allocator, string, print work

alloc.cpp

@@ -133,10 +133,10 @@ static void defaultAllocatorInit() {
     DefaultAllocatorInited = true;
 }
 
-Allocator* Allocator::GetDefault() {
+Allocator& Allocator :: GetDefault() {
     ULE_TYPES_H_FTAG;
     if (!DefaultAllocatorInited) defaultAllocatorInit();
-    return &DefaultAllocator;
+    return DefaultAllocator;
 }
 
 //================================================================================ 
@@ -174,13 +174,18 @@ void* Arena :: Alloc(u32 sizeInBytes) {
     ULE_TYPES_H_FTAG;
 
     u8* p = this->buffer + this->index;
+    u8* end = this->buffer + this->bufferSizeInBytes;
+    if ((p + sizeInBytes) >= end) {
+        return null;
+    }
     this->index += sizeInBytes;
-    //String::memset(p, 0, sizeInBytes);
+
     return (void*) p;
 
     /*
     u8* p = this->buffer + this->index;
     u32 offset = (u32) alignForward2((u64) p, 64);
+    //String::memset(p, 0, sizeInBytes);
 
     #pragma clang diagnostic push
     #pragma clang diagnostic ignored "-Wint-to-void-pointer-cast"
@@ -201,6 +206,11 @@ void Arena :: Clear() {
     ULE_TYPES_H_FTAG;
     this->index = 0;
 }
+
+float Arena :: PercentUsed() {
+    ULE_TYPES_H_FTAG;
+    return ((float)this->index)/((float)this->bufferSizeInBytes);
+}
 //================================================================================ 
 
 struct StackAllocator {
@@ -213,5 +223,3 @@ struct PoolAllocator {
     u8* buffer;
 };
 
-
-

alloc.h

@@ -20,6 +20,7 @@ typedef void  (*clearer)     (                void*);
 typedef void  (*destroyer)   (                void*);
 
 // operating system allocator wrappers
+// the overloads with the 'void*' are to make them conform to the allocator interface above.
 extern void* pMalloc(size_t size);
 extern void* pMalloc(size_t size, void* allocatorState);
 extern void* pCalloc(size_t maxNumOfElements, size_t elementSize);
@@ -33,15 +34,15 @@ extern void pFree(const void* ptr, void* allocatorState);
 
 
 struct Allocator {
-    void* state;
     mallocator mallocate;
     callocator callocate;
     reallocator reallocate;
-    freeer free;            // releases a specific piece of memory
+    freeer free;            // releases a specific piece/chunk of memory, identified by pointer.
     clearer clear;          // should release all the memory owned by this allocator at once.
     destroyer destroy;      // releases all the memory owned by this allocator, and also destroys the allocator.
+    void* state;
 
-    static Allocator* GetDefault();
+    static Allocator& GetDefault();
 
     Allocator() {
         this->state = null;
@@ -62,6 +63,24 @@ struct Allocator {
         this->clear = null;
         this->destroy = null;
     }
+
+    Allocator(
+        mallocator mallocate,
+        callocator callocate = null,
+        reallocator reallocate = null,
+        freeer free = null,
+        clearer clear = null,
+        destroyer destroy = null,
+        void* state = null
+    ) {
+        this->mallocate = mallocate;
+        this->callocate = callocate;
+        this->reallocate = reallocate;
+        this->free = free;
+        this->clear = clear;
+        this->destroy = destroy;
+        this->state = state;
+    }
 };
 
 struct Arena {
@@ -73,6 +92,7 @@ struct Arena {
 
     void* Alloc(u32 sizeInBytes);
     void Clear();
+    float PercentUsed();
 };
 
 #endif

array.hpp

@@ -1,4 +1,5 @@
 
+#pragma once
 #ifndef ULE_ARRAY_H
 #define ULE_ARRAY_H
 
@@ -7,7 +8,7 @@
 #include "config.h"
 #include "alloc.h" // allocators...
 #include "serialize.h" // serialization
-#include "string.h" // String::memcpy
+#include "string.h" // String::memcpy, String::memset
 #include "types.h" // type definitions
 
 
@@ -15,32 +16,38 @@
 // should work with any data type for T including primitive types
 // some initial |capacity| is heap-allocated and a pointer is stored to it as |data|
 // the |length| of the array, or number of filled slots is also tracked.
-//
-// it implements a single constructor, and operator new. no destructor, or operator overloading besides that.
-// remember to use ->data or .data to actually access the underlying array.
-//
-// overhead:
-//      size of the struct will be 128 bits on 64-bit platforms
-//      note that if you heap allocate this structure and store it on another struct, you will have to chase two pointers to get at the data.
-//      to avoid this, I often include this struct in other structs so there's only one pointer dereference,
-//      just like including a raw pointer array + length in your struct.
-//      because I like to do this, automatic destructors are not useful.
-//
 template <typename T>
 struct Array {
+    T* data;
     u32 length;
     u32 capacity;
-    T* data;
+    Allocator* allocator;
 
-    Array<T>(u32 _capacity = 8) {
+    Array<T>(u32 _capacity = 8, Allocator* allocator = &Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
+        this->data     = (T*) allocator->mallocate(sizeof(T) * _capacity, allocator->state);
+        String::memset(this->data, '\0', sizeof(T) * _capacity);
         this->length   = 0;
         this->capacity = _capacity;
-        this->data     = (T*) pCalloc(sizeof (T), _capacity);
+        this->allocator = allocator;
+    }
+    static Array<T>* Init(u32 _capacity = 8, Allocator* allocator = &Allocator::GetDefault()) {
+        Array<T>* array  = allocator->mallocate(sizeof(Array<T>), allocator->state);
+        array->allocator = allocator;
+        array->length    = 0;
+        array->capacity  = _capacity;
+        const size_t size = sizeof(T) * _capacity;
+        array->data      = (T*) allocator->mallocate(size, array->allocator->state);
+        String::memset(array->data, '\0', size);
+        return array;
     }
-    void* operator new(size_t size) {
+
+    // function call overhead + dev 'massert' bounds-checking overhead.
+    // you can use "->data[i]" if you know you can't be out of bounds.
+    T& Array::operator[](u32 index) const {
         ULE_TYPES_H_FTAG;
-        return pMalloc((u32) size);
+        massert(index >= 0 && index < this->length, "array access out of bounds!");
+        return this->data[index];
     }
 
     void checkIfShouldGrow() {
@@ -133,6 +140,10 @@ struct Array {
     // returns the next address into which you can store a T. makes sure there's enough room first.
     // it is irresponsible to call this and then not store a T in that address. this increments length,
     // reserving the next spot for you.
+    //
+    // C++ 'placement new' is the more safe/standard way to do this, but it requires your initialization
+    // logic is in a constructor somewhere, which isn't necessarily the case for you, and isn't for us.
+    //
     T* pushNextAddrPromise() {
         ULE_TYPES_H_FTAG;
         this->checkIfShouldGrow();
@@ -228,6 +239,21 @@ struct Array {
         return this->length == this->capacity;
     }
 
+    void sort(bool(*comparator)(T a, T b)) {
+        ULE_TYPES_H_FTAG;
+        massert(comparator != null, "can't call sort with a null comparator");
+        const auto length = this->length;
+        for (u32 i = 0; i < length; i++) {
+            for (u32 j = i + 1; j < length; j++) {
+                if (comparator(this->data[i], this->data[j])) {
+                    auto temp     = this->data[i];
+                    this->data[i] = this->data[j];
+                    this->data[j] = temp;
+                }
+            }
+        }
+    }
+
     void clear() {
         ULE_TYPES_H_FTAG;
         this->length = 0;
@@ -261,6 +287,15 @@ static void deserialize(char** buffer, Array<T>* array) {
     ULE_TYPES_H_FTAG;
     deserialize(buffer, &array->length);
     deserialize(buffer, &array->capacity);
+
+    // |array| should have already been allocated, including its |data| member,
+    // but this may not be sufficient to write in the data we now wish to.
+    // realloc to be sure.
+    array->data = (T*) pRealloc(array->data, sizeof(T) * array->capacity);
+    if (array->data == null) {
+        massert(false, "failed to realloc when deserializing an array.");
+    }
+
     for (u32 i = 0; i < array->length; i++) {
         deserialize(buffer, array->data + i);
     }

file.cpp

@@ -1,7 +1,8 @@
 
+#ifndef _CRT_SECURE_NO_WARNINGS
 #define _CRT_SECURE_NO_WARNINGS
 #include <stdio.h> // fopen, fseek, ftell, fclose
-#undef _CRT_SECURE_NO_WARNINGS
+#endif
 
 #include "alloc.h"
 #include "array.hpp"

print.cpp

@@ -48,6 +48,18 @@ void println(const char* format, ...) {
     va_end(args);
 }
 
+void printBitsLE(const size_t count, void* ptr) {
+    u8* b = (u8*) ptr;
+    u8 byte;
+    for (s32 i = count - 1; i >= 0; i--) {
+        for (s32 j = 7; j >= 0; j--) {
+            byte = (b[i] >> j) & 1;
+            print("%u", byte);
+        }
+    }
+    println();
+}
+
 /**
  * Prints a stack trace.
  * Implementation varies for Win32 vs. *nix
@@ -134,6 +146,7 @@ void writeMinidump(void* exceptionPointers) { // 'EXCEPTION_POINTERS*' actually
 }
 #else
 void writeMinidump(void* exceptionPointers) {} // stub... does nothing on Unix
+
 // OSX and Linux stacktrace stuff.
 #include <execinfo.h> // backtrace, backtrace_symbols
 #include <cxxabi.h> // abi::__cxa_demangle
@@ -232,14 +245,14 @@ void _debug(
 ) {
     ULE_TYPES_H_FTAG;
     if (format == null) {
-        print("[%s, %s] [%s:%d] %sdebug:%s null\n", date, time, filename, line, ANSI_BLUE, ANSI_RESET);
+        print("[%s, %s] [%s:%d] null\n", date, time, filename, line);
         return;
     }
 
     va_list args;
     va_start(args, format);
 
-    print("[%s, %s] [%s:%d] %sdebug:%s ", date, time, filename, line, ANSI_BLUE, ANSI_RESET);
+    print("[%s, %s] [%s:%d] ", date, time, filename, line);
     vprintln(format, args);
 
     va_end(args);

print.h

@@ -81,8 +81,13 @@ extern void vprintln(const char* format, va_list args);
 extern void print(const char* format, ...);
 extern void println(const char* format, ...);
 
+// prints the |count| * 8 little endian bits pointed to by |ptr|
+extern void printBitsLE(const size_t count, void* ptr);
+
 // Prints a stack trace, or concatenates the stack trace to |string| if it is not null.
 extern void trace(String* string = null);
+
+// on windows only
 extern void writeMinidump(void* exceptionPointers);
 
 // This ends the program and calls trace(). generally you should use 'massert' instead

string.h

@@ -7,20 +7,20 @@
 #include "types.h"
 #include "alloc.h"
 
-#include <string.h> // @TODO remove this
+#include <cstring> // memcpy, memset, memcmp
 
 #define STB_SPRINTF_IMPLEMENTATION
 #define STB_SPRINTF_STATIC
 #include <stb/stb_sprintf.h>
 
-#define STR_MEMALLOC  pMalloc
-#define STR_MEMFREE   pFree
-#define STR_ASSERT    assert
+#define STR_ASSERT assert
 #define STR_IMPLEMENTATION
 #define STR_SUPPORT_STD_STRING 0
 #define STR_DEFINE_STR32 0 // the type Str32, which would normally be available, conflicts with a type in MacTypes.h
 
 
+static Allocator* _stringClassAllocator = &Allocator::GetDefault(); // only used for the string class, string functions you pass an allocator if it allocates
+
 // 'String' is a datatype, but it also is a namespace for a bunch of static 'char*' operations that 
 // you would normally find in the <cstring> or <string.h> header
 // The datatype is a modified version of a string class developed by Omar Cornut: https://github.com/ocornut/str
@@ -99,7 +99,7 @@ public:
         return c == '0' || c == '1';
     }
 
-    static inline char* intToString(u64 integer) {
+    static inline char* intToString(u64 integer) { // @ALLOC
         ULE_TYPES_H_FTAG;
         u32 capacity = 10;
         u32* remainders = (u32*) pMalloc(sizeof (u32) * capacity);
@@ -191,23 +191,24 @@ public:
     }
 
     // heap allocates a copy of |string| and returns a pointer to it.
-    static inline char* cpy(const char* string, u32 length, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* cpy(const char* string, u32 length, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
-        char* buffer = (char*) allocator->mallocate(sizeof (char) * (length + 1), allocator->state);
+        assert(&allocator == &Allocator::GetDefault());
+        char* buffer = (char*) pMalloc(sizeof(char) * (length+1)); //allocator.mallocate(sizeof (char) * (length + 1), allocator.state);
 
         u32 i = 0;
         for (; i < length; i++) {
             buffer[i] = string[i];
         }
         buffer[i] = '\0';
+
         return buffer;
     }
 
     // heap allocates a copy of |string| and returns a pointer to it.
-    static inline char* cpy(const char* string, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* cpy(const char* string, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         u32 len = String::len(string);
-
         return String::cpy(string, len, allocator = Allocator::GetDefault());
     }
 
@@ -222,6 +223,30 @@ public:
         return memeq(m1, m2, l1);
     }
 
+    // assumes null termination.
+    static inline bool LexographicComparisonASCII(const char* str1, const char* str2) {
+        u32 min;
+
+        const u32 l1 = String::len(str1);
+        const u32 l2 = String::len(str2);
+        if (l1 > l2) {
+            min = l2;
+        } else {
+            min = l1;
+        }
+
+        for (u32 i = 0; i < min; i++) {
+            if (str1[i] < str2[i]) {
+                return true;
+
+            } else if (str1[i] > str2[i]) {
+                return false;
+            }
+        }
+
+        return l1 < l2;
+    }
+
     #ifdef _WIN32
     static inline size_t wcharToChar(wchar_t* wstring, char* buffer, size_t maxBufferLength) {
         ULE_TYPES_H_FTAG;
@@ -239,12 +264,16 @@ public:
 
     static inline void memcpy(void* dest, void* src, u32 size) {
         ULE_TYPES_H_FTAG;
-        u8* dest_ = (u8*) dest;
-        u8* src_ = (u8*) src;
+        // allowing c++ compilers to know we're invoking memcpy allows more aggressive optimizations, sometimes resulting in 8-9x speedup,
+        // when compared to the horrible, commented out loop below.
+        std::memcpy(dest, src, size); 
 
-        for (u32 i = 0; i < size; i++) {
-            dest_[i] = src_[i];
-        }
+        //u8* dest_ = (u8*) dest;
+        //u8* src_ = (u8*) src;
+
+        //for (u32 i = 0; i < size; i++) {
+        //    dest_[i] = src_[i];
+        //}
     }
 
     // replace all instances of |c1| in |string| with |c2|
@@ -340,7 +369,7 @@ public:
     // @TODO ALL OF THESE TRIMS
     //static inline char* trimStart(const char* str, u32 count);
     //static inline char* trimEnd(const char* str, u32 count);
-    static inline char* trim(const char* str, u32 count, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* trim(const char* str, u32 count, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         u32 length = String::len(str);
 
@@ -348,7 +377,7 @@ public:
             return (char*) "";
         }
 
-        char* buffer = (char*) allocator->mallocate(sizeof (char) * (length - 1), allocator->state);
+        char* buffer = (char*) allocator.mallocate(sizeof (char) * (length - 1), allocator.state);
 
         u32 i = 0;
         for (; i < (length - count); i++) {
@@ -360,10 +389,10 @@ public:
         return buffer;
     }
 
-    static inline char* asciiToLower(const char* str, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* asciiToLower(const char* str, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         u32 length = String::len(str);
-        char* buffer = (char*) allocator->mallocate(sizeof (char) * length + 1, allocator->state);
+        char* buffer = (char*) allocator.mallocate(sizeof (char) * length + 1, allocator.state);
         u32 i = 0;
         for (; i < length; i++) {
             buffer[i] = String::ASCII_LOWER[str[i]];
@@ -372,10 +401,10 @@ public:
         return buffer;
     }
 
-    static inline char* asciiToUpper(const char* str, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* asciiToUpper(const char* str, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         u32 length = String::len(str);
-        char* buffer = (char*) allocator->mallocate(sizeof (char) * length + 1, allocator->state);
+        char* buffer = (char*) allocator.mallocate(sizeof (char) * length + 1, allocator.state);
         u32 i = 0;
         for (; i < length; i++) {
             buffer[i] = String::ASCII_LOWER[str[i]];
@@ -384,12 +413,12 @@ public:
         return buffer;
     }
 
-    static inline char* concat(const char* str1, const char* str2, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* concat(const char* str1, const char* str2, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         u32 l1 = String::len(str1);
         u32 l2 = String::len(str2);
         u32 newLength = l1 + l2;
-        char* newBuffer = (char*) allocator->mallocate(sizeof (char) * newLength + 1, allocator->state);
+        char* newBuffer = (char*) allocator.mallocate(sizeof (char) * newLength + 1, allocator.state);
         u32 i = 0;
         for (; i < newLength; i++) {
             if (i < l1) {
@@ -419,9 +448,9 @@ public:
         return String::write(dest, src, length);
     }
 
-    static inline char* read(const char* buffer, u32 length, Allocator* allocator = Allocator::GetDefault()) {
+    static inline char* read(const char* buffer, u32 length, Allocator& allocator = Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
-        char* tk = (char*) allocator->mallocate(sizeof (char) * length + 1, allocator->state);
+        char* tk = (char*) allocator.mallocate(sizeof (char) * length + 1, allocator.state);
         u32 i = 0;
         while (i < length) {
             tk[i] = *(buffer + i);
@@ -431,6 +460,12 @@ public:
         return tk;
     }
 
+    //---------------------------------------------------------------
+    // Begin String class type
+    static inline void SetStringClassAllocator(Allocator* allocator) {
+        _stringClassAllocator = allocator;
+    }
+
     char*               Data;                   // Point to LocalBuf() or heap allocated
     int                 Capacity : 21;          // Max 2 MB
     int                 LocalBufSize : 10;      // Max 1023 bytes
@@ -469,7 +504,7 @@ public:
     inline void set_ref(const char* src) {
         ULE_TYPES_H_FTAG; 
         if (Owned && !is_using_local_buf())
-            STR_MEMFREE(Data);
+            _stringClassAllocator->free(Data, null);
         Data = src ? (char*)src : EmptyBuffer;
         Capacity = 0;
         Owned = 0;
@@ -513,7 +548,7 @@ public:
     inline void clear() {
         ULE_TYPES_H_FTAG; 
         if (Owned && !is_using_local_buf())
-            STR_MEMFREE(Data);
+            _stringClassAllocator->free(Data, null);
         if (LocalBufSize) {
             Data = local_buf();
             Data[0] = '\0';
@@ -539,7 +574,7 @@ public:
             new_capacity = LocalBufSize;
         } else {
             // Disowned or LocalBuf -> Heap
-            new_data = (char*)STR_MEMALLOC((size_t)new_capacity * sizeof(char));
+            new_data = (char*)_stringClassAllocator->mallocate((size_t)new_capacity * sizeof(char), null);
         }
 
         // string in Data might be longer than new_capacity if it wasn't owned, don't copy too much
@@ -551,7 +586,7 @@ public:
         new_data[new_capacity - 1] = 0;
 
         if (Owned && !is_using_local_buf())
-            STR_MEMFREE(Data);
+            _stringClassAllocator->free(Data, null);
 
         Data = new_data;
         Capacity = new_capacity;
@@ -565,7 +600,7 @@ public:
             return;
 
         if (Owned && !is_using_local_buf())
-            STR_MEMFREE(Data);
+            _stringClassAllocator->free(Data, null);
 
         if (new_capacity < LocalBufSize) {
             // Disowned -> LocalBuf
@@ -573,7 +608,7 @@ public:
             Capacity = LocalBufSize;
         } else {
             // Disowned or LocalBuf -> Heap
-            Data = (char*)STR_MEMALLOC((size_t)new_capacity * sizeof(char));
+            Data = (char*)_stringClassAllocator->mallocate((size_t)new_capacity * sizeof(char), null);
             Capacity = new_capacity;
         }
         Owned = 1;
@@ -585,9 +620,9 @@ public:
         int new_capacity = length() + 1;
         if (Capacity <= new_capacity) return;
 
-        char* new_data = (char*)STR_MEMALLOC((size_t)new_capacity * sizeof(char));
+        char* new_data = (char*)_stringClassAllocator->mallocate((size_t)new_capacity * sizeof(char), null);
         memcpy(new_data, Data, (size_t)new_capacity);
-        STR_MEMFREE(Data);
+        _stringClassAllocator->free(Data, null);
         Data = new_data;
         Capacity = new_capacity;
     }
@@ -725,7 +760,7 @@ public:
     inline ~String()
     {
         if (Owned && !is_using_local_buf())
-            STR_MEMFREE(Data);
+            if (_stringClassAllocator->free != null) ::_stringClassAllocator->free(Data, null);
     }
 
 protected:
@@ -831,12 +866,12 @@ public:
             new_data = (char*)this + sizeof(String); \
             new_capacity = LocalBufSize; \
         } else { \
-            new_data = (char*)STR_MEMALLOC((size_t)new_capacity * sizeof(char)); \
+            new_data = (char*)_stringClassAllocator->mallocate((size_t)new_capacity * sizeof(char), null); \
         } \
         strncpy(new_data, Data, (size_t)new_capacity - 1); \
         new_data[new_capacity - 1] = 0; \
         if (Owned && !is_using_local_buf()) \
-            STR_MEMFREE(Data); \
+            _stringClassAllocator->free(Data, null); \
         Data = new_data; \
         Capacity = new_capacity; \
         Owned = 1; \
@@ -846,14 +881,14 @@ public:
         if (new_capacity <= Capacity) \
             return; \
         if (Owned && !is_using_local_buf()) \
-            STR_MEMFREE(Data); \
+            _stringClassAllocator->free(Data, null); \
         if (new_capacity < LocalBufSize) { \
             Data = (char*)this + sizeof(String); \
             Capacity = LocalBufSize; \
         } else { \
             while (Capacity < new_capacity) { \
                 Capacity = (s32)(Capacity * 1.5f); \
-                Data = (char*) STR_MEMALLOC((size_t) Capacity * sizeof(char)); \
+                Data = (char*) _stringClassAllocator->mallocate((size_t) Capacity * sizeof(char), null); \
             } \
         } \
         Owned = 1; \

table.hpp

@@ -1,4 +1,5 @@
 
+#pragma once
 #ifndef ULE_TABLE_H
 #define ULE_TABLE_H
 
@@ -245,6 +246,11 @@ struct Table {
         return (Table<V>*) pMalloc(sizeof(Table<V>));
     }
 
+    float LoadFactor() {
+        ULE_TYPES_H_FTAG;
+        return ((float)this->length)/((float)this->lanes);
+    }
+
     V insert(const char* key, u32 keyLength, V value) {
         ULE_TYPES_H_FTAG;
         TableEntry<V>* entry = this->lookup(key, keyLength);
@@ -295,7 +301,7 @@ struct Table {
 
     // do not set |freeValues| to true unless the template parameter 'T' is a pointer,
     // and the table is responsible for freeing the memory.
-    void clear(bool freeValues = false) {
+    void clear(bool freeValues = false, bool freeKeys = true) {
         ULE_TYPES_H_FTAG;
         for (u32 i = 0; i < this->lanes; i++) {
             TableEntry<V>** lane = &this->entries[i];
@@ -304,7 +310,7 @@ struct Table {
             while (entry != null) {
                 auto next = entry->next;
 
-                pFree(entry->key);
+                if (freeKeys) pFree(entry->key);
                 if (freeValues) {
                     // @HACK - it's only relevant to free the value if it's an owned pointer
                     // (the table is effectively 'responsible' for that memory)
@@ -337,7 +343,8 @@ struct Table {
 
     void traverse(const std::function <void (TableEntry<V>*)>& entryCallback) {
         ULE_TYPES_H_FTAG;
-        for (u32 i = 0; i < this->lanes; i++) {
+        const auto lanes = this->lanes;
+        for (u32 i = 0; i < lanes; i++) {
             TableEntry<V>* entry = this->entries[i];
 
             while (entry != null) {
@@ -437,7 +444,7 @@ static void deserialize(char** buffer, Table<T>** table) {
 // a key size of 16, 32, or 64.
 //
 //================================================================================ 
-//#include <mmintrin.h>
+#include <mmintrin.h>
 template <size_t KEY_SIZE, typename std::enable_if<KEY_SIZE == 64>::type* = nullptr>
 static inline bool fixedKeySizeMemEq(u8* m1, u8* m2) {
     ULE_TYPES_H_FTAG;
@@ -515,21 +522,105 @@ struct FixedKeySizeTableEntry {
     V value;
 };
 
+template <typename V>
+struct DenseHashSetEntry {
+
+};
+
+template <typename V>
+struct DenseHashSet {
+    enum ControlByte {
+        EMPTY    = -128, // 0b10000000 // indicates a free slot, and a place to stop probing
+        DELETED  =   -2, // 0b11111110 // indicates there is no value here, but you should keep probing
+        SENTINEL =   -1, // 0b11111111 // indicates end of stream
+        //FULL =         // 0b0xxxxxxx // indicates an occupied slot
+    };
+
+    u8*        metadata; // parallel array to 'values', each byte is a 'ControlByte'.
+    u32        length;
+    u32        capacity;
+    V*         values;
+    Allocator* allocator;
+
+    // a 'group' represents a capacity for 16 elements in the table. the actual capacity will be |_numGroups * 16|
+    // we use some SSE instructions that operate on chunks of 16 at a time, so we need this constraint.
+    DenseHashSet<V>(u32 _numGroups = 4, Allocator* _allocator = &Allocator::GetDefault()) {
+        ULE_TYPES_H_FTAG;
+        this->length   = 0;
+        this->capacity = _numGroups;
+
+        const size_t valueSize    = sizeof(T)  * this->capacity * 16;
+        const size_t metadataSize = sizeof(u8) * this->capacity * 16;
+        this->values   = (T*)  _allocator->mallocate(valueSize,    _allocator->state);
+        this->metadata = (u8*) _allocator->mallocate(metadataSize, _allocator->state);
+
+        String::memset(this->values, '\0', valueSize);
+        String::memset(this->metadata, (u8) DenseHashSet::EMPTY, metadataSize);
+    }
+
+    float LoadFactor() {
+        ULE_TYPES_H_FTAG;
+        return ((float)this->length)/((float)this->lanes);
+    }
+
+    u32 H1(u32 h) {
+        ULE_TYPES_H_FTAG;
+        return h >> 7;
+    }
+
+    u8 H2(u32 h) {
+        ULE_TYPES_H_FTAG;
+        return (u8) (h & 0x7F);
+    }
+
+    u32 Match(u8 h2) const {
+        auto match = _mm_set1_epi8(h2);
+        return _mm_movemask_epi8(_mm_cmpeq_epi8(match, metadata));
+    }
+
+    V Insert(const char* key, size_t keyLength, V value) {
+        ULE_TYPES_H_FTAG;
+        // @TODO check if we need to resize
+    }
+
+    bool Lookup(const char* key, size_t keyLength, V* outValue) {
+        ULE_TYPES_H_FTAG;
+        u32 h = APHash(key, keyLength);
+        size_t group = this->H1(h) % this->capacity;
+        while (1) {
+            u8* g = this->metadata + group * 16;
+            u32 mask = this->Match(this->H2(h));
+            for (u32 i = 0; i < 32; i++) {
+                u32 idx = mask & (1 << i);
+            }
+            if (mask == 0) return false;
+            group = (group + 1) % this->capacity;
+        }
+    }
+};
+
 template <size_t KEY_SIZE, typename V>
 struct FixedKeySizeTable {
+    FixedKeySizeTableEntry<KEY_SIZE, V>** entries;
     u32 lanes;
     u32 length;
-    FixedKeySizeTableEntry<KEY_SIZE, V>** entries;
+    Allocator* allocator;
 
-    FixedKeySizeTable<KEY_SIZE, V>(u32 _lanes = 16) {
+    FixedKeySizeTable<KEY_SIZE, V>(u32 _lanes = 16, Allocator* allocator = &Allocator::GetDefault()) {
         ULE_TYPES_H_FTAG;
         this->lanes = _lanes;
         this->length = 0;
-        this->entries = (FixedKeySizeTableEntry<KEY_SIZE, V>**) pCalloc(sizeof(FixedKeySizeTableEntry<KEY_SIZE, V>*), this->lanes);
+        this->allocator = allocator;
+        this->entries = (FixedKeySizeTableEntry<KEY_SIZE, V>**) allocator->callocate(sizeof(FixedKeySizeTableEntry<KEY_SIZE, V>*), this->lanes, null);
     }
     void* operator new(size_t size) {
         ULE_TYPES_H_FTAG;
-        return (FixedKeySizeTable<KEY_SIZE, V>*) pMalloc(sizeof(FixedKeySizeTable<KEY_SIZE, V>));
+        return (FixedKeySizeTable<KEY_SIZE, V>*) this->allocator->mallocate(sizeof(FixedKeySizeTable<KEY_SIZE, V>), null);
+    }
+
+    float LoadFactor() {
+        ULE_TYPES_H_FTAG;
+        return ((float)this->length)/((float)this->lanes);
     }
 
     V insert(const char* key, V value) {
@@ -537,7 +628,7 @@ struct FixedKeySizeTable {
         FixedKeySizeTableEntry<KEY_SIZE, V>* entry = this->lookup(key);
 
         if (!entry) { // no entry with that key exists
-            entry = (FixedKeySizeTableEntry<KEY_SIZE, V>*) pCalloc(sizeof(FixedKeySizeTableEntry<KEY_SIZE, V>), 1);
+            entry = (FixedKeySizeTableEntry<KEY_SIZE, V>*) this->allocator->callocate(sizeof(FixedKeySizeTableEntry<KEY_SIZE, V>), 1, null);
             String::write((char*)entry->key, key, KEY_SIZE);
             entry->value = value;
 
@@ -604,12 +695,12 @@ struct FixedKeySizeTable {
 #pragma clang diagnostic push
 #pragma clang diagnostic ignored "-Wint-to-void-pointer-cast"
 #endif
-                    pFree((void*) entry->value);
+                    if (this->allocator->free) this->allocator->free((void*) entry->value, null);
 #ifdef __clang__
 #pragma clang diagnostic pop
 #endif
                 }
-                pFree(entry);
+                if (this->allocator->free) this->allocator->free(entry, null);
 
                 entry = next;
             }
@@ -622,9 +713,9 @@ struct FixedKeySizeTable {
 
     void traverse(const std::function <void (FixedKeySizeTableEntry<KEY_SIZE, V>*)>& entryCallback) {
         ULE_TYPES_H_FTAG;
-        for (u32 i = 0; i < this->lanes; i++) {
+        const auto lanes = this->lanes;
+        for (u32 i = 0; i < lanes; i++) {
             FixedKeySizeTableEntry<KEY_SIZE, V>* entry = this->entries[i];
-
             while (entry != null) {
                 entryCallback(entry);
                 entry = entry->next;