Implement Uniformbuffer objects for nodetree parameters

For users that means you can tweak shaders in the nodetree and things
are way faster. This is a huge improvement, particularly in
systems that have no shader cache.

From the code perspective it means we are no longer re-compiling the
shader every time a value is tweaked in the UI. We are using uniforms
for those values.

It would be slow to add that many uniforms for all the shaders. So
instead we are using UBO (Uniform Buffer Objects).

This fixes the main issue of T51467. However GWN_shaderinterface_create() still
needs to be improvedi. When opening a .blend all shaders are compiled once, so
optimizing it will bring a measurable impact.

========================================================================
NOTE: This breaks update of Cycles material upon nodetree nodes
tweaking. It will be fixed separately by depsgraph, once tackling T51925
(Animated Eevee values slowdown).

The idea is to make Depsgraph update more granular. The XXX TODO in
rna_nodetree.c will be tackled at that time as well.
========================================================================

Reviewers: sergey, brecht, fclem

Differential Revision: https://developer.blender.org/D2739

1 files changed, 134 insertions, 16 deletions

diff --git a/source/blender/gpu/intern/gpu_codegen.c b/source/blender/gpu/intern/gpu_codegen.c
index cd367c821ba..9fa78bfe5b9 100644
--- a/source/blender/gpu/intern/gpu_codegen.c
+++ b/source/blender/gpu/intern/gpu_codegen.c
@@ -36,6 +36,7 @@
 #include "DNA_customdata_types.h"
 #include "DNA_image_types.h"
 #include "DNA_material_types.h"
+#include "DNA_node_types.h"
 
 #include "BLI_blenlib.h"
 #include "BLI_utildefines.h"
@@ -47,6 +48,7 @@
 #include "GPU_material.h"
 #include "GPU_shader.h"
 #include "GPU_texture.h"
+#include "GPU_uniformbuffer.h"
 
 #include "BLI_sys_types.h" /* for intptr_t support */
 
@@ -508,12 +510,16 @@ static void codegen_set_unique_ids(ListBase *nodes)
 	BLI_ghash_free(definehash, NULL, NULL);
 }
 
-static int codegen_print_uniforms_functions(DynStr *ds, ListBase *nodes)
+/**
+ * It will create an UBO for GPUMaterial if there is any GPU_DYNAMIC_UBO.
+ */
+static int codegen_process_uniforms_functions(GPUMaterial *material, DynStr *ds, ListBase *nodes)
 {
 	GPUNode *node;
 	GPUInput *input;
 	const char *name;
 	int builtins = 0;
+	ListBase ubo_inputs = {NULL, NULL};
 
 	/* print uniforms */
 	for (node = nodes->first; node; node = node->next) {
@@ -545,7 +551,13 @@ static int codegen_print_uniforms_functions(DynStr *ds, ListBase *nodes)
 				}
 			}
 			else if (input->source == GPU_SOURCE_VEC_UNIFORM) {
-				if (input->dynamicvec) {
+				if (input->dynamictype == GPU_DYNAMIC_UBO) {
+					if (!input->link) {
+						/* We handle the UBOuniforms separately. */
+						BLI_addtail(&ubo_inputs, BLI_genericNodeN(input));
+					}
+				}
+				else if (input->dynamicvec) {
 					/* only create uniforms for dynamic vectors */
 					BLI_dynstr_appendf(ds, "uniform %s unf%d;\n",
 						GPU_DATATYPE_STR[input->type], input->id);
@@ -584,6 +596,22 @@ static int codegen_print_uniforms_functions(DynStr *ds, ListBase *nodes)
 		}
 	}
 
+	/* Handle the UBO block separately. */
+	if ((material != NULL) && !BLI_listbase_is_empty(&ubo_inputs)) {
+		GPU_material_create_uniform_buffer(material, &ubo_inputs);
+
+		/* Inputs are sorted */
+		BLI_dynstr_appendf(ds, "\nlayout (std140) uniform %s {\n", GPU_UBO_BLOCK_NAME);
+
+		for (LinkData *link = ubo_inputs.first; link; link = link->next) {
+			input = link->data;
+			BLI_dynstr_appendf(ds, "\t%s unf%d;\n",
+				GPU_DATATYPE_STR[input->type], input->id);
+		}
+		BLI_dynstr_append(ds, "};\n");
+		BLI_freelistN(&ubo_inputs);
+	}
+
 	BLI_dynstr_append(ds, "\n");
 
 	return builtins;
@@ -686,7 +714,7 @@ static void codegen_call_functions(DynStr *ds, ListBase *nodes, GPUOutput *final
 	BLI_dynstr_append(ds, ";\n");
 }
 
-static char *code_generate_fragment(ListBase *nodes, GPUOutput *output, bool use_new_shading)
+static char *code_generate_fragment(GPUMaterial *material, ListBase *nodes, GPUOutput *output, bool use_new_shading)
 {
 	DynStr *ds = BLI_dynstr_new();
 	char *code;
@@ -703,7 +731,7 @@ static char *code_generate_fragment(ListBase *nodes, GPUOutput *output, bool use
 #endif
 
 	codegen_set_unique_ids(nodes);
-	builtins = codegen_print_uniforms_functions(ds, nodes);
+	builtins = codegen_process_uniforms_functions(material, ds, nodes);
 
 #if 0
 	if (G.debug & G_DEBUG)
@@ -1397,7 +1425,7 @@ static void gpu_node_input_link(GPUNode *node, GPUNodeLink *link, const GPUType
 		name = outnode->name;
 		input = outnode->inputs.first;
 
-		if ((STREQ(name, "set_value") || STREQ(name, "set_rgb")) &&
+		if ((STREQ(name, "set_value") || STREQ(name, "set_rgb") || STREQ(name, "set_rgba")) &&
 		    (input->type == type))
 		{
 			input = MEM_dupallocN(outnode->inputs.first);
@@ -1514,15 +1542,84 @@ static void gpu_node_input_link(GPUNode *node, GPUNodeLink *link, const GPUType
 	BLI_addtail(&node->inputs, input);
 }
 
-static void gpu_node_input_socket(GPUNode *node, GPUNodeStack *sock)
+
+static const char *gpu_uniform_set_function_from_type(eNodeSocketDatatype type)
 {
-	GPUNodeLink *link;
+	switch (type) {
+		case SOCK_FLOAT:
+			return "set_value";
+		case SOCK_VECTOR:
+			return "set_rgb";
+		case SOCK_RGBA:
+			return "set_rgba";
+		default:
+			 BLI_assert(!"No gpu function for non-supported eNodeSocketDatatype");
+			 return NULL;
+	}
+}
+
+/**
+ * Link stack uniform buffer.
+ * This is called for the input/output sockets that are note connected.
+ */
+static GPUNodeLink *gpu_uniformbuffer_link(
+        GPUMaterial *mat, bNode *node, GPUNodeStack *stack, const int index, const eNodeSocketInOut in_out)
+{
+	bNodeSocket *socket;
+	if (in_out == SOCK_IN) {
+		socket = BLI_findlink(&node->original->inputs, index);
+	}
+	else {
+		socket = BLI_findlink(&node->original->outputs, index);
+	}
+
+	BLI_assert(socket != NULL);
+	BLI_assert(socket->in_out == in_out);
+
+	if ((socket->flag & SOCK_HIDE_VALUE) == 0) {
+		GPUNodeLink *link;
+		switch (socket->type) {
+			case SOCK_FLOAT:
+			{
+				bNodeSocketValueFloat *socket_data = socket->default_value;
+				link = GPU_uniform_buffer(&socket_data->value, GPU_FLOAT);
+				break;
+			}
+			case SOCK_VECTOR:
+			{
+				bNodeSocketValueRGBA *socket_data = socket->default_value;
+				link = GPU_uniform_buffer(socket_data->value, GPU_VEC3);
+				break;
+			}
+			case SOCK_RGBA:
+			{
+				bNodeSocketValueRGBA *socket_data = socket->default_value;
+				link = GPU_uniform_buffer(socket_data->value, GPU_VEC4);
+				break;
+			}
+			default:
+				return NULL;
+				break;
+		}
+
+		if (in_out == SOCK_IN) {
+			GPU_link(mat, gpu_uniform_set_function_from_type(socket->type), link, &stack->link);
+		}
+		return link;
+	}
+	return NULL;
+}
 
+static void gpu_node_input_socket(GPUMaterial *material, bNode *bnode, GPUNode *node, GPUNodeStack *sock, const int index)
+{
 	if (sock->link) {
 		gpu_node_input_link(node, sock->link, sock->type);
 	}
+	else if ((material != NULL) && (gpu_uniformbuffer_link(material, bnode, sock, index, SOCK_IN) != NULL)) {
+		gpu_node_input_link(node, sock->link, sock->type);
+	}
 	else {
-		link = GPU_node_link_create();
+		GPUNodeLink *link = GPU_node_link_create();
 		link->ptr1 = sock->vec;
 		gpu_node_input_link(node, link, sock->type);
 	}
@@ -1685,6 +1782,21 @@ GPUNodeLink *GPU_dynamic_uniform(float *num, GPUDynamicType dynamictype, void *d
 	return link;
 }
 
+/**
+ * Add uniform to UBO struct of GPUMaterial.
+ */
+GPUNodeLink *GPU_uniform_buffer(float *num, GPUType gputype)
+{
+	GPUNodeLink *link = GPU_node_link_create();
+	link->ptr1 = num;
+	link->ptr2 = NULL;
+	link->dynamic = true;
+	link->dynamictype = GPU_DYNAMIC_UBO;
+	link->type = gputype;
+
+	return link;
+}
+
 GPUNodeLink *GPU_image(Image *ima, ImageUser *iuser, bool is_data)
 {
 	GPUNodeLink *link = GPU_node_link_create();
@@ -1795,7 +1907,7 @@ bool GPU_link(GPUMaterial *mat, const char *name, ...)
 	return true;
 }
 
-bool GPU_stack_link(GPUMaterial *mat, const char *name, GPUNodeStack *in, GPUNodeStack *out, ...)
+bool GPU_stack_link(GPUMaterial *material, bNode *bnode, const char *name, GPUNodeStack *in, GPUNodeStack *out, ...)
 {
 	GPUNode *node;
 	GPUFunction *function;
@@ -1815,11 +1927,11 @@ bool GPU_stack_link(GPUMaterial *mat, const char *name, GPUNodeStack *in, GPUNod
 
 	if (in) {
 		for (i = 0; in[i].type != GPU_NONE; i++) {
-			gpu_node_input_socket(node, &in[i]);
+			gpu_node_input_socket(material, bnode, node, &in[i], i);
 			totin++;
 		}
 	}
-	
+
 	if (out) {
 		for (i = 0; out[i].type != GPU_NONE; i++) {
 			gpu_node_output(node, out[i].type, &out[i].link);
@@ -1841,7 +1953,7 @@ bool GPU_stack_link(GPUMaterial *mat, const char *name, GPUNodeStack *in, GPUNod
 			if (totin == 0) {
 				link = va_arg(params, GPUNodeLink *);
 				if (link->socket)
-					gpu_node_input_socket(node, link->socket);
+					gpu_node_input_socket(NULL, NULL, node, link->socket, -1);
 				else
 					gpu_node_input_link(node, link, function->paramtype[i]);
 			}
@@ -1851,8 +1963,8 @@ bool GPU_stack_link(GPUMaterial *mat, const char *name, GPUNodeStack *in, GPUNod
 	}
 	va_end(params);
 
-	gpu_material_add_node(mat, node);
-	
+	gpu_material_add_node(material, node);
+
 	return true;
 }
 
@@ -1877,6 +1989,11 @@ int GPU_link_changed(GPUNodeLink *link)
 		return 0;
 }
 
+GPUNodeLink *GPU_uniformbuffer_link_out(GPUMaterial *mat, bNode *node, GPUNodeStack *stack, const int index)
+{
+	return gpu_uniformbuffer_link(mat, node, stack, index, SOCK_OUT);
+}
+
 /* Pass create/free */
 
 static void gpu_nodes_tag(GPUNodeLink *link)
@@ -1917,6 +2034,7 @@ static void gpu_nodes_prune(ListBase *nodes, GPUNodeLink *outlink)
 }
 
 GPUPass *GPU_generate_pass_new(
+        struct GPUMaterial *material,
         ListBase *nodes, struct GPUNodeLink *frag_outlink,
         GPUVertexAttribs *attribs,
         const char *vert_code, const char *geom_code,
@@ -1933,7 +2051,7 @@ GPUPass *GPU_generate_pass_new(
 	gpu_nodes_get_vertex_attributes(nodes, attribs);
 
 	/* generate code and compile with opengl */
-	fragmentgen = code_generate_fragment(nodes, frag_outlink->output, true);
+	fragmentgen = code_generate_fragment(material, nodes, frag_outlink->output, true);
 	vertexgen = code_generate_vertex_new(nodes, vert_code, (geom_code != NULL));
 
 	tmp = BLI_strdupcat(frag_lib, glsl_material_library);
@@ -2012,7 +2130,7 @@ GPUPass *GPU_generate_pass(
 	gpu_nodes_get_builtin_flag(nodes, builtins);
 
 	/* generate code and compile with opengl */
-	fragmentcode = code_generate_fragment(nodes, outlink->output, false);
+	fragmentcode = code_generate_fragment(NULL, nodes, outlink->output, false);
 	vertexcode = code_generate_vertex(nodes, type);
 	geometrycode = code_generate_geometry(nodes, use_opensubdiv);