scalar_coeffmod(

native/src/seal/util/polyarithsmallmod.h

@@ -39,8 +39,13 @@ namespace seal
                 throw std::invalid_argument("modulus");
             }
 #endif
+
+#ifdef SEAL_USE_INTEL_HEXL
+            intel::hexl::EltwiseReduceMod(result, poly, coeff_count, modulus.value(), 0, 1);
+#else
             SEAL_ITERATE(
                 iter(poly, result), coeff_count, [&](auto I) { get<1>(I) = barrett_reduce_64(get<0>(I), modulus); });
+#endif
         }
 
         inline void modulo_poly_coeffs(
@@ -314,7 +319,11 @@ namespace seal
                 throw std::invalid_argument("result");
             }
 #endif
+
             const uint64_t modulus_value = modulus.value();
+#ifdef SEAL_USE_INTEL_HEXL
+            intel::hexl::EltwiseSubMod(result, operand1, operand2, coeff_count, modulus_value);
+#else
             SEAL_ITERATE(iter(operand1, operand2, result), coeff_count, [&](auto I) {
 #ifdef SEAL_DEBUG
                 if (get<0>(I) >= modulus_value)
@@ -330,6 +339,7 @@ namespace seal
                 std::int64_t borrow = sub_uint64(get<0>(I), get<1>(I), &temp_result);
                 get<2>(I) = temp_result + (modulus_value & static_cast<std::uint64_t>(-borrow));
             });
+#endif
         }
 
         inline void sub_poly_coeffmod(

native/tests/seal/modulus.cpp

@@ -104,7 +104,13 @@ namespace sealtest
         stringstream stream;
 
         Modulus mod;
+#ifdef SEAL_USE_ZLIB
         compr_mode_type compr_mode = compr_mode_type::zlib;
+#elif defined(SEAL_USE_ZSTD)
+        compr_mode_type compr_mode = compr_mode_type::zstd;
+#else
+        compr_mode_type compr_mode = compr_mode_type::none;
+#endif
         mod.save(stream, compr_mode);
 
         Modulus mod2;

native/tests/seal/serialization.cpp

@@ -52,11 +52,15 @@ namespace sealtest
         Serialization::SEALHeader header;
         ASSERT_TRUE(Serialization::IsValidHeader(header));
 
-        header.compr_mode = (compr_mode_type)0x01;
+#ifdef SEAL_USE_ZLIB
+        header.compr_mode = compr_mode_type::zlib;
         ASSERT_TRUE(Serialization::IsValidHeader(header));
+#endif
 
-        header.compr_mode = (compr_mode_type)0x02;
+#ifdef SEAL_USE_ZSTD
+        compr_mode_type compr_mode = compr_mode_type::zstd;
         ASSERT_TRUE(Serialization::IsValidHeader(header));
+#endif
 
         Serialization::SEALHeader invalid_header;
         invalid_header.magic = 0x1212;