paged_allocator.h

#pragma once
#include "gaia/config/config.h"
 
#include <cinttypes>
#include <cstdint>
#include <type_traits>
 
#include "gaia/cnt/fwd_llist.h"
#include "gaia/cnt/sarray.h"
#include "gaia/core/bit_utils.h"
#include "gaia/core/dyn_singleton.h"
#include "gaia/core/utility.h"
#include "gaia/mem/mem_alloc.h"
#include "gaia/meta/type_info.h"
#include "gaia/util/logging.h"
 
namespace gaia {
  namespace mem {
    static constexpr uint32_t MemoryBlockAlignment = 16;
    static constexpr uint32_t MemoryBlockBytesDefault = 32768;
    static constexpr uint32_t MemoryBlockUsableOffset = sizeof(uintptr_t);
 
    struct GAIA_API MemoryPageHeader {
      void* m_data;
 
      MemoryPageHeader(void* ptr): m_data(ptr) {}
    };
 
    template <typename T, uint32_t RequestedBlockSize>
    struct MemoryPage: MemoryPageHeader, cnt::fwd_llist_base<MemoryPage<T, RequestedBlockSize>> {
      static constexpr uint32_t next_multiple_of_alignment(uint32_t num) {
        return (num + (MemoryBlockAlignment - 1)) & uint32_t(-(int32_t)MemoryBlockAlignment);
      }
      static constexpr uint32_t calculate_block_size() {
        if constexpr (RequestedBlockSize == 0)
          return next_multiple_of_alignment(MemoryBlockBytesDefault);
        else
          return next_multiple_of_alignment(RequestedBlockSize);
      }
 
      static constexpr uint32_t MemoryBlockBytes = calculate_block_size();
      static constexpr uint16_t NBlocks = 48;
      static constexpr uint16_t NBlocks_Bits = (uint16_t)core::count_bits(NBlocks);
      static constexpr uint32_t InvalidBlockId = NBlocks + 1;
      static constexpr uint32_t BlockArrayBytes = ((uint32_t)NBlocks_Bits * (uint32_t)NBlocks + 7) / 8;
 
      using Page = MemoryPage<T, RequestedBlockSize>;
      using BlockArray = cnt::sarray<uint8_t, BlockArrayBytes>;
      using BitView = core::bit_view<NBlocks_Bits>;
 
      BlockArray m_blocks;
 
      uint32_t m_blockCnt : NBlocks_Bits;
      uint32_t m_usedBlocks : NBlocks_Bits;
      uint32_t m_nextFreeBlock : NBlocks_Bits;
      uint32_t m_freeBlocks : NBlocks_Bits;
      // uint32_t m_unused : 8;
 
      MemoryPage(void* ptr):
          MemoryPageHeader(ptr), m_blockCnt(0), m_usedBlocks(0), m_nextFreeBlock(0), m_freeBlocks(0) {
        // One cacheline long on x86. The point is for this to be as small as possible
        static_assert(sizeof(MemoryPage) <= 64);
      }
 
      void write_block_idx(uint32_t blockIdx, uint32_t value) {
        const uint32_t bitPosition = blockIdx * NBlocks_Bits;
 
        GAIA_ASSERT(bitPosition < NBlocks * NBlocks_Bits);
        GAIA_ASSERT(value <= InvalidBlockId);
 
        BitView{{(uint8_t*)m_blocks.data(), BlockArrayBytes}}.set(bitPosition, (uint8_t)value);
      }
 
      uint8_t read_block_idx(uint32_t blockIdx) const {
        const uint32_t bitPosition = blockIdx * NBlocks_Bits;
 
        GAIA_ASSERT(bitPosition < NBlocks * NBlocks_Bits);
 
        return BitView{{(uint8_t*)m_blocks.data(), BlockArrayBytes}}.get(bitPosition);
      }
 
      GAIA_NODISCARD void* alloc_block() {
        auto StoreBlockAddress = [&](uint32_t index) {
          // Encode info about the block's page in the memory block.
          // The actual pointer returned is offset by MemoryBlockUsableOffset bytes
          uint8_t* pMemoryBlock = (uint8_t*)m_data + (index * MemoryBlockBytes);
          GAIA_ASSERT((uintptr_t)pMemoryBlock % MemoryBlockAlignment == 0);
          mem::unaligned_ref<uintptr_t>{pMemoryBlock} = (uintptr_t)this;
          return (void*)(pMemoryBlock + MemoryBlockUsableOffset);
        };
 
        // We don't want to go out of range for new blocks
        GAIA_ASSERT(!full() && "Trying to allocate too many blocks!");
 
        if (m_freeBlocks == 0U) {
          const auto index = m_blockCnt;
          ++m_usedBlocks;
          ++m_blockCnt;
          write_block_idx(index, index);
 
          return StoreBlockAddress(index);
        }
 
        GAIA_ASSERT(m_nextFreeBlock < m_blockCnt && "Block allocator recycle list broken!");
 
        ++m_usedBlocks;
        --m_freeBlocks;
 
        const auto index = m_nextFreeBlock;
        m_nextFreeBlock = read_block_idx(m_nextFreeBlock);
 
        return StoreBlockAddress(index);
      }
 
      void free_block(void* pBlock) {
        GAIA_ASSERT(m_usedBlocks > 0);
        GAIA_ASSERT(m_freeBlocks <= NBlocks);
 
        auto ReadBlockAddress = [&](void* pMemory) {
          // Offset the chunk memory so we get the real block address
          const auto* pMemoryBlock = (uint8_t*)pMemory - MemoryBlockUsableOffset;
          // Page pointer is written to the start of the memory block
          [[maybe_unused]] const auto* pPage = (Page**)pMemoryBlock;
          GAIA_ASSERT(*pPage == this);
          const auto blckAddr = (uintptr_t)pMemoryBlock;
          GAIA_ASSERT(blckAddr % 16 == 0);
          const auto dataAddr = (uintptr_t)m_data;
          const auto blockIdx = (uint32_t)((blckAddr - dataAddr) / MemoryBlockBytes);
          return blockIdx;
        };
        const auto blockIdx = ReadBlockAddress(pBlock);
 
        // Update our implicit list
        if (m_freeBlocks == 0U)
          write_block_idx(blockIdx, InvalidBlockId);
        else
          write_block_idx(blockIdx, m_nextFreeBlock);
        m_nextFreeBlock = blockIdx;
 
        ++m_freeBlocks;
        --m_usedBlocks;
      }
 
      GAIA_NODISCARD uint32_t used_blocks_cnt() const {
        return m_usedBlocks;
      }
 
      GAIA_NODISCARD bool full() const {
        return used_blocks_cnt() >= NBlocks;
      }
 
      GAIA_NODISCARD bool empty() const {
        return used_blocks_cnt() == 0;
      }
    };
 
    template <typename T, uint32_t RequestedBlockSize>
    struct MemoryPageContainer {
      cnt::fwd_llist<MemoryPage<T, RequestedBlockSize>> pagesFree;
      cnt::fwd_llist<MemoryPage<T, RequestedBlockSize>> pagesFull;
 
      GAIA_NODISCARD bool empty() const {
        return pagesFree.empty() && pagesFull.empty();
      }
    };
 
    struct GAIA_API MemoryPageStats final {
      uint64_t mem_total;
      uint64_t mem_used;
      uint32_t num_pages;
      uint32_t num_pages_free;
    };
 
    namespace detail {
      template <typename T, uint32_t RequestedBlockSize>
      class PagedAllocatorImpl;
    }
 
    template <typename T, uint32_t RequestedBlockSize = 0>
    using PagedAllocator = core::dyn_singleton<detail::PagedAllocatorImpl<T, RequestedBlockSize>>;
 
    namespace detail {
 
      template <typename T, uint32_t RequestedBlockSize>
      class PagedAllocatorImpl {
        friend ::gaia::mem::PagedAllocator<T, RequestedBlockSize>;
 
        inline static char s_strPageData[256]{};
        inline static char s_strMemPage[256]{};
 
        using Page = MemoryPage<T, RequestedBlockSize>;
        using PageContainer = MemoryPageContainer<T, RequestedBlockSize>;
 
        PageContainer m_pages;
        bool m_isDone = false;
 
      private:
        PagedAllocatorImpl() {
          // PagedAllocatorImpl is only used as a singleton so the constructor is going to be called just once.
          // Therefore, the strings are only going to be set once.
          auto ct_name = meta::type_info::name<T>();
          const auto ct_name_len = (uint32_t)ct_name.size();
          GAIA_STRCPY(s_strPageData, 256, "PageData_");
          memcpy((void*)&s_strPageData[9], (const void*)ct_name.data(), ct_name_len);
          s_strPageData[9 + ct_name_len] = 0;
          GAIA_STRCPY(s_strMemPage, 256, "MemPage_");
          memcpy((void*)&s_strMemPage[8], (const void*)ct_name.data(), ct_name_len);
          s_strMemPage[8 + ct_name_len] = 0;
        }
 
        void on_delete() {
          flush();
 
          // Make sure there are no leaks
          auto memStats = stats();
          if (memStats.mem_total != 0) {
            GAIA_ASSERT2(false, "Paged allocator leaking memory");
            GAIA_LOG_W("Paged allocator leaking memory!");
            diag();
          }
        }
 
      public:
        ~PagedAllocatorImpl() {
          on_delete();
        }
 
        PagedAllocatorImpl(PagedAllocatorImpl&& world) = delete;
        PagedAllocatorImpl(const PagedAllocatorImpl& world) = delete;
        PagedAllocatorImpl& operator=(PagedAllocatorImpl&&) = delete;
        PagedAllocatorImpl& operator=(const PagedAllocatorImpl&) = delete;
 
        void* alloc([[maybe_unused]] uint32_t dummy) {
          void* pBlock = nullptr;
 
          // Find first page with available space
          auto* pPage = m_pages.pagesFree.first;
          GAIA_ASSERT(pPage == nullptr || !pPage->full());
          if (pPage == nullptr) {
            // Allocate a new page if no free page was found
            pPage = alloc_page();
            m_pages.pagesFree.link(pPage);
          }
 
          // Allocate a new chunk of memory
          pBlock = pPage->alloc_block();
 
          // Handle full pages
          if (pPage->full()) {
            // Remove the page from the open list
            m_pages.pagesFree.unlink(pPage);
            // Move our page to the full list
            m_pages.pagesFull.link(pPage);
          }
 
          return pBlock;
        }
 
        GAIA_CLANG_WARNING_PUSH()
        // Memory is aligned so we can silence this warning
        GAIA_CLANG_WARNING_DISABLE("-Wcast-align")
 
        void free(void* pBlock) {
          // Decode the page from the address
          const auto pageAddr = *(uintptr_t*)((uint8_t*)pBlock - MemoryBlockUsableOffset);
          GAIA_ASSERT(pageAddr % MemoryBlockAlignment == 0);
          auto* pPage = (Page*)pageAddr;
          const bool wasFull = pPage->full();
 
#if GAIA_ASSERT_ENABLED
          if (wasFull) {
            const auto res = m_pages.pagesFull.has(pPage);
            GAIA_ASSERT(res && "Memory page couldn't be found among full pages");
          } else {
            const auto res = m_pages.pagesFree.has(pPage);
            GAIA_ASSERT(res && "Memory page couldn't be found among free pages");
          }
#endif
 
          // Free the chunk
          pPage->free_block(pBlock);
 
          // Update lists
          if (wasFull) {
            // Our page is no longer full
            m_pages.pagesFull.unlink(pPage);
            // Move our page to the open list
            m_pages.pagesFree.link(pPage);
          }
 
          // Special handling for the allocator signaled to destroy itself
          if (m_isDone) {
            // Remove the page right away
            if (pPage->empty()) {
              GAIA_ASSERT(!m_pages.pagesFree.empty());
              m_pages.pagesFree.unlink(pPage);
            }
 
            try_delete_this();
          }
        }
 
        GAIA_CLANG_WARNING_POP()
 
        
        MemoryPageStats stats() const {
          MemoryPageStats stats{};
 
          stats.num_pages = (uint32_t)m_pages.pagesFree.size() + (uint32_t)m_pages.pagesFull.size();
          stats.num_pages_free = (uint32_t)m_pages.pagesFree.size();
          stats.mem_total = stats.num_pages * (size_t)Page::MemoryBlockBytes * Page::NBlocks;
          stats.mem_used = m_pages.pagesFull.size() * (size_t)Page::MemoryBlockBytes * Page::NBlocks;
          for (const auto& page: m_pages.pagesFree)
            stats.mem_used += page.used_blocks_cnt() * (size_t)Page::MemoryBlockBytes;
 
          return stats;
        }
 
        void flush() {
          for (auto it = m_pages.pagesFree.begin(); it != m_pages.pagesFree.end();) {
            auto* pPage = &(*it);
            ++it;
 
            // Skip non-empty pages
            if (!pPage->empty())
              continue;
 
            m_pages.pagesFree.unlink(pPage);
            free_page(pPage);
          }
        }
 
        void diag() const {
          auto memStats = stats();
          GAIA_LOG_N("PagedAllocator %p stats", (void*)this);
          GAIA_LOG_N("  Allocated: %" PRIu64 " B", memStats.mem_total);
          GAIA_LOG_N("  Used: %" PRIu64 " B", memStats.mem_total - memStats.mem_used);
          GAIA_LOG_N("  Overhead: %" PRIu64 " B", memStats.mem_used);
          GAIA_LOG_N(
              "  Utilization: %.1f%%",
              memStats.mem_total != 0 ? 100.0 * ((double)memStats.mem_used / (double)memStats.mem_total) : 0.0);
          GAIA_LOG_N("  Pages: %u", memStats.num_pages);
          GAIA_LOG_N("  Free pages: %u", memStats.num_pages_free);
        }
 
      private:
        static Page* alloc_page() {
          const uint32_t size = Page::NBlocks * Page::MemoryBlockBytes;
          auto* pPageData = mem::AllocHelper::alloc_alig<uint8_t>(&s_strPageData[0], MemoryBlockAlignment, size);
          auto* pMemoryPage = mem::AllocHelper::alloc<Page>(&s_strMemPage[0]);
          return new (pMemoryPage) Page(pPageData);
        }
 
        static void free_page(Page* pMemoryPage) {
          GAIA_ASSERT(pMemoryPage != nullptr);
 
          mem::AllocHelper::free_alig(&s_strPageData[0], pMemoryPage->m_data);
          pMemoryPage->~MemoryPage();
          mem::AllocHelper::free(&s_strMemPage[0], pMemoryPage);
        }
 
        void done() {
          m_isDone = true;
        }
 
        void try_delete_this() {
          // When there is nothing left, delete the allocator
          if (m_pages.empty())
            delete this;
        }
      };
 
    } // namespace detail
  } // namespace mem
} // namespace gaia