PBVH Pixel extractor.

This patch contains an initial pixel extractor for PBVH and an initial paint brush implementation.
PBVH is an accelleration structure blender uses internally to speed up 3d painting operations.
At this moment it is extensively used by sculpt, vertex painting and weight painting.

For the 3d texturing brush we will be using the PBVH for texture painting.
Currently PBVH is organized to work on geometry (vertices, polygons and triangles).
For texture painting this should be extended it to use pixels.

{F12995467}

Screen recording has been done on a Mac Mini with a 6 core 3.3 GHZ Intel processor.

# Scope

This patch only contains an extending uv seams to fix uv seams. This is not actually we want, but was easy to add
to make the brush usable.

Pixels are places in the PBVH_Leaf nodes. We want to introduce a special node for pixels, but that will be done
in a separate patch to keep the code review small. This reduces the painting performance when using
low and medium poly assets.

In workbench textures aren't forced to be shown. For now use Material/Rendered view.

# Rasterization process

The rasterization process will generate the pixel information for a leaf node. In the future those
leaf nodes will be split up into multiple leaf nodes to increase the performance when there
isn't enough geometry. For this patch this was left out of scope.

In order to do so every polygon should be uniquely assigned to a leaf node.

For each leaf node
   for each polygon
     If polygon not assigned
       assign polygon to node.

Polygons are to complicated to be used directly we have to split the polygons into triangles.

For each leaf node
  for each polygon
    extract triangles from polygon.

The list of triangles can be stored inside the leaf node. The list of polygons aren't needed anymore.
Each triangle has:

    poly_index.
    vert_indices
    delta barycentric coordinate between x steps.

Each triangle is rasterized in rows. Sequential pixels (in uv space) are stored in a single structure.

    image position
    barycentric coordinate of the first pixel
    number of pixels
    triangle index inside the leaf node.

During the performed experiments we used a fairly simple rasterization process by
finding the UV bounds of an triangle and calculate the barycentric coordinates per
pixel inside the bounds. Even for complex models and huge images this process is
normally finished within 0.5 second. It could be that we want to change this algorithm
to reduce hickups when nodes are initialized during a stroke.

Reviewed By: brecht

Maniphest Tasks: T96710

Differential Revision: https://developer.blender.org/D14504

6 files changed, 579 insertions, 24 deletions

diff --git a/source/blender/editors/include/ED_sculpt.h b/source/blender/editors/include/ED_sculpt.h
index 3100c135db4..54d67c50d5c 100644
--- a/source/blender/editors/include/ED_sculpt.h
+++ b/source/blender/editors/include/ED_sculpt.h
@@ -17,6 +17,9 @@ struct UndoType;
 struct ViewContext;
 struct bContext;
 struct rcti;
+struct wmMsgSubscribeKey;
+struct wmMsgSubscribeValue;
+struct wmRegionMessageSubscribeParams;
 
 /* sculpt.c */
 
diff --git a/source/blender/editors/sculpt_paint/CMakeLists.txt b/source/blender/editors/sculpt_paint/CMakeLists.txt
index b89cbcf87fa..abdae5c44f9 100644
--- a/source/blender/editors/sculpt_paint/CMakeLists.txt
+++ b/source/blender/editors/sculpt_paint/CMakeLists.txt
@@ -72,6 +72,7 @@ set(SRC
   sculpt_multiplane_scrape.c
   sculpt_ops.c
   sculpt_paint_color.c
+  sculpt_paint_image.cc
   sculpt_pose.c
   sculpt_smooth.c
   sculpt_transform.c
diff --git a/source/blender/editors/sculpt_paint/sculpt.c b/source/blender/editors/sculpt_paint/sculpt.c
index 24c5a9fce52..bbbed32e316 100644
--- a/source/blender/editors/sculpt_paint/sculpt.c
+++ b/source/blender/editors/sculpt_paint/sculpt.c
@@ -2188,7 +2188,8 @@ void SCULPT_calc_area_normal_and_center(
 static float brush_strength(const Sculpt *sd,
                             const StrokeCache *cache,
                             const float feather,
-                            const UnifiedPaintSettings *ups)
+                            const UnifiedPaintSettings *ups,
+                            const PaintModeSettings *UNUSED(paint_mode_settings))
 {
   const Scene *scene = cache->vc->scene;
   const Brush *brush = BKE_paint_brush((Paint *)&sd->paint);
@@ -2750,6 +2751,41 @@ static void update_brush_local_mat(Sculpt *sd, Object *ob)
 /** \} */
 
 /* -------------------------------------------------------------------- */
+/** \name Texture painting
+ * \{ */
+
+static bool sculpt_needs_pbvh_pixels(PaintModeSettings *paint_mode_settings,
+                                     const Brush *brush,
+                                     Object *ob)
+{
+  if (brush->sculpt_tool == SCULPT_TOOL_PAINT && U.experimental.use_sculpt_texture_paint) {
+    Image *image;
+    ImageUser *image_user;
+    return SCULPT_paint_image_canvas_get(paint_mode_settings, ob, &image, &image_user);
+  }
+
+  return false;
+}
+
+static void sculpt_pbvh_update_pixels(PaintModeSettings *paint_mode_settings,
+                                      SculptSession *ss,
+                                      Object *ob)
+{
+  BLI_assert(ob->type == OB_MESH);
+  Mesh *mesh = (Mesh *)ob->data;
+
+  Image *image;
+  ImageUser *image_user;
+  if (!SCULPT_paint_image_canvas_get(paint_mode_settings, ob, &image, &image_user)) {
+    return;
+  }
+
+  BKE_pbvh_build_pixels(ss->pbvh, mesh->mloop, &mesh->ldata, image, image_user);
+}
+
+/** \} */
+
+/* -------------------------------------------------------------------- */
 /** \name Generic Brush Plane & Symmetry Utilities
  * \{ */
 
@@ -3075,7 +3111,8 @@ void SCULPT_vertcos_to_key(Object *ob, KeyBlock *kb, const float (*vertCos)[3])
 static void sculpt_topology_update(Sculpt *sd,
                                    Object *ob,
                                    Brush *brush,
-                                   UnifiedPaintSettings *UNUSED(ups))
+                                   UnifiedPaintSettings *UNUSED(ups),
+                                   PaintModeSettings *UNUSED(paint_mode_settings))
 {
   SculptSession *ss = ob->sculpt;
 
@@ -3170,7 +3207,11 @@ static void do_brush_action_task_cb(void *__restrict userdata,
   }
 }
 
-static void do_brush_action(Sculpt *sd, Object *ob, Brush *brush, UnifiedPaintSettings *ups)
+static void do_brush_action(Sculpt *sd,
+                            Object *ob,
+                            Brush *brush,
+                            UnifiedPaintSettings *ups,
+                            PaintModeSettings *paint_mode_settings)
 {
   SculptSession *ss = ob->sculpt;
   int totnode;
@@ -3209,6 +3250,10 @@ static void do_brush_action(Sculpt *sd, Object *ob, Brush *brush, UnifiedPaintSe
     nodes = sculpt_pbvh_gather_generic(ob, sd, brush, use_original, radius_scale, &totnode);
   }
 
+  if (sculpt_needs_pbvh_pixels(paint_mode_settings, brush, ob)) {
+    sculpt_pbvh_update_pixels(paint_mode_settings, ss, ob);
+  }
+
   /* Draw Face Sets in draw mode makes a single undo push, in alt-smooth mode deforms the
    * vertices and uses regular coords undo. */
   /* It also assigns the paint_face_set here as it needs to be done regardless of the stroke type
@@ -3399,7 +3444,7 @@ static void do_brush_action(Sculpt *sd, Object *ob, Brush *brush, UnifiedPaintSe
       SCULPT_do_displacement_smear_brush(sd, ob, nodes, totnode);
       break;
     case SCULPT_TOOL_PAINT:
-      SCULPT_do_paint_brush(sd, ob, nodes, totnode);
+      SCULPT_do_paint_brush(paint_mode_settings, sd, ob, nodes, totnode);
       break;
     case SCULPT_TOOL_SMEAR:
       SCULPT_do_smear_brush(sd, ob, nodes, totnode);
@@ -3704,10 +3749,18 @@ void SCULPT_cache_calc_brushdata_symm(StrokeCache *cache,
   }
 }
 
-typedef void (*BrushActionFunc)(Sculpt *sd, Object *ob, Brush *brush, UnifiedPaintSettings *ups);
+typedef void (*BrushActionFunc)(Sculpt *sd,
+                                Object *ob,
+                                Brush *brush,
+                                UnifiedPaintSettings *ups,
+                                PaintModeSettings *paint_mode_settings);
 
-static void do_tiled(
-    Sculpt *sd, Object *ob, Brush *brush, UnifiedPaintSettings *ups, BrushActionFunc action)
+static void do_tiled(Sculpt *sd,
+                     Object *ob,
+                     Brush *brush,
+                     UnifiedPaintSettings *ups,
+                     PaintModeSettings *paint_mode_settings,
+                     BrushActionFunc action)
 {
   SculptSession *ss = ob->sculpt;
   StrokeCache *cache = ss->cache;
@@ -3741,7 +3794,7 @@ static void do_tiled(
 
   /* First do the "un-tiled" position to initialize the stroke for this location. */
   cache->tile_pass = 0;
-  action(sd, ob, brush, ups);
+  action(sd, ob, brush, ups, paint_mode_settings);
 
   /* Now do it for all the tiles. */
   copy_v3_v3_int(cur, start);
@@ -3760,7 +3813,7 @@ static void do_tiled(
           cache->plane_offset[dim] = cur[dim] * step[dim];
           cache->initial_location[dim] = cur[dim] * step[dim] + original_initial_location[dim];
         }
-        action(sd, ob, brush, ups);
+        action(sd, ob, brush, ups, paint_mode_settings);
       }
     }
   }
@@ -3770,6 +3823,7 @@ static void do_radial_symmetry(Sculpt *sd,
                                Object *ob,
                                Brush *brush,
                                UnifiedPaintSettings *ups,
+                               PaintModeSettings *paint_mode_settings,
                                BrushActionFunc action,
                                const char symm,
                                const int axis,
@@ -3781,7 +3835,7 @@ static void do_radial_symmetry(Sculpt *sd,
     const float angle = 2.0f * M_PI * i / sd->radial_symm[axis - 'X'];
     ss->cache->radial_symmetry_pass = i;
     SCULPT_cache_calc_brushdata_symm(ss->cache, symm, axis, angle);
-    do_tiled(sd, ob, brush, ups, action);
+    do_tiled(sd, ob, brush, ups, paint_mode_settings, action);
   }
 }
 
@@ -3803,7 +3857,8 @@ static void sculpt_fix_noise_tear(Sculpt *sd, Object *ob)
 static void do_symmetrical_brush_actions(Sculpt *sd,
                                          Object *ob,
                                          BrushActionFunc action,
-                                         UnifiedPaintSettings *ups)
+                                         UnifiedPaintSettings *ups,
+                                         PaintModeSettings *paint_mode_settings)
 {
   Brush *brush = BKE_paint_brush(&sd->paint);
   SculptSession *ss = ob->sculpt;
@@ -3812,7 +3867,7 @@ static void do_symmetrical_brush_actions(Sculpt *sd,
 
   float feather = calc_symmetry_feather(sd, ss->cache);
 
-  cache->bstrength = brush_strength(sd, cache, feather, ups);
+  cache->bstrength = brush_strength(sd, cache, feather, ups, paint_mode_settings);
   cache->symmetry = symm;
 
   /* `symm` is a bit combination of XYZ -
@@ -3825,11 +3880,11 @@ static void do_symmetrical_brush_actions(Sculpt *sd,
     cache->radial_symmetry_pass = 0;
 
     SCULPT_cache_calc_brushdata_symm(cache, i, 0, 0);
-    do_tiled(sd, ob, brush, ups, action);
+    do_tiled(sd, ob, brush, ups, paint_mode_settings, action);
 
-    do_radial_symmetry(sd, ob, brush, ups, action, i, 'X', feather);
-    do_radial_symmetry(sd, ob, brush, ups, action, i, 'Y', feather);
-    do_radial_symmetry(sd, ob, brush, ups, action, i, 'Z', feather);
+    do_radial_symmetry(sd, ob, brush, ups, paint_mode_settings, action, i, 'X', feather);
+    do_radial_symmetry(sd, ob, brush, ups, paint_mode_settings, action, i, 'Y', feather);
+    do_radial_symmetry(sd, ob, brush, ups, paint_mode_settings, action, i, 'Z', feather);
   }
 }
 
@@ -4609,7 +4664,8 @@ static bool sculpt_needs_connectivity_info(const Sculpt *sd,
           SCULPT_TOOL_NEEDS_COLOR(brush->sculpt_tool) ||
           (brush->sculpt_tool == SCULPT_TOOL_CLOTH) || (brush->sculpt_tool == SCULPT_TOOL_SMEAR) ||
           (brush->sculpt_tool == SCULPT_TOOL_DRAW_FACE_SETS) ||
-          (brush->sculpt_tool == SCULPT_TOOL_DISPLACEMENT_SMEAR));
+          (brush->sculpt_tool == SCULPT_TOOL_DISPLACEMENT_SMEAR) ||
+          (brush->sculpt_tool == SCULPT_TOOL_PAINT));
 }
 
 void SCULPT_stroke_modifiers_check(const bContext *C, Object *ob, const Brush *brush)
@@ -5057,6 +5113,15 @@ void SCULPT_flush_update_step(bContext *C, SculptUpdateType update_flags)
     multires_mark_as_modified(depsgraph, ob, MULTIRES_COORDS_MODIFIED);
   }
 
+  if ((update_flags & SCULPT_UPDATE_IMAGE) != 0) {
+    ED_region_tag_redraw(region);
+    if (update_flags == SCULPT_UPDATE_IMAGE) {
+      /* Early exit when only need to update the images. We don't want to tag any geometry updates
+       * that would rebuilt the PBVH. */
+      return;
+    }
+  }
+
   DEG_id_tag_update(&ob->id, ID_RECALC_SHADING);
 
   /* Only current viewport matters, slower update for all viewports will
@@ -5136,6 +5201,16 @@ void SCULPT_flush_update_done(const bContext *C, Object *ob, SculptUpdateType up
         }
       }
     }
+
+    if (update_flags & SCULPT_UPDATE_IMAGE) {
+      LISTBASE_FOREACH (ScrArea *, area, &screen->areabase) {
+        SpaceLink *sl = area->spacedata.first;
+        if (sl->spacetype != SPACE_IMAGE) {
+          continue;
+        }
+        ED_area_tag_redraw_regiontype(area, RGN_TYPE_WINDOW);
+      }
+    }
   }
 
   if (update_flags & SCULPT_UPDATE_COORDS) {
@@ -5227,6 +5302,7 @@ static void sculpt_stroke_update_step(bContext *C,
   Object *ob = CTX_data_active_object(C);
   SculptSession *ss = ob->sculpt;
   const Brush *brush = BKE_paint_brush(&sd->paint);
+  ToolSettings *tool_settings = CTX_data_tool_settings(C);
 
   SCULPT_stroke_modifiers_check(C, ob, brush);
   sculpt_update_cache_variants(C, sd, ob, itemptr);
@@ -5246,10 +5322,10 @@ static void sculpt_stroke_update_step(bContext *C,
   }
 
   if (SCULPT_stroke_is_dynamic_topology(ss, brush)) {
-    do_symmetrical_brush_actions(sd, ob, sculpt_topology_update, ups);
+    do_symmetrical_brush_actions(sd, ob, sculpt_topology_update, ups, &tool_settings->paint_mode);
   }
 
-  do_symmetrical_brush_actions(sd, ob, do_brush_action, ups);
+  do_symmetrical_brush_actions(sd, ob, do_brush_action, ups, &tool_settings->paint_mode);
   sculpt_combine_proxies(sd, ob);
 
   /* Hack to fix noise texture tearing mesh. */
@@ -5280,7 +5356,12 @@ static void sculpt_stroke_update_step(bContext *C,
     SCULPT_flush_update_step(C, SCULPT_UPDATE_MASK);
   }
   else if (ELEM(brush->sculpt_tool, SCULPT_TOOL_PAINT, SCULPT_TOOL_SMEAR)) {
-    SCULPT_flush_update_step(C, SCULPT_UPDATE_COLOR);
+    if (SCULPT_use_image_paint_brush(&tool_settings->paint_mode, ob)) {
+      SCULPT_flush_update_step(C, SCULPT_UPDATE_IMAGE);
+    }
+    else {
+      SCULPT_flush_update_step(C, SCULPT_UPDATE_COLOR);
+    }
   }
   else {
     SCULPT_flush_update_step(C, SCULPT_UPDATE_COORDS);
@@ -5302,6 +5383,7 @@ static void sculpt_stroke_done(const bContext *C, struct PaintStroke *UNUSED(str
   Object *ob = CTX_data_active_object(C);
   SculptSession *ss = ob->sculpt;
   Sculpt *sd = CTX_data_tool_settings(C)->sculpt;
+  ToolSettings *tool_settings = CTX_data_tool_settings(C);
 
   /* Finished. */
   if (!ss->cache) {
@@ -5335,6 +5417,11 @@ static void sculpt_stroke_done(const bContext *C, struct PaintStroke *UNUSED(str
   if (brush->sculpt_tool == SCULPT_TOOL_MASK) {
     SCULPT_flush_update_done(C, ob, SCULPT_UPDATE_MASK);
   }
+  else if (brush->sculpt_tool == SCULPT_TOOL_PAINT) {
+    if (SCULPT_use_image_paint_brush(&tool_settings->paint_mode, ob)) {
+      SCULPT_flush_update_done(C, ob, SCULPT_UPDATE_IMAGE);
+    }
+  }
   else {
     SCULPT_flush_update_done(C, ob, SCULPT_UPDATE_COORDS);
   }
diff --git a/source/blender/editors/sculpt_paint/sculpt_intern.h b/source/blender/editors/sculpt_paint/sculpt_intern.h
index 466432a35ec..3839c0e71e4 100644
--- a/source/blender/editors/sculpt_paint/sculpt_intern.h
+++ b/source/blender/editors/sculpt_paint/sculpt_intern.h
@@ -11,11 +11,13 @@
 #include "DNA_key_types.h"
 #include "DNA_listBase.h"
 #include "DNA_meshdata_types.h"
+#include "DNA_scene_types.h"
 #include "DNA_vec_types.h"
 
 #include "BKE_paint.h"
 #include "BKE_pbvh.h"
 #include "BLI_bitmap.h"
+#include "BLI_compiler_attrs.h"
 #include "BLI_compiler_compat.h"
 #include "BLI_gsqueue.h"
 #include "BLI_threads.h"
@@ -25,12 +27,13 @@ extern "C" {
 #endif
 
 struct AutomaskingCache;
+struct Image;
+struct ImageUser;
 struct KeyBlock;
 struct Object;
 struct SculptUndoNode;
 struct bContext;
-
-enum ePaintSymmetryFlags;
+struct PaintModeSettings;
 
 /* Updates */
 
@@ -43,6 +46,7 @@ typedef enum SculptUpdateType {
   SCULPT_UPDATE_MASK = 1 << 1,
   SCULPT_UPDATE_VISIBILITY = 1 << 2,
   SCULPT_UPDATE_COLOR = 1 << 3,
+  SCULPT_UPDATE_IMAGE = 1 << 4,
 } SculptUpdateType;
 
 typedef struct SculptCursorGeometryInfo {
@@ -1626,7 +1630,29 @@ void SCULPT_multiplane_scrape_preview_draw(uint gpuattr,
 void SCULPT_do_draw_face_sets_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode);
 
 /* Paint Brush. */
-void SCULPT_do_paint_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode);
+void SCULPT_do_paint_brush(struct PaintModeSettings *paint_mode_settings,
+                           Sculpt *sd,
+                           Object *ob,
+                           PBVHNode **nodes,
+                           int totnode) ATTR_NONNULL();
+
+/**
+ * @brief Get the image canvas for painting on the given object.
+ *
+ * @return #true if an image is found. The #r_image and #r_image_user fields are filled with the
+ * image and image user. Returns false when the image isn't found. In the later case the r_image
+ * and r_image_user are set to nullptr/NULL.
+ */
+bool SCULPT_paint_image_canvas_get(struct PaintModeSettings *paint_mode_settings,
+                                   struct Object *ob,
+                                   struct Image **r_image,
+                                   struct ImageUser **r_image_user) ATTR_NONNULL();
+void SCULPT_do_paint_brush_image(struct PaintModeSettings *paint_mode_settings,
+                                 Sculpt *sd,
+                                 Object *ob,
+                                 PBVHNode **nodes,
+                                 int totnode) ATTR_NONNULL();
+bool SCULPT_use_image_paint_brush(struct PaintModeSettings *settings, Object *ob) ATTR_NONNULL();
 
 /* Smear Brush. */
 void SCULPT_do_smear_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode);
diff --git a/source/blender/editors/sculpt_paint/sculpt_paint_color.c b/source/blender/editors/sculpt_paint/sculpt_paint_color.c
index e7a713efa14..7a8a6e8e484 100644
--- a/source/blender/editors/sculpt_paint/sculpt_paint_color.c
+++ b/source/blender/editors/sculpt_paint/sculpt_paint_color.c
@@ -240,8 +240,14 @@ static void sample_wet_paint_reduce(const void *__restrict UNUSED(userdata),
   add_v4_v4(join->color, swptd->color);
 }
 
-void SCULPT_do_paint_brush(Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
+void SCULPT_do_paint_brush(
+    PaintModeSettings *paint_mode_settings, Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
 {
+  if (SCULPT_use_image_paint_brush(paint_mode_settings, ob)) {
+    SCULPT_do_paint_brush_image(paint_mode_settings, sd, ob, nodes, totnode);
+    return;
+  }
+
   Brush *brush = BKE_paint_brush(&sd->paint);
   SculptSession *ss = ob->sculpt;
 
diff --git a/source/blender/editors/sculpt_paint/sculpt_paint_image.cc b/source/blender/editors/sculpt_paint/sculpt_paint_image.cc
new file mode 100644
index 00000000000..082ff6260c2
--- /dev/null
+++ b/source/blender/editors/sculpt_paint/sculpt_paint_image.cc
@@ -0,0 +1,432 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later
+ * Copyright 2022 Blender Foundation. All rights reserved. */
+
+#include "DNA_image_types.h"
+#include "DNA_material_types.h"
+#include "DNA_mesh_types.h"
+#include "DNA_node_types.h"
+#include "DNA_object_types.h"
+
+#include "ED_paint.h"
+#include "ED_uvedit.h"
+
+#include "BLI_math.h"
+#include "BLI_math_color_blend.h"
+#include "BLI_task.h"
+
+#include "IMB_colormanagement.h"
+#include "IMB_imbuf.h"
+
+#include "BKE_brush.h"
+#include "BKE_image_wrappers.hh"
+#include "BKE_material.h"
+#include "BKE_pbvh.h"
+#include "BKE_pbvh_pixels.hh"
+
+#include "bmesh.h"
+
+#include "NOD_shader.h"
+
+#include "sculpt_intern.h"
+
+namespace blender::ed::sculpt_paint::paint::image {
+
+using namespace blender::bke::pbvh::pixels;
+using namespace blender::bke::image;
+
+struct ImageData {
+  Image *image = nullptr;
+  ImageUser *image_user = nullptr;
+
+  ~ImageData()
+  {
+  }
+
+  static bool init_active_image(Object *ob,
+                                ImageData *r_image_data,
+                                PaintModeSettings *paint_mode_settings)
+  {
+    return BKE_paint_canvas_image_get(
+        paint_mode_settings, ob, &r_image_data->image, &r_image_data->image_user);
+  }
+};
+
+struct TexturePaintingUserData {
+  Object *ob;
+  Brush *brush;
+  PBVHNode **nodes;
+  ImageData image_data;
+};
+
+/** Reading and writing to image buffer with 4 float channels. */
+class ImageBufferFloat4 {
+ private:
+  int pixel_offset;
+
+ public:
+  void set_image_position(ImBuf *image_buffer, ushort2 image_pixel_position)
+  {
+    pixel_offset = int(image_pixel_position.y) * image_buffer->x + int(image_pixel_position.x);
+  }
+
+  void next_pixel()
+  {
+    pixel_offset += 1;
+  }
+
+  float4 read_pixel(ImBuf *image_buffer) const
+  {
+    return &image_buffer->rect_float[pixel_offset * 4];
+  }
+
+  void write_pixel(ImBuf *image_buffer, const float4 pixel_data) const
+  {
+    copy_v4_v4(&image_buffer->rect_float[pixel_offset * 4], pixel_data);
+  }
+
+  const char *get_colorspace_name(ImBuf *image_buffer)
+  {
+    return IMB_colormanagement_get_float_colorspace(image_buffer);
+  }
+};
+
+/** Reading and writing to image buffer with 4 byte channels. */
+class ImageBufferByte4 {
+ private:
+  int pixel_offset;
+
+ public:
+  void set_image_position(ImBuf *image_buffer, ushort2 image_pixel_position)
+  {
+    pixel_offset = int(image_pixel_position.y) * image_buffer->x + int(image_pixel_position.x);
+  }
+
+  void next_pixel()
+  {
+    pixel_offset += 1;
+  }
+
+  float4 read_pixel(ImBuf *image_buffer) const
+  {
+    float4 result;
+    rgba_uchar_to_float(result,
+                        static_cast<const uchar *>(
+                            static_cast<const void *>(&(image_buffer->rect[pixel_offset]))));
+    return result;
+  }
+
+  void write_pixel(ImBuf *image_buffer, const float4 pixel_data) const
+  {
+    rgba_float_to_uchar(
+        static_cast<uchar *>(static_cast<void *>(&image_buffer->rect[pixel_offset])), pixel_data);
+  }
+
+  const char *get_colorspace_name(ImBuf *image_buffer)
+  {
+    return IMB_colormanagement_get_rect_colorspace(image_buffer);
+  }
+};
+
+template<typename ImageBuffer> class PaintingKernel {
+  ImageBuffer image_accessor;
+
+  SculptSession *ss;
+  const Brush *brush;
+  const int thread_id;
+  const MVert *mvert;
+
+  float4 brush_color;
+  float brush_strength;
+
+  SculptBrushTestFn brush_test_fn;
+  SculptBrushTest test;
+  /* Pointer to the last used image buffer to detect when buffers are switched. */
+  void *last_used_image_buffer_ptr = nullptr;
+  const char *last_used_color_space = nullptr;
+
+ public:
+  explicit PaintingKernel(SculptSession *ss,
+                          const Brush *brush,
+                          const int thread_id,
+                          const MVert *mvert)
+      : ss(ss), brush(brush), thread_id(thread_id), mvert(mvert)
+  {
+    init_brush_strength();
+    init_brush_test();
+  }
+
+  bool paint(const Triangles &triangles, const PackedPixelRow &pixel_row, ImBuf *image_buffer)
+  {
+    image_accessor.set_image_position(image_buffer, pixel_row.start_image_coordinate);
+    const TrianglePaintInput triangle = triangles.get_paint_input(pixel_row.triangle_index);
+    float3 pixel_pos = get_start_pixel_pos(triangle, pixel_row);
+    const float3 delta_pixel_pos = get_delta_pixel_pos(triangle, pixel_row, pixel_pos);
+    bool pixels_painted = false;
+    for (int x = 0; x < pixel_row.num_pixels; x++) {
+      if (!brush_test_fn(&test, pixel_pos)) {
+        pixel_pos += delta_pixel_pos;
+        image_accessor.next_pixel();
+        continue;
+      }
+
+      float4 color = image_accessor.read_pixel(image_buffer);
+      const float3 normal(0.0f, 0.0f, 0.0f);
+      const float3 face_normal(0.0f, 0.0f, 0.0f);
+      const float mask = 0.0f;
+      const float falloff_strength = SCULPT_brush_strength_factor(
+          ss, brush, pixel_pos, sqrtf(test.dist), normal, face_normal, mask, 0, thread_id);
+      float4 paint_color = brush_color * falloff_strength * brush_strength;
+      float4 buffer_color;
+      blend_color_mix_float(buffer_color, color, paint_color);
+      buffer_color *= brush->alpha;
+      IMB_blend_color_float(color, color, buffer_color, static_cast<IMB_BlendMode>(brush->blend));
+      image_accessor.write_pixel(image_buffer, color);
+      pixels_painted = true;
+
+      image_accessor.next_pixel();
+      pixel_pos += delta_pixel_pos;
+    }
+    return pixels_painted;
+  }
+
+  void init_brush_color(ImBuf *image_buffer)
+  {
+    const char *to_colorspace = image_accessor.get_colorspace_name(image_buffer);
+    if (last_used_color_space == to_colorspace) {
+      return;
+    }
+    copy_v3_v3(brush_color,
+               ss->cache->invert ? BKE_brush_secondary_color_get(ss->scene, brush) :
+                                   BKE_brush_color_get(ss->scene, brush));
+    /* NOTE: Brush colors are stored in sRGB. We use math color to follow other areas that
+     * use brush colors. From there on we use IMB_colormanagement to convert the brush color to the
+     * colorspace of the texture. This isn't ideal, but would need more refactoring to make sure
+     * that brush colors are stored in scene linear by default. */
+    srgb_to_linearrgb_v3_v3(brush_color, brush_color);
+    brush_color[3] = 1.0f;
+
+    const char *from_colorspace = IMB_colormanagement_role_colorspace_name_get(
+        COLOR_ROLE_SCENE_LINEAR);
+    ColormanageProcessor *cm_processor = IMB_colormanagement_colorspace_processor_new(
+        from_colorspace, to_colorspace);
+    IMB_colormanagement_processor_apply_v4(cm_processor, brush_color);
+    IMB_colormanagement_processor_free(cm_processor);
+    last_used_color_space = to_colorspace;
+  }
+
+ private:
+  void init_brush_strength()
+  {
+    brush_strength = ss->cache->bstrength;
+  }
+  void init_brush_test()
+  {
+    brush_test_fn = SCULPT_brush_test_init_with_falloff_shape(ss, &test, brush->falloff_shape);
+  }
+
+  /**
+   * Extract the starting pixel position from the given encoded_pixels belonging to the triangle.
+   */
+  float3 get_start_pixel_pos(const TrianglePaintInput &triangle,
+                             const PackedPixelRow &encoded_pixels) const
+  {
+    return init_pixel_pos(triangle, encoded_pixels.start_barycentric_coord);
+  }
+
+  /**
+   * Extract the delta pixel position that will be used to advance a Pixel instance to the next
+   * pixel.
+   */
+  float3 get_delta_pixel_pos(const TrianglePaintInput &triangle,
+                             const PackedPixelRow &encoded_pixels,
+                             const float3 &start_pixel) const
+  {
+    float3 result = init_pixel_pos(
+        triangle, encoded_pixels.start_barycentric_coord + triangle.delta_barycentric_coord_u);
+    return result - start_pixel;
+  }
+
+  float3 init_pixel_pos(const TrianglePaintInput &triangle,
+                        const float2 &barycentric_weights) const
+  {
+    const int3 &vert_indices = triangle.vert_indices;
+    float3 result;
+    const float3 barycentric(barycentric_weights.x,
+                             barycentric_weights.y,
+                             1.0f - barycentric_weights.x - barycentric_weights.y);
+    interp_v3_v3v3v3(result,
+                     mvert[vert_indices[0]].co,
+                     mvert[vert_indices[1]].co,
+                     mvert[vert_indices[2]].co,
+                     barycentric);
+    return result;
+  }
+};
+
+static std::vector<bool> init_triangle_brush_test(SculptSession *ss,
+                                                  Triangles &triangles,
+                                                  const MVert *mvert)
+{
+  std::vector<bool> brush_test(triangles.size());
+  SculptBrushTest test;
+  SCULPT_brush_test_init(ss, &test);
+  float3 brush_min_bounds(test.location[0] - test.radius,
+                          test.location[1] - test.radius,
+                          test.location[2] - test.radius);
+  float3 brush_max_bounds(test.location[0] + test.radius,
+                          test.location[1] + test.radius,
+                          test.location[2] + test.radius);
+  for (int triangle_index = 0; triangle_index < triangles.size(); triangle_index++) {
+    TrianglePaintInput &triangle = triangles.get_paint_input(triangle_index);
+
+    float3 triangle_min_bounds(mvert[triangle.vert_indices[0]].co);
+    float3 triangle_max_bounds(triangle_min_bounds);
+    for (int i = 1; i < 3; i++) {
+      const float3 &pos = mvert[triangle.vert_indices[i]].co;
+      triangle_min_bounds.x = min_ff(triangle_min_bounds.x, pos.x);
+      triangle_min_bounds.y = min_ff(triangle_min_bounds.y, pos.y);
+      triangle_min_bounds.z = min_ff(triangle_min_bounds.z, pos.z);
+      triangle_max_bounds.x = max_ff(triangle_max_bounds.x, pos.x);
+      triangle_max_bounds.y = max_ff(triangle_max_bounds.y, pos.y);
+      triangle_max_bounds.z = max_ff(triangle_max_bounds.z, pos.z);
+    }
+    brush_test[triangle_index] = isect_aabb_aabb_v3(
+        brush_min_bounds, brush_max_bounds, triangle_min_bounds, triangle_max_bounds);
+  }
+  return brush_test;
+}
+
+static void do_paint_pixels(void *__restrict userdata,
+                            const int n,
+                            const TaskParallelTLS *__restrict tls)
+{
+  TexturePaintingUserData *data = static_cast<TexturePaintingUserData *>(userdata);
+  Object *ob = data->ob;
+  SculptSession *ss = ob->sculpt;
+  const Brush *brush = data->brush;
+  PBVHNode *node = data->nodes[n];
+
+  NodeData &node_data = BKE_pbvh_pixels_node_data_get(*node);
+  const int thread_id = BLI_task_parallel_thread_id(tls);
+  MVert *mvert = SCULPT_mesh_deformed_mverts_get(ss);
+
+  std::vector<bool> brush_test = init_triangle_brush_test(ss, node_data.triangles, mvert);
+
+  PaintingKernel<ImageBufferFloat4> kernel_float4(ss, brush, thread_id, mvert);
+  PaintingKernel<ImageBufferByte4> kernel_byte4(ss, brush, thread_id, mvert);
+
+  ImageUser image_user = *data->image_data.image_user;
+  bool pixels_updated = false;
+  for (UDIMTilePixels &tile_data : node_data.tiles) {
+    LISTBASE_FOREACH (ImageTile *, tile, &data->image_data.image->tiles) {
+      ImageTileWrapper image_tile(tile);
+      if (image_tile.get_tile_number() == tile_data.tile_number) {
+        image_user.tile = image_tile.get_tile_number();
+
+        ImBuf *image_buffer = BKE_image_acquire_ibuf(data->image_data.image, &image_user, nullptr);
+        if (image_buffer == nullptr) {
+          continue;
+        }
+
+        if (image_buffer->rect_float != nullptr) {
+          kernel_float4.init_brush_color(image_buffer);
+        }
+        else {
+          kernel_float4.init_brush_color(image_buffer);
+        }
+
+        for (const PackedPixelRow &pixel_row : tile_data.pixel_rows) {
+          if (!brush_test[pixel_row.triangle_index]) {
+            continue;
+          }
+          bool pixels_painted = false;
+          if (image_buffer->rect_float != nullptr) {
+            pixels_painted = kernel_float4.paint(node_data.triangles, pixel_row, image_buffer);
+          }
+          else {
+            pixels_painted = kernel_byte4.paint(node_data.triangles, pixel_row, image_buffer);
+          }
+
+          if (pixels_painted) {
+            tile_data.mark_dirty(pixel_row);
+          }
+        }
+
+        BKE_image_release_ibuf(data->image_data.image, image_buffer, nullptr);
+        pixels_updated |= tile_data.flags.dirty;
+        break;
+      }
+    }
+  }
+
+  node_data.flags.dirty |= pixels_updated;
+}
+
+static void do_mark_dirty_regions(void *__restrict userdata,
+                                  const int n,
+                                  const TaskParallelTLS *__restrict UNUSED(tls))
+{
+  TexturePaintingUserData *data = static_cast<TexturePaintingUserData *>(userdata);
+  PBVHNode *node = data->nodes[n];
+  BKE_pbvh_pixels_mark_image_dirty(*node, *data->image_data.image, *data->image_data.image_user);
+}
+
+}  // namespace blender::ed::sculpt_paint::paint::image
+
+extern "C" {
+
+using namespace blender::ed::sculpt_paint::paint::image;
+
+bool SCULPT_paint_image_canvas_get(PaintModeSettings *paint_mode_settings,
+                                   Object *ob,
+                                   Image **r_image,
+                                   ImageUser **r_image_user)
+{
+  BLI_assert(r_image);
+  BLI_assert(r_image_user);
+  ImageData image_data;
+  if (!ImageData::init_active_image(ob, &image_data, paint_mode_settings)) {
+    return false;
+  }
+
+  *r_image = image_data.image;
+  *r_image_user = image_data.image_user;
+  return true;
+}
+
+bool SCULPT_use_image_paint_brush(PaintModeSettings *settings, Object *ob)
+{
+  if (!U.experimental.use_sculpt_texture_paint) {
+    return false;
+  }
+  if (ob->type != OB_MESH) {
+    return false;
+  }
+  Image *image;
+  ImageUser *image_user;
+  return BKE_paint_canvas_image_get(settings, ob, &image, &image_user);
+}
+
+void SCULPT_do_paint_brush_image(
+    PaintModeSettings *paint_mode_settings, Sculpt *sd, Object *ob, PBVHNode **nodes, int totnode)
+{
+  Brush *brush = BKE_paint_brush(&sd->paint);
+
+  TexturePaintingUserData data = {nullptr};
+  data.ob = ob;
+  data.brush = brush;
+  data.nodes = nodes;
+
+  if (!ImageData::init_active_image(ob, &data.image_data, paint_mode_settings)) {
+    return;
+  }
+
+  TaskParallelSettings settings;
+  BKE_pbvh_parallel_range_settings(&settings, true, totnode);
+  BLI_task_parallel_range(0, totnode, &data, do_paint_pixels, &settings);
+
+  TaskParallelSettings settings_flush;
+  BKE_pbvh_parallel_range_settings(&settings_flush, false, totnode);
+  BLI_task_parallel_range(0, totnode, &data, do_mark_dirty_regions, &settings_flush);
+}
+}