Moved vectorblur code from render to compositor
authorJeroen Bakker <j.bakker@atmind.nl>
Thu, 12 Apr 2018 08:27:38 +0000 (10:27 +0200)
committerJeroen Bakker <j.bakker@atmind.nl>
Thu, 12 Apr 2018 08:27:38 +0000 (10:27 +0200)
In preparation of the removal of blender internal render we
moved the vectorblur code that was placed in the render package
(legacy) to the compositor. The compositor is only using this
code even the blender internal renderer did not use the code at
all.

source/blender/compositor/operations/COM_RenderLayersProg.h
source/blender/compositor/operations/COM_VectorBlurOperation.cpp
source/blender/render/extern/include/RE_pipeline.h
source/blender/render/extern/include/RE_render_ext.h
source/blender/render/intern/source/zbuf.c
source/blenderplayer/bad_level_call_stubs/stubs.c

index 1be159067701fa872d2b4d9d6d9b948b4bffd94c..7cfead7e98c906ba69e0c73014242d8c8e6142c1 100644 (file)
@@ -30,8 +30,6 @@
 #include "BKE_image.h"
 extern "C" {
 #  include "RE_pipeline.h"
-#  include "RE_shader_ext.h"
-#  include "RE_render_ext.h"
 #  include "MEM_guardedalloc.h"
 }
 
@@ -46,7 +44,7 @@ protected:
         * Reference to the scene object.
         */
        Scene *m_scene;
-       
+
        /**
         * layerId of the layer where this operation needs to get its data from
         */
@@ -61,24 +59,24 @@ protected:
         * cached instance to the float buffer inside the layer
         */
        float *m_inputBuffer;
-       
+
        /**
         * renderpass where this operation needs to get its data from
         */
        std::string m_passName;
-       
+
        int m_elementsize;
 
        /**
         * @brief render data used for active rendering
         */
        const RenderData *m_rd;
-       
+
        /**
         * Determine the output resolution. The resolution is retrieved from the Renderer
         */
        void determineResolution(unsigned int resolution[2], unsigned int preferredResolution[2]);
-       
+
        /**
         * retrieve the reference to the float buffer of the renderer.
         */
index 36f06e9243626e3fd2ed844f0c2820f7c4490eea..bccfe02508987a137e9a986d910c1461ce594d16 100644 (file)
  *             Monique Dewanchand
  */
 
-#include "COM_VectorBlurOperation.h"
-#include "BLI_math.h"
-
+#include <string.h>
 #include "MEM_guardedalloc.h"
-
-// use the implementation of blender internal renderer to calculate the vector blur.
+#include "BLI_math.h"
 extern "C" {
-#  include "RE_pipeline.h"
+       #include "BLI_jitter_2d.h"
 }
+#include "COM_VectorBlurOperation.h"
 
+
+/* Defined */
+#define PASS_VECTOR_MAX        10000.0f
+
+/* Forward declarations */
+struct ZSpan;
+struct DrawBufPixel;
+void zbuf_accumulate_vecblur(
+               NodeBlurData *nbd, int xsize, int ysize, float *newrect,
+               const float *imgrect, float *vecbufrect, const float *zbufrect);
+void zbuf_alloc_span(ZSpan *zspan, int rectx, int recty, float clipcrop);
+void zbuf_free_span(ZSpan *zspan);
+void antialias_tagbuf(int xsize, int ysize, char *rectmove);
+
+
+/* VectorBlurOperation */
 VectorBlurOperation::VectorBlurOperation() : NodeOperation()
 {
        this->addInputSocket(COM_DT_COLOR);
@@ -51,7 +65,7 @@ void VectorBlurOperation::initExecution()
        this->m_inputSpeedProgram = getInputSocketReader(2);
        this->m_cachedInstance = NULL;
        QualityStepHelper::initExecution(COM_QH_INCREASE);
-       
+
 }
 
 void VectorBlurOperation::executePixel(float output[4], int x, int y, void *data)
@@ -77,7 +91,7 @@ void *VectorBlurOperation::initializeTileData(rcti *rect)
        if (this->m_cachedInstance) {
                return this->m_cachedInstance;
        }
-       
+
        lockMutex();
        if (this->m_cachedInstance == NULL) {
                MemoryBuffer *tile = (MemoryBuffer *)this->m_inputImageProgram->initializeTileData(rect);
@@ -114,6 +128,690 @@ void VectorBlurOperation::generateVectorBlur(float *data, MemoryBuffer *inputIma
        blurdata.minspeed = this->m_settings->minspeed;
        blurdata.curved = this->m_settings->curved;
        blurdata.fac = this->m_settings->fac;
-       RE_zbuf_accumulate_vecblur(&blurdata, this->getWidth(), this->getHeight(), data, inputImage->getBuffer(), inputSpeed->getBuffer(), inputZ->getBuffer());
+       zbuf_accumulate_vecblur(&blurdata, this->getWidth(), this->getHeight(), data, inputImage->getBuffer(), inputSpeed->getBuffer(), inputZ->getBuffer());
        return;
 }
+
+/* ****************** Spans ******************************* */
+/* span fill in method, is also used to localize data for zbuffering */
+typedef struct ZSpan {
+       /* range for clipping */
+       int rectx, recty;
+
+       /* actual filled in range */
+       int miny1, maxy1, miny2, maxy2;
+       /* vertex pointers detect min/max range in */
+       const float *minp1, *maxp1, *minp2, *maxp2;
+       float *span1, *span2;
+
+       /* transform from hoco to zbuf co */
+       float zmulx, zmuly, zofsx, zofsy;
+
+       int *rectz;
+       DrawBufPixel* rectdraw;
+       float clipcrop;
+
+} ZSpan;
+
+/* each zbuffer has coordinates transformed to local rect coordinates, so we can simply clip */
+void zbuf_alloc_span(ZSpan *zspan, int rectx, int recty, float clipcrop)
+{
+       memset(zspan, 0, sizeof(ZSpan));
+
+       zspan->rectx= rectx;
+       zspan->recty= recty;
+
+       zspan->span1= (float*)MEM_mallocN(recty*sizeof(float), "zspan");
+       zspan->span2= (float*)MEM_mallocN(recty*sizeof(float), "zspan");
+
+       zspan->clipcrop= clipcrop;
+}
+
+void zbuf_free_span(ZSpan *zspan)
+{
+       if (zspan) {
+               if (zspan->span1) MEM_freeN(zspan->span1);
+               if (zspan->span2) MEM_freeN(zspan->span2);
+               zspan->span1= zspan->span2= NULL;
+       }
+}
+
+/* reset range for clipping */
+static void zbuf_init_span(ZSpan *zspan)
+{
+       zspan->miny1= zspan->miny2= zspan->recty+1;
+       zspan->maxy1= zspan->maxy2= -1;
+       zspan->minp1= zspan->maxp1= zspan->minp2= zspan->maxp2= NULL;
+}
+
+static void zbuf_add_to_span(ZSpan *zspan, const float v1[2], const float v2[2])
+{
+       const float *minv, *maxv;
+       float *span;
+       float xx1, dx0, xs0;
+       int y, my0, my2;
+
+       if (v1[1]<v2[1]) {
+               minv= v1; maxv= v2;
+       }
+       else {
+               minv= v2; maxv= v1;
+       }
+
+       my0= ceil(minv[1]);
+       my2= floor(maxv[1]);
+
+       if (my2<0 || my0>= zspan->recty) return;
+
+       /* clip top */
+       if (my2>=zspan->recty) my2= zspan->recty-1;
+       /* clip bottom */
+       if (my0<0) my0= 0;
+
+       if (my0>my2) return;
+       /* if (my0>my2) should still fill in, that way we get spans that skip nicely */
+
+       xx1= maxv[1]-minv[1];
+       if (xx1>FLT_EPSILON) {
+               dx0= (minv[0]-maxv[0])/xx1;
+               xs0= dx0*(minv[1]-my2) + minv[0];
+       }
+       else {
+               dx0 = 0.0f;
+               xs0 = min_ff(minv[0], maxv[0]);
+       }
+
+       /* empty span */
+       if (zspan->maxp1 == NULL) {
+               span= zspan->span1;
+       }
+       else {  /* does it complete left span? */
+               if ( maxv == zspan->minp1 || minv==zspan->maxp1) {
+                       span= zspan->span1;
+               }
+               else {
+                       span= zspan->span2;
+               }
+       }
+
+       if (span==zspan->span1) {
+//             printf("left span my0 %d my2 %d\n", my0, my2);
+               if (zspan->minp1==NULL || zspan->minp1[1] > minv[1] ) {
+                       zspan->minp1= minv;
+               }
+               if (zspan->maxp1==NULL || zspan->maxp1[1] < maxv[1] ) {
+                       zspan->maxp1= maxv;
+               }
+               if (my0<zspan->miny1) zspan->miny1= my0;
+               if (my2>zspan->maxy1) zspan->maxy1= my2;
+       }
+       else {
+//             printf("right span my0 %d my2 %d\n", my0, my2);
+               if (zspan->minp2==NULL || zspan->minp2[1] > minv[1] ) {
+                       zspan->minp2= minv;
+               }
+               if (zspan->maxp2==NULL || zspan->maxp2[1] < maxv[1] ) {
+                       zspan->maxp2= maxv;
+               }
+               if (my0<zspan->miny2) zspan->miny2= my0;
+               if (my2>zspan->maxy2) zspan->maxy2= my2;
+       }
+
+       for (y=my2; y>=my0; y--, xs0+= dx0) {
+               /* xs0 is the xcoord! */
+               span[y]= xs0;
+       }
+}
+
+/* ******************** VECBLUR ACCUM BUF ************************* */
+
+typedef struct DrawBufPixel {
+       const float *colpoin;
+       float alpha;
+} DrawBufPixel;
+
+
+static void zbuf_fill_in_rgba(ZSpan *zspan, DrawBufPixel *col, float *v1, float *v2, float *v3, float *v4)
+{
+       DrawBufPixel *rectpofs, *rp;
+       double zxd, zyd, zy0, zverg;
+       float x0, y0, z0;
+       float x1, y1, z1, x2, y2, z2, xx1;
+       const float *span1, *span2;
+       float *rectzofs, *rz;
+       int x, y;
+       int sn1, sn2, rectx, my0, my2;
+
+       /* init */
+       zbuf_init_span(zspan);
+
+       /* set spans */
+       zbuf_add_to_span(zspan, v1, v2);
+       zbuf_add_to_span(zspan, v2, v3);
+       zbuf_add_to_span(zspan, v3, v4);
+       zbuf_add_to_span(zspan, v4, v1);
+
+       /* clipped */
+       if (zspan->minp2==NULL || zspan->maxp2==NULL) return;
+
+       my0 = max_ii(zspan->miny1, zspan->miny2);
+       my2 = min_ii(zspan->maxy1, zspan->maxy2);
+
+       //      printf("my %d %d\n", my0, my2);
+       if (my2<my0) return;
+
+       /* ZBUF DX DY, in floats still */
+       x1= v1[0]- v2[0];
+       x2= v2[0]- v3[0];
+       y1= v1[1]- v2[1];
+       y2= v2[1]- v3[1];
+       z1= v1[2]- v2[2];
+       z2= v2[2]- v3[2];
+       x0= y1*z2-z1*y2;
+       y0= z1*x2-x1*z2;
+       z0= x1*y2-y1*x2;
+
+       if (z0==0.0f) return;
+
+       xx1= (x0*v1[0] + y0*v1[1])/z0 + v1[2];
+
+       zxd= -(double)x0/(double)z0;
+       zyd= -(double)y0/(double)z0;
+       zy0= ((double)my2)*zyd + (double)xx1;
+
+       /* start-offset in rect */
+       rectx= zspan->rectx;
+       rectzofs= (float *)(zspan->rectz + rectx*my2);
+       rectpofs= ((DrawBufPixel *)zspan->rectdraw) + rectx*my2;
+
+       /* correct span */
+       sn1= (my0 + my2)/2;
+       if (zspan->span1[sn1] < zspan->span2[sn1]) {
+               span1= zspan->span1+my2;
+               span2= zspan->span2+my2;
+       }
+       else {
+               span1= zspan->span2+my2;
+               span2= zspan->span1+my2;
+       }
+
+       for (y=my2; y>=my0; y--, span1--, span2--) {
+
+               sn1= floor(*span1);
+               sn2= floor(*span2);
+               sn1++;
+
+               if (sn2>=rectx) sn2= rectx-1;
+               if (sn1<0) sn1= 0;
+
+               if (sn2>=sn1) {
+                       zverg= (double)sn1*zxd + zy0;
+                       rz= rectzofs+sn1;
+                       rp= rectpofs+sn1;
+                       x= sn2-sn1;
+
+                       while (x>=0) {
+                               if (zverg < (double)*rz) {
+                                       *rz= zverg;
+                                       *rp= *col;
+                               }
+                               zverg+= zxd;
+                               rz++;
+                               rp++;
+                               x--;
+                       }
+               }
+
+               zy0-=zyd;
+               rectzofs-= rectx;
+               rectpofs-= rectx;
+       }
+}
+
+/* char value==255 is filled in, rest should be zero */
+/* returns alpha values, but sets alpha to 1 for zero alpha pixels that have an alpha value as neighbor */
+void antialias_tagbuf(int xsize, int ysize, char *rectmove)
+{
+       char *row1, *row2, *row3;
+       char prev, next;
+       int a, x, y, step;
+
+       /* 1: tag pixels to be candidate for AA */
+       for (y=2; y<ysize; y++) {
+               /* setup rows */
+               row1= rectmove + (y-2)*xsize;
+               row2= row1 + xsize;
+               row3= row2 + xsize;
+               for (x=2; x<xsize; x++, row1++, row2++, row3++) {
+                       if (row2[1]) {
+                               if (row2[0]==0 || row2[2]==0 || row1[1]==0 || row3[1]==0)
+                                       row2[1]= 128;
+                       }
+               }
+       }
+
+       /* 2: evaluate horizontal scanlines and calculate alphas */
+       row1= rectmove;
+       for (y=0; y<ysize; y++) {
+               row1++;
+               for (x=1; x<xsize; x++, row1++) {
+                       if (row1[0]==128 && row1[1]==128) {
+                               /* find previous color and next color and amount of steps to blend */
+                               prev= row1[-1];
+                               step= 1;
+                               while (x+step<xsize && row1[step]==128)
+                                       step++;
+
+                               if (x+step!=xsize) {
+                                       /* now we can blend values */
+                                       next= row1[step];
+
+                                       /* note, prev value can be next value, but we do this loop to clear 128 then */
+                                       for (a=0; a<step; a++) {
+                                               int fac, mfac;
+
+                                               fac= ((a+1)<<8)/(step+1);
+                                               mfac= 255-fac;
+
+                                               row1[a]= (prev*mfac + next*fac)>>8;
+                                       }
+                               }
+                       }
+               }
+       }
+
+       /* 3: evaluate vertical scanlines and calculate alphas */
+       /*    use for reading a copy of the original tagged buffer */
+       for (x=0; x<xsize; x++) {
+               row1= rectmove + x+xsize;
+
+               for (y=1; y<ysize; y++, row1+=xsize) {
+                       if (row1[0]==128 && row1[xsize]==128) {
+                               /* find previous color and next color and amount of steps to blend */
+                               prev= row1[-xsize];
+                               step= 1;
+                               while (y+step<ysize && row1[step*xsize]==128)
+                                       step++;
+
+                               if (y+step!=ysize) {
+                                       /* now we can blend values */
+                                       next= row1[step*xsize];
+                                       /* note, prev value can be next value, but we do this loop to clear 128 then */
+                                       for (a=0; a<step; a++) {
+                                               int fac, mfac;
+
+                                               fac= ((a+1)<<8)/(step+1);
+                                               mfac= 255-fac;
+
+                                               row1[a*xsize]= (prev*mfac + next*fac)>>8;
+                                       }
+                               }
+                       }
+               }
+       }
+
+       /* last: pixels with 0 we fill in zbuffer, with 1 we skip for mask */
+       for (y=2; y<ysize; y++) {
+               /* setup rows */
+               row1= rectmove + (y-2)*xsize;
+               row2= row1 + xsize;
+               row3= row2 + xsize;
+               for (x=2; x<xsize; x++, row1++, row2++, row3++) {
+                       if (row2[1]==0) {
+                               if (row2[0]>1 || row2[2]>1 || row1[1]>1 || row3[1]>1)
+                                       row2[1]= 1;
+                       }
+               }
+       }
+}
+
+/* in: two vectors, first vector points from origin back in time, 2nd vector points to future */
+/* we make this into 3 points, center point is (0, 0) */
+/* and offset the center point just enough to make curve go through midpoint */
+
+static void quad_bezier_2d(float *result, float *v1, float *v2, float *ipodata)
+{
+       float p1[2], p2[2], p3[2];
+
+       p3[0]= -v2[0];
+       p3[1]= -v2[1];
+
+       p1[0]= v1[0];
+       p1[1]= v1[1];
+
+       /* official formula 2*p2 - 0.5*p1 - 0.5*p3 */
+       p2[0]= -0.5f*p1[0] - 0.5f*p3[0];
+       p2[1]= -0.5f*p1[1] - 0.5f*p3[1];
+
+       result[0]= ipodata[0]*p1[0] + ipodata[1]*p2[0] + ipodata[2]*p3[0];
+       result[1]= ipodata[0]*p1[1] + ipodata[1]*p2[1] + ipodata[2]*p3[1];
+}
+
+static void set_quad_bezier_ipo(float fac, float *data)
+{
+       float mfac= (1.0f-fac);
+
+       data[0]= mfac*mfac;
+       data[1]= 2.0f*mfac*fac;
+       data[2]= fac*fac;
+}
+
+void zbuf_accumulate_vecblur(
+        NodeBlurData *nbd, int xsize, int ysize, float *newrect,
+        const float *imgrect, float *vecbufrect, const float *zbufrect)
+{
+       ZSpan zspan;
+       DrawBufPixel *rectdraw, *dr;
+       static float jit[256][2];
+       float v1[3], v2[3], v3[3], v4[3], fx, fy;
+       const float *dimg, *dz, *ro;
+       float *rectvz, *dvz, *dvec1, *dvec2, *dz1, *dz2, *rectz;
+       float *minvecbufrect= NULL, *rectweight, *rw, *rectmax, *rm;
+       float maxspeedsq= (float)nbd->maxspeed*nbd->maxspeed;
+       int y, x, step, maxspeed=nbd->maxspeed, samples= nbd->samples;
+       int tsktsk= 0;
+       static int firsttime= 1;
+       char *rectmove, *dm;
+
+       zbuf_alloc_span(&zspan, xsize, ysize, 1.0f);
+       zspan.zmulx=  ((float)xsize)/2.0f;
+       zspan.zmuly=  ((float)ysize)/2.0f;
+       zspan.zofsx= 0.0f;
+       zspan.zofsy= 0.0f;
+
+       /* the buffers */
+       rectz= (float*)MEM_mapallocN(sizeof(float)*xsize*ysize, "zbuf accum");
+       zspan.rectz= (int *)rectz;
+
+       rectmove= (char*)MEM_mapallocN(xsize*ysize, "rectmove");
+       rectdraw= (DrawBufPixel*)MEM_mapallocN(sizeof(DrawBufPixel)*xsize*ysize, "rect draw");
+       zspan.rectdraw= rectdraw;
+
+       rectweight= (float*)MEM_mapallocN(sizeof(float)*xsize*ysize, "rect weight");
+       rectmax= (float*)MEM_mapallocN(sizeof(float)*xsize*ysize, "rect max");
+
+       /* debug... check if PASS_VECTOR_MAX still is in buffers */
+       dvec1= vecbufrect;
+       for (x= 4*xsize*ysize; x>0; x--, dvec1++) {
+               if (dvec1[0]==PASS_VECTOR_MAX) {
+                       dvec1[0]= 0.0f;
+                       tsktsk= 1;
+               }
+       }
+       if (tsktsk) printf("Found uninitialized speed in vector buffer... fixed.\n");
+
+       /* min speed? then copy speedbuffer to recalculate speed vectors */
+       if (nbd->minspeed) {
+               float minspeed= (float)nbd->minspeed;
+               float minspeedsq= minspeed*minspeed;
+
+               minvecbufrect= (float*)MEM_mapallocN(4*sizeof(float)*xsize*ysize, "minspeed buf");
+
+               dvec1= vecbufrect;
+               dvec2= minvecbufrect;
+               for (x= 2*xsize*ysize; x>0; x--, dvec1+=2, dvec2+=2) {
+                       if (dvec1[0]==0.0f && dvec1[1]==0.0f) {
+                               dvec2[0]= dvec1[0];
+                               dvec2[1]= dvec1[1];
+                       }
+                       else {
+                               float speedsq= dvec1[0]*dvec1[0] + dvec1[1]*dvec1[1];
+                               if (speedsq <= minspeedsq) {
+                                       dvec2[0]= 0.0f;
+                                       dvec2[1]= 0.0f;
+                               }
+                               else {
+                                       speedsq = 1.0f - minspeed / sqrtf(speedsq);
+                                       dvec2[0]= speedsq*dvec1[0];
+                                       dvec2[1]= speedsq*dvec1[1];
+                               }
+                       }
+               }
+               SWAP(float *, minvecbufrect, vecbufrect);
+       }
+
+       /* make vertex buffer with averaged speed and zvalues */
+       rectvz= (float*)MEM_mapallocN(4*sizeof(float)*(xsize+1)*(ysize+1), "vertices");
+       dvz= rectvz;
+       for (y=0; y<=ysize; y++) {
+
+               if (y==0)
+                       dvec1= vecbufrect + 4*y*xsize;
+               else
+                       dvec1= vecbufrect + 4*(y-1)*xsize;
+
+               if (y==ysize)
+                       dvec2= vecbufrect + 4*(y-1)*xsize;
+               else
+                       dvec2= vecbufrect + 4*y*xsize;
+
+               for (x=0; x<=xsize; x++) {
+
+                       /* two vectors, so a step loop */
+                       for (step=0; step<2; step++, dvec1+=2, dvec2+=2, dvz+=2) {
+                               /* average on minimal speed */
+                               int div= 0;
+
+                               if (x!=0) {
+                                       if (dvec1[-4]!=0.0f || dvec1[-3]!=0.0f) {
+                                               dvz[0]= dvec1[-4];
+                                               dvz[1]= dvec1[-3];
+                                               div++;
+                                       }
+                                       if (dvec2[-4]!=0.0f || dvec2[-3]!=0.0f) {
+                                               if (div==0) {
+                                                       dvz[0]= dvec2[-4];
+                                                       dvz[1]= dvec2[-3];
+                                                       div++;
+                                               }
+                                               else if ( (ABS(dvec2[-4]) + ABS(dvec2[-3]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
+                                                       dvz[0]= dvec2[-4];
+                                                       dvz[1]= dvec2[-3];
+                                               }
+                                       }
+                               }
+
+                               if (x!=xsize) {
+                                       if (dvec1[0]!=0.0f || dvec1[1]!=0.0f) {
+                                               if (div==0) {
+                                                       dvz[0]= dvec1[0];
+                                                       dvz[1]= dvec1[1];
+                                                       div++;
+                                               }
+                                               else if ( (ABS(dvec1[0]) + ABS(dvec1[1]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
+                                                       dvz[0]= dvec1[0];
+                                                       dvz[1]= dvec1[1];
+                                               }
+                                       }
+                                       if (dvec2[0]!=0.0f || dvec2[1]!=0.0f) {
+                                               if (div==0) {
+                                                       dvz[0]= dvec2[0];
+                                                       dvz[1]= dvec2[1];
+                                               }
+                                               else if ( (ABS(dvec2[0]) + ABS(dvec2[1]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
+                                                       dvz[0]= dvec2[0];
+                                                       dvz[1]= dvec2[1];
+                                               }
+                                       }
+                               }
+                               if (maxspeed) {
+                                       float speedsq= dvz[0]*dvz[0] + dvz[1]*dvz[1];
+                                       if (speedsq > maxspeedsq) {
+                                               speedsq = (float)maxspeed / sqrtf(speedsq);
+                                               dvz[0]*= speedsq;
+                                               dvz[1]*= speedsq;
+                                       }
+                               }
+                       }
+               }
+       }
+
+       /* set border speeds to keep border speeds on border */
+       dz1= rectvz;
+       dz2= rectvz+4*(ysize)*(xsize+1);
+       for (x=0; x<=xsize; x++, dz1+=4, dz2+=4) {
+               dz1[1]= 0.0f;
+               dz2[1]= 0.0f;
+               dz1[3]= 0.0f;
+               dz2[3]= 0.0f;
+       }
+       dz1= rectvz;
+       dz2= rectvz+4*(xsize);
+       for (y=0; y<=ysize; y++, dz1+=4*(xsize+1), dz2+=4*(xsize+1)) {
+               dz1[0]= 0.0f;
+               dz2[0]= 0.0f;
+               dz1[2]= 0.0f;
+               dz2[2]= 0.0f;
+       }
+
+       /* tag moving pixels, only these faces we draw */
+       dm= rectmove;
+       dvec1= vecbufrect;
+       for (x=xsize*ysize; x>0; x--, dm++, dvec1+=4) {
+               if ((dvec1[0]!=0.0f || dvec1[1]!=0.0f || dvec1[2]!=0.0f || dvec1[3]!=0.0f))
+                       *dm= 255;
+       }
+
+       antialias_tagbuf(xsize, ysize, rectmove);
+
+       /* has to become static, the init-jit calls a random-seed, screwing up texture noise node */
+       if (firsttime) {
+               firsttime= 0;
+               BLI_jitter_init(jit, 256);
+       }
+
+       memset(newrect, 0, sizeof(float)*xsize*ysize*4);
+
+       /* accumulate */
+       samples/= 2;
+       for (step= 1; step<=samples; step++) {
+               float speedfac= 0.5f*nbd->fac*(float)step/(float)(samples+1);
+               int side;
+
+               for (side=0; side<2; side++) {
+                       float blendfac, ipodata[4];
+
+                       /* clear zbuf, if we draw future we fill in not moving pixels */
+                       if (0)
+                               for (x= xsize*ysize-1; x>=0; x--) rectz[x]= 10e16;
+                       else
+                               for (x= xsize*ysize-1; x>=0; x--) {
+                                       if (rectmove[x]==0)
+                                               rectz[x]= zbufrect[x];
+                                       else
+                                               rectz[x]= 10e16;
+                               }
+
+                       /* clear drawing buffer */
+                       for (x= xsize*ysize-1; x>=0; x--) rectdraw[x].colpoin= NULL;
+
+                       dimg= imgrect;
+                       dm= rectmove;
+                       dz= zbufrect;
+                       dz1= rectvz;
+                       dz2= rectvz + 4*(xsize + 1);
+
+                       if (side) {
+                               if (nbd->curved==0) {
+                                       dz1+= 2;
+                                       dz2+= 2;
+                               }
+                               speedfac= -speedfac;
+                       }
+
+                       set_quad_bezier_ipo(0.5f + 0.5f*speedfac, ipodata);
+
+                       for (fy= -0.5f+jit[step & 255][0], y=0; y<ysize; y++, fy+=1.0f) {
+                               for (fx= -0.5f+jit[step & 255][1], x=0; x<xsize; x++, fx+=1.0f, dimg+=4, dz1+=4, dz2+=4, dm++, dz++) {
+                                       if (*dm>1) {
+                                               float jfx = fx + 0.5f;
+                                               float jfy = fy + 0.5f;
+                                               DrawBufPixel col;
+
+                                               /* make vertices */
+                                               if (nbd->curved) {      /* curved */
+                                                       quad_bezier_2d(v1, dz1, dz1+2, ipodata);
+                                                       v1[0]+= jfx; v1[1]+= jfy; v1[2]= *dz;
+
+                                                       quad_bezier_2d(v2, dz1+4, dz1+4+2, ipodata);
+                                                       v2[0]+= jfx+1.0f; v2[1]+= jfy; v2[2]= *dz;
+
+                                                       quad_bezier_2d(v3, dz2+4, dz2+4+2, ipodata);
+                                                       v3[0]+= jfx+1.0f; v3[1]+= jfy+1.0f; v3[2]= *dz;
+
+                                                       quad_bezier_2d(v4, dz2, dz2+2, ipodata);
+                                                       v4[0]+= jfx; v4[1]+= jfy+1.0f; v4[2]= *dz;
+                                               }
+                                               else {
+                                                       v1[0]= speedfac*dz1[0]+jfx;      v1[1]= speedfac*dz1[1]+jfy;      v1[2]= *dz;
+                                                       v2[0]= speedfac*dz1[4]+jfx+1.0f; v2[1]= speedfac*dz1[5]+jfy;      v2[2]= *dz;
+                                                       v3[0]= speedfac*dz2[4]+jfx+1.0f; v3[1]= speedfac*dz2[5]+jfy+1.0f; v3[2]= *dz;
+                                                       v4[0]= speedfac*dz2[0]+jfx;      v4[1]= speedfac*dz2[1]+jfy+1.0f; v4[2]= *dz;
+                                               }
+                                               if (*dm==255) col.alpha= 1.0f;
+                                               else if (*dm<2) col.alpha= 0.0f;
+                                               else col.alpha= ((float)*dm)/255.0f;
+                                               col.colpoin= dimg;
+
+                                               zbuf_fill_in_rgba(&zspan, &col, v1, v2, v3, v4);
+                                       }
+                               }
+                               dz1+=4;
+                               dz2+=4;
+                       }
+
+                       /* blend with a falloff. this fixes the ugly effect you get with
+                        * a fast moving object. then it looks like a solid object overlayed
+                        * over a very transparent moving version of itself. in reality, the
+                        * whole object should become transparent if it is moving fast, be
+                        * we don't know what is behind it so we don't do that. this hack
+                        * overestimates the contribution of foreground pixels but looks a
+                        * bit better without a sudden cutoff. */
+                       blendfac= ((samples - step)/(float)samples);
+                       /* smoothstep to make it look a bit nicer as well */
+                       blendfac= 3.0f*pow(blendfac, 2.0f) - 2.0f*pow(blendfac, 3.0f);
+
+                       /* accum */
+                       rw= rectweight;
+                       rm= rectmax;
+                       for (dr= rectdraw, dz2=newrect, x= xsize*ysize-1; x>=0; x--, dr++, dz2+=4, rw++, rm++) {
+                               if (dr->colpoin) {
+                                       float bfac= dr->alpha*blendfac;
+
+                                       dz2[0] += bfac*dr->colpoin[0];
+                                       dz2[1] += bfac*dr->colpoin[1];
+                                       dz2[2] += bfac*dr->colpoin[2];
+                                       dz2[3] += bfac*dr->colpoin[3];
+
+                                       *rw += bfac;
+                                       *rm= MAX2(*rm, bfac);
+                               }
+                       }
+               }
+       }
+
+       /* blend between original images and accumulated image */
+       rw= rectweight;
+       rm= rectmax;
+       ro= imgrect;
+       dm= rectmove;
+       for (dz2=newrect, x= xsize*ysize-1; x>=0; x--, dz2+=4, ro+=4, rw++, rm++, dm++) {
+               float mfac = *rm;
+               float fac = (*rw == 0.0f)? 0.0f: mfac/(*rw);
+               float nfac = 1.0f - mfac;
+
+               dz2[0]= fac*dz2[0] + nfac*ro[0];
+               dz2[1]= fac*dz2[1] + nfac*ro[1];
+               dz2[2]= fac*dz2[2] + nfac*ro[2];
+               dz2[3]= fac*dz2[3] + nfac*ro[3];
+       }
+
+       MEM_freeN(rectz);
+       MEM_freeN(rectmove);
+       MEM_freeN(rectdraw);
+       MEM_freeN(rectvz);
+       MEM_freeN(rectweight);
+       MEM_freeN(rectmax);
+       if (minvecbufrect) MEM_freeN(vecbufrect);  /* rects were swapped! */
+       zbuf_free_span(&zspan);
+}
index de71b8d10236f17e3c685ccc2dd9b9e0b43343f0..8065de831ba2b1a7f3b2902843f2367cf2a7c6f6 100644 (file)
@@ -336,10 +336,6 @@ void RE_current_scene_update_cb(struct Render *re, void *handle, void (*f)(void
 
 /* should move to kernel once... still unsure on how/where */
 float RE_filter_value(int type, float x);
-/* vector blur zbuffer method */
-void RE_zbuf_accumulate_vecblur(
-        struct NodeBlurData *nbd, int xsize, int ysize, float *newrect,
-        const float *imgrect, float *vecbufrect, const float *zbufrect);
 
 int RE_seq_render_active(struct Scene *scene, struct RenderData *rd);
 
index 1779ccc5d19b9bd5f7f9e0e167638e0235730e77..971be381f970d0973a8d1d653888964c370b7568 100644 (file)
@@ -62,9 +62,6 @@ void RE_sample_material_color(
 /* imagetexture.c */
 void ibuf_sample(struct ImBuf *ibuf, float fx, float fy, float dx, float dy, float result[4]);
 
-/* zbuf.c */
-void antialias_tagbuf(int xsize, int ysize, char *rectmove);
-
 /* pointdensity.c */
 struct PointDensity;
 
@@ -89,4 +86,3 @@ void RE_point_density_sample(
 void RE_point_density_free(struct PointDensity *pd);
 
 #endif /* __RE_RENDER_EXT_H__ */
-
index 14184f4404bbeadb577eeb8a84c9ba5f65cd458e..aa073ad9a83530029b0108d54ae669dd00626a0b 100644 (file)
@@ -2642,559 +2642,6 @@ void zbuffer_sss(RenderPart *pa, unsigned int lay, void *handle, void (*func)(vo
        zbuf_free_span(&zspan);
 }
 
-/* ******************** VECBLUR ACCUM BUF ************************* */
-
-typedef struct DrawBufPixel {
-       const float *colpoin;
-       float alpha;
-} DrawBufPixel;
-
-
-static void zbuf_fill_in_rgba(ZSpan *zspan, DrawBufPixel *col, float *v1, float *v2, float *v3, float *v4)
-{
-       DrawBufPixel *rectpofs, *rp;
-       double zxd, zyd, zy0, zverg;
-       float x0, y0, z0;
-       float x1, y1, z1, x2, y2, z2, xx1;
-       const float *span1, *span2;
-       float *rectzofs, *rz;
-       int x, y;
-       int sn1, sn2, rectx, my0, my2;
-       
-       /* init */
-       zbuf_init_span(zspan);
-       
-       /* set spans */
-       zbuf_add_to_span(zspan, v1, v2);
-       zbuf_add_to_span(zspan, v2, v3);
-       zbuf_add_to_span(zspan, v3, v4);
-       zbuf_add_to_span(zspan, v4, v1);
-       
-       /* clipped */
-       if (zspan->minp2==NULL || zspan->maxp2==NULL) return;
-       
-       my0 = max_ii(zspan->miny1, zspan->miny2);
-       my2 = min_ii(zspan->maxy1, zspan->maxy2);
-       
-       //      printf("my %d %d\n", my0, my2);
-       if (my2<my0) return;
-       
-       /* ZBUF DX DY, in floats still */
-       x1= v1[0]- v2[0];
-       x2= v2[0]- v3[0];
-       y1= v1[1]- v2[1];
-       y2= v2[1]- v3[1];
-       z1= v1[2]- v2[2];
-       z2= v2[2]- v3[2];
-       x0= y1*z2-z1*y2;
-       y0= z1*x2-x1*z2;
-       z0= x1*y2-y1*x2;
-       
-       if (z0==0.0f) return;
-       
-       xx1= (x0*v1[0] + y0*v1[1])/z0 + v1[2];
-       
-       zxd= -(double)x0/(double)z0;
-       zyd= -(double)y0/(double)z0;
-       zy0= ((double)my2)*zyd + (double)xx1;
-       
-       /* start-offset in rect */
-       rectx= zspan->rectx;
-       rectzofs= (float *)(zspan->rectz + rectx*my2);
-       rectpofs= ((DrawBufPixel *)zspan->rectp) + rectx*my2;
-       
-       /* correct span */
-       sn1= (my0 + my2)/2;
-       if (zspan->span1[sn1] < zspan->span2[sn1]) {
-               span1= zspan->span1+my2;
-               span2= zspan->span2+my2;
-       }
-       else {
-               span1= zspan->span2+my2;
-               span2= zspan->span1+my2;
-       }
-       
-       for (y=my2; y>=my0; y--, span1--, span2--) {
-               
-               sn1= floor(*span1);
-               sn2= floor(*span2);
-               sn1++; 
-               
-               if (sn2>=rectx) sn2= rectx-1;
-               if (sn1<0) sn1= 0;
-               
-               if (sn2>=sn1) {
-                       zverg= (double)sn1*zxd + zy0;
-                       rz= rectzofs+sn1;
-                       rp= rectpofs+sn1;
-                       x= sn2-sn1;
-                       
-                       while (x>=0) {
-                               if (zverg < (double)*rz) {
-                                       *rz= zverg;
-                                       *rp= *col;
-                               }
-                               zverg+= zxd;
-                               rz++; 
-                               rp++; 
-                               x--;
-                       }
-               }
-               
-               zy0-=zyd;
-               rectzofs-= rectx;
-               rectpofs-= rectx;
-       }
-}
-
-/* char value==255 is filled in, rest should be zero */
-/* returns alpha values, but sets alpha to 1 for zero alpha pixels that have an alpha value as neighbor */
-void antialias_tagbuf(int xsize, int ysize, char *rectmove)
-{
-       char *row1, *row2, *row3;
-       char prev, next;
-       int a, x, y, step;
-       
-       /* 1: tag pixels to be candidate for AA */
-       for (y=2; y<ysize; y++) {
-               /* setup rows */
-               row1= rectmove + (y-2)*xsize;
-               row2= row1 + xsize;
-               row3= row2 + xsize;
-               for (x=2; x<xsize; x++, row1++, row2++, row3++) {
-                       if (row2[1]) {
-                               if (row2[0]==0 || row2[2]==0 || row1[1]==0 || row3[1]==0)
-                                       row2[1]= 128;
-                       }
-               }
-       }
-       
-       /* 2: evaluate horizontal scanlines and calculate alphas */
-       row1= rectmove;
-       for (y=0; y<ysize; y++) {
-               row1++;
-               for (x=1; x<xsize; x++, row1++) {
-                       if (row1[0]==128 && row1[1]==128) {
-                               /* find previous color and next color and amount of steps to blend */
-                               prev= row1[-1];
-                               step= 1;
-                               while (x+step<xsize && row1[step]==128)
-                                       step++;
-                               
-                               if (x+step!=xsize) {
-                                       /* now we can blend values */
-                                       next= row1[step];
-
-                                       /* note, prev value can be next value, but we do this loop to clear 128 then */
-                                       for (a=0; a<step; a++) {
-                                               int fac, mfac;
-                                               
-                                               fac= ((a+1)<<8)/(step+1);
-                                               mfac= 255-fac;
-                                               
-                                               row1[a]= (prev*mfac + next*fac)>>8; 
-                                       }
-                               }
-                       }
-               }
-       }
-       
-       /* 3: evaluate vertical scanlines and calculate alphas */
-       /*    use for reading a copy of the original tagged buffer */
-       for (x=0; x<xsize; x++) {
-               row1= rectmove + x+xsize;
-               
-               for (y=1; y<ysize; y++, row1+=xsize) {
-                       if (row1[0]==128 && row1[xsize]==128) {
-                               /* find previous color and next color and amount of steps to blend */
-                               prev= row1[-xsize];
-                               step= 1;
-                               while (y+step<ysize && row1[step*xsize]==128)
-                                       step++;
-                               
-                               if (y+step!=ysize) {
-                                       /* now we can blend values */
-                                       next= row1[step*xsize];
-                                       /* note, prev value can be next value, but we do this loop to clear 128 then */
-                                       for (a=0; a<step; a++) {
-                                               int fac, mfac;
-                                               
-                                               fac= ((a+1)<<8)/(step+1);
-                                               mfac= 255-fac;
-                                               
-                                               row1[a*xsize]= (prev*mfac + next*fac)>>8; 
-                                       }
-                               }
-                       }
-               }
-       }
-       
-       /* last: pixels with 0 we fill in zbuffer, with 1 we skip for mask */
-       for (y=2; y<ysize; y++) {
-               /* setup rows */
-               row1= rectmove + (y-2)*xsize;
-               row2= row1 + xsize;
-               row3= row2 + xsize;
-               for (x=2; x<xsize; x++, row1++, row2++, row3++) {
-                       if (row2[1]==0) {
-                               if (row2[0]>1 || row2[2]>1 || row1[1]>1 || row3[1]>1)
-                                       row2[1]= 1;
-                       }
-               }
-       }
-}
-
-/* in: two vectors, first vector points from origin back in time, 2nd vector points to future */
-/* we make this into 3 points, center point is (0, 0) */
-/* and offset the center point just enough to make curve go through midpoint */
-
-static void quad_bezier_2d(float *result, float *v1, float *v2, float *ipodata)
-{
-       float p1[2], p2[2], p3[2];
-       
-       p3[0]= -v2[0];
-       p3[1]= -v2[1];
-       
-       p1[0]= v1[0];
-       p1[1]= v1[1];
-       
-       /* official formula 2*p2 - 0.5*p1 - 0.5*p3 */
-       p2[0]= -0.5f*p1[0] - 0.5f*p3[0];
-       p2[1]= -0.5f*p1[1] - 0.5f*p3[1];
-       
-       result[0]= ipodata[0]*p1[0] + ipodata[1]*p2[0] + ipodata[2]*p3[0];
-       result[1]= ipodata[0]*p1[1] + ipodata[1]*p2[1] + ipodata[2]*p3[1];
-}
-
-static void set_quad_bezier_ipo(float fac, float *data)
-{
-       float mfac= (1.0f-fac);
-       
-       data[0]= mfac*mfac;
-       data[1]= 2.0f*mfac*fac;
-       data[2]= fac*fac;
-}
-
-void RE_zbuf_accumulate_vecblur(
-        NodeBlurData *nbd, int xsize, int ysize, float *newrect,
-        const float *imgrect, float *vecbufrect, const float *zbufrect)
-{
-       ZSpan zspan;
-       DrawBufPixel *rectdraw, *dr;
-       static float jit[256][2];
-       float v1[3], v2[3], v3[3], v4[3], fx, fy;
-       const float *dimg, *dz, *ro;
-       float *rectvz, *dvz, *dvec1, *dvec2, *dz1, *dz2, *rectz;
-       float *minvecbufrect= NULL, *rectweight, *rw, *rectmax, *rm;
-       float maxspeedsq= (float)nbd->maxspeed*nbd->maxspeed;
-       int y, x, step, maxspeed=nbd->maxspeed, samples= nbd->samples;
-       int tsktsk= 0;
-       static int firsttime= 1;
-       char *rectmove, *dm;
-       
-       zbuf_alloc_span(&zspan, xsize, ysize, 1.0f);
-       zspan.zmulx=  ((float)xsize)/2.0f;
-       zspan.zmuly=  ((float)ysize)/2.0f;
-       zspan.zofsx= 0.0f;
-       zspan.zofsy= 0.0f;
-       
-       /* the buffers */
-       rectz= MEM_mapallocN(sizeof(float)*xsize*ysize, "zbuf accum");
-       zspan.rectz= (int *)rectz;
-       
-       rectmove= MEM_mapallocN(xsize*ysize, "rectmove");
-       rectdraw= MEM_mapallocN(sizeof(DrawBufPixel)*xsize*ysize, "rect draw");
-       zspan.rectp= (int *)rectdraw;
-
-       rectweight= MEM_mapallocN(sizeof(float)*xsize*ysize, "rect weight");
-       rectmax= MEM_mapallocN(sizeof(float)*xsize*ysize, "rect max");
-       
-       /* debug... check if PASS_VECTOR_MAX still is in buffers */
-       dvec1= vecbufrect;
-       for (x= 4*xsize*ysize; x>0; x--, dvec1++) {
-               if (dvec1[0]==PASS_VECTOR_MAX) {
-                       dvec1[0]= 0.0f;
-                       tsktsk= 1;
-               }
-       }
-       if (tsktsk) printf("Found uninitialized speed in vector buffer... fixed.\n");
-       
-       /* min speed? then copy speedbuffer to recalculate speed vectors */
-       if (nbd->minspeed) {
-               float minspeed= (float)nbd->minspeed;
-               float minspeedsq= minspeed*minspeed;
-               
-               minvecbufrect= MEM_mapallocN(4*sizeof(float)*xsize*ysize, "minspeed buf");
-               
-               dvec1= vecbufrect;
-               dvec2= minvecbufrect;
-               for (x= 2*xsize*ysize; x>0; x--, dvec1+=2, dvec2+=2) {
-                       if (dvec1[0]==0.0f && dvec1[1]==0.0f) {
-                               dvec2[0]= dvec1[0];
-                               dvec2[1]= dvec1[1];
-                       }
-                       else {
-                               float speedsq= dvec1[0]*dvec1[0] + dvec1[1]*dvec1[1];
-                               if (speedsq <= minspeedsq) {
-                                       dvec2[0]= 0.0f;
-                                       dvec2[1]= 0.0f;
-                               }
-                               else {
-                                       speedsq = 1.0f - minspeed / sqrtf(speedsq);
-                                       dvec2[0]= speedsq*dvec1[0];
-                                       dvec2[1]= speedsq*dvec1[1];
-                               }
-                       }
-               }
-               SWAP(float *, minvecbufrect, vecbufrect);
-       }
-       
-       /* make vertex buffer with averaged speed and zvalues */
-       rectvz= MEM_mapallocN(4*sizeof(float)*(xsize+1)*(ysize+1), "vertices");
-       dvz= rectvz;
-       for (y=0; y<=ysize; y++) {
-               
-               if (y==0)
-                       dvec1= vecbufrect + 4*y*xsize;
-               else
-                       dvec1= vecbufrect + 4*(y-1)*xsize;
-               
-               if (y==ysize)
-                       dvec2= vecbufrect + 4*(y-1)*xsize;
-               else
-                       dvec2= vecbufrect + 4*y*xsize;
-               
-               for (x=0; x<=xsize; x++) {
-                       
-                       /* two vectors, so a step loop */
-                       for (step=0; step<2; step++, dvec1+=2, dvec2+=2, dvz+=2) {
-                               /* average on minimal speed */
-                               int div= 0;
-                               
-                               if (x!=0) {
-                                       if (dvec1[-4]!=0.0f || dvec1[-3]!=0.0f) {
-                                               dvz[0]= dvec1[-4];
-                                               dvz[1]= dvec1[-3];
-                                               div++;
-                                       }
-                                       if (dvec2[-4]!=0.0f || dvec2[-3]!=0.0f) {
-                                               if (div==0) {
-                                                       dvz[0]= dvec2[-4];
-                                                       dvz[1]= dvec2[-3];
-                                                       div++;
-                                               }
-                                               else if ( (ABS(dvec2[-4]) + ABS(dvec2[-3]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
-                                                       dvz[0]= dvec2[-4];
-                                                       dvz[1]= dvec2[-3];
-                                               }
-                                       }
-                               }
-
-                               if (x!=xsize) {
-                                       if (dvec1[0]!=0.0f || dvec1[1]!=0.0f) {
-                                               if (div==0) {
-                                                       dvz[0]= dvec1[0];
-                                                       dvz[1]= dvec1[1];
-                                                       div++;
-                                               }
-                                               else if ( (ABS(dvec1[0]) + ABS(dvec1[1]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
-                                                       dvz[0]= dvec1[0];
-                                                       dvz[1]= dvec1[1];
-                                               }
-                                       }
-                                       if (dvec2[0]!=0.0f || dvec2[1]!=0.0f) {
-                                               if (div==0) {
-                                                       dvz[0]= dvec2[0];
-                                                       dvz[1]= dvec2[1];
-                                               }
-                                               else if ( (ABS(dvec2[0]) + ABS(dvec2[1]))< (ABS(dvz[0]) + ABS(dvz[1])) ) {
-                                                       dvz[0]= dvec2[0];
-                                                       dvz[1]= dvec2[1];
-                                               }
-                                       }
-                               }
-                               if (maxspeed) {
-                                       float speedsq= dvz[0]*dvz[0] + dvz[1]*dvz[1];
-                                       if (speedsq > maxspeedsq) {
-                                               speedsq = (float)maxspeed / sqrtf(speedsq);
-                                               dvz[0]*= speedsq;
-                                               dvz[1]*= speedsq;
-                                       }
-                               }
-                       }
-               }
-       }
-       
-       /* set border speeds to keep border speeds on border */
-       dz1= rectvz;
-       dz2= rectvz+4*(ysize)*(xsize+1);
-       for (x=0; x<=xsize; x++, dz1+=4, dz2+=4) {
-               dz1[1]= 0.0f;
-               dz2[1]= 0.0f;
-               dz1[3]= 0.0f;
-               dz2[3]= 0.0f;
-       }
-       dz1= rectvz;
-       dz2= rectvz+4*(xsize);
-       for (y=0; y<=ysize; y++, dz1+=4*(xsize+1), dz2+=4*(xsize+1)) {
-               dz1[0]= 0.0f;
-               dz2[0]= 0.0f;
-               dz1[2]= 0.0f;
-               dz2[2]= 0.0f;
-       }
-       
-       /* tag moving pixels, only these faces we draw */
-       dm= rectmove;
-       dvec1= vecbufrect;
-       for (x=xsize*ysize; x>0; x--, dm++, dvec1+=4) {
-               if ((dvec1[0]!=0.0f || dvec1[1]!=0.0f || dvec1[2]!=0.0f || dvec1[3]!=0.0f))
-                       *dm= 255;
-       }
-       
-       antialias_tagbuf(xsize, ysize, rectmove);
-       
-       /* has to become static, the init-jit calls a random-seed, screwing up texture noise node */
-       if (firsttime) {
-               firsttime= 0;
-               BLI_jitter_init(jit, 256);
-       }
-       
-       memset(newrect, 0, sizeof(float)*xsize*ysize*4);
-
-       /* accumulate */
-       samples/= 2;
-       for (step= 1; step<=samples; step++) {
-               float speedfac= 0.5f*nbd->fac*(float)step/(float)(samples+1);
-               int side;
-               
-               for (side=0; side<2; side++) {
-                       float blendfac, ipodata[4];
-                       
-                       /* clear zbuf, if we draw future we fill in not moving pixels */
-                       if (0)
-                               for (x= xsize*ysize-1; x>=0; x--) rectz[x]= 10e16;
-                       else 
-                               for (x= xsize*ysize-1; x>=0; x--) {
-                                       if (rectmove[x]==0)
-                                               rectz[x]= zbufrect[x];
-                                       else
-                                               rectz[x]= 10e16;
-                               }
-                       
-                       /* clear drawing buffer */
-                       for (x= xsize*ysize-1; x>=0; x--) rectdraw[x].colpoin= NULL;
-                       
-                       dimg= imgrect;
-                       dm= rectmove;
-                       dz= zbufrect;
-                       dz1= rectvz;
-                       dz2= rectvz + 4*(xsize + 1);
-                       
-                       if (side) {
-                               if (nbd->curved==0) {
-                                       dz1+= 2;
-                                       dz2+= 2;
-                               }
-                               speedfac= -speedfac;
-                       }
-                       
-                       set_quad_bezier_ipo(0.5f + 0.5f*speedfac, ipodata);
-                       
-                       for (fy= -0.5f+jit[step & 255][0], y=0; y<ysize; y++, fy+=1.0f) {
-                               for (fx= -0.5f+jit[step & 255][1], x=0; x<xsize; x++, fx+=1.0f, dimg+=4, dz1+=4, dz2+=4, dm++, dz++) {
-                                       if (*dm>1) {
-                                               float jfx = fx + 0.5f;
-                                               float jfy = fy + 0.5f;
-                                               DrawBufPixel col;
-                                               
-                                               /* make vertices */
-                                               if (nbd->curved) {      /* curved */
-                                                       quad_bezier_2d(v1, dz1, dz1+2, ipodata);
-                                                       v1[0]+= jfx; v1[1]+= jfy; v1[2]= *dz;
-
-                                                       quad_bezier_2d(v2, dz1+4, dz1+4+2, ipodata);
-                                                       v2[0]+= jfx+1.0f; v2[1]+= jfy; v2[2]= *dz;
-
-                                                       quad_bezier_2d(v3, dz2+4, dz2+4+2, ipodata);
-                                                       v3[0]+= jfx+1.0f; v3[1]+= jfy+1.0f; v3[2]= *dz;
-                                                       
-                                                       quad_bezier_2d(v4, dz2, dz2+2, ipodata);
-                                                       v4[0]+= jfx; v4[1]+= jfy+1.0f; v4[2]= *dz;
-                                               }
-                                               else {
-                                                       v1[0]= speedfac*dz1[0]+jfx;                     v1[1]= speedfac*dz1[1]+jfy;                     v1[2]= *dz;
-                                                       v2[0]= speedfac*dz1[4]+jfx+1.0f;                v2[1]= speedfac*dz1[5]+jfy;                     v2[2]= *dz;
-                                                       v3[0]= speedfac*dz2[4]+jfx+1.0f;                v3[1]= speedfac*dz2[5]+jfy+1.0f;                v3[2]= *dz;
-                                                       v4[0]= speedfac*dz2[0]+jfx;                     v4[1]= speedfac*dz2[1]+jfy+1.0f;                v4[2]= *dz;
-                                               }
-                                               if (*dm==255) col.alpha= 1.0f;
-                                               else if (*dm<2) col.alpha= 0.0f;
-                                               else col.alpha= ((float)*dm)/255.0f;
-                                               col.colpoin= dimg;
-
-                                               zbuf_fill_in_rgba(&zspan, &col, v1, v2, v3, v4);
-                                       }
-                               }
-                               dz1+=4;
-                               dz2+=4;
-                       }
-
-                       /* blend with a falloff. this fixes the ugly effect you get with
-                        * a fast moving object. then it looks like a solid object overlayed
-                        * over a very transparent moving version of itself. in reality, the
-                        * whole object should become transparent if it is moving fast, be
-                        * we don't know what is behind it so we don't do that. this hack
-                        * overestimates the contribution of foreground pixels but looks a
-                        * bit better without a sudden cutoff. */
-                       blendfac= ((samples - step)/(float)samples);
-                       /* smoothstep to make it look a bit nicer as well */
-                       blendfac= 3.0f*pow(blendfac, 2.0f) - 2.0f*pow(blendfac, 3.0f);
-
-                       /* accum */
-                       rw= rectweight;
-                       rm= rectmax;
-                       for (dr= rectdraw, dz2=newrect, x= xsize*ysize-1; x>=0; x--, dr++, dz2+=4, rw++, rm++) {
-                               if (dr->colpoin) {
-                                       float bfac= dr->alpha*blendfac;
-                                       
-                                       dz2[0] += bfac*dr->colpoin[0];
-                                       dz2[1] += bfac*dr->colpoin[1];
-                                       dz2[2] += bfac*dr->colpoin[2];
-                                       dz2[3] += bfac*dr->colpoin[3];
-
-                                       *rw += bfac;
-                                       *rm= MAX2(*rm, bfac);
-                               }
-                       }
-               }
-       }
-       
-       /* blend between original images and accumulated image */
-       rw= rectweight;
-       rm= rectmax;
-       ro= imgrect;
-       dm= rectmove;
-       for (dz2=newrect, x= xsize*ysize-1; x>=0; x--, dz2+=4, ro+=4, rw++, rm++, dm++) {
-               float mfac = *rm;
-               float fac = (*rw == 0.0f)? 0.0f: mfac/(*rw);
-               float nfac = 1.0f - mfac;
-
-               dz2[0]= fac*dz2[0] + nfac*ro[0];
-               dz2[1]= fac*dz2[1] + nfac*ro[1];
-               dz2[2]= fac*dz2[2] + nfac*ro[2];
-               dz2[3]= fac*dz2[3] + nfac*ro[3];
-       }
-
-       MEM_freeN(rectz);
-       MEM_freeN(rectmove);
-       MEM_freeN(rectdraw);
-       MEM_freeN(rectvz);
-       MEM_freeN(rectweight);
-       MEM_freeN(rectmax);
-       if (minvecbufrect) MEM_freeN(vecbufrect);  /* rects were swapped! */
-       zbuf_free_span(&zspan);
-}
-
 /* ******************** ABUF ************************* */
 
 /**
index 98092124dd55f3f44a19d4f99c6a2ac75d13ec40..211006e997addbef53d2521c68ae0306efe13f53 100644 (file)
@@ -247,10 +247,6 @@ struct RenderPass *RE_pass_find_by_name(volatile struct RenderLayer *rl, const c
 struct RenderPass *RE_pass_find_by_type(volatile struct RenderLayer *rl, int passtype, const char *viewname) RET_NULL
 bool RE_HasCombinedLayer(RenderResult *res) RET_ZERO
 
-/* zbuf.c stub */
-void antialias_tagbuf(int xsize, int ysize, char *rectmove) RET_NONE
-void RE_zbuf_accumulate_vecblur(struct NodeBlurData *nbd, int xsize, int ysize, float *newrect, const float *imgrect, float *vecbufrect, const float *zbufrect) RET_NONE
-
 /* imagetexture.c stub */
 void ibuf_sample(struct ImBuf *ibuf, float fx, float fy, float dx, float dy, float *result) RET_NONE