GP: Refactor drawing engine to single VBO

This is part of T57829.

Reduce the number of batches used to only one by shader type.  This reduces GPU overhead and increase a lot the FPS. As the number of batches is small, the time to allocate and free memory was reduced in 90% or more.

Also the code has been simplified and all batch management has been removed because this is not necessary. Now, all shading groups are created after all vertex buffer data for all strokes has been created using DRW_shgroup_call_range_add().

All batch cache data has been moved to the Object runtime struct and not as before where some parts (derived data) were saved inside GPD datablock.

For particles, now the code is faster and cleaner and gets better FPS.

Thanks to Clément Foucault for his help and advices to improve speed.

1 files changed, 93 insertions, 153 deletions

diff --git a/source/blender/draw/engines/gpencil/gpencil_cache_utils.c b/source/blender/draw/engines/gpencil/gpencil_cache_utils.c
index 30887e3fb19..15ac3f37add 100644
--- a/source/blender/draw/engines/gpencil/gpencil_cache_utils.c
+++ b/source/blender/draw/engines/gpencil/gpencil_cache_utils.c
@@ -26,14 +26,14 @@
 #include "DRW_engine.h"
 #include "DRW_render.h"
 
-#include "BKE_global.h"
-
 #include "ED_gpencil.h"
 #include "ED_view3d.h"
 
 #include "DNA_gpencil_types.h"
 #include "DNA_view3d_types.h"
 
+#include "BKE_gpencil.h"
+
 #include "gpencil_engine.h"
 
 #include "draw_cache_impl.h"
@@ -41,63 +41,6 @@
 #include "DEG_depsgraph.h"
 #include "DEG_depsgraph_query.h"
 
-static bool gpencil_check_ob_duplicated(tGPencilObjectCache *cache_array,
-	int gp_cache_used, Object *ob, int *r_index)
-{
-	if (gp_cache_used == 0) {
-		return false;
-	}
-
-	for (int i = 0; i < gp_cache_used; i++) {
-		tGPencilObjectCache *cache_elem = &cache_array[i];
-		if (cache_elem->ob == ob) {
-			*r_index = cache_elem->data_idx;
-			return true;
-		}
-	}
-	return false;
-}
-
-static bool gpencil_check_datablock_duplicated(
-        tGPencilObjectCache *cache_array, int gp_cache_used,
-        Object *ob, bGPdata *gpd)
-{
-	if (gp_cache_used == 0) {
-		return false;
-	}
-
-	for (int i = 0; i < gp_cache_used; i++) {
-		tGPencilObjectCache *cache_elem = &cache_array[i];
-		if ((cache_elem->ob != ob) &&
-		    (cache_elem->gpd == gpd))
-		{
-			return true;
-		}
-	}
-	return false;
-}
-
-static int gpencil_len_datablock_duplicated(
-	tGPencilObjectCache *cache_array, int gp_cache_used,
-	Object *ob, bGPdata *gpd)
-{
-	int tot = 0;
-	if (gp_cache_used == 0) {
-		return 0;
-	}
-
-	for (int i = 0; i < gp_cache_used; i++) {
-		tGPencilObjectCache *cache_elem = &cache_array[i];
-		if ((cache_elem->ob != ob) &&
-		    (cache_elem->gpd == gpd) &&
-		    (!cache_elem->is_dup_ob))
-		{
-			tot++;
-		}
-	}
-	return tot;
-}
-
  /* add a gpencil object to cache to defer drawing */
 tGPencilObjectCache *gpencil_object_cache_add(
         tGPencilObjectCache *cache_array, Object *ob,
@@ -133,38 +76,15 @@ tGPencilObjectCache *gpencil_object_cache_add(
 	copy_m4_m4(cache_elem->obmat, ob->obmat);
 	cache_elem->idx = *gp_cache_used;
 
-	/* check if object is duplicated */
-	cache_elem->is_dup_ob = gpencil_check_ob_duplicated(
-	        cache_array,
-	        *gp_cache_used, ob_orig,
-	        &cache_elem->data_idx);
-
-	if (!cache_elem->is_dup_ob) {
-		/* check if object reuse datablock */
-		cache_elem->is_dup_data = gpencil_check_datablock_duplicated(
-			cache_array, *gp_cache_used,
-			ob_orig, cache_elem->gpd);
-		if (cache_elem->is_dup_data) {
-			cache_elem->data_idx = gpencil_len_datablock_duplicated(
-				cache_array, *gp_cache_used,
-				ob_orig, cache_elem->gpd);
-
-			cache_elem->gpd->flag |= GP_DATA_CACHE_IS_DIRTY;
-		}
-		else {
-			cache_elem->data_idx = 0;
-		}
-	}
-	else {
-		cache_elem->is_dup_data = false;
-	}
+	/* object is duplicated (particle) */
+	cache_elem->is_dup_ob = ob->base_flag & BASE_FROMDUPLI;
 
 	/* save FXs */
 	cache_elem->pixfactor = cache_elem->gpd->pixfactor;
 	cache_elem->shader_fx = ob_orig->shader_fx;
 
-	cache_elem->init_grp = 0;
-	cache_elem->end_grp = -1;
+	cache_elem->init_grp = NULL;
+	cache_elem->end_grp = NULL;
 
 	/* calculate zdepth from point of view */
 	float zdepth = 0.0;
@@ -198,6 +118,48 @@ tGPencilObjectCache *gpencil_object_cache_add(
 	return cache_array;
 }
 
+/* add a shading group to the cache to create later */
+GpencilBatchGroup *gpencil_group_cache_add(
+	GpencilBatchGroup *cache_array,
+	bGPDlayer *gpl, bGPDframe *gpf, bGPDstroke *gps,
+	const short type, const bool onion,
+	const int vertex_idx,
+	int *grp_size, int *grp_used)
+{
+	GpencilBatchGroup *cache_elem = NULL;
+	GpencilBatchGroup *p = NULL;
+
+	/* By default a cache is created with one block with a predefined number of free slots,
+	if the size is not enough, the cache is reallocated adding a new block of free slots.
+	This is done in order to keep cache small */
+	if (*grp_used + 1 > *grp_size) {
+		if ((*grp_size == 0) || (cache_array == NULL)) {
+			p = MEM_callocN(sizeof(struct GpencilBatchGroup) * GPENCIL_GROUPS_BLOCK_SIZE, "GpencilBatchGroup");
+			*grp_size = GPENCIL_GROUPS_BLOCK_SIZE;
+		}
+		else {
+			*grp_size += GPENCIL_GROUPS_BLOCK_SIZE;
+			p = MEM_recallocN(cache_array, sizeof(struct GpencilBatchGroup) * *grp_size);
+		}
+		cache_array = p;
+	}
+	/* zero out all data */
+	cache_elem = &cache_array[*grp_used];
+	memset(cache_elem, 0, sizeof(*cache_elem));
+
+	cache_elem->gpl = gpl;
+	cache_elem->gpf = gpf;
+	cache_elem->gps = gps;
+	cache_elem->type = type;
+	cache_elem->onion = onion;
+	cache_elem->vertex_idx = vertex_idx;
+
+	/* increase slots used in cache */
+	(*grp_used)++;
+
+	return cache_array;
+}
+
 /* get current cache data */
 static GpencilBatchCache *gpencil_batch_get_element(Object *ob)
 {
@@ -208,51 +170,33 @@ static GpencilBatchCache *gpencil_batch_get_element(Object *ob)
 /* verify if cache is valid */
 static bool gpencil_batch_cache_valid(GpencilBatchCache *cache, bGPdata *gpd, int cfra)
 {
+	bool valid = true;
 	if (cache == NULL) {
 		return false;
 	}
 
 	cache->is_editmode = GPENCIL_ANY_EDIT_MODE(gpd);
-
 	if (cfra != cache->cache_frame) {
-		return false;
+		valid = false;
 	}
-
-	if (gpd->flag & GP_DATA_CACHE_IS_DIRTY) {
-		return false;
+	else if (gpd->flag & GP_DATA_CACHE_IS_DIRTY) {
+		valid = false;
 	}
-
-	if (cache->is_editmode) {
-		return false;
+	else if (gpd->flag & GP_DATA_SHOW_ONIONSKINS) {
+		/* if onion, set as dirty always
+		 * This reduces performance, but avoid any crash in the multiple
+		 * overlay and multiwindow options and keep all windows working
+		 */
+		valid = false;
 	}
-
-	if (cache->is_dirty) {
-		return false;
+	else if (cache->is_editmode) {
+		valid = false;
 	}
-
-	return true;
-}
-
-/* resize the cache to the number of slots */
-static void gpencil_batch_cache_resize(GpencilBatchCache *cache, int slots)
-{
-	cache->cache_size = slots;
-	cache->batch_stroke = MEM_recallocN(cache->batch_stroke, sizeof(struct Gwn_Batch *) * slots);
-	cache->batch_fill = MEM_recallocN(cache->batch_fill, sizeof(struct Gwn_Batch *) * slots);
-	cache->batch_edit = MEM_recallocN(cache->batch_edit, sizeof(struct Gwn_Batch *) * slots);
-	cache->batch_edlin = MEM_recallocN(cache->batch_edlin, sizeof(struct Gwn_Batch *) * slots);
-}
-
-/* check size and increase if no free slots */
-void gpencil_batch_cache_check_free_slots(Object *ob)
-{
-	GpencilBatchCache *cache = gpencil_batch_get_element(ob);
-
-	/* the memory is reallocated by chunks, not for one slot only to improve speed */
-	if (cache->cache_idx >= cache->cache_size) {
-		cache->cache_size += GPENCIL_MIN_BATCH_SLOTS_CHUNK;
-		gpencil_batch_cache_resize(cache, cache->cache_size);
+	else if (cache->is_dirty) {
+		valid = false;
 	}
+
+	return valid;
 }
 
 /* cache init */
@@ -263,10 +207,6 @@ static GpencilBatchCache *gpencil_batch_cache_init(Object *ob, int cfra)
 
 	GpencilBatchCache *cache = gpencil_batch_get_element(ob);
 
-	if (G.debug_value >= 664) {
-		printf("gpencil_batch_cache_init: %s\n", ob->id.name);
-	}
-
 	if (!cache) {
 		cache = MEM_callocN(sizeof(*cache), __func__);
 		ob_orig->runtime.gpencil_cache = cache;
@@ -275,19 +215,17 @@ static GpencilBatchCache *gpencil_batch_cache_init(Object *ob, int cfra)
 		memset(cache, 0, sizeof(*cache));
 	}
 
-	cache->cache_size = GPENCIL_MIN_BATCH_SLOTS_CHUNK;
-	cache->batch_stroke = MEM_callocN(sizeof(struct Gwn_Batch *) * cache->cache_size, "Gpencil_Batch_Stroke");
-	cache->batch_fill = MEM_callocN(sizeof(struct Gwn_Batch *) * cache->cache_size, "Gpencil_Batch_Fill");
-	cache->batch_edit = MEM_callocN(sizeof(struct Gwn_Batch *) * cache->cache_size, "Gpencil_Batch_Edit");
-	cache->batch_edlin = MEM_callocN(sizeof(struct Gwn_Batch *) * cache->cache_size, "Gpencil_Batch_Edlin");
-
 	cache->is_editmode = GPENCIL_ANY_EDIT_MODE(gpd);
-	gpd->flag &= ~GP_DATA_CACHE_IS_DIRTY;
 
-	cache->cache_idx = 0;
 	cache->is_dirty = true;
+
 	cache->cache_frame = cfra;
 
+	/* create array of derived frames equal to number of layers */
+	cache->tot_layers = BLI_listbase_count(&gpd->layers);
+	CLAMP_MIN(cache->tot_layers, 1);
+	cache->derived_array = MEM_callocN(sizeof(struct bGPDframe) * cache->tot_layers, "Derived GPF");
+
 	return cache;
 }
 
@@ -298,24 +236,30 @@ static void gpencil_batch_cache_clear(GpencilBatchCache *cache)
 		return;
 	}
 
-	if (cache->cache_size == 0) {
-		return;
-	}
-
-	if (cache->cache_size > 0) {
-		for (int i = 0; i < cache->cache_size; i++) {
-			GPU_BATCH_DISCARD_SAFE(cache->batch_stroke[i]);
-			GPU_BATCH_DISCARD_SAFE(cache->batch_fill[i]);
-			GPU_BATCH_DISCARD_SAFE(cache->batch_edit[i]);
-			GPU_BATCH_DISCARD_SAFE(cache->batch_edlin[i]);
-		}
-		MEM_SAFE_FREE(cache->batch_stroke);
-		MEM_SAFE_FREE(cache->batch_fill);
-		MEM_SAFE_FREE(cache->batch_edit);
-		MEM_SAFE_FREE(cache->batch_edlin);
+	GPU_BATCH_DISCARD_SAFE(cache->b_stroke.batch);
+	GPU_BATCH_DISCARD_SAFE(cache->b_point.batch);
+	GPU_BATCH_DISCARD_SAFE(cache->b_fill.batch);
+	GPU_BATCH_DISCARD_SAFE(cache->b_edit.batch);
+	GPU_BATCH_DISCARD_SAFE(cache->b_edlin.batch);
+
+	MEM_SAFE_FREE(cache->b_stroke.batch);
+	MEM_SAFE_FREE(cache->b_point.batch);
+	MEM_SAFE_FREE(cache->b_fill.batch);
+	MEM_SAFE_FREE(cache->b_edit.batch);
+	MEM_SAFE_FREE(cache->b_edlin.batch);
+
+	MEM_SAFE_FREE(cache->grp_cache);
+	cache->grp_size = 0;
+	cache->grp_used = 0;
+
+	/* clear all frames derived data */
+	for (int i = 0; i < cache->tot_layers; i++) {
+		bGPDframe *derived_gpf = &cache->derived_array[i];
+		BKE_gpencil_free_frame_runtime_data(derived_gpf);
+		derived_gpf = NULL;
 	}
-
-	cache->cache_size = 0;
+	cache->tot_layers = 0;
+	MEM_SAFE_FREE(cache->derived_array);
 }
 
 /* get cache */
@@ -326,10 +270,6 @@ GpencilBatchCache *gpencil_batch_cache_get(Object *ob, int cfra)
 
 	GpencilBatchCache *cache = gpencil_batch_get_element(ob);
 	if (!gpencil_batch_cache_valid(cache, gpd, cfra)) {
-		if (G.debug_value >= 664) {
-			printf("gpencil_batch_cache: %s\n", gpd->id.name);
-		}
-
 		if (cache) {
 			gpencil_batch_cache_clear(cache);
 		}