Bump maximum threads number to 1024

This commit contains all the changes required for most optimal maximum threads
number bump. This is needed to avoid possibly unneeded initialization or data
allocation on systems with lower threads count.

TODO: Still need to review arrays in render data structures from render_types.h,

P.S. We might remove actual bump of max threads from this patch, so when we'll
be applying the patch we can do all the preparation work and then do actual
bump of max threads.

Reviewers: mont29, campbellbarton

Reviewed By: mont29, campbellbarton

Maniphest Tasks: T43306

Differential Revision: https://developer.blender.org/D1343

7 files changed, 24 insertions, 13 deletions

diff --git a/source/blender/blenlib/BLI_threads.h b/source/blender/blenlib/BLI_threads.h
index b4a465bbc74..0b1b4d8ee8c 100644
--- a/source/blender/blenlib/BLI_threads.h
+++ b/source/blender/blenlib/BLI_threads.h
@@ -42,7 +42,7 @@ extern "C" {
 #endif
 
 /* for tables, button in UI, etc */
-#define BLENDER_MAX_THREADS     64
+#define BLENDER_MAX_THREADS     1024
 
 struct ListBase;
 struct TaskScheduler;
diff --git a/source/blender/render/intern/include/render_types.h b/source/blender/render/intern/include/render_types.h
index 6de5da3795a..b3a5ccdae17 100644
--- a/source/blender/render/intern/include/render_types.h
+++ b/source/blender/render/intern/include/render_types.h
@@ -180,6 +180,7 @@ struct Render {
 	float jit[32][2];
 	float mblur_jit[32][2];
 	ListBase *qmcsamplers;
+	int num_qmc_samplers;
 	
 	/* shadow counter, detect shadow-reuse for shaders */
 	int shadowsamplenr[BLENDER_MAX_THREADS];
diff --git a/source/blender/render/intern/include/rendercore.h b/source/blender/render/intern/include/rendercore.h
index 308903c6c6d..7254fd25ee6 100644
--- a/source/blender/render/intern/include/rendercore.h
+++ b/source/blender/render/intern/include/rendercore.h
@@ -90,7 +90,7 @@ extern void ray_shadow(ShadeInput *shi, LampRen *lar, float shadfac[4]);
 extern void ray_trace(ShadeInput *shi, ShadeResult *);
 extern void ray_ao(ShadeInput *shi, float ao[3], float env[3]);
 extern void init_jitter_plane(LampRen *lar);
-extern void init_ao_sphere(struct World *wrld);
+extern void init_ao_sphere(Render *re, struct World *wrld);
 extern void init_render_qmcsampler(Render *re);
 extern void free_render_qmcsampler(Render *re);
 
diff --git a/source/blender/render/intern/source/convertblender.c b/source/blender/render/intern/source/convertblender.c
index 907974e20dc..ab828a0c04e 100644
--- a/source/blender/render/intern/source/convertblender.c
+++ b/source/blender/render/intern/source/convertblender.c
@@ -5230,7 +5230,7 @@ void RE_Database_FromScene(Render *re, Main *bmain, Scene *scene, unsigned int l
 
 		if (re->wrld.mode & (WO_AMB_OCC|WO_ENV_LIGHT|WO_INDIRECT_LIGHT))
 			if (re->wrld.ao_samp_method == WO_AOSAMP_CONSTANT)
-				init_ao_sphere(&re->wrld);
+				init_ao_sphere(re, &re->wrld);
 	}
 	
 	/* still bad... doing all */
@@ -5956,7 +5956,7 @@ void RE_Database_Baking(Render *re, Main *bmain, Scene *scene, unsigned int lay,
 		
 		if (re->wrld.mode & (WO_AMB_OCC|WO_ENV_LIGHT|WO_INDIRECT_LIGHT))
 			if (re->wrld.ao_samp_method == WO_AOSAMP_CONSTANT)
-				init_ao_sphere(&re->wrld);
+				init_ao_sphere(re, &re->wrld);
 	}
 	
 	/* still bad... doing all */
diff --git a/source/blender/render/intern/source/occlusion.c b/source/blender/render/intern/source/occlusion.c
index c5c3b6bbf94..b3d31e3b93a 100644
--- a/source/blender/render/intern/source/occlusion.c
+++ b/source/blender/render/intern/source/occlusion.c
@@ -115,6 +115,8 @@ typedef struct OcclusionTree {
 	int doindirect;
 
 	OcclusionCache *cache;
+
+	int num_threads;
 } OcclusionTree;
 
 typedef struct OcclusionThread {
@@ -641,6 +643,7 @@ static void occ_build_sh_normalize(OccNode *node)
 
 static OcclusionTree *occ_tree_build(Render *re)
 {
+	const int num_threads = re->r.threads;
 	OcclusionTree *tree;
 	ObjectInstanceRen *obi;
 	ObjectRen *obr;
@@ -679,7 +682,7 @@ static OcclusionTree *occ_tree_build(Render *re)
 	BLI_memarena_use_calloc(tree->arena);
 
 	if (re->wrld.aomode & WO_AOCACHE)
-		tree->cache = MEM_callocN(sizeof(OcclusionCache) * BLENDER_MAX_THREADS, "OcclusionCache");
+		tree->cache = MEM_callocN(sizeof(OcclusionCache) * num_threads, "OcclusionCache");
 
 	tree->face = MEM_callocN(sizeof(OccFace) * totface, "OcclusionFace");
 	tree->co = MEM_callocN(sizeof(float) * 3 * totface, "OcclusionCo");
@@ -730,9 +733,11 @@ static OcclusionTree *occ_tree_build(Render *re)
 	if (!(re->test_break(re->tbh)))
 		occ_build_sh_normalize(tree->root);
 
-	for (a = 0; a < BLENDER_MAX_THREADS; a++)
+	for (a = 0; a < num_threads; a++)
 		tree->stack[a] = MEM_callocN(sizeof(OccNode) * TOTCHILD * (tree->maxdepth + 1), "OccStack");
 
+	tree->num_threads = num_threads;
+
 	return tree;
 }
 
@@ -742,7 +747,7 @@ static void occ_free_tree(OcclusionTree *tree)
 
 	if (tree) {
 		if (tree->arena) BLI_memarena_free(tree->arena);
-		for (a = 0; a < BLENDER_MAX_THREADS; a++)
+		for (a = 0; a < tree->num_threads; a++)
 			if (tree->stack[a])
 				MEM_freeN(tree->stack[a]);
 		if (tree->occlusion) MEM_freeN(tree->occlusion);
diff --git a/source/blender/render/intern/source/rayshade.c b/source/blender/render/intern/source/rayshade.c
index 9aac5ed1f1d..26a0b0c71b4 100644
--- a/source/blender/render/intern/source/rayshade.c
+++ b/source/blender/render/intern/source/rayshade.c
@@ -174,10 +174,11 @@ void freeraytree(Render *re)
 	
 #ifdef RE_RAYCOUNTER
 	{
+		const int num_threads = re->r.threads;
 		RayCounter sum;
 		memset(&sum, 0, sizeof(sum));
 		int i;
-		for (i=0; i<BLENDER_MAX_THREADS; i++)
+		for (i=0; i<num_threads; i++)
 			RE_RC_MERGE(&sum, re_rc_counter+i);
 		RE_RC_INFO(&sum);
 	}
@@ -1186,7 +1187,9 @@ static void QMC_sampleHemiCosine(float vec[3], QMCSampler *qsa, int thread, int
 /* called from convertBlenderScene.c */
 void init_render_qmcsampler(Render *re)
 {
-	re->qmcsamplers= MEM_callocN(sizeof(ListBase)*BLENDER_MAX_THREADS, "QMCListBase");
+	const int num_threads = re->r.threads;
+	re->qmcsamplers= MEM_callocN(sizeof(ListBase)*num_threads, "QMCListBase");
+	re->num_qmc_samplers = num_threads;
 }
 
 static QMCSampler *get_thread_qmcsampler(Render *re, int thread, int type, int tot)
@@ -1220,7 +1223,7 @@ void free_render_qmcsampler(Render *re)
 	if (re->qmcsamplers) {
 		QMCSampler *qsa, *next;
 		int a;
-		for (a=0; a<BLENDER_MAX_THREADS; a++) {
+		for (a = 0; a < re->num_qmc_samplers; a++) {
 			for (qsa=re->qmcsamplers[a].first; qsa; qsa=next) {
 				next= qsa->next;
 				QMC_freeSampler(qsa);
@@ -1695,9 +1698,10 @@ static void DS_energy(float *sphere, int tot, float vec[3])
 /* called from convertBlenderScene.c */
 /* creates an equally distributed spherical sample pattern */
 /* and allocates threadsafe memory */
-void init_ao_sphere(World *wrld)
+void init_ao_sphere(Render *re, World *wrld)
 {
 	/* fixed random */
+	const int num_threads = re->r.threads;
 	RNG *rng;
 	float *fp;
 	int a, tot, iter= 16;
@@ -1721,7 +1725,7 @@ void init_ao_sphere(World *wrld)
 	}
 	
 	/* tables */
-	wrld->aotables= MEM_mallocN(BLENDER_MAX_THREADS*3*tot*sizeof(float), "AO tables");
+	wrld->aotables= MEM_mallocN(num_threads*3*tot*sizeof(float), "AO tables");
 
 	BLI_rng_free(rng);
 }
diff --git a/source/blender/render/intern/source/render_texture.c b/source/blender/render/intern/source/render_texture.c
index 513cfa6df7d..b4a14f5337d 100644
--- a/source/blender/render/intern/source/render_texture.c
+++ b/source/blender/render/intern/source/render_texture.c
@@ -3653,7 +3653,8 @@ void render_realtime_texture(ShadeInput *shi, Image *ima)
 	if (firsttime) {
 		BLI_lock_thread(LOCK_IMAGE);
 		if (firsttime) {
-			for (a=0; a<BLENDER_MAX_THREADS; a++) {
+			const int num_threads = BLI_system_thread_count();
+			for (a = 0; a < num_threads; a++) {
 				memset(&imatex[a], 0, sizeof(Tex));
 				BKE_texture_default(&imatex[a]);
 				imatex[a].type= TEX_IMAGE;