Merge branch 'blender2.8' into blender2.8-workbench

19 files changed, 221 insertions, 207 deletions

diff --git a/source/blender/draw/CMakeLists.txt b/source/blender/draw/CMakeLists.txt
index f96f496a0f9..3bf1acaf79e 100644
--- a/source/blender/draw/CMakeLists.txt
+++ b/source/blender/draw/CMakeLists.txt
@@ -234,6 +234,7 @@ data_to_c_simple(modes/shaders/object_empty_image_vert.glsl SRC)
 data_to_c_simple(modes/shaders/object_outline_resolve_frag.glsl SRC)
 data_to_c_simple(modes/shaders/object_outline_expand_frag.glsl SRC)
 data_to_c_simple(modes/shaders/object_outline_detect_frag.glsl SRC)
+data_to_c_simple(modes/shaders/object_outline_prepass_frag.glsl SRC)
 data_to_c_simple(modes/shaders/object_grid_frag.glsl SRC)
 data_to_c_simple(modes/shaders/object_grid_vert.glsl SRC)
 data_to_c_simple(modes/shaders/object_lightprobe_grid_vert.glsl SRC)
diff --git a/source/blender/draw/engines/eevee/eevee_motion_blur.c b/source/blender/draw/engines/eevee/eevee_motion_blur.c
index 9b19163c8d7..4bac8ba2ebd 100644
--- a/source/blender/draw/engines/eevee/eevee_motion_blur.c
+++ b/source/blender/draw/engines/eevee/eevee_motion_blur.c
@@ -68,16 +68,13 @@ static void eevee_motion_blur_camera_get_matrix_at_time(
 	cam_cpy.data = &camdata_cpy;
 
 	const DRWContextState *draw_ctx = DRW_context_state_get();
-	/* We will be modifying time, so we create copy of eval_ctx. */
-	EvaluationContext eval_ctx = draw_ctx->eval_ctx;
-	eval_ctx.ctime = time;
 
 	/* Past matrix */
 	/* FIXME : This is a temporal solution that does not take care of parent animations */
 	/* Recalc Anim manualy */
 	BKE_animsys_evaluate_animdata(scene, &cam_cpy.id, cam_cpy.adt, time, ADT_RECALC_ALL);
 	BKE_animsys_evaluate_animdata(scene, &camdata_cpy.id, camdata_cpy.adt, time, ADT_RECALC_ALL);
-	BKE_object_where_is_calc_time(&eval_ctx, scene, &cam_cpy, time);
+	BKE_object_where_is_calc_time(draw_ctx->depsgraph, scene, &cam_cpy, time);
 
 	/* Compute winmat */
 	CameraParams params;
diff --git a/source/blender/draw/intern/DRW_render.h b/source/blender/draw/intern/DRW_render.h
index 66ca8971a46..27f2d891cc0 100644
--- a/source/blender/draw/intern/DRW_render.h
+++ b/source/blender/draw/intern/DRW_render.h
@@ -187,6 +187,7 @@ typedef enum {
 	DRW_TEX_R_8,
 	DRW_TEX_R_16,
 	DRW_TEX_R_16I,
+	DRW_TEX_R_16U,
 	DRW_TEX_R_32,
 	DRW_TEX_DEPTH_16,
 	DRW_TEX_DEPTH_24,
@@ -506,7 +507,6 @@ typedef struct DRWContextState {
 
 	struct RenderEngineType *engine_type;
 
-	EvaluationContext eval_ctx;
 	struct Depsgraph *depsgraph;
 
 	eObjectMode object_mode;
diff --git a/source/blender/draw/intern/draw_armature.c b/source/blender/draw/intern/draw_armature.c
index c14fe70e0c3..fe87e7f17fd 100644
--- a/source/blender/draw/intern/draw_armature.c
+++ b/source/blender/draw/intern/draw_armature.c
@@ -1228,7 +1228,7 @@ static void draw_armature_edit(Object *ob)
 
 	const bool show_text = DRW_state_show_text();
 
-	for (eBone = arm->edbo->first, index = 0; eBone; eBone = eBone->next, index++) {
+	for (eBone = arm->edbo->first, index = ob->select_color; eBone; eBone = eBone->next, index += 0x10000) {
 		if (eBone->layer & arm->layer) {
 			if ((eBone->flag & BONE_HIDDEN_A) == 0) {
 				const int select_id = is_select ? index : (unsigned int)-1;
diff --git a/source/blender/draw/intern/draw_cache_impl_particles.c b/source/blender/draw/intern/draw_cache_impl_particles.c
index 0530d05c199..d7e8a6b71df 100644
--- a/source/blender/draw/intern/draw_cache_impl_particles.c
+++ b/source/blender/draw/intern/draw_cache_impl_particles.c
@@ -46,6 +46,8 @@
 
 #include "GPU_batch.h"
 
+#include "DEG_depsgraph_query.h"
+
 #include "draw_cache_impl.h"  /* own include */
 
 static void particle_batch_cache_clear(ParticleSystem *psys);
@@ -466,7 +468,7 @@ static void particle_batch_cache_ensure_pos(Object *object, ParticleSystem *psys
 	ParticleSimulationData sim = {NULL};
 	const DRWContextState *draw_ctx = DRW_context_state_get();
 
-	sim.eval_ctx = &draw_ctx->eval_ctx;
+	sim.depsgraph = draw_ctx->depsgraph;
 	sim.scene = draw_ctx->scene;
 	sim.ob = object;
 	sim.psys = psys;
@@ -494,7 +496,7 @@ static void particle_batch_cache_ensure_pos(Object *object, ParticleSystem *psys
 	GWN_vertbuf_data_alloc(cache->pos, psys->totpart);
 
 	for (curr_point = 0, i = 0, pa = psys->particles; i < psys->totpart; i++, pa++) {
-		state.time = draw_ctx->eval_ctx.ctime;
+		state.time = DEG_get_ctime(draw_ctx->depsgraph);
 		if (!psys_get_particle_state(&sim, curr_point, &state, 0)) {
 			continue;
 		}
diff --git a/source/blender/draw/intern/draw_manager.c b/source/blender/draw/intern/draw_manager.c
index b8ed341e2c5..7d4d27b7147 100644
--- a/source/blender/draw/intern/draw_manager.c
+++ b/source/blender/draw/intern/draw_manager.c
@@ -351,16 +351,6 @@ static void drw_viewport_cache_resize(void)
 	DRW_instance_data_list_resize(DST.idatalist);
 }
 
-static void drw_state_eval_ctx_init(DRWManager *dst)
-{
-	DRWContextState *draw_ctx = &dst->draw_ctx;
-	DEG_evaluation_context_init_from_scene(
-	        &draw_ctx->eval_ctx,
-	        draw_ctx->scene,
-	        draw_ctx->view_layer,
-	        DST.options.is_scene_render ? DAG_EVAL_RENDER : DAG_EVAL_VIEWPORT);
-}
-
 /* Not a viewport variable, we could split this out. */
 static void drw_context_state_init(void)
 {
@@ -389,8 +379,6 @@ static void drw_context_state_init(void)
 	else {
 		DST.draw_ctx.object_pose = NULL;
 	}
-
-	drw_state_eval_ctx_init(&DST);
 }
 
 /* It also stores viewport variable to an immutable place: DST
@@ -1123,15 +1111,14 @@ void DRW_notify_id_update(const DRWUpdateContext *update_ctx, ID *id)
  * for each relevant engine / mode engine. */
 void DRW_draw_view(const bContext *C)
 {
-	EvaluationContext eval_ctx;
-	CTX_data_eval_ctx(C, &eval_ctx);
+	Depsgraph *depsgraph = CTX_data_depsgraph(C);
 	RenderEngineType *engine_type = CTX_data_engine_type(C);
 	ARegion *ar = CTX_wm_region(C);
 	View3D *v3d = CTX_wm_view3d(C);
 
 	/* Reset before using it. */
 	drw_state_prepare_clean_for_draw(&DST);
-	DRW_draw_render_loop_ex(eval_ctx.depsgraph, engine_type, ar, v3d, C);
+	DRW_draw_render_loop_ex(depsgraph, engine_type, ar, v3d, C);
 }
 
 /**
@@ -1555,7 +1542,14 @@ void DRW_draw_select_loop(
 		drw_engines_cache_init();
 
 		if (use_obedit) {
+#if 0
 			drw_engines_cache_populate(obact);
+#else
+			FOREACH_OBJECT_IN_MODE_BEGIN (view_layer, obact->mode, ob_iter) {
+				drw_engines_cache_populate(ob_iter);
+			}
+			FOREACH_OBJECT_IN_MODE_END;
+#endif
 		}
 		else {
 			DEG_OBJECT_ITER_BEGIN(
diff --git a/source/blender/draw/intern/draw_manager.h b/source/blender/draw/intern/draw_manager.h
index dd7e84f67d4..4f322b11931 100644
--- a/source/blender/draw/intern/draw_manager.h
+++ b/source/blender/draw/intern/draw_manager.h
@@ -222,6 +222,7 @@ struct DRWShadingGroup {
 	int normalworld;
 	int orcotexfac;
 	int eye;
+	int callid;
 	uint16_t matflag; /* Matrices needed, same as DRWCall.flag */
 
 #ifndef NDEBUG
diff --git a/source/blender/draw/intern/draw_manager_data.c b/source/blender/draw/intern/draw_manager_data.c
index ae7854b436c..544129ea124 100644
--- a/source/blender/draw/intern/draw_manager_data.c
+++ b/source/blender/draw/intern/draw_manager_data.c
@@ -498,6 +498,7 @@ static void drw_shgroup_init(DRWShadingGroup *shgroup, GPUShader *shader)
 	shgroup->normalworld = GPU_shader_get_builtin_uniform(shader, GWN_UNIFORM_WORLDNORMAL);
 	shgroup->orcotexfac = GPU_shader_get_builtin_uniform(shader, GWN_UNIFORM_ORCO);
 	shgroup->eye = GPU_shader_get_builtin_uniform(shader, GWN_UNIFORM_EYE);
+	shgroup->callid = GPU_shader_get_builtin_uniform(shader, GWN_UNIFORM_CALLID);
 
 	shgroup->matflag = 0;
 	if (shgroup->modelinverse > -1)
diff --git a/source/blender/draw/intern/draw_manager_exec.c b/source/blender/draw/intern/draw_manager_exec.c
index 2a5a595c682..dcd976805fd 100644
--- a/source/blender/draw/intern/draw_manager_exec.c
+++ b/source/blender/draw/intern/draw_manager_exec.c
@@ -384,100 +384,53 @@ void DRW_state_clip_planes_reset(void)
 /** \name Clipping (DRW_clipping)
  * \{ */
 
-/* Extract the 8 corners (world space).
+/* Extract the 8 corners from a Projection Matrix.
  * Although less accurate, this solution can be simplified as follows:
  * BKE_boundbox_init_from_minmax(&bbox, (const float[3]){-1.0f, -1.0f, -1.0f}, (const float[3]){1.0f, 1.0f, 1.0f});
- * for (int i = 0; i < 8; i++) {mul_project_m4_v3(viewprojinv, bbox.vec[i]);}
+ * for (int i = 0; i < 8; i++) {mul_project_m4_v3(projinv, bbox.vec[i]);}
  */
-static void draw_frustum_boundbox_calc(
-        const float (*projmat)[4], const float (*viewinv)[4], BoundBox *r_bbox)
+static void draw_frustum_boundbox_calc(const float(*projmat)[4], BoundBox *r_bbox)
 {
-	float screenvecs[3][3], loc[3], near, far, w_half, h_half;
+	float near, far, left, right, bottom, top;
 	bool is_persp = projmat[3][3] == 0.0f;
-	copy_m3_m4(screenvecs, viewinv);
-	copy_v3_v3(loc, viewinv[3]);
 
-	/* get the values of the minimum and maximum clipping planes distances
-	 * and half the width and height of the nearplane rectangle. */
 	if (is_persp) {
-		near = projmat[3][2] / (projmat[2][2] - 1.0f);
-		far = projmat[3][2] / (projmat[2][2] + 1.0f);
-		w_half = near / projmat[0][0];
-		h_half = near / projmat[1][1];
+		near   = projmat[3][2] / (projmat[2][2] - 1.0f);
+		far    = projmat[3][2] / (projmat[2][2] + 1.0f);
+		left   = near * (projmat[2][0] - 1.0f) / projmat[0][0];
+		right  = near * (projmat[2][0] + 1.0f) / projmat[0][0];
+		bottom = near * (projmat[2][1] - 1.0f) / projmat[1][1];
+		top    = near * (projmat[2][1] + 1.0f) / projmat[1][1];
 	}
 	else {
-		near = (projmat[3][2] + 1.0f) / projmat[2][2];
-		far = (projmat[3][2] - 1.0f) / projmat[2][2];
-		w_half = 1.0f / projmat[0][0];
-		h_half = 1.0f / projmat[1][1];
+		near   = ( projmat[3][2] + 1.0f) / projmat[2][2];
+		far    = ( projmat[3][2] - 1.0f) / projmat[2][2];
+		left   = (-projmat[3][0] - 1.0f) / projmat[0][0];
+		right  = (-projmat[3][0] + 1.0f) / projmat[0][0];
+		bottom = (-projmat[3][1] - 1.0f) / projmat[1][1];
+		top    = (-projmat[3][1] + 1.0f) / projmat[1][1];
 	}
 
-	/* With vectors aligned to the screen, reconstruct
-	 * the near plane from the dimensions obtained earlier. */
-	float mid[3], hor[3], ver[3];
-	mul_v3_v3fl(hor, screenvecs[0], w_half);
-	mul_v3_v3fl(ver, screenvecs[1], h_half);
-	madd_v3_v3v3fl(mid, loc, screenvecs[2], -near);
-
-	/* The case below is for non-symmetric frustum. */
-	if (is_persp) {
-		madd_v3_v3fl(mid, hor, projmat[2][0]);
-		madd_v3_v3fl(mid, ver, projmat[2][1]);
-	}
-	else {
-		madd_v3_v3fl(mid, hor, -projmat[3][0]);
-		madd_v3_v3fl(mid, ver, -projmat[3][1]);
-	}
-
-	r_bbox->vec[0][0] = mid[0] - ver[0] - hor[0];
-	r_bbox->vec[0][1] = mid[1] - ver[1] - hor[1];
-	r_bbox->vec[0][2] = mid[2] - ver[2] - hor[2];
-
-	r_bbox->vec[3][0] = mid[0] + ver[0] - hor[0];
-	r_bbox->vec[3][1] = mid[1] + ver[1] - hor[1];
-	r_bbox->vec[3][2] = mid[2] + ver[2] - hor[2];
-
-	r_bbox->vec[7][0] = mid[0] + ver[0] + hor[0];
-	r_bbox->vec[7][1] = mid[1] + ver[1] + hor[1];
-	r_bbox->vec[7][2] = mid[2] + ver[2] + hor[2];
-
-	r_bbox->vec[4][0] = mid[0] - ver[0] + hor[0];
-	r_bbox->vec[4][1] = mid[1] - ver[1] + hor[1];
-	r_bbox->vec[4][2] = mid[2] - ver[2] + hor[2];
+	r_bbox->vec[0][2] = r_bbox->vec[3][2] = r_bbox->vec[7][2] = r_bbox->vec[4][2] = -near;
+	r_bbox->vec[0][0] = r_bbox->vec[3][0] = left;
+	r_bbox->vec[4][0] = r_bbox->vec[7][0] = right;
+	r_bbox->vec[0][1] = r_bbox->vec[4][1] = bottom;
+	r_bbox->vec[7][1] = r_bbox->vec[3][1] = top;
 
 	/* Get the coordinates of the far plane. */
 	if (is_persp) {
 		float sca_far = far / near;
-		mid[0] = mid[0] + (mid[0] - loc[0]) * sca_far;
-		mid[1] = mid[1] + (mid[1] - loc[1]) * sca_far;
-		mid[2] = mid[2] + (mid[2] - loc[2]) * sca_far;
-
-		mul_v3_fl(hor, sca_far);
-		mul_v3_fl(ver, sca_far);
+		left   *= sca_far;
+		bottom *= sca_far;
+		right  *= sca_far;
+		top    *= sca_far;
 	}
-	else {
-		madd_v3_v3v3fl(mid, loc, screenvecs[2], -far);
 
-		/* Non-symmetric frustum. */
-		madd_v3_v3fl(mid, hor, -projmat[3][0]);
-		madd_v3_v3fl(mid, ver, -projmat[3][1]);
-	}
-
-	r_bbox->vec[1][0] = mid[0] - ver[0] - hor[0];
-	r_bbox->vec[1][1] = mid[1] - ver[1] - hor[1];
-	r_bbox->vec[1][2] = mid[2] - ver[2] - hor[2];
-
-	r_bbox->vec[2][0] = mid[0] + ver[0] - hor[0];
-	r_bbox->vec[2][1] = mid[1] + ver[1] - hor[1];
-	r_bbox->vec[2][2] = mid[2] + ver[2] - hor[2];
-
-	r_bbox->vec[6][0] = mid[0] + ver[0] + hor[0];
-	r_bbox->vec[6][1] = mid[1] + ver[1] + hor[1];
-	r_bbox->vec[6][2] = mid[2] + ver[2] + hor[2];
-
-	r_bbox->vec[5][0] = mid[0] - ver[0] + hor[0];
-	r_bbox->vec[5][1] = mid[1] - ver[1] + hor[1];
-	r_bbox->vec[5][2] = mid[2] - ver[2] + hor[2];
+	r_bbox->vec[1][2] = r_bbox->vec[2][2] = r_bbox->vec[6][2] = r_bbox->vec[5][2] = -far;
+	r_bbox->vec[1][0] = r_bbox->vec[2][0] = left;
+	r_bbox->vec[6][0] = r_bbox->vec[5][0] = right;
+	r_bbox->vec[1][1] = r_bbox->vec[5][1] = bottom;
+	r_bbox->vec[2][1] = r_bbox->vec[6][1] = top;
 }
 
 static void draw_clipping_setup_from_view(void)
@@ -495,11 +448,15 @@ static void draw_clipping_setup_from_view(void)
 #if 0 /* It has accuracy problems. */
 	BKE_boundbox_init_from_minmax(&bbox, (const float[3]){-1.0f, -1.0f, -1.0f}, (const float[3]){1.0f, 1.0f, 1.0f});
 	for (int i = 0; i < 8; i++) {
-		mul_project_m4_v3(DST.view_data.matstate.mat[DRW_MAT_PERSINV], bbox.vec[i]);
+		mul_project_m4_v3(projinv, bbox.vec[i]);
 	}
 #else
-	draw_frustum_boundbox_calc(projmat, viewinv, &bbox);
+	draw_frustum_boundbox_calc(projmat, &bbox);
 #endif
+	/* Transform into world space. */
+	for (int i = 0; i < 8; i++) {
+		mul_m4_v3(viewinv, bbox.vec[i]);
+	}
 
 	/* Compute clip planes using the world space frustum corners. */
 	for (int p = 0; p < 6; p++) {
@@ -527,9 +484,6 @@ static void draw_clipping_setup_from_view(void)
 		float *nearpoint = bbox.vec[0];
 		float *farpoint = bbox.vec[6];
 
-		mul_project_m4_v3(projinv, nearpoint);
-		mul_project_m4_v3(projinv, farpoint);
-
 		/* just use median point */
 		mid_v3_v3v3(bsphere->center, farpoint, nearpoint);
 		bsphere->radius = len_v3v3(bsphere->center, farpoint);
@@ -553,7 +507,6 @@ static void draw_clipping_setup_from_view(void)
 		float h_sq = len_squared_v3v3(mid_min, mid_max);
 
 		float fac = (4 * h_sq + b_sq - a_sq) / (8 * h_sq);
-		BLI_assert(fac >= 0.0f);
 
 		/* The goal is to get the smallest sphere,
 		 * not the sphere that passes through each corner */
@@ -1010,7 +963,7 @@ static void draw_shgroup(DRWShadingGroup *shgroup, DRWState pass_state)
 #  define GPU_SELECT_LOAD_IF_PICKSEL_LIST_END(start, count)
 #  define GPU_SELECT_LOAD_IF_PICKSEL_LIST(_shgroup, _start, _count) \
 	_start = 0;                                                     \
-	_count = _shgroup->interface.instance_count;
+	_count = _shgroup->instance_count;
 
 #endif
 
@@ -1052,6 +1005,7 @@ static void draw_shgroup(DRWShadingGroup *shgroup, DRWState pass_state)
 	}
 	else {
 		bool prev_neg_scale = false;
+		int callid = 0;
 		for (DRWCall *call = shgroup->calls.first; call; call = call->next) {
 
 			/* OPTI/IDEA(clem): Do this preparation in another thread. */
@@ -1060,6 +1014,12 @@ static void draw_shgroup(DRWShadingGroup *shgroup, DRWState pass_state)
 			if ((call->state->flag & DRW_CALL_CULLED) != 0)
 				continue;
 
+			/* XXX small exception/optimisation for outline rendering. */
+			if (shgroup->callid != -1) {
+				GPU_shader_uniform_vector_int(shgroup->shader, shgroup->callid, 1, 1, &callid);
+				callid += 1;
+			}
+
 			/* Negative scale objects */
 			bool neg_scale = call->state->flag & DRW_CALL_NEGSCALE;
 			if (neg_scale != prev_neg_scale) {
diff --git a/source/blender/draw/intern/draw_manager_texture.c b/source/blender/draw/intern/draw_manager_texture.c
index 65856a6bf5c..f083e399ac2 100644
--- a/source/blender/draw/intern/draw_manager_texture.c
+++ b/source/blender/draw/intern/draw_manager_texture.c
@@ -36,6 +36,7 @@ void drw_texture_get_format(
 			 * Generally they are multiple of 16bit. */
 			case DRW_TEX_R_16:
 			case DRW_TEX_R_16I:
+			case DRW_TEX_R_16U:
 			case DRW_TEX_R_32:
 			case DRW_TEX_RG_8:
 			case DRW_TEX_RG_16:
@@ -72,6 +73,7 @@ void drw_texture_get_format(
 		case DRW_TEX_R_8: *r_data_type = GPU_R8; break;
 		case DRW_TEX_R_16: *r_data_type = GPU_R16F; break;
 		case DRW_TEX_R_16I: *r_data_type = GPU_R16I; break;
+		case DRW_TEX_R_16U: *r_data_type = GPU_R16UI; break;
 		case DRW_TEX_R_32: *r_data_type = GPU_R32F; break;
 #if 0
 		case DRW_TEX_RGB_8: *r_data_type = GPU_RGB8; break;
diff --git a/source/blender/draw/modes/edit_curve_mode.c b/source/blender/draw/modes/edit_curve_mode.c
index 73a4fb1e9e6..b33ebd8ba60 100644
--- a/source/blender/draw/modes/edit_curve_mode.c
+++ b/source/blender/draw/modes/edit_curve_mode.c
@@ -28,6 +28,8 @@
 
 #include "DNA_curve_types.h"
 
+#include "BKE_object.h"
+
 /* If builtin shaders are needed */
 #include "GPU_shader.h"
 #include "GPU_batch.h"
@@ -233,7 +235,12 @@ static void EDIT_CURVE_cache_populate(void *vedata, Object *ob)
 	UNUSED_VARS(psl, stl);
 
 	if (ob->type == OB_CURVE) {
-		if (ob == draw_ctx->object_edit) {
+#if 0
+		if (ob == draw_ctx->object_edit)
+#else
+		if ((ob == draw_ctx->object_edit) || BKE_object_is_in_editmode_and_selected(ob))
+#endif
+		{
 			Curve *cu = ob->data;
 			/* Get geometry cache */
 			struct Gwn_Batch *geom;
diff --git a/source/blender/draw/modes/edit_lattice_mode.c b/source/blender/draw/modes/edit_lattice_mode.c
index 0268f4eb453..e8628711ffd 100644
--- a/source/blender/draw/modes/edit_lattice_mode.c
+++ b/source/blender/draw/modes/edit_lattice_mode.c
@@ -26,6 +26,8 @@
 #include "DRW_engine.h"
 #include "DRW_render.h"
 
+#include "BKE_object.h"
+
 /* If builtin shaders are needed */
 #include "GPU_shader.h"
 
@@ -192,7 +194,7 @@ static void EDIT_LATTICE_cache_populate(void *vedata, Object *ob)
 	UNUSED_VARS(psl);
 
 	if (ob->type == OB_LATTICE) {
-		if (ob == draw_ctx->object_edit) {
+		if ((ob == draw_ctx->object_edit) || BKE_object_is_in_editmode_and_selected(ob)) {
 			/* Get geometry cache */
 			struct Gwn_Batch *geom;
 
diff --git a/source/blender/draw/modes/edit_mesh_mode.c b/source/blender/draw/modes/edit_mesh_mode.c
index 4bd69941809..c465fa38f04 100644
--- a/source/blender/draw/modes/edit_mesh_mode.c
+++ b/source/blender/draw/modes/edit_mesh_mode.c
@@ -37,6 +37,8 @@
 
 #include "edit_mesh_mode_intern.h" /* own include */
 
+#include "BKE_object.h"
+
 extern struct GPUUniformBuffer *globals_ubo; /* draw_common.c */
 extern struct GlobalsUboStorage ts; /* draw_common.c */
 
@@ -448,7 +450,7 @@ static void EDIT_MESH_cache_populate(void *vedata, Object *ob)
 	struct Gwn_Batch *geom;
 
 	if (ob->type == OB_MESH) {
-		if (ob == draw_ctx->object_edit) {
+		if ((ob == draw_ctx->object_edit) || BKE_object_is_in_editmode_and_selected(ob)) {
 			const Mesh *me = ob->data;
 			IDProperty *ces_mode_ed = BKE_layer_collection_engine_evaluated_get(ob, COLLECTION_MODE_EDIT, "");
 			bool do_occlude_wire = BKE_collection_engine_property_value_get_bool(ces_mode_ed, "show_occlude_wire");
diff --git a/source/blender/draw/modes/edit_metaball_mode.c b/source/blender/draw/modes/edit_metaball_mode.c
index bcabeef5bc3..f7b7113a4d6 100644
--- a/source/blender/draw/modes/edit_metaball_mode.c
+++ b/source/blender/draw/modes/edit_metaball_mode.c
@@ -28,6 +28,7 @@
 
 #include "DNA_meta_types.h"
 
+#include "BKE_object.h"
 #include "BKE_mball.h"
 
 /* If builtin shaders are needed */
@@ -171,7 +172,7 @@ static void EDIT_METABALL_cache_populate(void *vedata, Object *ob)
 		const DRWContextState *draw_ctx = DRW_context_state_get();
 		DRWShadingGroup *group = stl->g_data->group;
 
-		if (ob == draw_ctx->object_edit) {
+		if ((ob == draw_ctx->object_edit) || BKE_object_is_in_editmode_and_selected(ob)) {
 			MetaBall *mb = ob->data;
 
 			const bool is_select = DRW_state_is_select();
diff --git a/source/blender/draw/modes/object_mode.c b/source/blender/draw/modes/object_mode.c
index 939507a4f21..0d9e46d1268 100644
--- a/source/blender/draw/modes/object_mode.c
+++ b/source/blender/draw/modes/object_mode.c
@@ -40,6 +40,8 @@
 
 #include "BIF_gl.h"
 
+#include "BLI_string_utils.h"
+
 #include "BKE_anim.h"
 #include "BKE_camera.h"
 #include "BKE_curve.h"
@@ -69,6 +71,7 @@ extern struct GPUUniformBuffer *globals_ubo; /* draw_common.c */
 extern struct GPUTexture *globals_ramp; /* draw_common.c */
 extern GlobalsUboStorage ts;
 
+extern char datatoc_object_outline_prepass_frag_glsl[];
 extern char datatoc_object_outline_resolve_frag_glsl[];
 extern char datatoc_object_outline_detect_frag_glsl[];
 extern char datatoc_object_outline_expand_frag_glsl[];
@@ -223,6 +226,13 @@ typedef struct OBJECT_PrivateData {
 	DRWShadingGroup *points_select;
 	DRWShadingGroup *points_select_group;
 	DRWShadingGroup *points_transform;
+
+	/* Outlines id offset */
+	int id_ofs_active;
+	int id_ofs_active_group;
+	int id_ofs_select;
+	int id_ofs_select_group;
+	int id_ofs_transform;
 } OBJECT_PrivateData; /* Transient data */
 
 static struct {
@@ -232,6 +242,7 @@ static struct {
 	struct Gwn_VertFormat *empty_image_wire_format;
 
 	/* fullscreen shaders */
+	GPUShader *outline_prepass_sh;
 	GPUShader *outline_resolve_sh;
 	GPUShader *outline_resolve_aa_sh;
 	GPUShader *outline_detect_sh;
@@ -259,6 +270,7 @@ static struct {
 	bool draw_grid;
 	/* Temp buffer textures */
 	struct GPUTexture *outlines_depth_tx;
+	struct GPUTexture *outlines_id_tx;
 	struct GPUTexture *outlines_color_tx;
 	struct GPUTexture *outlines_blur_tx;
 } e_data = {NULL}; /* Engine data */
@@ -288,14 +300,20 @@ static void OBJECT_engine_init(void *vedata)
 	if (DRW_state_is_fbo()) {
 		e_data.outlines_depth_tx = DRW_texture_pool_query_2D(size[0], size[1], DRW_TEX_DEPTH_24,
 		                                                     &draw_engine_object_type);
-		e_data.outlines_color_tx = DRW_texture_pool_query_2D(size[0], size[1], DRW_TEX_RGBA_8,
-		                                                     &draw_engine_object_type);
+		/* XXX TODO DRW_TEX_R_16U can overflow, it would cause no harm
+		 * (only bad colored or missing outlines) but we should
+		 * use 32bits only if the scene have that many objects */
+		e_data.outlines_id_tx = DRW_texture_pool_query_2D(size[0], size[1], DRW_TEX_R_16U,
+		                                                  &draw_engine_object_type);
 
 		GPU_framebuffer_ensure_config(&fbl->outlines_fb, {
 			GPU_ATTACHMENT_TEXTURE(e_data.outlines_depth_tx),
-			GPU_ATTACHMENT_TEXTURE(e_data.outlines_color_tx)
+			GPU_ATTACHMENT_TEXTURE(e_data.outlines_id_tx)
 		});
 
+		e_data.outlines_color_tx = DRW_texture_pool_query_2D(size[0], size[1], DRW_TEX_RGBA_8,
+		                                                     &draw_engine_object_type);
+
 		GPU_framebuffer_ensure_config(&fbl->expand_fb, {
 			GPU_ATTACHMENT_NONE,
 			GPU_ATTACHMENT_TEXTURE(e_data.outlines_color_tx)
@@ -310,66 +328,58 @@ static void OBJECT_engine_init(void *vedata)
 		});
 	}
 
+	/* Shaders */
 	if (!e_data.outline_resolve_sh) {
+		/* Outline */
+		e_data.outline_prepass_sh = DRW_shader_create_3D(datatoc_object_outline_prepass_frag_glsl, NULL);
+
 		e_data.outline_resolve_sh = DRW_shader_create_fullscreen(datatoc_object_outline_resolve_frag_glsl, NULL);
-	}
 
-	if (!e_data.outline_resolve_aa_sh) {
 		e_data.outline_resolve_aa_sh = DRW_shader_create_with_lib(
 		            datatoc_common_fullscreen_vert_glsl, NULL,
 		            datatoc_object_outline_resolve_frag_glsl,
 		            datatoc_common_fxaa_lib_glsl,
 		            "#define FXAA_ALPHA\n"
 		            "#define USE_FXAA\n");
-	}
 
-	if (!e_data.outline_detect_sh) {
-		e_data.outline_detect_sh = DRW_shader_create_fullscreen(datatoc_object_outline_detect_frag_glsl, NULL);
-	}
+		e_data.outline_detect_sh = DRW_shader_create_with_lib(
+		            datatoc_common_fullscreen_vert_glsl, NULL,
+		            datatoc_object_outline_detect_frag_glsl,
+		            datatoc_common_globals_lib_glsl,
+		            NULL);
 
-	if (!e_data.outline_fade_sh) {
 		e_data.outline_fade_sh = DRW_shader_create_fullscreen(datatoc_object_outline_expand_frag_glsl, NULL);
-	}
 
-	if (!e_data.object_empty_image_sh) {
+		/* Empty images */
 		e_data.object_empty_image_sh = DRW_shader_create_with_lib(
 		           datatoc_object_empty_image_vert_glsl, NULL,
 		           datatoc_object_empty_image_frag_glsl,
 		           datatoc_common_globals_lib_glsl, NULL);
-	}
 
-	if (!e_data.object_empty_image_wire_sh) {
 		e_data.object_empty_image_wire_sh = DRW_shader_create_with_lib(
 		           datatoc_object_empty_image_vert_glsl, NULL,
 		           datatoc_object_empty_image_frag_glsl,
 		           datatoc_common_globals_lib_glsl,
 		           "#define USE_WIRE\n");
-	}
 
-	if (!e_data.grid_sh) {
+		/* Grid */
 		e_data.grid_sh = DRW_shader_create_with_lib(
 		        datatoc_object_grid_vert_glsl, NULL,
 		        datatoc_object_grid_frag_glsl,
 		        datatoc_common_globals_lib_glsl, NULL);
-	}
 
-	if (!e_data.part_prim_sh) {
+		/* Particles */
 		e_data.part_prim_sh = DRW_shader_create(
 		        datatoc_object_particle_prim_vert_glsl, NULL, datatoc_gpu_shader_flat_color_frag_glsl, NULL);
-	}
 
-	if (!e_data.part_axis_sh) {
 		e_data.part_axis_sh = DRW_shader_create(
 		        datatoc_object_particle_prim_vert_glsl, NULL, datatoc_gpu_shader_flat_color_frag_glsl,
 		        "#define USE_AXIS\n");
-	}
 
-	if (!e_data.part_dot_sh) {
 		e_data.part_dot_sh = DRW_shader_create(
 		        datatoc_object_particle_dot_vert_glsl, NULL, datatoc_object_particle_dot_frag_glsl, NULL);
-	}
 
-	if (!e_data.lightprobe_grid_sh) {
+		/* Lightprobes */
 		e_data.lightprobe_grid_sh = DRW_shader_create(
 		        datatoc_object_lightprobe_grid_vert_glsl, NULL, datatoc_gpu_shader_uniform_color_frag_glsl, NULL);
 	}
@@ -558,6 +568,7 @@ static void OBJECT_engine_free(void)
 	MEM_SAFE_FREE(e_data.particle_format);
 	MEM_SAFE_FREE(e_data.empty_image_format);
 	MEM_SAFE_FREE(e_data.empty_image_wire_format);
+	DRW_SHADER_FREE_SAFE(e_data.outline_prepass_sh);
 	DRW_SHADER_FREE_SAFE(e_data.outline_resolve_sh);
 	DRW_SHADER_FREE_SAFE(e_data.outline_resolve_aa_sh);
 	DRW_SHADER_FREE_SAFE(e_data.outline_detect_sh);
@@ -571,10 +582,10 @@ static void OBJECT_engine_free(void)
 	DRW_SHADER_FREE_SAFE(e_data.lightprobe_grid_sh);
 }
 
-static DRWShadingGroup *shgroup_outline(DRWPass *pass, const float col[4], GPUShader *sh)
+static DRWShadingGroup *shgroup_outline(DRWPass *pass, const int *ofs, GPUShader *sh)
 {
 	DRWShadingGroup *grp = DRW_shgroup_create(sh, pass);
-	DRW_shgroup_uniform_vec4(grp, "color", col, 1);
+	DRW_shgroup_uniform_int(grp, "baseId", ofs, 1);
 
 	return grp;
 }
@@ -602,12 +613,19 @@ static DRWShadingGroup *shgroup_theme_id_to_outline_or(
 {
 	switch (theme_id) {
 		case TH_ACTIVE:
+			stl->g_data->id_ofs_active++;
 			return stl->g_data->outlines_active;
 		case TH_SELECT:
+			stl->g_data->id_ofs_select++;
 			return stl->g_data->outlines_select;
-		case TH_GROUP_ACTIVE:
+		case TH_GROUP:
+			stl->g_data->id_ofs_select_group++;
 			return stl->g_data->outlines_select_group;
+		case TH_GROUP_ACTIVE:
+			stl->g_data->id_ofs_active_group++;
+			return stl->g_data->outlines_active_group;
 		case TH_TRANSFORM:
+			stl->g_data->id_ofs_transform++;
 			return stl->g_data->outlines_transform;
 		default:
 			return fallback;
@@ -622,8 +640,10 @@ static DRWShadingGroup *shgroup_theme_id_to_wire_or(
 			return stl->g_data->wire_active;
 		case TH_SELECT:
 			return stl->g_data->wire_select;
-		case TH_GROUP_ACTIVE:
+		case TH_GROUP:
 			return stl->g_data->wire_select_group;
+		case TH_GROUP_ACTIVE:
+			return stl->g_data->wire_active_group;
 		case TH_TRANSFORM:
 			return stl->g_data->wire_transform;
 		default:
@@ -639,8 +659,10 @@ static DRWShadingGroup *shgroup_theme_id_to_point_or(
 			return stl->g_data->points_active;
 		case TH_SELECT:
 			return stl->g_data->points_select;
-		case TH_GROUP_ACTIVE:
+		case TH_GROUP:
 			return stl->g_data->points_select_group;
+		case TH_GROUP_ACTIVE:
+			return stl->g_data->points_active_group;
 		case TH_TRANSFORM:
 			return stl->g_data->points_transform;
 		default:
@@ -786,18 +808,24 @@ static void OBJECT_cache_init(void *vedata)
 		DRWState state = DRW_STATE_WRITE_COLOR | DRW_STATE_WRITE_DEPTH | DRW_STATE_DEPTH_LESS | DRW_STATE_WIRE;
 		psl->outlines = DRW_pass_create("Outlines Depth Pass", state);
 
-		GPUShader *sh = GPU_shader_get_builtin_shader(GPU_SHADER_3D_UNIFORM_COLOR);
+		GPUShader *sh = e_data.outline_prepass_sh;
 
 		/* Select */
-		stl->g_data->outlines_select = shgroup_outline(psl->outlines, ts.colorSelect, sh);
-		stl->g_data->outlines_select_group = shgroup_outline(psl->outlines, ts.colorGroupActive, sh);
+		stl->g_data->outlines_select = shgroup_outline(psl->outlines, &stl->g_data->id_ofs_select, sh);
+		stl->g_data->outlines_select_group = shgroup_outline(psl->outlines, &stl->g_data->id_ofs_select_group, sh);
 
 		/* Transform */
-		stl->g_data->outlines_transform = shgroup_outline(psl->outlines, ts.colorTransform, sh);
+		stl->g_data->outlines_transform = shgroup_outline(psl->outlines, &stl->g_data->id_ofs_transform, sh);
 
 		/* Active */
-		stl->g_data->outlines_active = shgroup_outline(psl->outlines, ts.colorActive, sh);
-		stl->g_data->outlines_active_group = shgroup_outline(psl->outlines, ts.colorGroupActive, sh);
+		stl->g_data->outlines_active = shgroup_outline(psl->outlines, &stl->g_data->id_ofs_active, sh);
+		stl->g_data->outlines_active_group = shgroup_outline(psl->outlines, &stl->g_data->id_ofs_active_group, sh);
+
+		stl->g_data->id_ofs_select = 0;
+		stl->g_data->id_ofs_select_group = 0;
+		stl->g_data->id_ofs_active = 0;
+		stl->g_data->id_ofs_active_group = 0;
+		stl->g_data->id_ofs_transform = 0;
 	}
 
 	{
@@ -822,10 +850,12 @@ static void OBJECT_cache_init(void *vedata)
 		psl->outlines_search = DRW_pass_create("Outlines Detect Pass", state);
 
 		DRWShadingGroup *grp = DRW_shgroup_create(e_data.outline_detect_sh, psl->outlines_search);
-		DRW_shgroup_uniform_texture_ref(grp, "outlineColor", &e_data.outlines_color_tx);
+		DRW_shgroup_uniform_texture_ref(grp, "outlineId", &e_data.outlines_id_tx);
 		DRW_shgroup_uniform_texture_ref(grp, "outlineDepth", &e_data.outlines_depth_tx);
 		DRW_shgroup_uniform_texture_ref(grp, "sceneDepth", &dtxl->depth);
+		DRW_shgroup_uniform_block(grp, "globalsBlock", globals_ubo);
 		DRW_shgroup_uniform_float(grp, "alphaOcclu", &alphaOcclu, 1);
+		DRW_shgroup_uniform_int(grp, "idOffsets", &stl->g_data->id_ofs_active, 5);
 		DRW_shgroup_call_add(grp, quad, NULL);
 
 		psl->outlines_expand = DRW_pass_create("Outlines Expand Pass", state);
@@ -2038,6 +2068,7 @@ static void OBJECT_draw_scene(void *vedata)
 	OBJECT_PassList *psl = ((OBJECT_Data *)vedata)->psl;
 	OBJECT_StorageList *stl = ((OBJECT_Data *)vedata)->stl;
 	OBJECT_FramebufferList *fbl = ((OBJECT_Data *)vedata)->fbl;
+	OBJECT_PrivateData *g_data = stl->g_data;
 	DefaultFramebufferList *dfbl = DRW_viewport_framebuffer_list_get();
 	DefaultTextureList *dtxl = DRW_viewport_texture_list_get();
 
@@ -2045,7 +2076,18 @@ static void OBJECT_draw_scene(void *vedata)
 
 	if (DRW_state_is_fbo()) {
 		DRW_stats_group_start("Outlines");
-		
+
+		int id_ct_select = g_data->id_ofs_select;
+		int id_ct_select_group = g_data->id_ofs_select_group;
+		int id_ct_active = g_data->id_ofs_active;
+		int id_ct_active_group = g_data->id_ofs_active_group;
+
+		g_data->id_ofs_active = 1;
+		g_data->id_ofs_active_group = g_data->id_ofs_active + id_ct_active + 1;
+		g_data->id_ofs_select = g_data->id_ofs_active_group + id_ct_active_group + 1;
+		g_data->id_ofs_select_group = g_data->id_ofs_select + id_ct_select + 1;
+		g_data->id_ofs_transform = g_data->id_ofs_select_group + id_ct_select_group + 1;
+
 		/* Render filled polygon on a separate framebuffer */
 		GPU_framebuffer_bind(fbl->outlines_fb);
 		GPU_framebuffer_clear_color_depth(fbl->outlines_fb, clearcol, 1.0f);
diff --git a/source/blender/draw/modes/pose_mode.c b/source/blender/draw/modes/pose_mode.c
index 749c3e71368..1d3d31ab54d 100644
--- a/source/blender/draw/modes/pose_mode.c
+++ b/source/blender/draw/modes/pose_mode.c
@@ -139,7 +139,9 @@ bool DRW_pose_mode_armature(Object *ob, Object *active_ob)
 	const DRWContextState *draw_ctx = DRW_context_state_get();
 
 	/* Pode armature is handled by pose mode engine. */
-	if ((ob == active_ob) && ((draw_ctx->object_mode & OB_MODE_POSE) != 0)) {
+	if (((ob == active_ob) || (ob->base_flag & BASE_SELECTED)) &&
+	    ((draw_ctx->object_mode & OB_MODE_POSE) != 0))
+	{
 		return true;
 	}
 
diff --git a/source/blender/draw/modes/sculpt_mode.c b/source/blender/draw/modes/sculpt_mode.c
index 65f4653591f..4fb4464df06 100644
--- a/source/blender/draw/modes/sculpt_mode.c
+++ b/source/blender/draw/modes/sculpt_mode.c
@@ -202,7 +202,7 @@ static void SCULPT_cache_populate(void *vedata, Object *ob)
 				 * but this avoids waiting on first stroke) */
 				Scene *scene = draw_ctx->scene;
 
-				BKE_sculpt_update_mesh_elements(&draw_ctx->eval_ctx, scene, scene->toolsettings->sculpt, ob, false, false);
+				BKE_sculpt_update_mesh_elements(draw_ctx->depsgraph, scene, scene->toolsettings->sculpt, ob, false, false);
 			}
 
 			PBVH *pbvh = ob->sculpt->pbvh;
diff --git a/source/blender/draw/modes/shaders/object_outline_detect_frag.glsl b/source/blender/draw/modes/shaders/object_outline_detect_frag.glsl
index dc0ea938436..6b0f66d7c94 100644
--- a/source/blender/draw/modes/shaders/object_outline_detect_frag.glsl
+++ b/source/blender/draw/modes/shaders/object_outline_detect_frag.glsl
@@ -3,77 +3,67 @@ in vec4 uvcoordsvar;
 
 out vec4 FragColor;
 
-uniform sampler2D outlineColor;
+uniform usampler2D outlineId;
 uniform sampler2D outlineDepth;
 uniform sampler2D sceneDepth;
 
+uniform int idOffsets[5];
+
 uniform float alphaOcclu;
 uniform vec2 viewportSize;
 
-void search_outline(ivec2 uv, vec4 ref_col, inout bool ref_occlu, inout bool outline)
+vec4 convert_id_to_color(int id)
 {
-	if (!outline) {
-		vec4 color = texelFetch(outlineColor, uv, 0).rgba;
-		if (color != ref_col) {
-			outline = true;
-		}
-		else {
-			float depth = texelFetch(outlineDepth, uv, 0).r;
-			float scene_depth = texelFetch(sceneDepth, uv, 0).r;
-			bool occlu = (depth > scene_depth);
-
-			if (occlu != ref_occlu && !ref_occlu) {
-				outline = true;
-			}
-		}
+	if (id == 0) {
+		return vec4(0.0);
+	}
+	if (id < idOffsets[1]) {
+		return colorActive;
+	}
+	else if (id < idOffsets[2]) {
+		return colorGroupActive;
+	}
+	else if (id < idOffsets[3]) {
+		return colorSelect;
+	}
+	else if (id < idOffsets[4]) {
+		return colorGroup;
+	}
+	else {
+		return colorTransform;
 	}
 }
 
+const ivec2 ofs[4] = ivec2[4](
+    ivec2( 1,  0), ivec2( 0,  1),
+    ivec2(-1,  0), ivec2( 0, -1)
+);
+
 void main()
 {
-	ivec2 uv = ivec2(gl_FragCoord.xy);
-
-	vec4 color[4];
-	/* Idea : Use a 16bit ID to identify the color
-	 * and store the colors in a UBO. And fetch all ids
-	 * for discontinuity check with one textureGather \o/ */
-	vec4 ref_col = texelFetch(outlineColor, uv, 0).rgba;
-	color[0] = texelFetchOffset(outlineColor, uv, 0, ivec2( 1,  0)).rgba;
-	color[1] = texelFetchOffset(outlineColor, uv, 0, ivec2( 0,  1)).rgba;
-	color[2] = texelFetchOffset(outlineColor, uv, 0, ivec2(-1,  0)).rgba;
-	color[3] = texelFetchOffset(outlineColor, uv, 0, ivec2( 0, -1)).rgba;
-
-	/* TODO GATHER */
-	vec4 depths;
-	float depth = texelFetch(outlineDepth, uv, 0).r;
-	depths.x = texelFetchOffset(outlineDepth, uv, 0, ivec2( 1,  0)).r;
-	depths.y = texelFetchOffset(outlineDepth, uv, 0, ivec2( 0,  1)).r;
-	depths.z = texelFetchOffset(outlineDepth, uv, 0, ivec2(-1,  0)).r;
-	depths.w = texelFetchOffset(outlineDepth, uv, 0, ivec2( 0, -1)).r;
-
-	vec4 scene_depths;
-	float scene_depth = texelFetch(sceneDepth, uv, 0).r;
-	scene_depths.x = texelFetchOffset(sceneDepth, uv, 0, ivec2( 1,  0)).r;
-	scene_depths.y = texelFetchOffset(sceneDepth, uv, 0, ivec2( 0,  1)).r;
-	scene_depths.z = texelFetchOffset(sceneDepth, uv, 0, ivec2(-1,  0)).r;
-	scene_depths.w = texelFetchOffset(sceneDepth, uv, 0, ivec2( 0, -1)).r;
+	ivec2 texel = ivec2(gl_FragCoord.xy);
+	vec2 uv = gl_FragCoord.xy / vec2(textureSize(outlineId, 0).xy);
 
-	bool ref_occlu = (depth > scene_depth);
-	bool outline = false;
-
-#if 1
-	bvec4 occlu = (!ref_occlu) ? notEqual(greaterThan(depths, scene_depths), bvec4(ref_occlu)) : bvec4(false);
-	outline = (!outline) ? (color[0] != ref_col) || occlu.x : true;
-	outline = (!outline) ? (color[1] != ref_col) || occlu.y : true;
-	outline = (!outline) ? (color[2] != ref_col) || occlu.z : true;
-	outline = (!outline) ? (color[3] != ref_col) || occlu.w : true;
+	uvec4 id;
+	uint ref_id = texelFetch(outlineId, texel, 0).r;
+#if 0 /* commented out until being tested */
+	id = textureGatherOffsets(outlineId, uv, ofs);
 #else
-	search_outline(uv + ivec2( 1,  0), ref_col, ref_occlu, outline);
-	search_outline(uv + ivec2( 0,  1), ref_col, ref_occlu, outline);
-	search_outline(uv + ivec2(-1,  0), ref_col, ref_occlu, outline);
-	search_outline(uv + ivec2( 0, -1), ref_col, ref_occlu, outline);
+	id.x = texelFetchOffset(outlineId, texel, 0, ofs[0]).r;
+	id.y = texelFetchOffset(outlineId, texel, 0, ofs[1]).r;
+	id.z = texelFetchOffset(outlineId, texel, 0, ofs[2]).r;
+	id.w = texelFetchOffset(outlineId, texel, 0, ofs[3]).r;
 #endif
 
-	FragColor = ref_col;
-	FragColor.a *= (outline) ? (ref_occlu) ? alphaOcclu : 1.0 : 0.0;
+	float ref_depth = texelFetch(outlineDepth, texel, 0).r;
+	float scene_depth = texelFetch(sceneDepth, texel, 0).r;
+
+	/* Avoid bad cases of zfighting for occlusion only. */
+	const float epsilon = 3.0 / 8388608.0;
+	bool occluded = (ref_depth > scene_depth + epsilon);
+	bool outline = any(notEqual(id, uvec4(ref_id)));
+
+	FragColor = convert_id_to_color(int(ref_id));
+	FragColor.a *= (occluded) ? alphaOcclu : 1.0;
+	FragColor.a = (outline) ? FragColor.a : 0.0;
 }
diff --git a/source/blender/draw/modes/shaders/object_outline_prepass_frag.glsl b/source/blender/draw/modes/shaders/object_outline_prepass_frag.glsl
new file mode 100644
index 00000000000..776adb787ad
--- /dev/null
+++ b/source/blender/draw/modes/shaders/object_outline_prepass_frag.glsl
@@ -0,0 +1,10 @@
+uniform int callId;
+uniform int baseId;
+
+/* using uint because 16bit uint can contain more ids than int. */
+out uint outId;
+
+void main()
+{
+	outId = uint(baseId + callId);
+}