Welcome to mirror list, hosted at ThFree Co, Russian Federation.

git.blender.org/blender.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat (limited to 'source/blender/gpu/intern/gpu_buffers.c')
-rw-r--r--source/blender/gpu/intern/gpu_buffers.c248
1 files changed, 148 insertions, 100 deletions
diff --git a/source/blender/gpu/intern/gpu_buffers.c b/source/blender/gpu/intern/gpu_buffers.c
index e622943597a..62843f0905f 100644
--- a/source/blender/gpu/intern/gpu_buffers.c
+++ b/source/blender/gpu/intern/gpu_buffers.c
@@ -383,8 +383,9 @@ void GPU_buffer_free(GPUBuffer *buffer)
BLI_mutex_unlock(&buffer_mutex);
}
-/* currently unused */
-// #define USE_GPU_POINT_LINK
+#if 0 /* currently unused */
+# define USE_GPU_POINT_LINK
+#endif
typedef struct GPUVertPointLink {
#ifdef USE_GPU_POINT_LINK
@@ -452,7 +453,7 @@ static void gpu_drawobject_init_vert_points(GPUDrawObject *gdo, MFace *f, int to
int i, *mat_orig_to_new;
mat_orig_to_new = MEM_callocN(sizeof(*mat_orig_to_new) * totmat,
- "GPUDrawObject.mat_orig_to_new");
+ "GPUDrawObject.mat_orig_to_new");
/* allocate the array and space for links */
gdo->vert_points = MEM_mallocN(sizeof(GPUVertPointLink) * gdo->totvert,
"GPUDrawObject.vert_points");
@@ -939,17 +940,62 @@ static void GPU_buffer_copy_mcol(DerivedMesh *dm, float *varray_, int *index, in
static void GPU_buffer_copy_edge(DerivedMesh *dm, float *varray_, int *UNUSED(index), int *UNUSED(mat_orig_to_new), void *UNUSED(user))
{
- MEdge *medge;
+ MEdge *medge, *medge_base;
unsigned int *varray = (unsigned int *)varray_;
- int i, totedge;
+ int i, totedge, iloose, inorm, iloosehidden, inormhidden;
+ int tot_loose_hidden = 0, tot_loose = 0;
+ int tot_hidden = 0, tot = 0;
- medge = dm->getEdgeArray(dm);
+ medge_base = medge = dm->getEdgeArray(dm);
totedge = dm->getNumEdges(dm);
for (i = 0; i < totedge; i++, medge++) {
- varray[i * 2] = dm->drawObject->vert_points[medge->v1].point_index;
- varray[i * 2 + 1] = dm->drawObject->vert_points[medge->v2].point_index;
+ if (medge->flag & ME_EDGEDRAW) {
+ if (medge->flag & ME_LOOSEEDGE) tot_loose++;
+ else tot++;
+ }
+ else {
+ if (medge->flag & ME_LOOSEEDGE) tot_loose_hidden++;
+ else tot_hidden++;
+ }
+ }
+
+ inorm = 0;
+ inormhidden = tot;
+ iloose = tot + tot_hidden;
+ iloosehidden = iloose + tot_loose;
+
+ medge = medge_base;
+ for (i = 0; i < totedge; i++, medge++) {
+ if (medge->flag & ME_EDGEDRAW) {
+ if (medge->flag & ME_LOOSEEDGE) {
+ varray[iloose * 2] = dm->drawObject->vert_points[medge->v1].point_index;
+ varray[iloose * 2 + 1] = dm->drawObject->vert_points[medge->v2].point_index;
+ iloose++;
+ }
+ else {
+ varray[inorm * 2] = dm->drawObject->vert_points[medge->v1].point_index;
+ varray[inorm * 2 + 1] = dm->drawObject->vert_points[medge->v2].point_index;
+ inorm++;
+ }
+ }
+ else {
+ if (medge->flag & ME_LOOSEEDGE) {
+ varray[iloosehidden * 2] = dm->drawObject->vert_points[medge->v1].point_index;
+ varray[iloosehidden * 2 + 1] = dm->drawObject->vert_points[medge->v2].point_index;
+ iloosehidden++;
+ }
+ else {
+ varray[inormhidden * 2] = dm->drawObject->vert_points[medge->v1].point_index;
+ varray[inormhidden * 2 + 1] = dm->drawObject->vert_points[medge->v2].point_index;
+ inormhidden++;
+ }
+ }
}
+
+ dm->drawObject->tot_loose_edge_drawn = tot_loose;
+ dm->drawObject->loose_edge_offset = tot + tot_hidden;
+ dm->drawObject->tot_edge_drawn = tot;
}
static void GPU_buffer_copy_uvedge(DerivedMesh *dm, float *varray, int *UNUSED(index), int *UNUSED(mat_orig_to_new), void *UNUSED(user))
@@ -1007,7 +1053,7 @@ const GPUBufferTypeSettings gpu_buffer_type_settings[] = {
{GPU_buffer_copy_normal, GL_ARRAY_BUFFER_ARB, 3},
{GPU_buffer_copy_mcol, GL_ARRAY_BUFFER_ARB, 3},
{GPU_buffer_copy_uv, GL_ARRAY_BUFFER_ARB, 2},
- {GPU_buffer_copy_uv_texpaint, GL_ARRAY_BUFFER_ARB, 4},
+ {GPU_buffer_copy_uv_texpaint, GL_ARRAY_BUFFER_ARB, 4},
{GPU_buffer_copy_edge, GL_ELEMENT_ARRAY_BUFFER_ARB, 2},
{GPU_buffer_copy_uvedge, GL_ELEMENT_ARRAY_BUFFER_ARB, 4}
};
@@ -1498,6 +1544,7 @@ struct GPU_PBVH_Buffers {
bool use_matcaps;
float diffuse_color[4];
};
+
typedef enum {
VBO_ENABLED,
VBO_DISABLED
@@ -1577,8 +1624,8 @@ static void gpu_color_from_mask_quad_set(const CCGKey *key,
}
void GPU_update_mesh_pbvh_buffers(GPU_PBVH_Buffers *buffers, MVert *mvert,
- int *vert_indices, int totvert, const float *vmask,
- int (*face_vert_indices)[4], bool show_diffuse_color)
+ int *vert_indices, int totvert, const float *vmask,
+ int (*face_vert_indices)[4], bool show_diffuse_color)
{
VertexBufferFormat *vert_data;
int i, j, k;
@@ -1715,9 +1762,9 @@ void GPU_update_mesh_pbvh_buffers(GPU_PBVH_Buffers *buffers, MVert *mvert,
}
GPU_PBVH_Buffers *GPU_build_mesh_pbvh_buffers(int (*face_vert_indices)[4],
- MFace *mface, MVert *mvert,
- int *face_indices,
- int totface)
+ MFace *mface, MVert *mvert,
+ int *face_indices,
+ int totface)
{
GPU_PBVH_Buffers *buffers;
unsigned short *tri_data;
@@ -1807,8 +1854,8 @@ GPU_PBVH_Buffers *GPU_build_mesh_pbvh_buffers(int (*face_vert_indices)[4],
}
void GPU_update_grid_pbvh_buffers(GPU_PBVH_Buffers *buffers, CCGElem **grids,
- const DMFlagMat *grid_flag_mats, int *grid_indices,
- int totgrid, const CCGKey *key, bool show_diffuse_color)
+ const DMFlagMat *grid_flag_mats, int *grid_indices,
+ int totgrid, const CCGKey *key, bool show_diffuse_color)
{
VertexBufferFormat *vert_data;
int i, j, k, x, y;
@@ -1921,51 +1968,51 @@ void GPU_update_grid_pbvh_buffers(GPU_PBVH_Buffers *buffers, CCGElem **grids,
/* Build the element array buffer of grid indices using either
* unsigned shorts or unsigned ints. */
#define FILL_QUAD_BUFFER(type_, tot_quad_, buffer_) \
- { \
- type_ *tri_data; \
- int offset = 0; \
- int i, j, k; \
- \
- glBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, \
- sizeof(type_) * (tot_quad_) * 6, NULL, \
- GL_STATIC_DRAW_ARB); \
- \
- /* Fill the buffer */ \
- tri_data = glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, \
- GL_WRITE_ONLY_ARB); \
- if (tri_data) { \
- for (i = 0; i < totgrid; ++i) { \
- BLI_bitmap *gh = NULL; \
- if (grid_hidden) \
- gh = grid_hidden[(grid_indices)[i]]; \
- \
- for (j = 0; j < gridsize - 1; ++j) { \
- for (k = 0; k < gridsize - 1; ++k) { \
- /* Skip hidden grid face */ \
- if (gh && \
- paint_is_grid_face_hidden(gh, \
- gridsize, k, j)) \
- continue; \
- \
- *(tri_data++) = offset + j * gridsize + k + 1; \
- *(tri_data++) = offset + j * gridsize + k; \
- *(tri_data++) = offset + (j + 1) * gridsize + k; \
- \
- *(tri_data++) = offset + (j + 1) * gridsize + k + 1; \
- *(tri_data++) = offset + j * gridsize + k + 1; \
- *(tri_data++) = offset + (j + 1) * gridsize + k; \
- } \
- } \
- \
- offset += gridsize * gridsize; \
- } \
- glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); \
- } \
- else { \
- glDeleteBuffersARB(1, &(buffer_)); \
- (buffer_) = 0; \
- } \
- } (void)0
+ { \
+ type_ *tri_data; \
+ int offset = 0; \
+ int i, j, k; \
+ \
+ glBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, \
+ sizeof(type_) * (tot_quad_) * 6, NULL, \
+ GL_STATIC_DRAW_ARB); \
+ \
+ /* Fill the buffer */ \
+ tri_data = glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, \
+ GL_WRITE_ONLY_ARB); \
+ if (tri_data) { \
+ for (i = 0; i < totgrid; ++i) { \
+ BLI_bitmap *gh = NULL; \
+ if (grid_hidden) \
+ gh = grid_hidden[(grid_indices)[i]]; \
+ \
+ for (j = 0; j < gridsize - 1; ++j) { \
+ for (k = 0; k < gridsize - 1; ++k) { \
+ /* Skip hidden grid face */ \
+ if (gh && \
+ paint_is_grid_face_hidden(gh, \
+ gridsize, k, j)) \
+ continue; \
+ \
+ *(tri_data++) = offset + j * gridsize + k + 1; \
+ *(tri_data++) = offset + j * gridsize + k; \
+ *(tri_data++) = offset + (j + 1) * gridsize + k; \
+ \
+ *(tri_data++) = offset + (j + 1) * gridsize + k + 1; \
+ *(tri_data++) = offset + j * gridsize + k + 1; \
+ *(tri_data++) = offset + (j + 1) * gridsize + k; \
+ } \
+ } \
+ \
+ offset += gridsize * gridsize; \
+ } \
+ glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); \
+ } \
+ else { \
+ glDeleteBuffersARB(1, &(buffer_)); \
+ (buffer_) = 0; \
+ } \
+ } (void)0
/* end FILL_QUAD_BUFFER */
static GLuint gpu_get_grid_buffer(int gridsize, GLenum *index_type, unsigned *totquad)
@@ -2021,7 +2068,7 @@ static GLuint gpu_get_grid_buffer(int gridsize, GLenum *index_type, unsigned *to
}
GPU_PBVH_Buffers *GPU_build_grid_pbvh_buffers(int *grid_indices, int totgrid,
- BLI_bitmap **grid_hidden, int gridsize)
+ BLI_bitmap **grid_hidden, int gridsize)
{
GPU_PBVH_Buffers *buffers;
int totquad;
@@ -2103,7 +2150,6 @@ static void gpu_bmesh_vert_to_buffer_copy(BMVert *v,
BM_ELEM_CD_GET_FLOAT(v, cd_vert_mask_offset),
diffuse_color,
vd->color);
-
/* Assign index for use in the triangle index buffer */
/* note: caller must set: bm->elem_index_dirty |= BM_VERT; */
@@ -2153,11 +2199,11 @@ static int gpu_bmesh_face_visible_count(GSet *bm_faces)
/* Creates a vertex buffer (coordinate, normal, color) and, if smooth
* shading, an element index buffer. */
void GPU_update_bmesh_pbvh_buffers(GPU_PBVH_Buffers *buffers,
- BMesh *bm,
- GSet *bm_faces,
- GSet *bm_unique_verts,
- GSet *bm_other_verts,
- bool show_diffuse_color)
+ BMesh *bm,
+ GSet *bm_faces,
+ GSet *bm_unique_verts,
+ GSet *bm_other_verts,
+ bool show_diffuse_color)
{
VertexBufferFormat *vert_data;
void *tri_data;
@@ -2204,8 +2250,8 @@ void GPU_update_bmesh_pbvh_buffers(GPU_PBVH_Buffers *buffers,
/* Initialize vertex buffer */
glBindBufferARB(GL_ARRAY_BUFFER_ARB, buffers->vert_buf);
glBufferDataARB(GL_ARRAY_BUFFER_ARB,
- sizeof(VertexBufferFormat) * totvert,
- NULL, GL_STATIC_DRAW_ARB);
+ sizeof(VertexBufferFormat) * totvert,
+ NULL, GL_STATIC_DRAW_ARB);
/* Fill vertex buffer */
vert_data = glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
@@ -2246,7 +2292,9 @@ void GPU_update_bmesh_pbvh_buffers(GPU_PBVH_Buffers *buffers,
float fmask = 0;
int i;
- // BM_iter_as_array(bm, BM_VERTS_OF_FACE, f, (void**)v, 3);
+#if 0
+ BM_iter_as_array(bm, BM_VERTS_OF_FACE, f, (void**)v, 3);
+#endif
BM_face_as_array_vert_tri(f, v);
/* Average mask value */
@@ -2666,7 +2714,7 @@ bool GPU_pbvh_buffers_diffuse_changed(GPU_PBVH_Buffers *buffers, GSet *bm_faces,
GPU_material_diffuse_get(f->mat_nr + 1, diffuse_color);
}
else if (buffers->use_bmesh) {
- /* due to dynamc nature of dyntopo, only get first material */
+ /* due to dynamic nature of dyntopo, only get first material */
if (BLI_gset_size(bm_faces) > 0) {
GSetIterator gs_iter;
BMFace *f;
@@ -2717,7 +2765,7 @@ static void gpu_pbvh_buffer_free_intern(GLuint id)
if (pool->maxpbvhsize == pool->totpbvhbufids) {
pool->maxpbvhsize += MAX_FREE_GPU_BUFF_IDS;
pool->pbvhbufids = MEM_reallocN(pool->pbvhbufids,
- sizeof(*pool->pbvhbufids) * pool->maxpbvhsize);
+ sizeof(*pool->pbvhbufids) * pool->maxpbvhsize);
}
/* insert the buffer into the beginning of the pool */
@@ -2742,33 +2790,33 @@ void GPU_free_pbvh_buffers(GPU_PBVH_Buffers *buffers)
void GPU_draw_pbvh_BB(float min[3], float max[3], bool leaf)
{
const float quads[4][4][3] = {
- {
- {min[0], min[1], min[2]},
- {max[0], min[1], min[2]},
- {max[0], min[1], max[2]},
- {min[0], min[1], max[2]}
- },
-
- {
- {min[0], min[1], min[2]},
- {min[0], max[1], min[2]},
- {min[0], max[1], max[2]},
- {min[0], min[1], max[2]}
- },
-
- {
- {max[0], max[1], min[2]},
- {max[0], min[1], min[2]},
- {max[0], min[1], max[2]},
- {max[0], max[1], max[2]}
- },
-
- {
- {max[0], max[1], min[2]},
- {min[0], max[1], min[2]},
- {min[0], max[1], max[2]},
- {max[0], max[1], max[2]}
- },
+ {
+ {min[0], min[1], min[2]},
+ {max[0], min[1], min[2]},
+ {max[0], min[1], max[2]},
+ {min[0], min[1], max[2]}
+ },
+
+ {
+ {min[0], min[1], min[2]},
+ {min[0], max[1], min[2]},
+ {min[0], max[1], max[2]},
+ {min[0], min[1], max[2]}
+ },
+
+ {
+ {max[0], max[1], min[2]},
+ {max[0], min[1], min[2]},
+ {max[0], min[1], max[2]},
+ {max[0], max[1], max[2]}
+ },
+
+ {
+ {max[0], max[1], min[2]},
+ {min[0], max[1], min[2]},
+ {min[0], max[1], max[2]},
+ {max[0], max[1], max[2]}
+ },
};
if (leaf)