changed ll_aligned_(malloc|free) to take alignment size as a template argument

8 files changed, 43 insertions, 41 deletions

diff --git a/indra/llcommon/llalignedarray.h b/indra/llcommon/llalignedarray.h
index 6f18bfe25c..b68e9e0f82 100644
--- a/indra/llcommon/llalignedarray.h
+++ b/indra/llcommon/llalignedarray.h
@@ -60,7 +60,7 @@ LLAlignedArray<T, alignment>::LLAlignedArray()
 template <class T, U32 alignment>
 LLAlignedArray<T, alignment>::~LLAlignedArray()
 {
-	ll_aligned_free(alignment, mArray);
+	ll_aligned_free<alignment>(mArray);
 	mArray = NULL;
 	mElementCount = 0;
 	mCapacity = 0;
@@ -74,7 +74,7 @@ void LLAlignedArray<T, alignment>::push_back(const T& elem)
 	{
 		mCapacity++;
 		mCapacity *= 2;
-		T* new_buf = (T*) ll_aligned_malloc(alignment, mCapacity*sizeof(T));
+		T* new_buf = (T*) ll_aligned_malloc<alignment>(mCapacity*sizeof(T));
 		if (mArray)
 		{
 			ll_memcpy_nonaliased_aligned_16((char*)new_buf, (char*)mArray, sizeof(T)*mElementCount);
@@ -86,7 +86,7 @@ void LLAlignedArray<T, alignment>::push_back(const T& elem)
 	mArray[mElementCount++] = elem;
 
 	//delete old array here to prevent error on a.push_back(a[0])
-	ll_aligned_free(alignment, old_buf);
+	ll_aligned_free<alignment>(old_buf);
 }
 
 template <class T, U32 alignment>
@@ -95,11 +95,11 @@ void LLAlignedArray<T, alignment>::resize(U32 size)
 	if (mCapacity < size)
 	{
 		mCapacity = size+mCapacity*2;
-		T* new_buf = mCapacity > 0 ? (T*) ll_aligned_malloc(alignment, mCapacity*sizeof(T)) : NULL;
+		T* new_buf = mCapacity > 0 ? (T*) ll_aligned_malloc<alignment>(mCapacity*sizeof(T)) : NULL;
 		if (mArray)
 		{
 			ll_memcpy_nonaliased_aligned_16((char*) new_buf, (char*) mArray, sizeof(T)*mElementCount);
-			ll_aligned_free(alignment, mArray);
+			ll_aligned_free<alignment>(mArray);
 		}
 
 		/*for (U32 i = mElementCount; i < mCapacity; ++i)
diff --git a/indra/llcommon/llmemory.h b/indra/llcommon/llmemory.h
index 5ed4ea7d9e..3de59350db 100755
--- a/indra/llcommon/llmemory.h
+++ b/indra/llcommon/llmemory.h
@@ -204,37 +204,39 @@ inline void ll_aligned_free_32(void *p)
 }
 
 // general purpose dispatch functions that are forced inline so they can compile down to a single call
-LL_FORCE_INLINE void* ll_aligned_malloc(size_t alignment, size_t size)
+template<size_t ALIGNMENT>
+LL_FORCE_INLINE void* ll_aligned_malloc(size_t size)
 {
-	if (LL_DEFAULT_HEAP_ALIGN % alignment == 0)
+	if (LL_DEFAULT_HEAP_ALIGN % ALIGNMENT == 0)
 	{
 		return malloc(size);
 	}
-	else if (alignment == 16)
+	else if (ALIGNMENT == 16)
 	{
 		return ll_aligned_malloc_16(size);
 	}
-	else if (alignment == 32)
+	else if (ALIGNMENT == 32)
 	{
 		return ll_aligned_malloc_32(size);
 	}
 	else
 	{
-		return ll_aligned_malloc_fallback(size, alignment);
+		return ll_aligned_malloc_fallback(size, ALIGNMENT);
 	}
 }
 
-LL_FORCE_INLINE void ll_aligned_free(size_t alignment, void* ptr)
+template<size_t ALIGNMENT>
+LL_FORCE_INLINE void ll_aligned_free(void* ptr)
 {
-	if (alignment == LL_DEFAULT_HEAP_ALIGN)
+	if (ALIGNMENT == LL_DEFAULT_HEAP_ALIGN)
 	{
 		free(ptr);
 	}
-	else if (alignment == 16)
+	else if (ALIGNMENT == 16)
 	{
 		ll_aligned_free_16(ptr);
 	}
-	else if (alignment == 32)
+	else if (ALIGNMENT == 32)
 	{
 		return ll_aligned_free_32(ptr);
 	}
diff --git a/indra/llcommon/lltrace.h b/indra/llcommon/lltrace.h
index 325112b9b1..2f4390e4d1 100644
--- a/indra/llcommon/lltrace.h
+++ b/indra/llcommon/lltrace.h
@@ -371,40 +371,40 @@ public:
 	void* operator new(size_t size) 
 	{
 		claim_alloc(sMemStat, size);
-		return ll_aligned_malloc(ALIGNMENT, size);
+		return ll_aligned_malloc<ALIGNMENT>(size);
 	}
 
 	template<int CUSTOM_ALIGNMENT>
 	static void* aligned_new(size_t size)
 	{
 		claim_alloc(sMemStat, size);
-		return ll_aligned_malloc(CUSTOM_ALIGNMENT, size);
+		return ll_aligned_malloc<CUSTOM_ALIGNMENT>(size);
 	}
 
 	void operator delete(void* ptr, size_t size)
 	{
 		disclaim_alloc(sMemStat, size);
-		ll_aligned_free(ALIGNMENT, ptr);
+		ll_aligned_free<ALIGNMENT>(ptr);
 	}
 
 	template<int CUSTOM_ALIGNMENT>
 	static void aligned_delete(void* ptr, size_t size)
 	{
 		disclaim_alloc(sMemStat, size);
-		ll_aligned_free(CUSTOM_ALIGNMENT, ptr);
+		ll_aligned_free<CUSTOM_ALIGNMENT>(ptr);
 	}
 
 
 	void* operator new [](size_t size)
 	{
 		claim_alloc(sMemStat, size);
-		return ll_aligned_malloc(ALIGNMENT, size);
+		return ll_aligned_malloc<ALIGNMENT>(size);
 	}
 
 	void operator delete[](void* ptr, size_t size)
 	{
 		disclaim_alloc(sMemStat, size);
-		ll_aligned_free(ALIGNMENT, ptr);
+		ll_aligned_free<ALIGNMENT>(ptr);
 	}
 
 	// claim memory associated with other objects/data as our own, adding to our calculated footprint
diff --git a/indra/llcommon/lltraceaccumulators.h b/indra/llcommon/lltraceaccumulators.h
index 85873d469a..77370629d3 100644
--- a/indra/llcommon/lltraceaccumulators.h
+++ b/indra/llcommon/lltraceaccumulators.h
@@ -441,12 +441,12 @@ namespace LLTrace
 		// arrays are allocated with 32 byte alignment
 		void *operator new [](size_t size)
 		{
-			return ll_aligned_malloc(32, size);
+			return ll_aligned_malloc<32>(size);
 		}
 
 		void operator delete[](void* ptr, size_t size)
 		{
-			ll_aligned_free(32, ptr);
+			ll_aligned_free<32>(ptr);
 		}
 
 		TimeBlockAccumulator();
diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp
index 699eeb0b89..26d2ae2963 100755
--- a/indra/llmath/llvolume.cpp
+++ b/indra/llmath/llvolume.cpp
@@ -4667,7 +4667,7 @@ LLVolumeFace::~LLVolumeFace()
 
 void LLVolumeFace::freeData()
 {
-	ll_aligned_free(64, mPositions);
+	ll_aligned_free<64>(mPositions);
 	mPositions = NULL;
 
 	//normals and texture coordinates are part of the same buffer as mPositions, do not free them separately
@@ -5245,7 +5245,7 @@ void LLVolumeFace::cacheOptimize()
 	//allocate space for new buffer
 	S32 num_verts = mNumVertices;
 	S32 size = ((num_verts*sizeof(LLVector2)) + 0xF) & ~0xF;
-	LLVector4a* pos = (LLVector4a*) ll_aligned_malloc(64, sizeof(LLVector4a)*2*num_verts+size);
+	LLVector4a* pos = (LLVector4a*) ll_aligned_malloc<64>(sizeof(LLVector4a)*2*num_verts+size);
 	LLVector4a* norm = pos + num_verts;
 	LLVector2* tc = (LLVector2*) (norm + num_verts);
 
@@ -5295,7 +5295,7 @@ void LLVolumeFace::cacheOptimize()
 		mIndices[i] = new_idx[mIndices[i]];
 	}
 	
-	ll_aligned_free(64, mPositions);
+	ll_aligned_free<64>(mPositions);
 	// DO NOT free mNormals and mTexCoords as they are part of mPositions buffer
 	ll_aligned_free_16(mWeights);
 	ll_aligned_free_16(mTangents);
@@ -6023,7 +6023,7 @@ void LLVolumeFace::createTangents()
 
 void LLVolumeFace::resizeVertices(S32 num_verts)
 {
-	ll_aligned_free(64, mPositions);
+	ll_aligned_free<64>(mPositions);
 	//DO NOT free mNormals and mTexCoords as they are part of mPositions buffer
 	ll_aligned_free_16(mTangents);
 
@@ -6034,7 +6034,7 @@ void LLVolumeFace::resizeVertices(S32 num_verts)
 		//pad texture coordinate block end to allow for QWORD reads
 		S32 size = ((num_verts*sizeof(LLVector2)) + 0xF) & ~0xF;
 
-		mPositions = (LLVector4a*) ll_aligned_malloc(64, sizeof(LLVector4a)*2*num_verts+size);
+		mPositions = (LLVector4a*) ll_aligned_malloc<64>(sizeof(LLVector4a)*2*num_verts+size);
 		mNormals = mPositions+num_verts;
 		mTexCoords = (LLVector2*) (mNormals+num_verts);
 
@@ -6074,7 +6074,7 @@ void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, con
 
 		LLVector4a* old_buf = mPositions;
 
-		mPositions = (LLVector4a*) ll_aligned_malloc(64, new_size);
+		mPositions = (LLVector4a*) ll_aligned_malloc<64>(new_size);
 		mNormals = mPositions+new_verts;
 		mTexCoords = (LLVector2*) (mNormals+new_verts);
 
@@ -6090,7 +6090,7 @@ void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, con
 	//just clear tangents
 	ll_aligned_free_16(mTangents);
 	mTangents = NULL;
-		ll_aligned_free(64, old_buf);
+		ll_aligned_free<64>(old_buf);
 
 		mNumAllocatedVertices = new_verts;
 
@@ -6191,7 +6191,7 @@ void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMat
 
 	//allocate new buffer space
 	LLVector4a* old_buf = mPositions;
-	mPositions = (LLVector4a*) ll_aligned_malloc(64, new_size);
+	mPositions = (LLVector4a*) ll_aligned_malloc<64>(new_size);
 	mNormals = mPositions + new_count;
 	mTexCoords = (LLVector2*) (mNormals+new_count);
 
diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp
index 2a36a3b96f..3bbbccbad1 100755
--- a/indra/llrender/llvertexbuffer.cpp
+++ b/indra/llrender/llvertexbuffer.cpp
@@ -202,7 +202,7 @@ volatile U8* LLVBOPool::allocate(U32& name, U32 size, bool for_seed)
 			glBufferDataARB(mType, size, 0, mUsage);
 			if (mUsage != GL_DYNAMIC_COPY_ARB)
 			{ //data will be provided by application
-				ret = (U8*) ll_aligned_malloc(64, size);
+				ret = (U8*) ll_aligned_malloc<64>(size);
 			}
 		}
 		else
@@ -310,7 +310,7 @@ void LLVBOPool::cleanup()
 			
 			if (r.mClientData)
 			{
-				ll_aligned_free(64, (void*) r.mClientData);
+				ll_aligned_free<64>((void*) r.mClientData);
 			}
 
 			l.pop_front();
diff --git a/indra/llvfs/llvfile.cpp b/indra/llvfs/llvfile.cpp
index 9b57246ea1..add88fe0a3 100755
--- a/indra/llvfs/llvfile.cpp
+++ b/indra/llvfs/llvfile.cpp
@@ -137,12 +137,12 @@ U8* LLVFile::readFile(LLVFS *vfs, const LLUUID &uuid, LLAssetType::EType type, S
 	}
 	else
 	{		
-		data = (U8*) ll_aligned_malloc(16, file_size);
+		data = (U8*) ll_aligned_malloc<16>(file_size);
 		file.read(data, file_size);	/* Flawfinder: ignore */ 
 		
 		if (file.getLastBytesRead() != (S32)file_size)
 		{
-			ll_aligned_free(16, data);
+			ll_aligned_free<16>(data);
 			data = NULL;
 			file_size = 0;
 		}
diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp
index 6b5d06376f..818232a5c6 100755
--- a/indra/newview/llvovolume.cpp
+++ b/indra/newview/llvovolume.cpp
@@ -4433,13 +4433,13 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group)
 
 	const U32 MAX_FACE_COUNT = 4096;
 	
-	static LLFace** fullbright_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** bump_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** simple_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** norm_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** spec_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** normspec_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
-	static LLFace** alpha_faces = (LLFace**) ll_aligned_malloc(64, MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** fullbright_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** bump_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** simple_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** norm_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** spec_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** normspec_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
+	static LLFace** alpha_faces = (LLFace**) ll_aligned_malloc<64>(MAX_FACE_COUNT*sizeof(LLFace*));
 	
 	U32 fullbright_count = 0;
 	U32 bump_count = 0;