Make irrArray backed by std::vector (#101)

This commit is contained in:
paradust7 2022-05-21 14:56:36 -07:00 committed by GitHub
parent 593103a261
commit 3e81f38098
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
10 changed files with 147 additions and 426 deletions

@ -100,7 +100,7 @@ namespace scene
private:
//! Internal members used by CSkinnedMesh
friend class CSkinnedMesh;
bool *Moved;
char *Moved;
core::vector3df StaticPos;
core::vector3df StaticNormal;
};

@ -1,70 +0,0 @@
// Copyright (C) 2002-2012 Nikolaus Gebhardt
// This file is part of the "Irrlicht Engine".
// For conditions of distribution and use, see copyright notice in irrlicht.h
#ifndef __IRR_HEAPSORT_H_INCLUDED__
#define __IRR_HEAPSORT_H_INCLUDED__
#include "irrTypes.h"
namespace irr
{
namespace core
{
//! Sinks an element into the heap.
template<class T>
inline void heapsink(T*array, s32 element, s32 max)
{
while ((element<<1) < max) // there is a left child
{
s32 j = (element<<1);
if (j+1 < max && array[j] < array[j+1])
j = j+1; // take right child
if (array[element] < array[j])
{
T t = array[j]; // swap elements
array[j] = array[element];
array[element] = t;
element = j;
}
else
return;
}
}
//! Sorts an array with size 'size' using heapsort.
template<class T>
inline void heapsort(T* array_, s32 size)
{
// for heapsink we pretend this is not c++, where
// arrays start with index 0. So we decrease the array pointer,
// the maximum always +2 and the element always +1
T* virtualArray = array_ - 1;
s32 virtualSize = size + 2;
s32 i;
// build heap
for (i=((size-1)/2); i>=0; --i)
heapsink(virtualArray, i+1, virtualSize-1);
// sort array, leave out the last element (0)
for (i=size-1; i>0; --i)
{
T t = array_[0];
array_[0] = array_[i];
array_[i] = t;
heapsink(virtualArray, 1, i + 1);
}
}
} // end namespace core
} // end namespace irr
#endif

@ -5,9 +5,11 @@
#ifndef __IRR_ARRAY_H_INCLUDED__
#define __IRR_ARRAY_H_INCLUDED__
#include <algorithm>
#include <iterator>
#include <vector>
#include "irrTypes.h"
#include "heapsort.h"
#include "irrAllocator.h"
#include "irrMath.h"
namespace irr
@ -18,44 +20,27 @@ namespace core
//! Self reallocating template array (like stl vector) with additional features.
/** Some features are: Heap sorting, binary search methods, easier debugging.
*/
template <class T, typename TAlloc = irrAllocator<T> >
template <class T>
class array
{
public:
static_assert(!std::is_same<T, bool>::value,
"irr::core::array<T> with T = bool not supported. Use std::vector instead.");
//! Default constructor for empty array.
array() : data(0), allocated(0), used(0),
strategy(ALLOC_STRATEGY_DOUBLE), free_when_destroyed(true), is_sorted(true)
{
}
array() : m_data(), is_sorted(true)
{ }
//! Constructs an array and allocates an initial chunk of memory.
/** \param start_count Amount of elements to pre-allocate. */
explicit array(u32 start_count) : data(0), allocated(0), used(0),
strategy(ALLOC_STRATEGY_DOUBLE),
free_when_destroyed(true), is_sorted(true)
explicit array(u32 start_count) : m_data(), is_sorted(true)
{
reallocate(start_count);
m_data.reserve(start_count);
}
//! Copy constructor
array(const array<T, TAlloc>& other) : data(0)
{
*this = other;
}
//! Destructor.
/** Frees allocated memory, if set_free_when_destroyed was not set to
false by the user before. */
~array()
{
clear();
}
array(const array<T>& other) : m_data(other.m_data), is_sorted(other.is_sorted)
{ }
//! Reallocates the array, make it bigger or smaller.
/** \param new_size New size of array.
@ -65,52 +50,28 @@ public:
*/
void reallocate(u32 new_size, bool canShrink=true)
{
if (allocated==new_size)
return;
if (!canShrink && (new_size < allocated))
return;
T* old_data = data;
data = allocator.allocate(new_size); //new T[new_size];
allocated = new_size;
// copy old data
const s32 end = used < new_size ? used : new_size;
for (s32 i=0; i<end; ++i)
{
// data[i] = old_data[i];
allocator.construct(&data[i], old_data[i]);
size_t allocated = m_data.capacity();
if (new_size < allocated) {
if (canShrink)
m_data.resize(new_size);
} else {
m_data.reserve(new_size);
}
// destruct old data
for (u32 j=0; j<used; ++j)
allocator.destruct(&old_data[j]);
if (allocated < used)
used = allocated;
allocator.deallocate(old_data); //delete [] old_data;
}
//! set a new allocation strategy
/** if the maximum size of the array is unknown, you can define how big the
allocation should happen.
\param newStrategy New strategy to apply to this array. */
void setAllocStrategy ( eAllocStrategy newStrategy = ALLOC_STRATEGY_DOUBLE )
{
strategy = newStrategy;
}
//! Adds an element at back of array.
/** If the array is too small to add this new element it is made bigger.
\param element: Element to add at the back of the array. */
void push_back(const T& element)
{
insert(element, used);
m_data.push_back(element);
is_sorted = false;
}
void push_back(T&& element)
{
m_data.push_back(std::move(element));
is_sorted = false;
}
@ -121,7 +82,14 @@ public:
\param element Element to add at the back of the array. */
void push_front(const T& element)
{
insert(element);
m_data.insert(m_data.begin(), element);
is_sorted = false;
}
void push_front(T&& element)
{
m_data.insert(m_data.begin(), std::move(element));
is_sorted = false;
}
@ -131,106 +99,21 @@ public:
\param index: Where position to insert the new element. */
void insert(const T& element, u32 index=0)
{
_IRR_DEBUG_BREAK_IF(index>used) // access violation
if (used + 1 > allocated)
{
// this doesn't work if the element is in the same
// array. So we'll copy the element first to be sure
// we'll get no data corruption
const T e(element);
// increase data block
u32 newAlloc;
switch ( strategy )
{
case ALLOC_STRATEGY_DOUBLE:
newAlloc = used + 5 + (allocated < 500 ? used : used >> 2);
break;
default:
case ALLOC_STRATEGY_SAFE:
newAlloc = used + 1;
break;
}
reallocate( newAlloc);
// move array content and construct new element
// first move end one up
for (u32 i=used; i>index; --i)
{
if (i<used)
allocator.destruct(&data[i]);
allocator.construct(&data[i], data[i-1]); // data[i] = data[i-1];
}
// then add new element
if (used > index)
allocator.destruct(&data[index]);
allocator.construct(&data[index], e); // data[index] = e;
}
else
{
// element inserted not at end
if ( used > index )
{
// create one new element at the end
allocator.construct(&data[used], data[used-1]);
// move the rest of the array content
for (u32 i=used-1; i>index; --i)
{
data[i] = data[i-1];
}
// insert the new element
data[index] = element;
}
else
{
// insert the new element to the end
allocator.construct(&data[index], element);
}
}
// set to false as we don't know if we have the comparison operators
_IRR_DEBUG_BREAK_IF(index > m_data.size()) // access violation
auto pos = std::next(m_data.begin(), index);
m_data.insert(pos, element);
is_sorted = false;
++used;
}
//! Clears the array and deletes all allocated memory.
void clear()
{
if (free_when_destroyed)
{
for (u32 i=0; i<used; ++i)
allocator.destruct(&data[i]);
allocator.deallocate(data); // delete [] data;
}
data = 0;
used = 0;
allocated = 0;
// vector::clear() reduces the size to 0, but doesn't free memory.
// This swap is guaranteed to delete the allocated memory.
std::vector<T>().swap(m_data);
is_sorted = true;
}
//! Sets pointer to new array, using this as new workspace.
/** Make sure that set_free_when_destroyed is used properly.
\param newPointer: Pointer to new array of elements.
\param size: Size of the new array.
\param _is_sorted Flag which tells whether the new array is already
sorted.
\param _free_when_destroyed Sets whether the new memory area shall be
freed by the array upon destruction, or if this will be up to the user
application. */
void set_pointer(T* newPointer, u32 size, bool _is_sorted=false, bool _free_when_destroyed=true)
{
clear();
data = newPointer;
allocated = size;
used = size;
is_sorted = _is_sorted;
free_when_destroyed=_free_when_destroyed;
}
//! Set (copy) data from given memory block
/** \param newData data to set, must have newSize elements
\param newSize Amount of elements in newData
@ -240,12 +123,11 @@ public:
*/
void set_data(const T* newData, u32 newSize, bool newDataIsSorted=false, bool canShrink=false)
{
reallocate(newSize, canShrink);
set_used(newSize);
for ( u32 i=0; i<newSize; ++i)
{
data[i] = newData[i];
m_data.resize(newSize);
if (canShrink) {
m_data.shrink_to_fit();
}
std::copy(newData, newData + newSize, m_data.begin());
is_sorted = newDataIsSorted;
}
@ -255,85 +137,51 @@ public:
\param size Amount of elements in otherData */
bool equals(const T* otherData, u32 size) const
{
if (used != size)
if (m_data.size() != size)
return false;
for (u32 i=0; i<size; ++i)
if (data[i] != otherData[i])
return false;
return true;
return std::equal(m_data.begin(), m_data.end(), otherData);
}
//! Sets if the array should delete the memory it uses upon destruction.
/** Also clear and set_pointer will only delete the (original) memory
area if this flag is set to true, which is also the default. The
methods reallocate, set_used, push_back, push_front, insert, and erase
will still try to deallocate the original memory, which might cause
troubles depending on the intended use of the memory area.
\param f If true, the array frees the allocated memory in its
destructor, otherwise not. The default is true. */
void set_free_when_destroyed(bool f)
{
free_when_destroyed = f;
}
//! Sets the size of the array and allocates new elements if necessary.
/** Please note: This is only secure when using it with simple types,
because no default constructor will be called for the added elements.
\param usedNow Amount of elements now used. */
/** \param usedNow Amount of elements now used. */
void set_used(u32 usedNow)
{
if (allocated < usedNow)
reallocate(usedNow);
used = usedNow;
m_data.resize(usedNow);
}
//! Assignment operator
const array<T, TAlloc>& operator=(const array<T, TAlloc>& other)
const array<T>& operator=(const array<T>& other)
{
if (this == &other)
return *this;
strategy = other.strategy;
// (TODO: we could probably avoid re-allocations of data when (allocated < other.allocated)
if (data)
clear();
used = other.used;
free_when_destroyed = true;
m_data = other.m_data;
is_sorted = other.is_sorted;
allocated = other.allocated;
if (other.allocated == 0)
{
data = 0;
}
else
{
data = allocator.allocate(other.allocated); // new T[other.allocated];
for (u32 i=0; i<other.used; ++i)
allocator.construct(&data[i], other.data[i]); // data[i] = other.data[i];
}
return *this;
}
array<T>& operator=(const std::vector<T> &other)
{
m_data = other;
is_sorted = false;
return *this;
}
array<T>& operator=(std::vector<T> &&other)
{
m_data = std::move(other);
is_sorted = false;
return *this;
}
//! Equality operator
bool operator == (const array<T, TAlloc>& other) const
bool operator == (const array<T>& other) const
{
return equals(other.const_pointer(), other.size());
}
//! Inequality operator
bool operator != (const array<T, TAlloc>& other) const
bool operator != (const array<T>& other) const
{
return !(*this==other);
}
@ -342,36 +190,36 @@ public:
//! Direct access operator
T& operator [](u32 index)
{
_IRR_DEBUG_BREAK_IF(index>=used) // access violation
_IRR_DEBUG_BREAK_IF(index >= m_data.size()) // access violation
return data[index];
return m_data[index];
}
//! Direct const access operator
const T& operator [](u32 index) const
{
_IRR_DEBUG_BREAK_IF(index>=used) // access violation
_IRR_DEBUG_BREAK_IF(index >= m_data.size()) // access violation
return data[index];
return m_data[index];
}
//! Gets last element.
T& getLast()
{
_IRR_DEBUG_BREAK_IF(!used) // access violation
_IRR_DEBUG_BREAK_IF(m_data.empty()) // access violation
return data[used-1];
return m_data.back();
}
//! Gets last element
const T& getLast() const
{
_IRR_DEBUG_BREAK_IF(!used) // access violation
_IRR_DEBUG_BREAK_IF(m_data.empty()) // access violation
return data[used-1];
return m_data.back();
}
@ -379,7 +227,7 @@ public:
/** \return Pointer to the array. */
T* pointer()
{
return data;
return &m_data[0];
}
@ -387,7 +235,7 @@ public:
/** \return Pointer to the array. */
const T* const_pointer() const
{
return data;
return &m_data[0];
}
@ -395,7 +243,7 @@ public:
/** \return Size of elements in the array which are actually occupied. */
u32 size() const
{
return used;
return m_data.size();
}
@ -404,7 +252,7 @@ public:
allocated would be allocated_size() * sizeof(ElementTypeUsed); */
u32 allocated_size() const
{
return allocated;
return m_data.capacity();
}
@ -412,7 +260,7 @@ public:
/** \return True if the array is empty false if not. */
bool empty() const
{
return used == 0;
return m_data.empty();
}
@ -421,10 +269,11 @@ public:
O(n*log n) in worst case. */
void sort()
{
if (!is_sorted && used>1)
heapsort(data, used);
if (!is_sorted) {
std::sort(m_data.begin(), m_data.end());
is_sorted = true;
}
}
//! Performs a binary search for an element, returns -1 if not found.
@ -437,10 +286,9 @@ public:
s32 binary_search(const T& element)
{
sort();
return binary_search(element, 0, used-1);
return binary_search(element, 0, (s32)m_data.size() - 1);
}
//! Performs a binary search for an element if possible, returns -1 if not found.
/** This method is for const arrays and so cannot call sort(), if the array is
not sorted then linear_search will be used instead. Potentially very slow!
@ -450,12 +298,11 @@ public:
s32 binary_search(const T& element) const
{
if (is_sorted)
return binary_search(element, 0, used-1);
return binary_search(element, 0, (s32)m_data.size() - 1);
else
return linear_search(element);
}
//! Performs a binary search for an element, returns -1 if not found.
/** \param element: Element to search for.
\param left First left index
@ -464,31 +311,15 @@ public:
is returned. */
s32 binary_search(const T& element, s32 left, s32 right) const
{
if (!used)
if (left > right)
return -1;
s32 m;
do
{
m = (left+right)>>1;
if (element < data[m])
right = m - 1;
else
left = m + 1;
} while((element < data[m] || data[m] < element) && left<=right);
// this last line equals to:
// " while((element != array[m]) && left<=right);"
// but we only want to use the '<' operator.
// the same in next line, it is "(element == array[m])"
if (!(element < data[m]) && !(data[m] < element))
return m;
auto lpos = std::next(m_data.begin(), left);
auto rpos = std::next(m_data.begin(), right);
auto it = std::lower_bound(lpos, rpos, element);
// *it = first element in [first, last) that is >= element, or last if not found.
if (*it < element || element < *it)
return -1;
return it - m_data.begin();
}
@ -503,25 +334,11 @@ public:
s32 binary_search_multi(const T& element, s32 &last)
{
sort();
s32 index = binary_search(element, 0, used-1);
if ( index < 0 )
return index;
// The search can be somewhere in the middle of the set
// look linear previous and past the index
last = index;
while ( index > 0 && !(element < data[index - 1]) && !(data[index - 1] < element) )
{
index -= 1;
}
// look linear up
while ( last < (s32) used - 1 && !(element < data[last + 1]) && !(data[last + 1] < element) )
{
last += 1;
}
return index;
auto iters = std::equal_range(m_data.begin(), m_data.end(), element);
if (iters.first == iters.second)
return -1;
last = (iters.second - m_data.begin()) - 1;
return iters.first - m_data.begin();
}
@ -533,11 +350,10 @@ public:
is returned. */
s32 linear_search(const T& element) const
{
for (u32 i=0; i<used; ++i)
if (element == data[i])
return (s32)i;
auto it = std::find(m_data.begin(), m_data.end(), element);
if (it == m_data.end())
return -1;
return it - m_data.begin();
}
@ -549,11 +365,11 @@ public:
is returned. */
s32 linear_reverse_search(const T& element) const
{
for (s32 i=used-1; i>=0; --i)
if (data[i] == element)
return i;
auto it = std::find(m_data.rbegin(), m_data.rend(), element);
if (it == m_data.rend())
return -1;
size_t offset = it - m_data.rbegin();
return m_data.size() - offset - 1;
}
@ -563,17 +379,9 @@ public:
\param index: Index of element to be erased. */
void erase(u32 index)
{
_IRR_DEBUG_BREAK_IF(index>=used) // access violation
for (u32 i=index+1; i<used; ++i)
{
allocator.destruct(&data[i-1]);
allocator.construct(&data[i-1], data[i]); // data[i-1] = data[i];
}
allocator.destruct(&data[used-1]);
--used;
_IRR_DEBUG_BREAK_IF(index >= m_data.size()) // access violation
auto it = std::next(m_data.begin(), index);
m_data.erase(it);
}
@ -584,30 +392,14 @@ public:
\param count: Amount of elements to be erased. */
void erase(u32 index, s32 count)
{
if (index>=used || count<1)
if (index >= m_data.size() || count < 1)
return;
if (index+count>used)
count = used-index;
u32 i;
for (i=index; i<index+count; ++i)
allocator.destruct(&data[i]);
for (i=index+count; i<used; ++i)
{
if (i-count >= index+count) // not already destructed before loop
allocator.destruct(&data[i-count]);
allocator.construct(&data[i-count], data[i]); // data[i-count] = data[i];
if (i >= used-count) // those which are not overwritten
allocator.destruct(&data[i]);
count = std::min(count, (s32)m_data.size() - (s32)index);
auto first = std::next(m_data.begin(), index);
auto last = std::next(first, count);
m_data.erase(first, last);
}
used-= count;
}
//! Sets if the array is sorted
void set_sorted(bool _is_sorted)
{
@ -619,38 +411,30 @@ public:
/** Afterward this object will contain the content of the other object and the other
object will contain the content of this object.
\param other Swap content with this object */
void swap(array<T, TAlloc>& other)
void swap(array<T>& other)
{
core::swap(data, other.data);
core::swap(allocated, other.allocated);
core::swap(used, other.used);
core::swap(allocator, other.allocator); // memory is still released by the same allocator used for allocation
eAllocStrategy helper_strategy(strategy); // can't use core::swap with bitfields
strategy = other.strategy;
other.strategy = helper_strategy;
bool helper_free_when_destroyed(free_when_destroyed);
free_when_destroyed = other.free_when_destroyed;
other.free_when_destroyed = helper_free_when_destroyed;
bool helper_is_sorted(is_sorted);
is_sorted = other.is_sorted;
other.is_sorted = helper_is_sorted;
m_data.swap(other.m_data);
std::swap(is_sorted, other.is_sorted);
}
//! Pull the contents of this array as a vector.
// The array is left empty.
std::vector<T> steal()
{
std::vector<T> ret = std::move(m_data);
m_data.clear();
is_sorted = true;
return ret;
}
typedef TAlloc allocator_type;
typedef T value_type;
typedef u32 size_type;
private:
T* data;
u32 allocated;
u32 used;
TAlloc allocator;
eAllocStrategy strategy:4;
bool free_when_destroyed:1;
bool is_sorted:1;
std::vector<T> m_data;
bool is_sorted;
};
} // end namespace core
} // end namespace irr

@ -50,7 +50,6 @@
#include "EMessageBoxFlags.h"
#include "ESceneNodeTypes.h"
#include "fast_atof.h"
#include "heapsort.h"
#include "IAnimatedMesh.h"
#include "IAnimatedMeshSceneNode.h"
#include "IAttributes.h"

@ -70,9 +70,9 @@ IAnimatedMesh* COBJMeshFileLoader::createMesh(io::IReadFile* file)
const u32 WORD_BUFFER_LENGTH = 512;
core::array<core::vector3df, core::irrAllocatorFast<core::vector3df> > vertexBuffer(1000);
core::array<core::vector3df, core::irrAllocatorFast<core::vector3df> > normalsBuffer(1000);
core::array<core::vector2df, core::irrAllocatorFast<core::vector2df> > textureCoordBuffer(1000);
core::array<core::vector3df> vertexBuffer(1000);
core::array<core::vector3df> normalsBuffer(1000);
core::array<core::vector2df> textureCoordBuffer(1000);
SObjMtl * currMtl = new SObjMtl();
Materials.push_back(currMtl);

@ -112,12 +112,12 @@ bool COGLES1Driver::genericDriverInit(const core::dimension2d<u32>& screenSize,
glPixelStorei(GL_PACK_ALIGNMENT, 1);
UserClipPlane.reallocate(MaxUserClipPlanes);
UserClipPlaneEnabled.reallocate(MaxUserClipPlanes);
UserClipPlaneEnabled.resize(MaxUserClipPlanes);
for (s32 i = 0; i < MaxUserClipPlanes; ++i)
{
UserClipPlane.push_back(core::plane3df());
UserClipPlaneEnabled.push_back(false);
UserClipPlaneEnabled[i] = false;
}
for (s32 i = 0; i < ETS_COUNT; ++i)

@ -363,7 +363,7 @@ namespace video
SMaterial Material, LastMaterial;
core::array<core::plane3df> UserClipPlane;
core::array<bool> UserClipPlaneEnabled;
std::vector<bool> UserClipPlaneEnabled;
core::stringc VendorName;

@ -231,6 +231,9 @@ namespace scene
struct DefaultNodeEntry
{
DefaultNodeEntry()
{ }
DefaultNodeEntry(ISceneNode* n) :
Node(n), TextureValue(0)
{
@ -251,6 +254,9 @@ namespace scene
//! sort on distance (center) to camera
struct TransparentNodeEntry
{
TransparentNodeEntry()
{ }
TransparentNodeEntry(ISceneNode* n, const core::vector3df& camera)
: Node(n)
{

@ -1042,7 +1042,7 @@ void CSkinnedMesh::finalize()
for (i=0; i<LocalBuffers.size(); ++i)
{
Vertices_Moved.push_back( core::array<bool>() );
Vertices_Moved.push_back( core::array<char>() );
Vertices_Moved[i].set_used(LocalBuffers[i]->getVertexCount());
}

@ -196,7 +196,9 @@ private:
core::array<SJoint*> AllJoints;
core::array<SJoint*> RootJoints;
core::array< core::array<bool> > Vertices_Moved;
// bool can't be used here because std::vector<bool>
// doesn't allow taking a reference to individual elements.
core::array< core::array<char> > Vertices_Moved;
core::aabbox3d<f32> BoundingBox;