3D View: new nethod of opengl selection

Intended to replace legacy GL_SELECT, without the limitations of
sample queries which can't access depth information.

This commit adds VIEW3D_SELECT_PICK_NEAREST and VIEW3D_SELECT_PICK_ALL
which access the depth buffers to detect whats under the pointer,
so initial selection is always the closest item.

The performance of this method depends a lot on the OpenGL
implementations glReadPixels.

Since reading depth can be slow, buffers are cached for object picking
so selecting re-uses depth data, performing 1 draw instead of 3
(for 24, 18, 10 px regions, picking with many items under the pointer).

Occlusion queries draw twice when picking nearest,
so worst case 6x draw calls per selection.

Even with these improvements occlusion queries is faster on AMD hardware.

Depth selection is disabled by default, toggle option under select method.
May enable by default if this works well on different hardware.

Reviewed as D2543

1 files changed, 209 insertions, 0 deletions

diff --git a/source/blender/gpu/intern/gpu_select_sample_query.c b/source/blender/gpu/intern/gpu_select_sample_query.c
new file mode 100644
index 00000000000..5576367edd9
--- /dev/null
+++ b/source/blender/gpu/intern/gpu_select_sample_query.c
@@ -0,0 +1,209 @@
+/*
+ * ***** BEGIN GPL LICENSE BLOCK *****
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ *
+ * The Original Code is Copyright (C) 2014 Blender Foundation.
+ * All rights reserved.
+ *
+ * Contributor(s): Antony Riakiotakis.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ */
+
+/** \file blender/gpu/intern/gpu_select.c
+ *  \ingroup gpu
+ *
+ * Interface for accessing gpu-related methods for selection. The semantics will be
+ * similar to glRenderMode(GL_SELECT) since the goal is to maintain compatibility.
+ */
+
+#include <stdlib.h>
+
+#include "GPU_select.h"
+#include "GPU_extensions.h"
+#include "GPU_glew.h"
+ 
+#include "MEM_guardedalloc.h"
+
+#include "BLI_rect.h"
+
+#include "BLI_utildefines.h"
+
+#include "gpu_select_private.h"
+
+
+/* Ad hoc number of queries to allocate to skip doing many glGenQueries */
+#define ALLOC_QUERIES 200
+
+typedef struct GPUQueryState {
+	/* Tracks whether a query has been issued so that gpu_load_id can end the previous one */
+	bool query_issued;
+	/* array holding the OpenGL query identifiers */
+	unsigned int *queries;
+	/* array holding the id corresponding to each query */
+	unsigned int *id;
+	/* number of queries in *queries and *id */
+	unsigned int num_of_queries;
+	/* index to the next query to start */
+	unsigned int active_query;
+	/* cache on initialization */
+	unsigned int (*buffer)[4];
+	/* buffer size (stores number of integers, for actual size multiply by sizeof integer)*/
+	unsigned int bufsize;
+	/* mode of operation */
+	char mode;
+	unsigned int index;
+	int oldhits;
+} GPUQueryState;
+
+static GPUQueryState g_query_state = {0};
+
+
+void gpu_select_query_begin(
+        unsigned int (*buffer)[4], unsigned int bufsize,
+        const rcti *input, char mode,
+        int oldhits)
+{
+	float viewport[4];
+
+	g_query_state.query_issued = false;
+	g_query_state.active_query = 0;
+	g_query_state.num_of_queries = 0;
+	g_query_state.bufsize = bufsize;
+	g_query_state.buffer = buffer;
+	g_query_state.mode = mode;
+	g_query_state.index = 0;
+	g_query_state.oldhits = oldhits;
+
+	g_query_state.num_of_queries = ALLOC_QUERIES;
+
+	g_query_state.queries = MEM_mallocN(g_query_state.num_of_queries * sizeof(*g_query_state.queries), "gpu selection queries");
+	g_query_state.id = MEM_mallocN(g_query_state.num_of_queries * sizeof(*g_query_state.id), "gpu selection ids");
+	glGenQueries(g_query_state.num_of_queries, g_query_state.queries);
+
+	glPushAttrib(GL_DEPTH_BUFFER_BIT | GL_VIEWPORT_BIT);
+	/* disable writing to the framebuffer */
+	glColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+	/* In order to save some fill rate we minimize the viewport using rect.
+	 * We need to get the region of the scissor so that our geometry doesn't
+	 * get rejected before the depth test. Should probably cull rect against
+	 * scissor for viewport but this is a rare case I think */
+	glGetFloatv(GL_SCISSOR_BOX, viewport);
+	glViewport(viewport[0], viewport[1], BLI_rcti_size_x(input), BLI_rcti_size_y(input));
+
+	/* occlusion queries operates on fragments that pass tests and since we are interested on all
+	 * objects in the view frustum independently of their order, we need to disable the depth test */
+	if (mode == GPU_SELECT_ALL) {
+		glDisable(GL_DEPTH_TEST);
+		glDepthMask(GL_FALSE);
+	}
+	else if (mode == GPU_SELECT_NEAREST_FIRST_PASS) {
+		glClear(GL_DEPTH_BUFFER_BIT);
+		glEnable(GL_DEPTH_TEST);
+		glDepthMask(GL_TRUE);
+		glDepthFunc(GL_LEQUAL);
+	}
+	else if (mode == GPU_SELECT_NEAREST_SECOND_PASS) {
+		glEnable(GL_DEPTH_TEST);
+		glDepthMask(GL_FALSE);
+		glDepthFunc(GL_EQUAL);
+	}
+}
+
+bool gpu_select_query_load_id(unsigned int id)
+{
+	if (g_query_state.query_issued) {
+		glEndQuery(GL_SAMPLES_PASSED);
+	}
+	/* if required, allocate extra queries */
+	if (g_query_state.active_query == g_query_state.num_of_queries) {
+		g_query_state.num_of_queries += ALLOC_QUERIES;
+		g_query_state.queries = MEM_reallocN(g_query_state.queries, g_query_state.num_of_queries * sizeof(*g_query_state.queries));
+		g_query_state.id = MEM_reallocN(g_query_state.id, g_query_state.num_of_queries * sizeof(*g_query_state.id));
+		glGenQueries(ALLOC_QUERIES, &g_query_state.queries[g_query_state.active_query]);
+	}
+
+	glBeginQuery(GL_SAMPLES_PASSED, g_query_state.queries[g_query_state.active_query]);
+	g_query_state.id[g_query_state.active_query] = id;
+	g_query_state.active_query++;
+	g_query_state.query_issued = true;
+
+	if (g_query_state.mode == GPU_SELECT_NEAREST_SECOND_PASS && g_query_state.index < g_query_state.oldhits) {
+		if (g_query_state.buffer[g_query_state.index][3] == id) {
+			g_query_state.index++;
+			return true;
+		}
+		else {
+			return false;
+		}
+	}
+
+	return true;
+}
+
+unsigned int gpu_select_query_end(void)
+{
+	int i;
+
+	unsigned int hits = 0;
+	const unsigned int maxhits = g_query_state.bufsize;
+
+	if (g_query_state.query_issued) {
+		glEndQuery(GL_SAMPLES_PASSED);
+	}
+
+	for (i = 0; i < g_query_state.active_query; i++) {
+		unsigned int result;
+		glGetQueryObjectuiv(g_query_state.queries[i], GL_QUERY_RESULT, &result);
+		if (result > 0) {
+			if (g_query_state.mode != GPU_SELECT_NEAREST_SECOND_PASS) {
+
+				if (hits < maxhits) {
+					g_query_state.buffer[hits][0] = 1;
+					g_query_state.buffer[hits][1] = 0xFFFF;
+					g_query_state.buffer[hits][2] = 0xFFFF;
+					g_query_state.buffer[hits][3] = g_query_state.id[i];
+
+					hits++;
+				}
+				else {
+					hits = -1;
+					break;
+				}
+			}
+			else {
+				int j;
+				/* search in buffer and make selected object first */
+				for (j = 0; j < g_query_state.oldhits; j++) {
+					if (g_query_state.buffer[j][3] == g_query_state.id[i]) {
+						g_query_state.buffer[j][1] = 0;
+						g_query_state.buffer[j][2] = 0;
+					}
+				}
+				break;
+			}
+		}
+	}
+
+	glDeleteQueries(g_query_state.num_of_queries, g_query_state.queries);
+	MEM_freeN(g_query_state.queries);
+	MEM_freeN(g_query_state.id);
+	glPopAttrib();
+	glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+
+	return hits;
+}