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:
authormano-wii <germano.costa@ig.com.br>2018-01-10 02:06:10 +0300
committerCampbell Barton <ideasman42@gmail.com>2018-01-10 02:11:32 +0300
commit3c6e19b30b4f2c9cf4809dfcd50097d3aeda3998 (patch)
treefe603acf39e8ad0e4c8b58d2514e0aecf89b75bd /source/blender/draw/intern/draw_cache_impl_displist.c
parenta0655ed487a838a5d2c1d39e62db9a39a4100715 (diff)
Curve/Surface Display: UV support
Diffstat (limited to 'source/blender/draw/intern/draw_cache_impl_displist.c')
-rw-r--r--source/blender/draw/intern/draw_cache_impl_displist.c194
1 files changed, 194 insertions, 0 deletions
diff --git a/source/blender/draw/intern/draw_cache_impl_displist.c b/source/blender/draw/intern/draw_cache_impl_displist.c
index fd095c0ce4f..627fb38d9d6 100644
--- a/source/blender/draw/intern/draw_cache_impl_displist.c
+++ b/source/blender/draw/intern/draw_cache_impl_displist.c
@@ -202,3 +202,197 @@ Gwn_IndexBuf **DRW_displist_indexbuf_calc_triangles_in_order_split_by_material(L
return shaded_triangles_in_order;
}
+
+static void displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ Gwn_VertBufRaw *pos_step, Gwn_VertBufRaw *nor_step, Gwn_VertBufRaw *uv_step,
+ const float v1[3], const float v2[3], const float v3[3],
+ const float n1[3], const float n2[3], const float n3[3],
+ const float uv1[2], const float uv2[2], const float uv3[2])
+{
+ copy_v3_v3(GWN_vertbuf_raw_step(pos_step), v1);
+ copy_v3_v3(GWN_vertbuf_raw_step(nor_step), n1);
+ copy_v2_v2(GWN_vertbuf_raw_step(uv_step), uv1);
+
+ copy_v3_v3(GWN_vertbuf_raw_step(pos_step), v2);
+ copy_v3_v3(GWN_vertbuf_raw_step(nor_step), n2);
+ copy_v2_v2(GWN_vertbuf_raw_step(uv_step), uv2);
+
+ copy_v3_v3(GWN_vertbuf_raw_step(pos_step), v3);
+ copy_v3_v3(GWN_vertbuf_raw_step(nor_step), n3);
+ copy_v2_v2(GWN_vertbuf_raw_step(uv_step), uv3);
+}
+
+Gwn_Batch **DRW_displist_batch_calc_tri_pos_normals_and_uv_split_by_material(ListBase *lb, uint gpumat_array_len)
+{
+ static Gwn_VertFormat shaded_triangles_format = { 0 };
+ static struct { uint pos, nor, uv; } attr_id;
+
+ if (shaded_triangles_format.attrib_ct == 0) {
+ /* initialize vertex format */
+ attr_id.pos = GWN_vertformat_attr_add(&shaded_triangles_format, "pos", GWN_COMP_F32, 3, GWN_FETCH_FLOAT);
+ attr_id.nor = GWN_vertformat_attr_add(&shaded_triangles_format, "nor", GWN_COMP_F32, 3, GWN_FETCH_FLOAT);
+ attr_id.uv = GWN_vertformat_attr_add(&shaded_triangles_format, "u", GWN_COMP_F32, 2, GWN_FETCH_FLOAT);
+ }
+
+ Gwn_Batch **shaded_triangles = MEM_mallocN(sizeof(*shaded_triangles) * gpumat_array_len, __func__);
+
+ Gwn_VertBuf **vbo = BLI_array_alloca(vbo, gpumat_array_len);
+ uint *vbo_len_capacity = BLI_array_alloca(vbo_len_capacity, gpumat_array_len);
+
+ Gwn_VertBufRaw *pos_step, *nor_step, *uv_step;
+ pos_step = BLI_array_alloca(pos_step, gpumat_array_len);
+ nor_step = BLI_array_alloca(nor_step, gpumat_array_len);
+ uv_step = BLI_array_alloca(uv_step, gpumat_array_len);
+
+ /* Create each vertex buffer */
+ for (int i = 0; i < gpumat_array_len; i++) {
+ vbo[i] = GWN_vertbuf_create_with_format(&shaded_triangles_format);
+ vbo_len_capacity[i] = 0;
+ }
+
+ /* Calc `vbo_len_capacity` */
+ for (const DispList *dl = lb->first; dl; dl = dl->next) {
+ vbo_len_capacity[dl->col] += dl_tri_len(dl) * 3;
+ }
+
+ /* Alloc each vertex buffer and get each raw data */
+ for (int i = 0; i < gpumat_array_len; i++) {
+ GWN_vertbuf_data_alloc(vbo[i], vbo_len_capacity[i]);
+ GWN_vertbuf_attr_get_raw_data(vbo[i], attr_id.pos, &pos_step[i]);
+ GWN_vertbuf_attr_get_raw_data(vbo[i], attr_id.nor, &nor_step[i]);
+ GWN_vertbuf_attr_get_raw_data(vbo[i], attr_id.uv, &uv_step[i]);
+ }
+
+ BKE_displist_normals_add(lb);
+
+ for (const DispList *dl = lb->first; dl; dl = dl->next) {
+ if (ELEM(dl->type, DL_INDEX3, DL_INDEX4, DL_SURF)) {
+ const int col = dl->col;
+ const float(*verts)[3] = (float(*)[3])dl->verts;
+ const float(*nors)[3] = (float(*)[3])dl->nors;
+ const int *idx = dl->index;
+ float uv[4][2];
+
+ if (dl->type == DL_INDEX3) {
+ const float x_max = (float)(dl->nr - 1);
+ uv[0][1] = uv[1][1] = uv[2][1] = 0.0f;
+ const int i_end = dl->parts;
+ for (int i = 0; i < i_end; i++, idx += 3) {
+ uv[0][0] = idx[0] / x_max;
+ uv[1][0] = idx[2] / x_max;
+ uv[2][0] = idx[1] / x_max;
+
+ displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ &pos_step[col], &nor_step[col], &uv_step[col],
+ verts[idx[0]], verts[idx[2]], verts[idx[1]],
+ nors[idx[0]], nors[idx[2]], nors[idx[1]],
+ uv[0], uv[1], uv[2]);
+ }
+ }
+ else if (dl->type == DL_SURF) {
+ uint quad[4];
+ for (int a = 0; a < dl->parts; a++) {
+ if ((dl->flag & DL_CYCL_V) == 0 && a == dl->parts - 1) {
+ break;
+ }
+
+ int b;
+ if (dl->flag & DL_CYCL_U) {
+ quad[0] = dl->nr * a;
+ quad[3] = quad[0] + dl->nr - 1;
+ quad[1] = quad[0] + dl->nr;
+ quad[2] = quad[3] + dl->nr;
+ b = 0;
+ }
+ else {
+ quad[3] = dl->nr * a;
+ quad[0] = quad[3] + 1;
+ quad[2] = quad[3] + dl->nr;
+ quad[1] = quad[0] + dl->nr;
+ b = 1;
+ }
+ if ((dl->flag & DL_CYCL_V) && a == dl->parts - 1) {
+ quad[1] -= dl->parts * dl->nr;
+ quad[2] -= dl->parts * dl->nr;
+ }
+
+ for (; b < dl->nr; b++) {
+ int orco_sizeu = dl->nr - 1;
+ int orco_sizev = dl->parts - 1;
+
+ /* exception as handled in convertblender.c too */
+ if (dl->flag & DL_CYCL_U) {
+ orco_sizeu++;
+ }
+ if (dl->flag & DL_CYCL_V) {
+ orco_sizev++;
+ }
+
+ for (int i = 0; i < 4; i++) {
+ /* find uv based on vertex index into grid array */
+ uv[i][0] = (quad[i] / dl->nr) / (float)orco_sizev;
+ uv[i][1] = (quad[i] % dl->nr) / (float)orco_sizeu;
+
+ /* cyclic correction */
+ if ((i == 1 || i == 2) && uv[i][0] == 0.0f) {
+ uv[i][0] = 1.0f;
+ }
+ if ((i == 0 || i == 1) && uv[i][1] == 0.0f) {
+ uv[i][1] = 1.0f;
+ }
+ }
+
+ displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ &pos_step[col], &nor_step[col], &uv_step[col],
+ verts[quad[0]], verts[quad[1]], verts[quad[2]],
+ nors[quad[0]], nors[quad[1]], nors[quad[2]],
+ uv[0], uv[1], uv[2]);
+
+ displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ &pos_step[col], &nor_step[col], &uv_step[col],
+ verts[quad[0]], verts[quad[2]], verts[quad[3]],
+ nors[quad[0]], nors[quad[2]], nors[quad[3]],
+ uv[0], uv[2], uv[3]);
+
+ quad[2] = quad[1];
+ quad[1]++;
+ quad[3] = quad[0];
+ quad[0]++;
+ }
+ }
+ }
+ else {
+ BLI_assert(dl->type == DL_INDEX4);
+ uv[0][0] = uv[0][1] = uv[1][0] = uv[3][1] = 0.0f;
+ uv[1][1] = uv[2][0] = uv[2][1] = uv[3][0] = 1.0f;
+
+ const int i_end = dl->parts;
+ for (int i = 0; i < i_end; i++, idx += 4) {
+ displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ &pos_step[col], &nor_step[col], &uv_step[col],
+ verts[idx[0]], verts[idx[1]], verts[idx[2]],
+ nors[idx[0]], nors[idx[1]], nors[idx[2]],
+ uv[0], uv[1], uv[2]);
+
+ if (idx[2] != idx[3]) {
+ displist_vertbuf_attr_set_tri_pos_normals_and_uv(
+ &pos_step[col], &nor_step[col], &uv_step[col],
+ verts[idx[0]], verts[idx[2]], verts[idx[3]],
+ nors[idx[0]], nors[idx[2]], nors[idx[3]],
+ uv[0], uv[2], uv[3]);
+ }
+ }
+ }
+ }
+ }
+
+ for (int i = 0; i < gpumat_array_len; i++) {
+ uint vbo_len_used = GWN_vertbuf_raw_used(&pos_step[i]);
+ if (vbo_len_capacity[i] != vbo_len_used) {
+ GWN_vertbuf_data_resize(vbo[i], vbo_len_used);
+ }
+ shaded_triangles[i] = GWN_batch_create_ex(GWN_PRIM_TRIS, vbo[i], NULL, GWN_BATCH_OWNS_VBO);
+ }
+
+ return shaded_triangles;
+}