michael@0: /* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
michael@0: /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0: 
michael@0: #include "WebGLElementArrayCache.h"
michael@0: 
michael@0: #include "nsTArray.h"
michael@0: #include "mozilla/Assertions.h"
michael@0: #include "mozilla/MemoryReporting.h"
michael@0: 
michael@0: #include <cstdlib>
michael@0: #include <cstring>
michael@0: #include <limits>
michael@0: #include <algorithm>
michael@0: 
michael@0: namespace mozilla {
michael@0: 
michael@0: static void
michael@0: SetUpperBound(uint32_t* out_upperBound, uint32_t newBound)
michael@0: {
michael@0:   if (!out_upperBound)
michael@0:       return;
michael@0: 
michael@0:   *out_upperBound = newBound;
michael@0: }
michael@0: 
michael@0: static void
michael@0: UpdateUpperBound(uint32_t* out_upperBound, uint32_t newBound)
michael@0: {
michael@0:   if (!out_upperBound)
michael@0:       return;
michael@0: 
michael@0:   *out_upperBound = std::max(*out_upperBound, newBound);
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * WebGLElementArrayCacheTree contains most of the implementation of WebGLElementArrayCache,
michael@0:  * which performs WebGL element array buffer validation for drawElements.
michael@0:  *
michael@0:  * Attention: Here lie nontrivial data structures, bug-prone algorithms, and non-canonical tweaks!
michael@0:  * Whence the explanatory comments, and compiled unit test.
michael@0:  *
michael@0:  * *** What problem are we solving here? ***
michael@0:  *
michael@0:  * WebGL::DrawElements has to validate that the elements are in range wrt the current vertex attribs.
michael@0:  * This boils down to the problem, given an array of integers, of computing the maximum in an arbitrary
michael@0:  * sub-array. The naive algorithm has linear complexity; this has been a major performance problem,
michael@0:  * see bug 569431. In that bug, we took the approach of caching the max for the whole array, which
michael@0:  * does cover most cases (DrawElements typically consumes the whole element array buffer) but doesn't
michael@0:  * help in other use cases:
michael@0:  *  - when doing "partial DrawElements" i.e. consuming only part of the element array buffer
michael@0:  *  - when doing frequent "partial buffer updates" i.e. bufferSubData calls updating parts of the
michael@0:  *    element array buffer
michael@0:  *
michael@0:  * *** The solution: a binary tree ***
michael@0:  *
michael@0:  * The solution implemented here is to use a binary tree as the cache data structure. Each tree node
michael@0:  * contains the max of its two children nodes. In this way, finding the maximum in any contiguous sub-array
michael@0:  * has log complexity instead of linear complexity.
michael@0:  *
michael@0:  * Simplistically, if the element array is
michael@0:  *
michael@0:  *     1   4   3   2
michael@0:  *
michael@0:  * then the corresponding tree is
michael@0:  *
michael@0:  *           4
michael@0:  *         _/ \_
michael@0:  *       4       3
michael@0:  *      / \     / \
michael@0:  *     1   4   3   2
michael@0:  *
michael@0:  * In practice, the bottom-most levels of the tree are both the largest to store (because they
michael@0:  * have more nodes), and the least useful performance-wise (because each node in the bottom
michael@0:  * levels concerns only few entries in the elements array buffer, it is cheap to compute).
michael@0:  *
michael@0:  * For this reason, we stop the tree a few levels above, so that each tree leaf actually corresponds
michael@0:  * to more than one element array entry.
michael@0:  *
michael@0:  * The number of levels that we "drop" is |sSkippedBottomTreeLevels| and the number of element array entries
michael@0:  * that each leaf corresponds to, is |sElementsPerLeaf|. This being a binary tree, we have
michael@0:  *
michael@0:  *   sElementsPerLeaf = 2 ^ sSkippedBottomTreeLevels.
michael@0:  *
michael@0:  * *** Storage layout of the binary tree ***
michael@0:  *
michael@0:  * We take advantage of the specifics of the situation to avoid generalist tree storage and instead
michael@0:  * store the tree entries in a vector, mTreeData.
michael@0:  *
michael@0:  * The number of leaves is given by mNumLeaves, and mTreeData is always a vector of length
michael@0:  *
michael@0:  *    2 * mNumLeaves.
michael@0:  *
michael@0:  * Its data layout is as follows: mTreeData[0] is unused, mTreeData[1] is the root node,
michael@0:  * then at offsets 2..3 is the tree level immediately below the root node, then at offsets 4..7
michael@0:  * is the tree level below that, etc.
michael@0:  *
michael@0:  * The figure below illustrates this by writing at each tree node the offset into mTreeData at
michael@0:  * which it is stored:
michael@0:  *
michael@0:  *           1
michael@0:  *         _/ \_
michael@0:  *       2       3
michael@0:  *      / \     / \
michael@0:  *     4   5   6   7
michael@0:  *    ...
michael@0:  *
michael@0:  * Thus, under the convention that the root level is level 0, we see that level N is stored at offsets
michael@0:  *
michael@0:  *    [ 2^n .. 2^(n+1) - 1 ]
michael@0:  *
michael@0:  * in mTreeData. Likewise, all the usual tree operations have simple mathematical expressions in
michael@0:  * terms of mTreeData offsets, see all the methods such as ParentNode, LeftChildNode, etc.
michael@0:  *
michael@0:  * *** Design constraint: element types aren't known at buffer-update time ***
michael@0:  *
michael@0:  * Note that a key constraint that we're operating under, is that we don't know the types of the elements
michael@0:  * by the time WebGL bufferData/bufferSubData methods are called. The type of elements is only
michael@0:  * specified in the drawElements call. This means that we may potentially have to store caches for
michael@0:  * multiple element types, for the same element array buffer. Since we don't know yet how many
michael@0:  * element types we'll eventually support (extensions add more), the concern about memory usage is serious.
michael@0:  * This is addressed by sSkippedBottomTreeLevels as explained above. Of course, in the typical
michael@0:  * case where each element array buffer is only ever used with one type, this is also addressed
michael@0:  * by having WebGLElementArrayCache lazily create trees for each type only upon first use.
michael@0:  *
michael@0:  * Another consequence of this constraint is that when invalidating the trees, we have to invalidate
michael@0:  * all existing trees. So if trees for types uint8_t, uint16_t and uint32_t have ever been constructed for this buffer,
michael@0:  * every subsequent invalidation will have to invalidate all trees even if one of the types is never
michael@0:  * used again. This implies that it is important to minimize the cost of invalidation i.e.
michael@0:  * do lazy updates upon use as opposed to immediately updating invalidated trees. This poses a problem:
michael@0:  * it is nontrivial to keep track of the part of the tree that's invalidated. The current solution
michael@0:  * can only keep track of an invalidated interval, from |mFirstInvalidatedLeaf| to |mLastInvalidatedLeaf|.
michael@0:  * The problem is that if one does two small, far-apart partial buffer updates, the resulting invalidated
michael@0:  * area is very large even though only a small part of the array really needed to be invalidated.
michael@0:  * The real solution to this problem would be to use a smarter data structure to keep track of the
michael@0:  * invalidated area, probably an interval tree. Meanwhile, we can probably live with the current situation
michael@0:  * as the unfavorable case seems to be a small corner case: in order to run into performance issues,
michael@0:  * the number of bufferSubData in between two consecutive draws must be small but greater than 1, and
michael@0:  * the partial buffer updates must be small and far apart. Anything else than this corner case
michael@0:  * should run fast in the current setting.
michael@0:  */
michael@0: template<typename T>
michael@0: struct WebGLElementArrayCacheTree
michael@0: {
michael@0:   // A too-high sSkippedBottomTreeLevels would harm the performance of small drawElements calls
michael@0:   // A too-low sSkippedBottomTreeLevels would cause undue memory usage.
michael@0:   // The current value has been validated by some benchmarking. See bug 732660.
michael@0:   static const size_t sSkippedBottomTreeLevels = 3;
michael@0:   static const size_t sElementsPerLeaf = 1 << sSkippedBottomTreeLevels;
michael@0:   static const size_t sElementsPerLeafMask = sElementsPerLeaf - 1; // sElementsPerLeaf is POT
michael@0: 
michael@0: private:
michael@0:   WebGLElementArrayCache& mParent;
michael@0:   FallibleTArray<T> mTreeData;
michael@0:   size_t mNumLeaves;
michael@0:   bool mInvalidated;
michael@0:   size_t mFirstInvalidatedLeaf;
michael@0:   size_t mLastInvalidatedLeaf;
michael@0: 
michael@0: public:
michael@0:   WebGLElementArrayCacheTree(WebGLElementArrayCache& p)
michael@0:     : mParent(p)
michael@0:     , mNumLeaves(0)
michael@0:     , mInvalidated(false)
michael@0:     , mFirstInvalidatedLeaf(0)
michael@0:     , mLastInvalidatedLeaf(0)
michael@0:   {
michael@0:     ResizeToParentSize();
michael@0:   }
michael@0: 
michael@0:   T GlobalMaximum() const {
michael@0:     MOZ_ASSERT(!mInvalidated);
michael@0:     return mTreeData[1];
michael@0:   }
michael@0: 
michael@0:   // returns the index of the parent node; if treeIndex=1 (the root node),
michael@0:   // the return value is 0.
michael@0:   static size_t ParentNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return treeIndex >> 1;
michael@0:   }
michael@0: 
michael@0:   static bool IsRightNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return treeIndex & 1;
michael@0:   }
michael@0: 
michael@0:   static bool IsLeftNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return !IsRightNode(treeIndex);
michael@0:   }
michael@0: 
michael@0:   static size_t SiblingNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return treeIndex ^ 1;
michael@0:   }
michael@0: 
michael@0:   static size_t LeftChildNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex);
michael@0:     return treeIndex << 1;
michael@0:   }
michael@0: 
michael@0:   static size_t RightChildNode(size_t treeIndex) {
michael@0:     MOZ_ASSERT(treeIndex);
michael@0:     return SiblingNode(LeftChildNode(treeIndex));
michael@0:   }
michael@0: 
michael@0:   static size_t LeftNeighborNode(size_t treeIndex, size_t distance = 1) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return treeIndex - distance;
michael@0:   }
michael@0: 
michael@0:   static size_t RightNeighborNode(size_t treeIndex, size_t distance = 1) {
michael@0:     MOZ_ASSERT(treeIndex > 1);
michael@0:     return treeIndex + distance;
michael@0:   }
michael@0: 
michael@0:   size_t LeafForElement(size_t element) {
michael@0:     size_t leaf = element / sElementsPerLeaf;
michael@0:     MOZ_ASSERT(leaf < mNumLeaves);
michael@0:     return leaf;
michael@0:   }
michael@0: 
michael@0:   size_t LeafForByte(size_t byte) {
michael@0:     return LeafForElement(byte / sizeof(T));
michael@0:   }
michael@0: 
michael@0:   // Returns the index, into the tree storage, where a given leaf is stored
michael@0:   size_t TreeIndexForLeaf(size_t leaf) {
michael@0:     // See above class comment. The tree storage is an array of length 2*mNumLeaves.
michael@0:     // The leaves are stored in its second half.
michael@0:     return leaf + mNumLeaves;
michael@0:   }
michael@0: 
michael@0:   static size_t LastElementUnderSameLeaf(size_t element) {
michael@0:     return element | sElementsPerLeafMask;
michael@0:   }
michael@0: 
michael@0:   static size_t FirstElementUnderSameLeaf(size_t element) {
michael@0:     return element & ~sElementsPerLeafMask;
michael@0:   }
michael@0: 
michael@0:   static size_t NextMultipleOfElementsPerLeaf(size_t numElements) {
michael@0:     return ((numElements - 1) | sElementsPerLeafMask) + 1;
michael@0:   }
michael@0: 
michael@0:   bool Validate(T maxAllowed, size_t firstLeaf, size_t lastLeaf,
michael@0:                 uint32_t* out_upperBound)
michael@0:   {
michael@0:     MOZ_ASSERT(!mInvalidated);
michael@0: 
michael@0:     size_t firstTreeIndex = TreeIndexForLeaf(firstLeaf);
michael@0:     size_t lastTreeIndex  = TreeIndexForLeaf(lastLeaf);
michael@0: 
michael@0:     while (true) {
michael@0:       // given that we tweak these values in nontrivial ways, it doesn't hurt to do
michael@0:       // this sanity check
michael@0:       MOZ_ASSERT(firstTreeIndex <= lastTreeIndex);
michael@0: 
michael@0:       // final case where there is only 1 node to validate at the current tree level
michael@0:       if (lastTreeIndex == firstTreeIndex) {
michael@0:         const T& curData = mTreeData[firstTreeIndex];
michael@0:         UpdateUpperBound(out_upperBound, curData);
michael@0:         return curData <= maxAllowed;
michael@0:       }
michael@0: 
michael@0:       // if the first node at current tree level is a right node, handle it individually
michael@0:       // and replace it with its right neighbor, which is a left node
michael@0:       if (IsRightNode(firstTreeIndex)) {
michael@0:         const T& curData = mTreeData[firstTreeIndex];
michael@0:         UpdateUpperBound(out_upperBound, curData);
michael@0:         if (curData > maxAllowed)
michael@0:           return false;
michael@0:         firstTreeIndex = RightNeighborNode(firstTreeIndex);
michael@0:       }
michael@0: 
michael@0:       // if the last node at current tree level is a left node, handle it individually
michael@0:       // and replace it with its left neighbor, which is a right node
michael@0:       if (IsLeftNode(lastTreeIndex)) {
michael@0:         const T& curData = mTreeData[lastTreeIndex];
michael@0:         UpdateUpperBound(out_upperBound, curData);
michael@0:         if (curData > maxAllowed)
michael@0:           return false;
michael@0:         lastTreeIndex = LeftNeighborNode(lastTreeIndex);
michael@0:       }
michael@0: 
michael@0:       // at this point it can happen that firstTreeIndex and lastTreeIndex "crossed" each
michael@0:       // other. That happens if firstTreeIndex was a right node and lastTreeIndex was its
michael@0:       // right neighor: in that case, both above tweaks happened and as a result, they ended
michael@0:       // up being swapped: lastTreeIndex is now the _left_ neighbor of firstTreeIndex.
michael@0:       // When that happens, there is nothing left to validate.
michael@0:       if (lastTreeIndex == LeftNeighborNode(firstTreeIndex)) {
michael@0:         return true;
michael@0:       }
michael@0: 
michael@0:       // walk up 1 level
michael@0:       firstTreeIndex = ParentNode(firstTreeIndex);
michael@0:       lastTreeIndex = ParentNode(lastTreeIndex);
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   template<typename U>
michael@0:   static U NextPowerOfTwo(U x) {
michael@0:     U result = 1;
michael@0:     while (result < x)
michael@0:       result <<= 1;
michael@0:     MOZ_ASSERT(result >= x);
michael@0:     MOZ_ASSERT((result & (result - 1)) == 0);
michael@0:     return result;
michael@0:   }
michael@0: 
michael@0:   bool ResizeToParentSize()
michael@0:   {
michael@0:     size_t numberOfElements = mParent.ByteSize() / sizeof(T);
michael@0:     size_t requiredNumLeaves = (numberOfElements + sElementsPerLeaf - 1) / sElementsPerLeaf;
michael@0: 
michael@0:     size_t oldNumLeaves = mNumLeaves;
michael@0:     mNumLeaves = NextPowerOfTwo(requiredNumLeaves);
michael@0:     Invalidate(0, mParent.ByteSize() - 1);
michael@0: 
michael@0:     // see class comment for why we the tree storage size is 2 * mNumLeaves
michael@0:     if (!mTreeData.SetLength(2 * mNumLeaves)) {
michael@0:       return false;
michael@0:     }
michael@0:     if (mNumLeaves != oldNumLeaves) {
michael@0:       memset(mTreeData.Elements(), 0, mTreeData.Length() * sizeof(mTreeData[0]));
michael@0:     }
michael@0:     return true;
michael@0:   }
michael@0: 
michael@0:   void Invalidate(size_t firstByte, size_t lastByte);
michael@0: 
michael@0:   void Update();
michael@0: 
michael@0:   size_t SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf) const
michael@0:   {
michael@0:     return aMallocSizeOf(this) + mTreeData.SizeOfExcludingThis(aMallocSizeOf);
michael@0:   }
michael@0: };
michael@0: 
michael@0: // TreeForType: just a template helper to select the right tree object for a given
michael@0: // element type.
michael@0: template<typename T>
michael@0: struct TreeForType {};
michael@0: 
michael@0: template<>
michael@0: struct TreeForType<uint8_t>
michael@0: {
michael@0:   static WebGLElementArrayCacheTree<uint8_t>*& Run(WebGLElementArrayCache *b) { return b->mUint8Tree; }
michael@0: };
michael@0: 
michael@0: template<>
michael@0: struct TreeForType<uint16_t>
michael@0: {
michael@0:   static WebGLElementArrayCacheTree<uint16_t>*& Run(WebGLElementArrayCache *b) { return b->mUint16Tree; }
michael@0: };
michael@0: 
michael@0: template<>
michael@0: struct TreeForType<uint32_t>
michael@0: {
michael@0:   static WebGLElementArrayCacheTree<uint32_t>*& Run(WebGLElementArrayCache *b) { return b->mUint32Tree; }
michael@0: };
michael@0: 
michael@0: // When the buffer gets updated from firstByte to lastByte,
michael@0: // calling this method will notify the tree accordingly
michael@0: template<typename T>
michael@0: void WebGLElementArrayCacheTree<T>::Invalidate(size_t firstByte, size_t lastByte)
michael@0: {
michael@0:   lastByte = std::min(lastByte, mNumLeaves * sElementsPerLeaf * sizeof(T) - 1);
michael@0:   if (firstByte > lastByte) {
michael@0:     return;
michael@0:   }
michael@0: 
michael@0:   size_t firstLeaf = LeafForByte(firstByte);
michael@0:   size_t lastLeaf = LeafForByte(lastByte);
michael@0: 
michael@0:   if (mInvalidated) {
michael@0:     mFirstInvalidatedLeaf = std::min(firstLeaf, mFirstInvalidatedLeaf);
michael@0:     mLastInvalidatedLeaf = std::max(lastLeaf, mLastInvalidatedLeaf);
michael@0:   } else {
michael@0:     mInvalidated = true;
michael@0:     mFirstInvalidatedLeaf = firstLeaf;
michael@0:     mLastInvalidatedLeaf = lastLeaf;
michael@0:   }
michael@0: }
michael@0: 
michael@0: 
michael@0: // When tree has been partially invalidated, from mFirstInvalidatedLeaf to
michael@0: // mLastInvalidatedLeaf, calling this method will 1) update the leaves in this interval
michael@0: // from the raw buffer data, and 2) propagate this update up the tree
michael@0: template<typename T>
michael@0: void WebGLElementArrayCacheTree<T>::Update()
michael@0: {
michael@0:   if (!mInvalidated) {
michael@0:     return;
michael@0:   }
michael@0: 
michael@0:   MOZ_ASSERT(mLastInvalidatedLeaf < mNumLeaves);
michael@0: 
michael@0:   size_t firstTreeIndex = TreeIndexForLeaf(mFirstInvalidatedLeaf);
michael@0:   size_t lastTreeIndex = TreeIndexForLeaf(mLastInvalidatedLeaf);
michael@0: 
michael@0:   // Step #1: initialize the tree leaves from plain buffer data.
michael@0:   // That is, each tree leaf must be set to the max of the |sElementsPerLeaf| corresponding
michael@0:   // buffer entries.
michael@0:   // condition-less scope to prevent leaking this scope's variables into the code below
michael@0:   {
michael@0:     // treeIndex is the index of the tree leaf we're writing, i.e. the destination index
michael@0:     size_t treeIndex = firstTreeIndex;
michael@0:     // srcIndex is the index in the source buffer
michael@0:     size_t srcIndex = mFirstInvalidatedLeaf * sElementsPerLeaf;
michael@0:     size_t numberOfElements = mParent.ByteSize() / sizeof(T);
michael@0:     while (treeIndex <= lastTreeIndex) {
michael@0:       T m = 0;
michael@0:       size_t a = srcIndex;
michael@0:       size_t srcIndexNextLeaf = std::min(a + sElementsPerLeaf, numberOfElements);
michael@0:       for (; srcIndex < srcIndexNextLeaf; srcIndex++) {
michael@0:         m = std::max(m, mParent.Element<T>(srcIndex));
michael@0:       }
michael@0:       mTreeData[treeIndex] = m;
michael@0:       treeIndex++;
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   // Step #2: propagate the values up the tree. This is simply a matter of walking up
michael@0:   // the tree and setting each node to the max of its two children.
michael@0:   while (firstTreeIndex > 1) {
michael@0: 
michael@0:     // move up 1 level
michael@0:     firstTreeIndex = ParentNode(firstTreeIndex);
michael@0:     lastTreeIndex = ParentNode(lastTreeIndex);
michael@0: 
michael@0:     // fast-exit case where only one node is invalidated at the current level
michael@0:     if (firstTreeIndex == lastTreeIndex) {
michael@0:       mTreeData[firstTreeIndex] = std::max(mTreeData[LeftChildNode(firstTreeIndex)], mTreeData[RightChildNode(firstTreeIndex)]);
michael@0:       continue;
michael@0:     }
michael@0: 
michael@0:     // initialize local iteration variables: child and parent.
michael@0:     size_t child = LeftChildNode(firstTreeIndex);
michael@0:     size_t parent = firstTreeIndex;
michael@0: 
michael@0:     // the unrolling makes this look more complicated than it is; the plain non-unrolled
michael@0:     // version is in the second while loop below
michael@0:     const int unrollSize = 8;
michael@0:     while (RightNeighborNode(parent, unrollSize - 1) <= lastTreeIndex)
michael@0:     {
michael@0:       for (int unroll = 0; unroll < unrollSize; unroll++)
michael@0:       {
michael@0:         T a = mTreeData[child];
michael@0:         child = RightNeighborNode(child);
michael@0:         T b = mTreeData[child];
michael@0:         child = RightNeighborNode(child);
michael@0:         mTreeData[parent] = std::max(a, b);
michael@0:         parent = RightNeighborNode(parent);
michael@0:       }
michael@0:     }
michael@0:     // plain non-unrolled version, used to terminate the job after the last unrolled iteration
michael@0:     while (parent <= lastTreeIndex)
michael@0:     {
michael@0:       T a = mTreeData[child];
michael@0:       child = RightNeighborNode(child);
michael@0:       T b = mTreeData[child];
michael@0:       child = RightNeighborNode(child);
michael@0:       mTreeData[parent] = std::max(a, b);
michael@0:       parent = RightNeighborNode(parent);
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   mInvalidated = false;
michael@0: }
michael@0: 
michael@0: WebGLElementArrayCache::~WebGLElementArrayCache() {
michael@0:   delete mUint8Tree;
michael@0:   delete mUint16Tree;
michael@0:   delete mUint32Tree;
michael@0:   free(mUntypedData);
michael@0: }
michael@0: 
michael@0: bool WebGLElementArrayCache::BufferData(const void* ptr, size_t byteSize) {
michael@0:   mByteSize = byteSize;
michael@0:   if (mUint8Tree)
michael@0:     if (!mUint8Tree->ResizeToParentSize())
michael@0:       return false;
michael@0:   if (mUint16Tree)
michael@0:     if (!mUint16Tree->ResizeToParentSize())
michael@0:       return false;
michael@0:   if (mUint32Tree)
michael@0:     if (!mUint32Tree->ResizeToParentSize())
michael@0:       return false;
michael@0:   mUntypedData = realloc(mUntypedData, byteSize);
michael@0:   if (!mUntypedData)
michael@0:     return false;
michael@0:   BufferSubData(0, ptr, byteSize);
michael@0:   return true;
michael@0: }
michael@0: 
michael@0: void WebGLElementArrayCache::BufferSubData(size_t pos, const void* ptr, size_t updateByteSize) {
michael@0:   if (!updateByteSize) return;
michael@0:   if (ptr)
michael@0:       memcpy(static_cast<uint8_t*>(mUntypedData) + pos, ptr, updateByteSize);
michael@0:   else
michael@0:       memset(static_cast<uint8_t*>(mUntypedData) + pos, 0, updateByteSize);
michael@0:   InvalidateTrees(pos, pos + updateByteSize - 1);
michael@0: }
michael@0: 
michael@0: void WebGLElementArrayCache::InvalidateTrees(size_t firstByte, size_t lastByte)
michael@0: {
michael@0:   if (mUint8Tree)
michael@0:     mUint8Tree->Invalidate(firstByte, lastByte);
michael@0:   if (mUint16Tree)
michael@0:     mUint16Tree->Invalidate(firstByte, lastByte);
michael@0:   if (mUint32Tree)
michael@0:     mUint32Tree->Invalidate(firstByte, lastByte);
michael@0: }
michael@0: 
michael@0: template<typename T>
michael@0: bool
michael@0: WebGLElementArrayCache::Validate(uint32_t maxAllowed, size_t firstElement,
michael@0:                                  size_t countElements, uint32_t* out_upperBound)
michael@0: {
michael@0:   SetUpperBound(out_upperBound, 0);
michael@0: 
michael@0:   // if maxAllowed is >= the max T value, then there is no way that a T index could be invalid
michael@0:   uint32_t maxTSize = std::numeric_limits<T>::max();
michael@0:   if (maxAllowed >= maxTSize) {
michael@0:     SetUpperBound(out_upperBound, maxTSize);
michael@0:     return true;
michael@0:   }
michael@0: 
michael@0:   T maxAllowedT(maxAllowed);
michael@0: 
michael@0:   // integer overflow must have been handled earlier, so we assert that maxAllowedT
michael@0:   // is exactly the max allowed value.
michael@0:   MOZ_ASSERT(uint32_t(maxAllowedT) == maxAllowed);
michael@0: 
michael@0:   if (!mByteSize || !countElements)
michael@0:     return true;
michael@0: 
michael@0:   WebGLElementArrayCacheTree<T>*& tree = TreeForType<T>::Run(this);
michael@0:   if (!tree) {
michael@0:     tree = new WebGLElementArrayCacheTree<T>(*this);
michael@0:   }
michael@0: 
michael@0:   size_t lastElement = firstElement + countElements - 1;
michael@0: 
michael@0:   tree->Update();
michael@0: 
michael@0:   // fast exit path when the global maximum for the whole element array buffer
michael@0:   // falls in the allowed range
michael@0:   T globalMax = tree->GlobalMaximum();
michael@0:   if (globalMax <= maxAllowedT)
michael@0:   {
michael@0:     SetUpperBound(out_upperBound, globalMax);
michael@0:     return true;
michael@0:   }
michael@0: 
michael@0:   const T* elements = Elements<T>();
michael@0: 
michael@0:   // before calling tree->Validate, we have to validate ourselves the boundaries of the elements span,
michael@0:   // to round them to the nearest multiple of sElementsPerLeaf.
michael@0:   size_t firstElementAdjustmentEnd = std::min(lastElement,
michael@0:                                             tree->LastElementUnderSameLeaf(firstElement));
michael@0:   while (firstElement <= firstElementAdjustmentEnd) {
michael@0:     const T& curData = elements[firstElement];
michael@0:     UpdateUpperBound(out_upperBound, curData);
michael@0:     if (curData > maxAllowedT)
michael@0:       return false;
michael@0:     firstElement++;
michael@0:   }
michael@0:   size_t lastElementAdjustmentEnd = std::max(firstElement,
michael@0:                                            tree->FirstElementUnderSameLeaf(lastElement));
michael@0:   while (lastElement >= lastElementAdjustmentEnd) {
michael@0:     const T& curData = elements[lastElement];
michael@0:     UpdateUpperBound(out_upperBound, curData);
michael@0:     if (curData > maxAllowedT)
michael@0:       return false;
michael@0:     lastElement--;
michael@0:   }
michael@0: 
michael@0:   // at this point, for many tiny validations, we're already done.
michael@0:   if (firstElement > lastElement)
michael@0:     return true;
michael@0: 
michael@0:   // general case
michael@0:   return tree->Validate(maxAllowedT,
michael@0:                         tree->LeafForElement(firstElement),
michael@0:                         tree->LeafForElement(lastElement),
michael@0:                         out_upperBound);
michael@0: }
michael@0: 
michael@0: bool
michael@0: WebGLElementArrayCache::Validate(GLenum type, uint32_t maxAllowed,
michael@0:                                  size_t firstElement, size_t countElements,
michael@0:                                  uint32_t* out_upperBound)
michael@0: {
michael@0:   if (type == LOCAL_GL_UNSIGNED_BYTE)
michael@0:     return Validate<uint8_t>(maxAllowed, firstElement, countElements, out_upperBound);
michael@0:   if (type == LOCAL_GL_UNSIGNED_SHORT)
michael@0:     return Validate<uint16_t>(maxAllowed, firstElement, countElements, out_upperBound);
michael@0:   if (type == LOCAL_GL_UNSIGNED_INT)
michael@0:     return Validate<uint32_t>(maxAllowed, firstElement, countElements, out_upperBound);
michael@0: 
michael@0:   MOZ_ASSERT(false, "Invalid type.");
michael@0:   return false;
michael@0: }
michael@0: 
michael@0: size_t
michael@0: WebGLElementArrayCache::SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf) const
michael@0: {
michael@0:   size_t uint8TreeSize  = mUint8Tree  ? mUint8Tree->SizeOfIncludingThis(aMallocSizeOf) : 0;
michael@0:   size_t uint16TreeSize = mUint16Tree ? mUint16Tree->SizeOfIncludingThis(aMallocSizeOf) : 0;
michael@0:   size_t uint32TreeSize = mUint32Tree ? mUint32Tree->SizeOfIncludingThis(aMallocSizeOf) : 0;
michael@0:   return aMallocSizeOf(this) +
michael@0:           mByteSize +
michael@0:           uint8TreeSize +
michael@0:           uint16TreeSize +
michael@0:           uint32TreeSize;
michael@0: }
michael@0: 
michael@0: } // end namespace mozilla