Render Monster support: (part 1)

Removed all limitations from render code for maximum threads. The only
define for this now is in BLI_threads.h, and currently set to 8.
Note that each thread renders an entire tile, and also allocates the
buffers for the tiles, so; more threads might work better with smaller
tiles.

IMPORTANT: node system won't work yet with more than 2 threads! So, don't
try material nodes or compositing with over 2 threads. That I'll commit
later today.

What does work (should work :) is AO and soft shadow now.

7 files changed, 121 insertions, 144 deletions

diff --git a/source/blender/render/intern/include/render_types.h b/source/blender/render/intern/include/render_types.h
index 2269f6ef801..3ac56e73099 100644
--- a/source/blender/render/intern/include/render_types.h
+++ b/source/blender/render/intern/include/render_types.h
@@ -37,6 +37,8 @@
 #include "DNA_object_types.h"
 #include "DNA_vec_types.h"
 
+#include "BLI_threads.h"
+
 #include "RE_pipeline.h"
 #include "RE_shader_ext.h"	/* TexResult, ShadeResult, ShadeInput */
 
@@ -48,8 +50,6 @@ struct GHash;
 
 #define TABLEINITSIZE 1024
 #define LAMPINITSIZE 256
-		/* hardcoded maximum now, for optimize tables */
-#define RE_MAXTHREAD	2
 
 typedef struct SampleTables
 {
@@ -77,6 +77,7 @@ typedef struct RenderPart
 	short sample, nr;				/* sample can be used by zbuffers, nr is partnr */
 	short thread;					/* thread id */
 	
+	char *clipflag;					/* clipflags for part zbuffering */
 } RenderPart;
 
 typedef struct Octree {
@@ -209,7 +210,7 @@ typedef struct ShadBuf {
 	ListBase buffers;
 	
 	/* irregular shadowbufer, result stored per thread */
-	struct ISBData *isb_result[RE_MAXTHREAD];
+	struct ISBData *isb_result[BLENDER_MAX_THREADS];
 } ShadBuf;
 
 /* ------------------------------------------------------------------------- */
@@ -326,7 +327,7 @@ typedef struct LampRen
 	float bias;
 	
 	short ray_samp, ray_sampy, ray_sampz, ray_samp_type, area_shape, ray_totsamp;
-	short xold1, yold1, xold2, yold2;	/* last jitter table for area lights */
+	short xold[BLENDER_MAX_THREADS], yold[BLENDER_MAX_THREADS];	/* last jitter table for area lights */
 	float area_size, area_sizey, area_sizez;
 	
 	struct ShadBuf *shb;
@@ -347,7 +348,7 @@ typedef struct LampRen
 	short YF_glowtype;
 	
 	/* ray optim */
-	VlakRen *vlr_last[RE_MAXTHREAD];
+	VlakRen *vlr_last[BLENDER_MAX_THREADS];
 	
 	struct MTex *mtex[MAX_MTEX];
 } LampRen;
diff --git a/source/blender/render/intern/include/rendercore.h b/source/blender/render/intern/include/rendercore.h
index dacb9e0b5f9..e70913fbbde 100644
--- a/source/blender/render/intern/include/rendercore.h
+++ b/source/blender/render/intern/include/rendercore.h
@@ -46,6 +46,7 @@
 struct HaloRen;
 struct ShadeInput;
 struct ShadeResult;
+struct World;
 
 /* ------------------------------------------------------------------------- */
 
@@ -109,7 +110,7 @@ extern void ray_shadow(ShadeInput *, LampRen *, float *);
 extern void ray_trace(ShadeInput *, ShadeResult *);
 extern void ray_ao(ShadeInput *, float *);
 extern void init_jitter_plane(LampRen *lar);
-extern void init_ao_sphere(float *sphere, int tot, int iter);
+extern void init_ao_sphere(struct World *wrld);
 
 #endif /* RENDER_EXT_H */
 
diff --git a/source/blender/render/intern/source/convertblender.c b/source/blender/render/intern/source/convertblender.c
index 89d4698f93b..992ed98c0eb 100644
--- a/source/blender/render/intern/source/convertblender.c
+++ b/source/blender/render/intern/source/convertblender.c
@@ -2974,6 +2974,11 @@ void RE_Database_Free(Render *re)
 		re->wrld.aosphere= NULL;
 		re->scene->world->aosphere= NULL;
 	}
+	if(re->wrld.aotables) {
+		MEM_freeN(re->wrld.aotables);
+		re->wrld.aotables= NULL;
+		re->scene->world->aotables= NULL;
+	}
 	
 	if(re->r.mode & R_RAYTRACE) freeoctree(re);
 	
@@ -3220,11 +3225,8 @@ void RE_Database_FromScene(Render *re, Scene *scene, int use_camera_view)
 	}
 	
 	init_render_world(re);	/* do first, because of ambient. also requires re->osa set correct */
-	if( (re->wrld.mode & WO_AMB_OCC) && (re->r.mode & R_RAYTRACE) ) {
-		re->wrld.aosphere= MEM_mallocN(2*3*re->wrld.aosamp*re->wrld.aosamp*sizeof(float), "AO sphere");
-		/* we make twice the amount of samples, because only a hemisphere is used */
-		init_ao_sphere(re->wrld.aosphere, 2*re->wrld.aosamp*re->wrld.aosamp, 16);
-	}
+	if( (re->wrld.mode & WO_AMB_OCC) && (re->r.mode & R_RAYTRACE) )
+		init_ao_sphere(&re->wrld);
 	
 	/* still bad... doing all */
 	init_render_textures(re);
@@ -3894,11 +3896,8 @@ void RE_Database_Baking(Render *re, Scene *scene, int type)
 	}
 	
 	init_render_world(re);	/* do first, because of ambient. also requires re->osa set correct */
-	if( (re->wrld.mode & WO_AMB_OCC) && (re->r.mode & R_RAYTRACE) ) {
-		re->wrld.aosphere= MEM_mallocN(2*3*re->wrld.aosamp*re->wrld.aosamp*sizeof(float), "AO sphere");
-		/* we make twice the amount of samples, because only a hemisphere is used */
-		init_ao_sphere(re->wrld.aosphere, 2*re->wrld.aosamp*re->wrld.aosamp, 16);
-	}
+	if( (re->wrld.mode & WO_AMB_OCC) && (re->r.mode & R_RAYTRACE) )
+		init_ao_sphere(&re->wrld);
 	
 	/* still bad... doing all */
 	init_render_textures(re);
diff --git a/source/blender/render/intern/source/pipeline.c b/source/blender/render/intern/source/pipeline.c
index 5db575d3550..cd954b8c363 100644
--- a/source/blender/render/intern/source/pipeline.c
+++ b/source/blender/render/intern/source/pipeline.c
@@ -1100,7 +1100,7 @@ static void threaded_tile_processor(Render *re)
 	RenderPart *pa, *nextpa;
 	RenderResult *rr;
 	rctf viewplane= re->viewplane;
-	int maxthreads, rendering=1, counter= 1, drawtimer=0, hasdrawn, minx=0;
+	int rendering=1, counter= 1, drawtimer=0, hasdrawn, minx=0;
 	
 	/* first step; the entire render result, or prepare exr buffer saving */
 	free_render_result(re->result);
@@ -1123,11 +1123,7 @@ static void threaded_tile_processor(Render *re)
 		IMB_exrtile_begin_write(rr->exrhandle, str, rr->rectx, rr->recty, rr->rectx/re->xparts, rr->recty/re->yparts);
 	}
 	
-	if(re->r.mode & R_THREADS) 
-		maxthreads= RE_MAXTHREAD;	/* should become button value too */
-	else maxthreads= 1;
-	
-	BLI_init_threads(&threads, do_part_thread, maxthreads);
+	BLI_init_threads(&threads, do_part_thread, re->r.threads);
 	
 	/* assuming no new data gets added to dbase... */
 	R= *re;
@@ -1154,7 +1150,7 @@ static void threaded_tile_processor(Render *re)
 			nextpa= find_next_part(re, minx);
 		}
 		else if(re->r.mode & R_PANORAMA) {
-			if(nextpa==NULL && BLI_available_threads(&threads)==maxthreads)
+			if(nextpa==NULL && BLI_available_threads(&threads)==re->r.threads)
 				nextpa= find_next_pano_slice(re, &minx, &viewplane);
 			else {
 				PIL_sleep_ms(50);
@@ -1195,7 +1191,7 @@ static void threaded_tile_processor(Render *re)
 			drawtimer= 0;
 
 		/* on break, wait for all slots to get freed */
-		if( (g_break=re->test_break()) && BLI_available_threads(&threads)==maxthreads)
+		if( (g_break=re->test_break()) && BLI_available_threads(&threads)==re->r.threads)
 			rendering= 0;
 		
 	}
@@ -1223,10 +1219,8 @@ void RE_TileProcessor(Render *re, int firsttile)
 
 	re->i.starttime= PIL_check_seconds_timer();
 
-	//if(re->r.mode & R_THREADS) 
 	//	threaded_tile_processor(re);
-	//else
-		render_tile_processor(re, firsttile);
+	render_tile_processor(re, firsttile);
 		
 	re->i.lastframetime= PIL_check_seconds_timer()- re->i.starttime;
 	re->stats_draw(&re->i);
diff --git a/source/blender/render/intern/source/ray.c b/source/blender/render/intern/source/ray.c
index 1de850edf83..972e4a8afc8 100644
--- a/source/blender/render/intern/source/ray.c
+++ b/source/blender/render/intern/source/ray.c
@@ -1762,7 +1762,10 @@ void init_jitter_plane(LampRen *lar)
 	float *fp;
 	int x, iter=12, tot= lar->ray_totsamp;
 	
-	fp=lar->jitter= MEM_mallocN(4*tot*2*sizeof(float), "lamp jitter tab");
+	/* at least 4, or max threads+1 tables */
+	if(BLENDER_MAX_THREADS < 4) x= 4;
+	else x= BLENDER_MAX_THREADS+1;
+	fp= lar->jitter= MEM_mallocN(x*tot*2*sizeof(float), "lamp jitter tab");
 	
 	/* set per-lamp fixed seed */
 	BLI_srandom(tot);
@@ -1780,7 +1783,7 @@ void init_jitter_plane(LampRen *lar)
 		}
 	}
 	
-	/* create the dithered tables */
+	/* create the dithered tables (could just check lamp type!) */
 	jitter_plane_offset(lar->jitter, lar->jitter+2*tot, tot, lar->area_size, lar->area_sizey, 0.5, 0.0);
 	jitter_plane_offset(lar->jitter, lar->jitter+4*tot, tot, lar->area_size, lar->area_sizey, 0.5, 0.5);
 	jitter_plane_offset(lar->jitter, lar->jitter+6*tot, tot, lar->area_size, lar->area_sizey, 0.0, 0.5);
@@ -1795,20 +1798,13 @@ static float *give_jitter_plane(LampRen *lar, int thread, int xs, int ys)
 			
 	if(lar->ray_samp_type & LA_SAMP_JITTER) {
 		/* made it threadsafe */
-		if(thread & 1) {
-			if(lar->xold1!=xs || lar->yold1!=ys) {
-				jitter_plane_offset(lar->jitter, lar->jitter+2*tot, tot, lar->area_size, lar->area_sizey, BLI_thread_frand(1), BLI_thread_frand(1));
-				lar->xold1= xs; lar->yold1= ys;
-			}
-			return lar->jitter+2*tot;
-		}
-		else {
-			if(lar->xold2!=xs || lar->yold2!=ys) {
-				jitter_plane_offset(lar->jitter, lar->jitter+4*tot, tot, lar->area_size, lar->area_sizey, BLI_thread_frand(0), BLI_thread_frand(0));
-				lar->xold2= xs; lar->yold2= ys;
-			}
-			return lar->jitter+4*tot;
+		
+		if(lar->xold[thread]!=xs || lar->yold[thread]!=ys) {
+			jitter_plane_offset(lar->jitter, lar->jitter+2*(thread+1)*tot, tot, lar->area_size, lar->area_sizey, BLI_thread_frand(thread), BLI_thread_frand(thread));
+			lar->xold[thread]= xs; 
+			lar->yold[thread]= ys;
 		}
+		return lar->jitter+2*(thread+1)*tot;
 	}
 	if(lar->ray_samp_type & LA_SAMP_DITHER) {
 		return lar->jitter + 2*tot*((xs & 1)+2*(ys & 1));
@@ -2029,45 +2025,52 @@ static void DS_energy(float *sphere, int tot, float *vec)
 
 /* called from convertBlenderScene.c */
 /* creates an equally distributed spherical sample pattern */
-void init_ao_sphere(float *sphere, int tot, int iter)
+/* and allocates threadsafe memory */
+void init_ao_sphere(World *wrld)
 {
 	float *fp;
-	int a;
+	int a, tot, iter= 16;
 
+	/* we make twice the amount of samples, because only a hemisphere is used */
+	tot= 2*wrld->aosamp*wrld->aosamp;
+	
+	wrld->aosphere= MEM_mallocN(3*tot*sizeof(float), "AO sphere");
+	
+	/* fixed random */
 	BLI_srandom(tot);
 	
 	/* init */
-	fp= sphere;
+	fp= wrld->aosphere;
 	for(a=0; a<tot; a++, fp+= 3) {
 		RandomSpherical(fp);
 	}
 	
 	while(iter--) {
-		for(a=0, fp= sphere; a<tot; a++, fp+= 3) {
-			DS_energy(sphere, tot, fp);
+		for(a=0, fp= wrld->aosphere; a<tot; a++, fp+= 3) {
+			DS_energy(wrld->aosphere, tot, fp);
 		}
 	}
+	
+	/* tables */
+	wrld->aotables= MEM_mallocN(BLENDER_MAX_THREADS*3*tot*sizeof(float), "AO tables");
 }
 
-
-static float *threadsafe_table_sphere(int test, int thread, int xs, int ys)
+/* give per thread a table, we have to compare xs ys because of way OSA works... */
+static float *threadsafe_table_sphere(int test, int thread, int xs, int ys, int tot)
 {
-	static float sphere1[2*3*256];
-	static float sphere2[2*3*256];
-	static int xs1=-1, xs2=-1, ys1=-1, ys2=-1;
-	
-	if(thread & 1) {
-		if(xs==xs1 && ys==ys1) return sphere1;
-		if(test) return NULL;
-		xs1= xs; ys1= ys;
-		return sphere1;
-	}
-	else  {
-		if(xs==xs2 && ys==ys2) return sphere2;
-		if(test) return NULL;
-		xs2= xs; ys2= ys;
-		return sphere2;
+	static int xso[BLENDER_MAX_THREADS], yso[BLENDER_MAX_THREADS];
+	static int firsttime= 1;
+	
+	if(firsttime) {
+		memset(xso, 255, sizeof(xso));
+		memset(yso, 255, sizeof(yso));
+		firsttime= 0;
 	}
+	
+	if(xs==xso[thread] && ys==yso[thread]) return R.wrld.aotables+ thread*tot*3;
+	if(test) return NULL;
+	xso[thread]= xs; yso[thread]= ys;
+	return R.wrld.aotables+ thread*tot*3;
 }
 
 static float *sphere_sampler(int type, int resol, int thread, int xs, int ys)
@@ -2097,9 +2100,9 @@ static float *sphere_sampler(int type, int resol, int thread, int xs, int ys)
 		float ang, *vec1;
 		int a;
 		
-		sphere= threadsafe_table_sphere(1, thread, xs, ys);	// returns table if xs and ys were equal to last call
+		sphere= threadsafe_table_sphere(1, thread, xs, ys, tot);	// returns table if xs and ys were equal to last call
 		if(sphere==NULL) {
-			sphere= threadsafe_table_sphere(0, thread, xs, ys);
+			sphere= threadsafe_table_sphere(0, thread, xs, ys, tot);
 			
 			// random rotation
 			ang= BLI_thread_frand(thread);
@@ -2239,7 +2242,7 @@ void ray_shadow(ShadeInput *shi, LampRen *lar, float *shadfac)
 
 	/* only when not mir tracing, first hit optimm */
 	if(shi->depth==0) 
-		isec.vlr_last= lar->vlr_last[shi->thread & 1];
+		isec.vlr_last= lar->vlr_last[shi->thread];
 	else 
 		isec.vlr_last= NULL;
 	
@@ -2352,7 +2355,7 @@ void ray_shadow(ShadeInput *shi, LampRen *lar, float *shadfac)
 
 	/* for first hit optim, set last interesected shadow face */
 	if(shi->depth==0) 
-		lar->vlr_last[shi->thread & 1]= isec.vlr_last;
+		lar->vlr_last[shi->thread]= isec.vlr_last;
 
 }
 
diff --git a/source/blender/render/intern/source/rendercore.c b/source/blender/render/intern/source/rendercore.c
index 6109bd75ea3..23777cea0f2 100644
--- a/source/blender/render/intern/source/rendercore.c
+++ b/source/blender/render/intern/source/rendercore.c
@@ -3315,6 +3315,7 @@ void zbufshadeDA_tile(RenderPart *pa)
 	/* free all */
 	MEM_freeN(pa->rectp); pa->rectp= NULL;
 	MEM_freeN(pa->rectz); pa->rectz= NULL;
+	MEM_freeN(pa->clipflag); pa->clipflag= NULL;
 	
 	/* display active layer */
 	rr->renrect.ymin=rr->renrect.ymax= 0;
@@ -3462,6 +3463,7 @@ void zbufshade_tile(RenderPart *pa)
 	
 	MEM_freeN(pa->rectp); pa->rectp= NULL;
 	MEM_freeN(pa->rectz); pa->rectz= NULL;
+	MEM_freeN(pa->clipflag); pa->clipflag= NULL;
 }
 
 /* ------------------------------------------------------------------------ */
@@ -3900,10 +3902,10 @@ static void *do_bake_thread(void *bs_v)
 /* returns 0 if nothing was handled */
 int RE_bake_shade_all_selected(Render *re, int type)
 {
-	BakeShade handles[RE_MAXTHREAD];
+	BakeShade handles[BLENDER_MAX_THREADS];
 	ListBase threads;
 	Image *ima;
-	int a, vdone=0, maxthreads= 1;
+	int a, vdone=0;
 	
 	/* initialize static vars */
 	get_next_bake_face(NULL);
@@ -3916,14 +3918,10 @@ int RE_bake_shade_all_selected(Render *re, int type)
 	R= *re;
 	R.bakebuf= NULL;
 	
-	if(re->r.mode & R_THREADS) 
-		maxthreads= RE_MAXTHREAD;	/* should become button value too */
-	else maxthreads= 1;
-
-	BLI_init_threads(&threads, do_bake_thread, maxthreads);
+	BLI_init_threads(&threads, do_bake_thread, re->r.threads);
 
 	/* get the threads running */
-	for(a=0; a<maxthreads; a++) {
+	for(a=0; a<re->r.threads; a++) {
 		/* set defaults in handles */
 		memset(&handles[a], 0, sizeof(BakeShade));
 		handles[a].shi.lay= re->scene->lay;
@@ -3935,11 +3933,11 @@ int RE_bake_shade_all_selected(Render *re, int type)
 	
 	/* wait for everything to be done */
 	a= 0;
-	while(a!=maxthreads) {
+	while(a!=re->r.threads) {
 		
 		PIL_sleep_ms(50);
 
-		for(a=0; a<maxthreads; a++)
+		for(a=0; a<re->r.threads; a++)
 			if(handles[a].ready==0)
 				break;
 	}
@@ -3955,7 +3953,7 @@ int RE_bake_shade_all_selected(Render *re, int type)
 	}
 	
 	/* calculate return value */
-	for(a=0; a<maxthreads; a++) {
+	for(a=0; a<re->r.threads; a++) {
 		vdone+= handles[a].vdone;
 		
 		zbuf_free_span(handles[a].zspan);
diff --git a/source/blender/render/intern/source/zbuf.c b/source/blender/render/intern/source/zbuf.c
index b724aeb704f..43119960cce 100644
--- a/source/blender/render/intern/source/zbuf.c
+++ b/source/blender/render/intern/source/zbuf.c
@@ -1685,68 +1685,42 @@ void set_part_zbuf_clipflag(RenderPart *pa)
 {
 	VertRen *ver=NULL;
 	float minx, miny, maxx, maxy, wco;
-	unsigned short clipclear;
 	int v;
+	char *clipflag;
 
+	/* flags stored in part now */
+	clipflag= pa->clipflag= MEM_mallocN(R.totvert+1, "part clipflags");
+	
 	minx= (2*pa->disprect.xmin - R.winx-1)/(float)R.winx;
 	maxx= (2*pa->disprect.xmax - R.winx+1)/(float)R.winx;
 	miny= (2*pa->disprect.ymin - R.winy-1)/(float)R.winy;
 	maxy= (2*pa->disprect.ymax - R.winy+1)/(float)R.winy;
 	
-	/* supports up to 4 threads this way */
-	clipclear= ~(15 << 4*(pa->thread & 3));
-	
-	/* extra security to prevent access to same data */
-	BLI_lock_thread(LOCK_CUSTOM1);
-	
-	for(v=0; v<R.totvert; v++) {
+	for(v=0; v<R.totvert; v++, clipflag++) {
 		if((v & 255)==0)
 			ver= RE_findOrAddVert(&R, v);
 		else ver++;
 		
 		wco= ver->ho[3];
-		ver->flag &= clipclear;
-		
-		switch(pa->thread & 3) {
-			case 0:
-				if( ver->ho[0] > maxx*wco) ver->flag |= 1;
-				else if( ver->ho[0]< minx*wco) ver->flag |= 2;
-				if( ver->ho[1] > maxy*wco) ver->flag |= 4;
-				else if( ver->ho[1]< miny*wco) ver->flag |= 8;
-				break;
-			case 1:
-				if( ver->ho[0] > maxx*wco) ver->flag |= 16;
-				else if( ver->ho[0]< minx*wco) ver->flag |= 32;
-				if( ver->ho[1] > maxy*wco) ver->flag |= 64;
-				else if( ver->ho[1]< miny*wco) ver->flag |= 128;
-				break;
-			case 2:
-				if( ver->ho[0] > maxx*wco) ver->flag |= 256;
-				else if( ver->ho[0]< minx*wco) ver->flag |= 512;
-				if( ver->ho[1] > maxy*wco) ver->flag |= 1024;
-				else if( ver->ho[1]< miny*wco) ver->flag |= 2048;
-				break;
-			case 3:
-				if( ver->ho[0] > maxx*wco) ver->flag |= 4096;
-				else if( ver->ho[0]< minx*wco) ver->flag |= 8192;
-				if( ver->ho[1] > maxy*wco) ver->flag |= 16384;
-				else if( ver->ho[1]< miny*wco) ver->flag |= 32768;
-				break;
-		}
+
+		*clipflag= 0;
+		if( ver->ho[0] > maxx*wco) *clipflag |= 1;
+		else if( ver->ho[0]< minx*wco) *clipflag |= 2;
+		if( ver->ho[1] > maxy*wco) *clipflag |= 4;
+		else if( ver->ho[1]< miny*wco) *clipflag |= 8;
 	}
-	
-	BLI_unlock_thread(LOCK_CUSTOM1);
 }
 
 void zbuffer_solid(RenderPart *pa, unsigned int lay, short layflag)
 {
 	ZSpan zspan;
 	VlakRen *vlr= NULL;
+	VertRen *v1, *v2, *v3, *v4;
 	Material *ma=0;
 	int v, zvlnr;
-	unsigned short clipmask;
 	short nofill=0, env=0, wire=0, all_z= layflag & SCE_LAY_ALL_Z;
-
+	char *clipflag= pa->clipflag;
+	
 	zbuf_alloc_span(&zspan, pa->rectx, pa->recty);
 	
 	/* needed for transform from hoco to zbuffer co */
@@ -1779,9 +1753,6 @@ void zbuffer_solid(RenderPart *pa, unsigned int lay, short layflag)
 	zspan.zbuffunc= zbuffillGL4;
 	zspan.zbuflinefunc= zbufline;
 
-	/* part clipflag, threaded */
-	clipmask= (15 << 4*(pa->thread & 3));
-
 	for(v=0; v<R.totvlak; v++) {
 
 		if((v & 255)==0) vlr= R.blovl[v>>8];
@@ -1813,11 +1784,17 @@ void zbuffer_solid(RenderPart *pa, unsigned int lay, short layflag)
 			if(nofill==0) {
 				unsigned short partclip;
 				
+				v1= vlr->v1;
+				v2= vlr->v2;
+				v3= vlr->v3;
+				v4= vlr->v4;
+				
 				/* partclipping doesn't need viewplane clipping */
-				if(vlr->v4) partclip= vlr->v1->flag & vlr->v2->flag & vlr->v3->flag & vlr->v4->flag;
-				else partclip= vlr->v1->flag & vlr->v2->flag & vlr->v3->flag;
+				partclip= clipflag[v1->index] & clipflag[v2->index] & clipflag[v3->index];
+				if(v4)
+					partclip &= clipflag[v4->index];
 				
-				if((partclip & clipmask)==0) {
+				if(partclip==0) {
 					
 					if(env) zvlnr= -1;
 					else zvlnr= v+1;
@@ -1825,14 +1802,14 @@ void zbuffer_solid(RenderPart *pa, unsigned int lay, short layflag)
 					if(wire) zbufclipwire(&zspan, zvlnr, vlr);
 					else {
 						/* strands allow to be filled in as quad */
-						if(vlr->v4 && (vlr->flag & R_STRAND)) {
-							zbufclip4(&zspan, zvlnr, vlr->v1->ho, vlr->v2->ho, vlr->v3->ho, vlr->v4->ho, vlr->v1->clip, vlr->v2->clip, vlr->v3->clip, vlr->v4->clip);
+						if(v4 && (vlr->flag & R_STRAND)) {
+							zbufclip4(&zspan, zvlnr, v1->ho, v2->ho, v3->ho, v4->ho, v1->clip, v2->clip, v3->clip, v4->clip);
 						}
 						else {
-							zbufclip(&zspan, zvlnr, vlr->v1->ho, vlr->v2->ho, vlr->v3->ho, vlr->v1->clip, vlr->v2->clip, vlr->v3->clip);
-							if(vlr->v4) {
+							zbufclip(&zspan, zvlnr, v1->ho, v2->ho, v3->ho, v1->clip, v2->clip, v3->clip);
+							if(v4) {
 								if(zvlnr>0) zvlnr+= RE_QUAD_OFFS;
-								zbufclip(&zspan, zvlnr, vlr->v1->ho, vlr->v3->ho, vlr->v4->ho, vlr->v1->clip, vlr->v3->clip, vlr->v4->clip);
+								zbufclip(&zspan, zvlnr, v1->ho, v3->ho, v4->ho, v1->clip, v3->clip, v4->clip);
 							}
 						}
 					}
@@ -2561,9 +2538,10 @@ static void zbuffer_abuf(RenderPart *pa, APixstr *APixbuf, ListBase *apsmbase, u
 	ZSpan zspan;
 	Material *ma=NULL;
 	VlakRen *vlr=NULL;
+	VertRen *v1, *v2, *v3, *v4;
 	float vec[3], hoco[4], mul, zval, fval;
 	int v, zvlnr, zsample, dofill= 0;
-	unsigned short clipmask;
+	char *clipflag= pa->clipflag;
 	
 	zbuf_alloc_span(&zspan, pa->rectx, pa->recty);
 	
@@ -2580,9 +2558,6 @@ static void zbuffer_abuf(RenderPart *pa, APixstr *APixbuf, ListBase *apsmbase, u
 	zspan.zbuffunc= zbuffillAc4;
 	zspan.zbuflinefunc= zbuflineAc;
 	
-	/* part clipflag, 4 threads */
-	clipmask= (15 << 4*(pa->thread & 3));
-	
 	for(zsample=0; zsample<R.osa || R.osa==0; zsample++) {
 		
 		copyto_abufz(pa, zspan.arectz, zsample);	/* init zbuffer */
@@ -2618,17 +2593,23 @@ static void zbuffer_abuf(RenderPart *pa, APixstr *APixbuf, ListBase *apsmbase, u
 				if((vlr->flag & R_VISIBLE) && (vlr->lay & lay)) {
 					unsigned short partclip;
 					
+					v1= vlr->v1;
+					v2= vlr->v2;
+					v3= vlr->v3;
+					v4= vlr->v4;
+					
 					/* partclipping doesn't need viewplane clipping */
-					if(vlr->v4) partclip= vlr->v1->flag & vlr->v2->flag & vlr->v3->flag & vlr->v4->flag;
-					else partclip= vlr->v1->flag & vlr->v2->flag & vlr->v3->flag;
+					partclip= clipflag[v1->index] & clipflag[v2->index] & clipflag[v3->index];
+					if(v4)
+						partclip &= clipflag[v4->index];
 					
-					if((partclip & clipmask)==0) {
+					if(partclip==0) {
 						/* a little advantage for transp rendering (a z offset) */
 						if( ma->zoffs != 0.0) {
 							mul= 0x7FFFFFFF;
-							zval= mul*(1.0+vlr->v1->ho[2]/vlr->v1->ho[3]);
+							zval= mul*(1.0+v1->ho[2]/v1->ho[3]);
 
-							VECCOPY(vec, vlr->v1->co);
+							VECCOPY(vec, v1->co);
 							/* z is negative, otherwise its being clipped */ 
 							vec[2]-= ma->zoffs;
 							projectverto(vec, R.winmat, hoco);
@@ -2642,14 +2623,14 @@ static void zbuffer_abuf(RenderPart *pa, APixstr *APixbuf, ListBase *apsmbase, u
 			
 						if(ma->mode & (MA_WIRE)) zbufclipwire(&zspan, zvlnr, vlr);
 						else {
-							if(vlr->v4 && (vlr->flag & R_STRAND)) {
-								zbufclip4(&zspan, zvlnr, vlr->v1->ho, vlr->v2->ho, vlr->v3->ho, vlr->v4->ho, vlr->v1->clip, vlr->v2->clip, vlr->v3->clip, vlr->v4->clip);
+							if(v4 && (vlr->flag & R_STRAND)) {
+								zbufclip4(&zspan, zvlnr, v1->ho, v2->ho, v3->ho, v4->ho, v1->clip, v2->clip, v3->clip, v4->clip);
 							}
 							else {
-								zbufclip(&zspan, zvlnr, vlr->v1->ho, vlr->v2->ho, vlr->v3->ho, vlr->v1->clip, vlr->v2->clip, vlr->v3->clip);
-								if(vlr->v4) {
+								zbufclip(&zspan, zvlnr, v1->ho, v2->ho, v3->ho, v1->clip, v2->clip, v3->clip);
+								if(v4) {
 									zvlnr+= RE_QUAD_OFFS;
-									zbufclip(&zspan, zvlnr, vlr->v1->ho, vlr->v3->ho, vlr->v4->ho, vlr->v1->clip, vlr->v3->clip, vlr->v4->clip);
+									zbufclip(&zspan, zvlnr, v1->ho, v3->ho, v4->ho, v1->clip, v3->clip, v4->clip);
 								}
 							}
 						}