summaryrefslogtreecommitdiff
path: root/indra/llrender
diff options
context:
space:
mode:
Diffstat (limited to 'indra/llrender')
-rw-r--r--indra/llrender/llimagegl.cpp4
-rw-r--r--indra/llrender/llvertexbuffer.cpp287
-rw-r--r--indra/llrender/llvertexbuffer.h15
3 files changed, 238 insertions, 68 deletions
diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp
index 2f02ccf30b..2579bad0b6 100644
--- a/indra/llrender/llimagegl.cpp
+++ b/indra/llrender/llimagegl.cpp
@@ -1813,7 +1813,7 @@ BOOL LLImageGL::getMask(const LLVector2 &tc)
{
LL_WARNS_ONCE("render") << "Ugh, non-finite u/v in mask pick" << LL_ENDL;
u = v = 0.f;
- llassert(false);
+ //llassert(false);
}
if (LL_UNLIKELY(u < 0.f || u > 1.f ||
@@ -1821,7 +1821,7 @@ BOOL LLImageGL::getMask(const LLVector2 &tc)
{
LL_WARNS_ONCE("render") << "Ugh, u/v out of range in image mask pick" << LL_ENDL;
u = v = 0.f;
- llassert(false);
+ //llassert(false);
}
S32 x = llfloor(u * mPickMaskWidth);
diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp
index 7fa47cd171..7f14a8d5ac 100644
--- a/indra/llrender/llvertexbuffer.cpp
+++ b/indra/llrender/llvertexbuffer.cpp
@@ -39,6 +39,7 @@
#include "llglheaders.h"
#include "llmemtype.h"
#include "llrender.h"
+#include "llvector4a.h"
//============================================================================
@@ -66,6 +67,27 @@ S32 LLVertexBuffer::sWeight4Loc = -1;
std::vector<U32> LLVertexBuffer::sDeleteList;
+#define LL_ALIGNED_VB 1
+
+#if LL_ALIGNED_VB
+
+S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] =
+{
+ sizeof(LLVector4), // TYPE_VERTEX,
+ sizeof(LLVector4), // TYPE_NORMAL,
+ sizeof(LLVector2), // TYPE_TEXCOORD0,
+ sizeof(LLVector2), // TYPE_TEXCOORD1,
+ sizeof(LLVector2), // TYPE_TEXCOORD2,
+ sizeof(LLVector2), // TYPE_TEXCOORD3,
+ sizeof(LLColor4U), // TYPE_COLOR,
+ sizeof(LLVector4), // TYPE_BINORMAL,
+ sizeof(F32), // TYPE_WEIGHT,
+ sizeof(LLVector4), // TYPE_WEIGHT4,
+ sizeof(LLVector4), // TYPE_CLOTHWEIGHT,
+};
+
+#else
+
S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] =
{
sizeof(LLVector3), // TYPE_VERTEX,
@@ -81,6 +103,8 @@ S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] =
sizeof(LLVector4), // TYPE_CLOTHWEIGHT,
};
+#endif
+
U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] =
{
GL_TRIANGLES,
@@ -237,10 +261,8 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask)
}
}
-void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const
+void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_offset) const
{
- llassert(mRequestedNumVerts >= 0);
-
if (start >= (U32) mRequestedNumVerts ||
end >= (U32) mRequestedNumVerts)
{
@@ -255,6 +277,25 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi
llerrs << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << llendl;
}
+ if (gDebugGL && !useVBOs())
+ {
+ U16* idx = ((U16*) getIndicesPointer())+indices_offset;
+ for (U32 i = 0; i < count; ++i)
+ {
+ if (idx[i] < start || idx[i] > end)
+ {
+ llerrs << "Index out of range: " << idx[i] << " not in [" << start << ", " << end << "]" << llendl;
+ }
+ }
+ }
+}
+
+void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const
+{
+ validateRange(start, end, count, indices_offset);
+
+ llassert(mRequestedNumVerts >= 0);
+
if (mGLIndices != sGLRenderIndices)
{
llerrs << "Wrong index buffer bound." << llendl;
@@ -273,17 +314,6 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi
U16* idx = ((U16*) getIndicesPointer())+indices_offset;
- if (gDebugGL && !useVBOs())
- {
- for (U32 i = 0; i < count; ++i)
- {
- if (idx[i] < start || idx[i] > end)
- {
- llerrs << "Index out of range: " << idx[i] << " not in [" << start << ", " << end << "]" << llendl;
- }
- }
- }
-
stop_glerror();
glDrawRangeElements(sGLMode[mode], start, end, count, GL_UNSIGNED_SHORT,
idx);
@@ -428,11 +458,42 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) :
mTypeMask = typemask;
mStride = stride;
+ mAlignedOffset = 0;
+ mAlignedIndexOffset = 0;
+
sCount++;
}
+#if LL_ALIGNED_VB
+//static
+S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets, S32 num_vertices)
+{
+ S32 offset = 0;
+ for (S32 i=0; i<TYPE_MAX; i++)
+ {
+ U32 mask = 1<<i;
+ if (typemask & mask)
+ {
+ if (offsets)
+ {
+ offsets[i] = offset;
+ offset += LLVertexBuffer::sTypeOffsets[i]*num_vertices;
+ offset = (offset + 0xF) & ~0xF;
+ }
+ }
+ }
+
+ return offset+16;
+}
+
+S32 LLVertexBuffer::getSize() const
+{
+ return mStride;
+}
+
+#else
//static
-S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets)
+S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets, S32 num_vertices)
{
S32 stride = 0;
for (S32 i=0; i<TYPE_MAX; i++)
@@ -451,6 +512,12 @@ S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets)
return stride;
}
+S32 LLVertexBuffer::getSize() const
+{
+ return mNumVerts*mStride;
+}
+
+#endif
// protected, use unref()
//virtual
LLVertexBuffer::~LLVertexBuffer()
@@ -560,7 +627,7 @@ void LLVertexBuffer::createGLBuffer()
{
static int gl_buffer_idx = 0;
mGLBuffer = ++gl_buffer_idx;
- mMappedData = new U8[size];
+ mMappedData = (U8*) _mm_malloc(size, 16);
memset(mMappedData, 0, size);
}
}
@@ -582,16 +649,20 @@ void LLVertexBuffer::createGLIndices()
mEmpty = TRUE;
+ //pad by 16 bytes for aligned copies
+ size += 16;
+
if (useVBOs())
{
+ //pad by another 16 bytes for VBO pointer adjustment
+ size += 16;
mMappedIndexData = NULL;
genIndices();
mResized = TRUE;
}
else
{
- mMappedIndexData = new U8[size];
- memset(mMappedIndexData, 0, size);
+ mMappedIndexData = (U8*) _mm_malloc(size, 16);
static int gl_buffer_idx = 0;
mGLIndices = ++gl_buffer_idx;
}
@@ -612,7 +683,7 @@ void LLVertexBuffer::destroyGLBuffer()
}
else
{
- delete [] mMappedData;
+ _mm_free(mMappedData);
mMappedData = NULL;
mEmpty = TRUE;
}
@@ -639,7 +710,7 @@ void LLVertexBuffer::destroyGLIndices()
}
else
{
- delete [] mMappedIndexData;
+ _mm_free(mMappedIndexData);
mMappedIndexData = NULL;
mEmpty = TRUE;
}
@@ -664,7 +735,7 @@ void LLVertexBuffer::updateNumVerts(S32 nverts)
}
mRequestedNumVerts = nverts;
-
+
if (!mDynamicSize)
{
mNumVerts = nverts;
@@ -679,6 +750,9 @@ void LLVertexBuffer::updateNumVerts(S32 nverts)
}
mNumVerts = nverts;
}
+#if LL_ALIGNED_VB
+ mStride = calcStride(mTypeMask, mOffsets, mNumVerts);
+#endif
}
@@ -754,9 +828,6 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
{
sAllocatedBytes -= getSize() + getIndicesSize();
- S32 oldsize = getSize();
- S32 old_index_size = getIndicesSize();
-
updateNumVerts(newnverts);
updateNumIndices(newnindices);
@@ -773,26 +844,10 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
}
else
{
- //delete old buffer, keep GL buffer for now
if (!useVBOs())
{
- U8* old = mMappedData;
- mMappedData = new U8[newsize];
- if (old)
- {
- memcpy(mMappedData, old, llmin(newsize, oldsize));
- if (newsize > oldsize)
- {
- memset(mMappedData+oldsize, 0, newsize-oldsize);
- }
-
- delete [] old;
- }
- else
- {
- memset(mMappedData, 0, newsize);
- mEmpty = TRUE;
- }
+ _mm_free(mMappedData);
+ mMappedData = (U8*) _mm_malloc(newsize, 16);
}
mResized = TRUE;
}
@@ -812,24 +867,8 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices)
{
if (!useVBOs())
{
- //delete old buffer, keep GL buffer for now
- U8* old = mMappedIndexData;
- mMappedIndexData = new U8[new_index_size];
-
- if (old)
- {
- memcpy(mMappedIndexData, old, llmin(new_index_size, old_index_size));
- if (new_index_size > old_index_size)
- {
- memset(mMappedIndexData+old_index_size, 0, new_index_size - old_index_size);
- }
- delete [] old;
- }
- else
- {
- memset(mMappedIndexData, 0, new_index_size);
- mEmpty = TRUE;
- }
+ _mm_free(mMappedIndexData);
+ mMappedIndexData = (U8*) _mm_malloc(new_index_size, 16);
}
mResized = TRUE;
}
@@ -886,12 +925,19 @@ U8* LLVertexBuffer::mapBuffer(S32 access)
setBuffer(0);
mLocked = TRUE;
stop_glerror();
- mMappedData = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+
+ U8* src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+ mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(src);
+ mAlignedOffset = mMappedData - src;
+
stop_glerror();
}
{
LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES);
- mMappedIndexData = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+ U8* src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
+ mMappedIndexData = LL_NEXT_ALIGNED_ADDRESS<U8>(src);
+ mAlignedIndexOffset = mMappedIndexData - src;
+
stop_glerror();
}
@@ -975,6 +1021,45 @@ void LLVertexBuffer::unmapBuffer()
//----------------------------------------------------------------------------
+#if LL_ALIGNED_VB
+
+template <class T,S32 type> struct VertexBufferStrider
+{
+ typedef LLStrider<T> strider_t;
+ static bool get(LLVertexBuffer& vbo,
+ strider_t& strider,
+ S32 index)
+ {
+ if (vbo.mapBuffer() == NULL)
+ {
+ llwarns << "mapBuffer failed!" << llendl;
+ return FALSE;
+ }
+
+ if (type == LLVertexBuffer::TYPE_INDEX)
+ {
+ S32 stride = sizeof(T);
+ strider = (T*)(vbo.getMappedIndices() + index*stride);
+ strider.setStride(0);
+ return TRUE;
+ }
+ else if (vbo.hasDataType(type))
+ {
+ S32 stride = LLVertexBuffer::sTypeOffsets[type];
+ strider = (T*)(vbo.getMappedData() + vbo.getOffset(type)+index*stride);
+ strider.setStride(stride);
+ return TRUE;
+ }
+ else
+ {
+ llerrs << "VertexBufferStrider could not find valid vertex data." << llendl;
+ }
+ return FALSE;
+ }
+};
+
+#else
+
template <class T,S32 type> struct VertexBufferStrider
{
typedef LLStrider<T> strider_t;
@@ -1010,6 +1095,7 @@ template <class T,S32 type> struct VertexBufferStrider
}
};
+#endif
bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector3>& strider, S32 index)
{
@@ -1272,6 +1358,81 @@ void LLVertexBuffer::setBuffer(U32 data_mask)
}
}
+#if LL_ALIGNED_VB
+
+// virtual (default)
+void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
+{
+ LLMemType mt2(LLMemType::MTYPE_VERTEX_SETUP_VERTEX_BUFFER);
+ stop_glerror();
+ U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData;
+
+ if ((data_mask & mTypeMask) != data_mask)
+ {
+ llerrs << "LLVertexBuffer::setupVertexBuffer missing required components for supplied data mask." << llendl;
+ }
+
+ if (data_mask & MAP_NORMAL)
+ {
+ glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL]));
+ }
+ if (data_mask & MAP_TEXCOORD3)
+ {
+ glClientActiveTextureARB(GL_TEXTURE3_ARB);
+ glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD3], (void*)(base + mOffsets[TYPE_TEXCOORD3]));
+ glClientActiveTextureARB(GL_TEXTURE0_ARB);
+ }
+ if (data_mask & MAP_TEXCOORD2)
+ {
+ glClientActiveTextureARB(GL_TEXTURE2_ARB);
+ glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD2], (void*)(base + mOffsets[TYPE_TEXCOORD2]));
+ glClientActiveTextureARB(GL_TEXTURE0_ARB);
+ }
+ if (data_mask & MAP_TEXCOORD1)
+ {
+ glClientActiveTextureARB(GL_TEXTURE1_ARB);
+ glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1]));
+ glClientActiveTextureARB(GL_TEXTURE0_ARB);
+ }
+ if (data_mask & MAP_BINORMAL)
+ {
+ glClientActiveTextureARB(GL_TEXTURE2_ARB);
+ glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL]));
+ glClientActiveTextureARB(GL_TEXTURE0_ARB);
+ }
+ if (data_mask & MAP_TEXCOORD0)
+ {
+ glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0]));
+ }
+ if (data_mask & MAP_COLOR)
+ {
+ glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeOffsets[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR]));
+ }
+
+ if (data_mask & MAP_WEIGHT)
+ {
+ glVertexAttribPointerARB(1, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT], (void*)(base + mOffsets[TYPE_WEIGHT]));
+ }
+
+ if (data_mask & MAP_WEIGHT4 && sWeight4Loc != -1)
+ {
+ glVertexAttribPointerARB(sWeight4Loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT4], (void*)(base+mOffsets[TYPE_WEIGHT4]));
+ }
+
+ if (data_mask & MAP_CLOTHWEIGHT)
+ {
+ glVertexAttribPointerARB(4, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeOffsets[TYPE_CLOTHWEIGHT], (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]));
+ }
+ if (data_mask & MAP_VERTEX)
+ {
+ glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_VERTEX], (void*)(base + 0));
+ }
+
+ llglassertok();
+}
+
+#else
+
// virtual (default)
void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
{
@@ -1344,6 +1505,8 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const
llglassertok();
}
+#endif
+
void LLVertexBuffer::markDirty(U32 vert_index, U32 vert_count, U32 indices_index, U32 indices_count)
{
// TODO: use GL_APPLE_flush_buffer_range here
diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h
index d1700aa54a..47146a5ec4 100644
--- a/indra/llrender/llvertexbuffer.h
+++ b/indra/llrender/llvertexbuffer.h
@@ -98,7 +98,7 @@ public:
//if offsets is not NULL, its contents will be filled
//with the offset of each vertex component in the buffer,
// indexed by the following enum
- static S32 calcStride(const U32& typemask, S32* offsets = NULL);
+ static S32 calcStride(const U32& typemask, S32* offsets = NULL, S32 num_vertices = 0);
enum {
TYPE_VERTEX,
@@ -187,12 +187,12 @@ public:
S32 getRequestedVerts() const { return mRequestedNumVerts; }
S32 getRequestedIndices() const { return mRequestedNumIndices; }
- U8* getIndicesPointer() const { return useVBOs() ? NULL : mMappedIndexData; }
+ U8* getIndicesPointer() const { return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; }
U8* getVerticesPointer() const { return useVBOs() ? NULL : mMappedData; }
S32 getStride() const { return mStride; }
U32 getTypeMask() const { return mTypeMask; }
BOOL hasDataType(S32 type) const { return ((1 << type) & getTypeMask()) ? TRUE : FALSE; }
- S32 getSize() const { return mNumVerts*mStride; }
+ S32 getSize() const;
S32 getIndicesSize() const { return mNumIndices * sizeof(U16); }
U8* getMappedData() const { return mMappedData; }
U8* getMappedIndices() const { return mMappedIndexData; }
@@ -207,12 +207,19 @@ public:
void drawArrays(U32 mode, U32 offset, U32 count) const;
void drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const;
+ //for debugging, validate data in given range is valid
+ void validateRange(U32 start, U32 end, U32 count, U32 offset) const;
+
+
+
protected:
S32 mNumVerts; // Number of vertices allocated
S32 mNumIndices; // Number of indices allocated
S32 mRequestedNumVerts; // Number of vertices requested
S32 mRequestedNumIndices; // Number of indices requested
+ ptrdiff_t mAlignedOffset;
+ ptrdiff_t mAlignedIndexOffset;
S32 mStride;
U32 mTypeMask;
S32 mUsage; // GL usage
@@ -227,7 +234,7 @@ protected:
S32 mOffsets[TYPE_MAX];
BOOL mResized; // if TRUE, client buffer has been resized and GL buffer has not
BOOL mDynamicSize; // if TRUE, buffer has been resized at least once (and should be padded)
-
+
class DirtyRegion
{
public:
generated by cgit v1.2.3 (git 2.25.1) at 2025-07-26 05:53:06 +0000