diff options
author | Lukas Steiblys <imbusy@imbusy.org> | 2011-02-15 22:04:00 +0300 |
---|---|---|
committer | Lukas Steiblys <imbusy@imbusy.org> | 2011-02-15 22:04:00 +0300 |
commit | 7fba9eab470783775561308683a839f0aa64f9fb (patch) | |
tree | 2b6991fbcf41a1cc3c68179b2ca93a94944ab600 /source/blender/gpu | |
parent | 15570631fd329ba4df1d60c7a919ad6fa1619ef2 (diff) |
added more code comments and removed some magic constants
Diffstat (limited to 'source/blender/gpu')
-rw-r--r-- | source/blender/gpu/GPU_buffers.h | 4 | ||||
-rw-r--r-- | source/blender/gpu/intern/gpu_buffers.c | 27 |
2 files changed, 19 insertions, 12 deletions
diff --git a/source/blender/gpu/GPU_buffers.h b/source/blender/gpu/GPU_buffers.h index 342cc6b72a3..4c6e24cde62 100644 --- a/source/blender/gpu/GPU_buffers.h +++ b/source/blender/gpu/GPU_buffers.h @@ -69,6 +69,8 @@ typedef struct GPUBuffer unsigned int id; /* used with vertex buffer objects */ } GPUBuffer; +/* stores deleted buffers so that new buffers wouldn't have to +be recreated that often. */ typedef struct GPUBufferPool { int size; /* number of allocated buffers stored */ @@ -114,6 +116,7 @@ typedef struct GPUDrawObject } GPUDrawObject; +/* used for GLSL materials */ typedef struct GPUAttrib { int index; @@ -173,6 +176,7 @@ void GPU_buffer_draw_elements( GPUBuffer *elements, unsigned int mode, int start /* called after drawing */ void GPU_buffer_unbind(void); +/* used to check whether to use the old (without buffers) code */ int GPU_buffer_legacy( struct DerivedMesh *dm ); #endif diff --git a/source/blender/gpu/intern/gpu_buffers.c b/source/blender/gpu/intern/gpu_buffers.c index d8be5274633..81d29a01075 100644 --- a/source/blender/gpu/intern/gpu_buffers.c +++ b/source/blender/gpu/intern/gpu_buffers.c @@ -59,6 +59,9 @@ #define MAX_GPU_ATTRIB_DATA 32 +/* material number is an 16-bit short and the range of short is from -16383 to 16383 (assume material number is non-negative) */ +#define MAX_MATERIALS 16384 + /* -1 - undefined, 0 - vertex arrays, 1 - VBOs */ static int useVBOs = -1; static GPUBufferPool *globalPool = 0; @@ -270,8 +273,8 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) { GPUDrawObject *object; MFace *mface; - int numverts[32768]; /* material number is an 16-bit short so there's at most 32768 materials */ - int redir[32768]; /* material number is an 16-bit short so there's at most 32768 materials */ + int numverts[MAX_MATERIALS]; + int redir[MAX_MATERIALS]; int *index; int i; int curmat, curverts, numfaces; @@ -288,19 +291,19 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) object->indices[i].next = 0; } /*object->legacy = 1;*/ - memset(numverts,0,sizeof(int)*32768); + memset(numverts,0,sizeof(int)*MAX_MATERIALS); mface = dm->getFaceArray(dm); numfaces= dm->getNumFaces(dm); for( i=0; i < numfaces; i++ ) { if( mface[i].v4 ) - numverts[mface[i].mat_nr+16383] += 6; /* split every quad into two triangles */ + numverts[mface[i].mat_nr] += 6; /* split every quad into two triangles */ else - numverts[mface[i].mat_nr+16383] += 3; + numverts[mface[i].mat_nr] += 3; } - for( i = 0; i < 32768; i++ ) { + for( i = 0; i < MAX_MATERIALS; i++ ) { if( numverts[i] > 0 ) { object->nmaterials++; object->nelements += numverts[i]; @@ -310,9 +313,9 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) index = MEM_mallocN(sizeof(int)*object->nmaterials,"GPU_drawobject_new_index"); curmat = curverts = 0; - for( i = 0; i < 32768; i++ ) { + for( i = 0; i < MAX_MATERIALS; i++ ) { if( numverts[i] > 0 ) { - object->materials[curmat].mat_nr = i-16383; + object->materials[curmat].mat_nr = i; object->materials[curmat].start = curverts; index[curmat] = curverts/3; object->materials[curmat].end = curverts+numverts[i]; @@ -322,7 +325,7 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) } object->faceRemap = MEM_mallocN(sizeof(int)*object->nelements/3,"GPU_drawobject_new_faceRemap"); for( i = 0; i < object->nmaterials; i++ ) { - redir[object->materials[i].mat_nr+16383] = i; /* material number -> material index */ + redir[object->materials[i].mat_nr] = i; /* material number -> material index */ } object->indexMem = MEM_callocN(sizeof(IndexLink)*object->nelements,"GPU_drawobject_new_indexMem"); @@ -340,7 +343,7 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) } for( i=0; i < numfaces; i++ ) { - int curInd = index[redir[mface[i].mat_nr+16383]]; + int curInd = index[redir[mface[i].mat_nr]]; object->faceRemap[curInd] = i; ADDLINK( mface[i].v1, curInd*3 ); ADDLINK( mface[i].v2, curInd*3+1 ); @@ -351,10 +354,10 @@ GPUDrawObject *GPU_drawobject_new( DerivedMesh *dm ) ADDLINK( mface[i].v4, curInd*3+4 ); ADDLINK( mface[i].v1, curInd*3+5 ); - index[redir[mface[i].mat_nr+16383]]+=2; + index[redir[mface[i].mat_nr]]+=2; } else { - index[redir[mface[i].mat_nr+16383]]++; + index[redir[mface[i].mat_nr]]++; } } |