/* FRACSUBR.C contains subroutines which belong primarily to CALCFRAC.C and FRACTALS.C, i.e. which are non-fractal-specific fractal engine subroutines. */ #ifndef XFRACT #include #include #else #include #endif #include #include /* see Fractint.c for a description of the "include" hierarchy */ #include "port.h" #include "prototyp.h" #include "fractype.h" /* routines in this module */ static void _fastcall adjust_to_limits(double); static void _fastcall smallest_add(double *); static int _fastcall ratio_bad(double,double); static void _fastcall plotdorbit(double,double,int); static int _fastcall combine_worklist(void); static void _fastcall adjust_to_limitsbf(double); static void _fastcall smallest_add_bf(bf_t); int resume_len; /* length of resume info */ static int resume_offset; /* offset in resume info gets */ int taborhelp; /* kludge for sound and tab or help key press */ void set_grid_pointers() { dx0 = MK_FP(extraseg,0); dy1 = (dx1 = (dy0 = dx0 + xdots) + ydots) + ydots; set_pixel_calc_functions(); } void fill_dx_array(void) { int i; if(use_grid) { dx0[0] = xxmin; /* fill up the x, y grids */ dy0[0] = yymax; dx1[0] = dy1[0] = 0; for (i = 1; i < xdots; i++ ) { dx0[i] = (double)(dx0[0] + i*delxx); dy1[i] = (double)(dy1[0] - i*delyy2); } for (i = 1; i < ydots; i++ ) { dy0[i] = (double)(dy0[0] - i*delyy); dx1[i] = (double)(dx1[0] + i*delxx2); } } } void fractal_floattobf(void) { int i; init_bf_dec(getprecdbl(CURRENTREZ)); floattobf(bfxmin,xxmin); floattobf(bfxmax,xxmax); floattobf(bfymin,yymin); floattobf(bfymax,yymax); floattobf(bfx3rd,xx3rd); floattobf(bfy3rd,yy3rd); for (i = 0; i < MAXPARAMS; i++) if(typehasparm(fractype,i,NULL)) floattobf(bfparms[i],param[i]); calc_status = 0; } #ifdef _MSC_VER #if _MSC_VER == 800 /* MSC8 doesn't correctly calculate the address of certain arrays here */ #pragma optimize( "", off ) #endif #endif int use_grid; void calcfracinit(void) /* initialize a *pile* of stuff for fractal calculation */ { int tries = 0; int i, gotprec; double ftemp; coloriter=oldcoloriter = 0L; for(i=0;i<10;i++) rhombus_stack[i] = 0; /* set up grid array compactly leaving space at end */ if(xdots > OLDMAXPIXELS || ydots > OLDMAXPIXELS || debugflag==3800) { use_grid=0; } else use_grid=1; set_grid_pointers(); #ifdef FIXINTEGER if(!(curfractalspecific->flags & BF_MATH)) { int tofloat; if((tofloat=curfractalspecific->tofloat) == NOFRACTAL) bf_math = 0; else if(!(fractalspecific[tofloat].flags & BF_MATH)) bf_math = 0; else if(bf_math) { curfractalspecific = &fractalspecific[tofloat]; fractype = tofloat; } } #endif /* switch back to double when zooming out if using arbitrary precision */ if(bf_math) { gotprec=getprecbf(CURRENTREZ); if((gotprec <= DBL_DIG+1 && debugflag != 3200) || math_tol[1] >= 1.0) { bfcornerstofloat(); bf_math = 0; } else init_bf_dec(gotprec); } else if(fractype==MANDELFP && debugflag==3200) { curfractalspecific = &fractalspecific[MANDELFP]; fractal_floattobf(); } else if(fractype==JULIAFP && debugflag==3200) { curfractalspecific = &fractalspecific[JULIAFP]; fractal_floattobf(); } else if(fractype==FPMANDELZPOWER && debugflag==3200) { curfractalspecific = &fractalspecific[FPMANDELZPOWER]; fractal_floattobf(); } else if(fractype==FPJULIAZPOWER && debugflag==3200) { curfractalspecific = &fractalspecific[FPJULIAZPOWER]; fractal_floattobf(); } else free_bf_vars(); if (usr_stdcalcmode == 's') { if (fractype != MANDELFP) usr_stdcalcmode = '1'; } init_restart: /* the following variables may be forced to a different setting due to calc routine constraints; usr_xxx is what the user last said is wanted, xxx is what we actually do in the current situation */ stdcalcmode = usr_stdcalcmode; periodicitycheck = usr_periodicitycheck; distest = usr_distest; biomorph = usr_biomorph; potflag = 0; if (potparam[0] != 0.0 && colors >= 64 && (curfractalspecific->calctype == StandardFractal || curfractalspecific->calctype == calcmandfp)) { potflag = 1; distest = usr_distest = 0; /* can't do distest too */ } curfractalspecific = &fractalspecific[fractype]; if (potflag && potparam[2] != 0.0) rqlim = potparam[2]; else if (bailout) /* user input bailout */ rqlim = bailout; else if (biomorph != -1) /* biomorph benefits from larger bailout */ rqlim = 100; else rqlim = curfractalspecific->orbit_bailout; if ((curfractalspecific->flags&NOROTATE) != 0) { /* ensure min xxmax) { ftemp = xxmax; xxmax = xxmin; xxmin = ftemp; } if (yymin > yymax) { ftemp = yymax; yymax = yymin; yymin = ftemp; } xx3rd = xxmin; yy3rd = yymin; } f_at_rad = 1.0/32768L; /* now setup arrays of real coordinates corresponding to each pixel */ if(bf_math) adjust_to_limitsbf(1.0); /* make sure all corners in valid range */ else { adjust_to_limits(1.0); /* make sure all corners in valid range */ delxx = (LDBL)(xxmax - xx3rd) / (LDBL)dxsize; /* calculate stepsizes */ delyy = (LDBL)(yymax - yy3rd) / (LDBL)dysize; delxx2 = (LDBL)(xx3rd - xxmin) / (LDBL)dysize; delyy2 = (LDBL)(yy3rd - yymin) / (LDBL)dxsize; fill_dx_array(); } /* skip this if plasma to avoid 3d problems */ /* skip if bf_math to avoid extraseg conflict with dx0 arrays */ /* skip if ifs, ifs3d, or lsystem to avoid crash when mathtolerance */ /* is set. These types don't auto switch between float and integer math */ if (fractype != PLASMA && bf_math == 0 && fractype != IFS && fractype != IFS3D && fractype != LSYSTEM) { double dx0,dy0,dx1,dy1; /* set up dx0 and dy0 analogs of lx0 and ly0 */ /* put fractal parameters in doubles */ dx0 = xxmin; /* fill up the x, y grids */ dy0 = yymax; dx1 = dy1 = 0; /* this way of defining the dx and dy arrays is not the most accurate, but it is kept because it is used to determine the limit of resolution */ for (i = 1; i < xdots; i++ ) { dx0 = (double)(dx0 + (double)delxx); dy1 = (double)(dy1 - (double)delyy2); } for (i = 1; i < ydots; i++ ) { dy0 = (double)(dy0 - (double)delyy); dx1 = (double)(dx1 + (double)delxx2); } if(bf_math == 0) /* redundant test, leave for now */ { double testx_try, testx_exact; double testy_try, testy_exact; /* Following is the old logic for detecting failure of double precision. It has two advantages: it is independent of the representation of numbers, and it is sensitive to resolution (allows depper zooms at lower resolution. However it fails for rotations of exactly 90 degrees, so we added a safety net by using the magnification. */ if(++tries < 2) /* for safety */ { static FCODE err[] = {"precision-detection error"}; if(tries > 1) stopmsg(0, err); /* Previously there were four tests of distortions in the zoom box used to detect precision limitations. In some cases of rotated/skewed zoom boxs, this causes the algorithm to bail out to arbitrary precision too soon. The logic now only tests the larger of the two deltas in an attempt to repair this bug. This should never make the transition to arbitrary precision sooner, but always later.*/ if(fabs(xxmax-xx3rd) > fabs(xx3rd-xxmin)) { testx_exact = xxmax-xx3rd; testx_try = dx0-xxmin; } else { testx_exact = xx3rd-xxmin; testx_try = dx1; } if(fabs(yy3rd-yymax) > fabs(yymin-yy3rd)) { testy_exact = yy3rd-yymax; testy_try = dy0-yymax; } else { testy_exact = yymin-yy3rd; testy_try = dy1; } if(ratio_bad(testx_try,testx_exact) || ratio_bad(testy_try,testy_exact)) { if(curfractalspecific->flags & BF_MATH) { fractal_floattobf(); goto init_restart; } adjust_to_limits(2.0); /* double the size */ if (calc_status == 2) /* due to restore of an old file? */ calc_status = 0; /* whatever, it isn't resumable */ goto init_restart; } /* end if ratio_bad etc. */ } /* end if tries < 2 */ } /* end if bf_math == 0 */ /* if long double available, this is more accurate */ fill_dx_array(); /* fill up the x, y grids */ /* re-set corners to match reality */ xxmax = (double)(xxmin + (xdots-1)*delxx + (ydots-1)*delxx2); yymin = (double)(yymax - (ydots-1)*delyy - (xdots-1)*delyy2); xx3rd = (double)(xxmin + (ydots-1)*delxx2); yy3rd = (double)(yymax - (ydots-1)*delyy); } /* end if not plasma */ /* for periodicity close-enough, and for unity: */ /* min(max(delx,delx2),max(dely,dely2) */ ddelmin = fabs((double)delxx); if (fabs((double)delxx2) > ddelmin) ddelmin = fabs((double)delxx2); if (fabs((double)delyy) > fabs((double)delyy2)) { if (fabs((double)delyy) < ddelmin) ddelmin = fabs((double)delyy); } else if (fabs((double)delyy2) < ddelmin) ddelmin = fabs((double)delyy2); /* calculate factors which plot real values to screen co-ords */ /* calcfrac.c plot_orbit routines have comments about this */ ftemp = (double)((0.0-delyy2) * delxx2 * dxsize * dysize - (xxmax-xx3rd) * (yy3rd-yymax)); if(ftemp != 0) { plotmx1 = (double)(delxx2 * dxsize * dysize / ftemp); plotmx2 = (yy3rd-yymax) * dxsize / ftemp; plotmy1 = (double)((0.0-delyy2) * dxsize * dysize / ftemp); plotmy2 = (xxmax-xx3rd) * dysize / ftemp; } if(bf_math == 0) free_bf_vars(); else { /* zap all of extraseg except high area to flush out bugs */ /* in production version this code can be deleted */ char far *extra; extra = (char far *)MK_FP(extraseg,0); far_memset(extra,0,(unsigned int)(0x10000l-(bflength+2)*22U)); } } #ifdef _MSC_VER #if _MSC_VER == 800 #pragma optimize( "", on ) /* restore optimization options */ #endif #endif void adjust_cornerbf(void) { /* make edges very near vert/horiz exact, to ditch rounding errs and */ /* to avoid problems when delta per axis makes too large a ratio */ double ftemp; double Xmagfactor, Rotation, Skew; LDBL Magnification; bf_t bftemp, bftemp2; bf_t btmp1; int saved; saved = save_stack(); bftemp = alloc_stack(rbflength+2); bftemp2 = alloc_stack(rbflength+2); btmp1 = alloc_stack(rbflength+2); /* While we're at it, let's adjust the Xmagfactor as well */ /* use bftemp, bftemp2 as bfXctr, bfYctr */ cvtcentermagbf(bftemp, bftemp2, &Magnification, &Xmagfactor, &Rotation, &Skew); ftemp = fabs(Xmagfactor); if (ftemp != 1 && ftemp >= (1-aspectdrift) && ftemp <= (1+aspectdrift)) { Xmagfactor = sign(Xmagfactor); cvtcornersbf(bftemp, bftemp2, Magnification, Xmagfactor, Rotation, Skew); } /* ftemp=fabs(xx3rd-xxmin); */ abs_a_bf(sub_bf(bftemp,bfx3rd,bfxmin)); /* ftemp2=fabs(xxmax-xx3rd);*/ abs_a_bf(sub_bf(bftemp2,bfxmax,bfx3rd)); /* if( (ftemp=fabs(xx3rd-xxmin)) < (ftemp2=fabs(xxmax-xx3rd)) ) */ if(cmp_bf(bftemp,bftemp2) < 0) { /* if (ftemp*10000 < ftemp2 && yy3rd != yymax) */ if (cmp_bf(mult_bf_int(btmp1,bftemp,10000),bftemp2) < 0 && cmp_bf(bfy3rd,bfymax) != 0 ) /* xx3rd = xxmin; */ copy_bf(bfx3rd, bfxmin); } /* else if (ftemp2*10000 < ftemp && yy3rd != yymin) */ if (cmp_bf(mult_bf_int(btmp1,bftemp2,10000),bftemp) < 0 && cmp_bf(bfy3rd,bfymin) != 0 ) /* xx3rd = xxmax; */ copy_bf(bfx3rd, bfxmax); /* ftemp=fabs(yy3rd-yymin); */ abs_a_bf(sub_bf(bftemp,bfy3rd,bfymin)); /* ftemp2=fabs(yymax-yy3rd); */ abs_a_bf(sub_bf(bftemp2,bfymax,bfy3rd)); /* if( (ftemp=fabs(yy3rd-yymin)) < (ftemp2=fabs(yymax-yy3rd)) ) */ if(cmp_bf(bftemp,bftemp2) < 0) { /* if (ftemp*10000 < ftemp2 && xx3rd != xxmax) */ if (cmp_bf(mult_bf_int(btmp1,bftemp,10000),bftemp2) < 0 && cmp_bf(bfx3rd,bfxmax) != 0 ) /* yy3rd = yymin; */ copy_bf(bfy3rd, bfymin); } /* else if (ftemp2*10000 < ftemp && xx3rd != xxmin) */ if (cmp_bf(mult_bf_int(btmp1,bftemp2,10000),bftemp) < 0 && cmp_bf(bfx3rd,bfxmin) != 0 ) /* yy3rd = yymax; */ copy_bf(bfy3rd, bfymax); restore_stack(saved); } void adjust_corner(void) { /* make edges very near vert/horiz exact, to ditch rounding errs and */ /* to avoid problems when delta per axis makes too large a ratio */ double ftemp,ftemp2; double Xctr, Yctr, Xmagfactor, Rotation, Skew; LDBL Magnification; /* While we're at it, let's adjust the Xmagfactor as well */ cvtcentermag(&Xctr, &Yctr, &Magnification, &Xmagfactor, &Rotation, &Skew); ftemp = fabs(Xmagfactor); if (ftemp != 1 && ftemp >= (1-aspectdrift) && ftemp <= (1+aspectdrift)) { Xmagfactor = sign(Xmagfactor); cvtcorners(Xctr, Yctr, Magnification, Xmagfactor, Rotation, Skew); } if( (ftemp=fabs(xx3rd-xxmin)) < (ftemp2=fabs(xxmax-xx3rd)) ) { if (ftemp*10000 < ftemp2 && yy3rd != yymax) xx3rd = xxmin; } if (ftemp2*10000 < ftemp && yy3rd != yymin) xx3rd = xxmax; if( (ftemp=fabs(yy3rd-yymin)) < (ftemp2=fabs(yymax-yy3rd)) ) { if (ftemp*10000 < ftemp2 && xx3rd != xxmax) yy3rd = yymin; } if (ftemp2*10000 < ftemp && xx3rd != xxmin) yy3rd = yymax; } static void _fastcall adjust_to_limitsbf(double expand) { LDBL limit; bf_t bcornerx[4],bcornery[4]; bf_t blowx,bhighx,blowy,bhighy,blimit,bftemp; bf_t bcenterx,bcentery,badjx,badjy,btmp1,btmp2; bf_t bexpand; int i; int saved; saved = save_stack(); bcornerx[0] = alloc_stack(rbflength+2); bcornerx[1] = alloc_stack(rbflength+2); bcornerx[2] = alloc_stack(rbflength+2); bcornerx[3] = alloc_stack(rbflength+2); bcornery[0] = alloc_stack(rbflength+2); bcornery[1] = alloc_stack(rbflength+2); bcornery[2] = alloc_stack(rbflength+2); bcornery[3] = alloc_stack(rbflength+2); blowx = alloc_stack(rbflength+2); bhighx = alloc_stack(rbflength+2); blowy = alloc_stack(rbflength+2); bhighy = alloc_stack(rbflength+2); blimit = alloc_stack(rbflength+2); bftemp = alloc_stack(rbflength+2); bcenterx = alloc_stack(rbflength+2); bcentery = alloc_stack(rbflength+2); badjx = alloc_stack(rbflength+2); badjy = alloc_stack(rbflength+2); btmp1 = alloc_stack(rbflength+2); btmp2 = alloc_stack(rbflength+2); bexpand = alloc_stack(rbflength+2); limit = 32767.99; /* if (bitshift >= 24) limit = 31.99; if (bitshift >= 29) limit = 3.99; */ floattobf(blimit,limit); floattobf(bexpand,expand); add_bf(bcenterx,bfxmin,bfxmax); half_a_bf(bcenterx); /* centery = (yymin+yymax)/2; */ add_bf(bcentery,bfymin,bfymax); half_a_bf(bcentery); /* if (xxmin == centerx) { */ if (cmp_bf(bfxmin,bcenterx)==0) { /* ohoh, infinitely thin, fix it */ smallest_add_bf(bfxmax); /* bfxmin -= bfxmax-centerx; */ sub_a_bf(bfxmin,sub_bf(btmp1,bfxmax,bcenterx)); } /* if (bfymin == centery) */ if (cmp_bf(bfymin,bcentery)==0) { smallest_add_bf(bfymax); /* bfymin -= bfymax-centery; */ sub_a_bf(bfymin,sub_bf(btmp1,bfymax,bcentery)); } /* if (bfx3rd == centerx) */ if (cmp_bf(bfx3rd,bcenterx)==0) smallest_add_bf(bfx3rd); /* if (bfy3rd == centery) */ if (cmp_bf(bfy3rd,bcentery)==0) smallest_add_bf(bfy3rd); /* setup array for easier manipulation */ /* cornerx[0] = xxmin; */ copy_bf(bcornerx[0],bfxmin); /* cornerx[1] = xxmax; */ copy_bf(bcornerx[1],bfxmax); /* cornerx[2] = xx3rd; */ copy_bf(bcornerx[2],bfx3rd); /* cornerx[3] = xxmin+(xxmax-xx3rd); */ sub_bf(bcornerx[3],bfxmax,bfx3rd); add_a_bf(bcornerx[3],bfxmin); /* cornery[0] = yymax; */ copy_bf(bcornery[0],bfymax); /* cornery[1] = yymin; */ copy_bf(bcornery[1],bfymin); /* cornery[2] = yy3rd; */ copy_bf(bcornery[2],bfy3rd); /* cornery[3] = yymin+(yymax-yy3rd); */ sub_bf(bcornery[3],bfymax,bfy3rd); add_a_bf(bcornery[3],bfymin); /* if caller wants image size adjusted, do that first */ if (expand != 1.0) { for (i=0; i<4; ++i) { /* cornerx[i] = centerx + (cornerx[i]-centerx)*expand; */ sub_bf(btmp1,bcornerx[i],bcenterx); mult_bf(bcornerx[i],btmp1,bexpand); add_a_bf(bcornerx[i],bcenterx); /* cornery[i] = centery + (cornery[i]-centery)*expand; */ sub_bf(btmp1,bcornery[i],bcentery); mult_bf(bcornery[i],btmp1,bexpand); add_a_bf(bcornery[i],bcentery); } } /* get min/max x/y values */ /* lowx = highx = cornerx[0]; */ copy_bf(blowx,bcornerx[0]); copy_bf(bhighx,bcornerx[0]); /* lowy = highy = cornery[0]; */ copy_bf(blowy,bcornery[0]); copy_bf(bhighy,bcornery[0]); for (i=1; i<4; ++i) { /* if (cornerx[i] < lowx) lowx = cornerx[i]; */ if (cmp_bf(bcornerx[i],blowx) < 0) copy_bf(blowx,bcornerx[i]); /* if (cornerx[i] > highx) highx = cornerx[i]; */ if (cmp_bf(bcornerx[i],bhighx) > 0) copy_bf(bhighx,bcornerx[i]); /* if (cornery[i] < lowy) lowy = cornery[i]; */ if (cmp_bf(bcornery[i],blowy) < 0) copy_bf(blowy,bcornery[i]); /* if (cornery[i] > highy) highy = cornery[i]; */ if (cmp_bf(bcornery[i],bhighy) > 0) copy_bf(bhighy,bcornery[i]); } /* if image is too large, downsize it maintaining center */ /* ftemp = highx-lowx; */ sub_bf(bftemp,bhighx,blowx); /* if (highy-lowy > ftemp) ftemp = highy-lowy; */ if (cmp_bf(sub_bf(btmp1,bhighy,blowy),bftemp) > 0) copy_bf(bftemp,btmp1); /* if image is too large, downsize it maintaining center */ floattobf(btmp1,limit*2.0); copy_bf(btmp2,bftemp); div_bf(bftemp,btmp1,btmp2); floattobf(btmp1,1.0); if (cmp_bf(bftemp,btmp1) < 0) for (i=0; i<4; ++i) { /* cornerx[i] = centerx + (cornerx[i]-centerx)*ftemp; */ sub_bf(btmp1,bcornerx[i],bcenterx); mult_bf(bcornerx[i],btmp1,bftemp); add_a_bf(bcornerx[i],bcenterx); /* cornery[i] = centery + (cornery[i]-centery)*ftemp; */ sub_bf(btmp1,bcornery[i],bcentery); mult_bf(bcornery[i],btmp1,bftemp); add_a_bf(bcornery[i],bcentery); } /* if any corner has x or y past limit, move the image */ /* adjx = adjy = 0; */ clear_bf(badjx); clear_bf(badjy); for (i=0; i<4; ++i) { /* if (cornerx[i] > limit && (ftemp = cornerx[i] - limit) > adjx) adjx = ftemp; */ if (cmp_bf(bcornerx[i],blimit) > 0 && cmp_bf(sub_bf(bftemp,bcornerx[i],blimit),badjx) > 0) copy_bf(badjx,bftemp); /* if (cornerx[i] < 0.0-limit && (ftemp = cornerx[i] + limit) < adjx) adjx = ftemp; */ if (cmp_bf(bcornerx[i],neg_bf(btmp1,blimit)) < 0 && cmp_bf(add_bf(bftemp,bcornerx[i],blimit),badjx) < 0) copy_bf(badjx,bftemp); /* if (cornery[i] > limit && (ftemp = cornery[i] - limit) > adjy) adjy = ftemp; */ if (cmp_bf(bcornery[i],blimit) > 0 && cmp_bf(sub_bf(bftemp,bcornery[i],blimit),badjy) > 0) copy_bf(badjy,bftemp); /* if (cornery[i] < 0.0-limit && (ftemp = cornery[i] + limit) < adjy) adjy = ftemp; */ if (cmp_bf(bcornery[i],neg_bf(btmp1,blimit)) < 0 && cmp_bf(add_bf(bftemp,bcornery[i],blimit),badjy) < 0) copy_bf(badjy,bftemp); } /* if (calc_status == 2 && (adjx != 0 || adjy != 0) && (zwidth == 1.0)) calc_status = 0; */ if (calc_status == 2 && (is_bf_not_zero(badjx)|| is_bf_not_zero(badjy)) && (zwidth == 1.0)) calc_status = 0; /* xxmin = cornerx[0] - adjx; */ sub_bf(bfxmin,bcornerx[0],badjx); /* xxmax = cornerx[1] - adjx; */ sub_bf(bfxmax,bcornerx[1],badjx); /* xx3rd = cornerx[2] - adjx; */ sub_bf(bfx3rd,bcornerx[2],badjx); /* yymax = cornery[0] - adjy; */ sub_bf(bfymax,bcornery[0],badjy); /* yymin = cornery[1] - adjy; */ sub_bf(bfymin,bcornery[1],badjy); /* yy3rd = cornery[2] - adjy; */ sub_bf(bfy3rd,bcornery[2],badjy); adjust_cornerbf(); /* make 3rd corner exact if very near other co-ords */ restore_stack(saved); } static void _fastcall adjust_to_limits(double expand) { double cornerx[4],cornery[4]; double lowx,highx,lowy,highy,limit,ftemp; double centerx,centery,adjx,adjy; int i; limit = 32767.99; centerx = (xxmin+xxmax)/2; centery = (yymin+yymax)/2; if (xxmin == centerx) { /* ohoh, infinitely thin, fix it */ smallest_add(&xxmax); xxmin -= xxmax-centerx; } if (yymin == centery) { smallest_add(&yymax); yymin -= yymax-centery; } if (xx3rd == centerx) smallest_add(&xx3rd); if (yy3rd == centery) smallest_add(&yy3rd); /* setup array for easier manipulation */ cornerx[0] = xxmin; cornerx[1] = xxmax; cornerx[2] = xx3rd; cornerx[3] = xxmin+(xxmax-xx3rd); cornery[0] = yymax; cornery[1] = yymin; cornery[2] = yy3rd; cornery[3] = yymin+(yymax-yy3rd); /* if caller wants image size adjusted, do that first */ if (expand != 1.0) { for (i=0; i<4; ++i) { cornerx[i] = centerx + (cornerx[i]-centerx)*expand; cornery[i] = centery + (cornery[i]-centery)*expand; } } /* get min/max x/y values */ lowx = highx = cornerx[0]; lowy = highy = cornery[0]; for (i=1; i<4; ++i) { if (cornerx[i] < lowx) lowx = cornerx[i]; if (cornerx[i] > highx) highx = cornerx[i]; if (cornery[i] < lowy) lowy = cornery[i]; if (cornery[i] > highy) highy = cornery[i]; } /* if image is too large, downsize it maintaining center */ ftemp = highx-lowx; if (highy-lowy > ftemp) ftemp = highy-lowy; /* if image is too large, downsize it maintaining center */ if ((ftemp = limit*2/ftemp) < 1.0) { for (i=0; i<4; ++i) { cornerx[i] = centerx + (cornerx[i]-centerx)*ftemp; cornery[i] = centery + (cornery[i]-centery)*ftemp; } } /* if any corner has x or y past limit, move the image */ adjx = adjy = 0; for (i=0; i<4; ++i) { if (cornerx[i] > limit && (ftemp = cornerx[i] - limit) > adjx) adjx = ftemp; if (cornerx[i] < 0.0-limit && (ftemp = cornerx[i] + limit) < adjx) adjx = ftemp; if (cornery[i] > limit && (ftemp = cornery[i] - limit) > adjy) adjy = ftemp; if (cornery[i] < 0.0-limit && (ftemp = cornery[i] + limit) < adjy) adjy = ftemp; } if (calc_status == 2 && (adjx != 0 || adjy != 0) && (zwidth == 1.0)) calc_status = 0; xxmin = cornerx[0] - adjx; xxmax = cornerx[1] - adjx; xx3rd = cornerx[2] - adjx; yymax = cornery[0] - adjy; yymin = cornery[1] - adjy; yy3rd = cornery[2] - adjy; adjust_corner(); /* make 3rd corner exact if very near other co-ords */ } static void _fastcall smallest_add(double *num) { *num += *num * 5.0e-16; } static void _fastcall smallest_add_bf(bf_t num) { bf_t btmp1; int saved; saved = save_stack(); btmp1 = alloc_stack(bflength+2); mult_bf(btmp1,floattobf(btmp1, 5.0e-16),num); add_a_bf(num,btmp1); restore_stack(saved); } static int _fastcall ratio_bad(double actual, double desired) { double ftemp, tol; tol = math_tol[1]; if(tol <= 0.0) return(1); else if(tol >= 1.0) return(0); ftemp = 0; if (desired != 0 && debugflag != 3400) ftemp = actual / desired; if (desired != 0 && debugflag != 3400) if ((ftemp = actual / desired) < (1.0-tol) || ftemp > (1.0+tol)) return(1); return(0); } /* Save/resume stuff: Engines which aren't resumable can simply ignore all this. Before calling the (per_image,calctype) routines (engine), calcfract sets: "resuming" to 0 if new image, nonzero if resuming a partially done image "calc_status" to 1 If an engine is interrupted and wants to be able to resume it must: store whatever status info it needs to be able to resume later set calc_status to 2 and return If subsequently called with resuming!=0, the engine must restore status info and continue from where it left off. Since the info required for resume can get rather large for some types, it is not stored directly in save_info. Instead, memory is dynamically allocated as required, and stored in .fra files as a separate block. To save info for later resume, an engine routine can use: alloc_resume(maxsize,version) Maxsize must be >= max bytes subsequently saved + 2; over-allocation is harmless except for possibility of insufficient mem available; undersize is not checked and probably causes serious misbehaviour. Version is an arbitrary number so that subsequent revisions of the engine can be made backward compatible. Alloc_resume sets calc_status to 2 (resumable) if it succeeds; to 3 if it cannot allocate memory (and issues warning to user). put_resume({bytes,&argument,} ... 0) Can be called as often as required to store the info. Arguments must not be far addresses. Is not protected against calls which use more space than allocated. To reload info when resuming, use: start_resume() initializes and returns version number get_resume({bytes,&argument,} ... 0) inverse of store_resume end_resume() optional, frees the memory area sooner than would happen otherwise Example, save info: alloc_resume(sizeof(parmarray)+100,2); put_resume(sizeof(row),&row, sizeof(col),&col, sizeof(parmarray),parmarray, 0); restore info: vsn = start_resume(); get_resume(sizeof(row),&row, sizeof(col),&col, 0); if (vsn >= 2) get_resume(sizeof(parmarray),parmarray,0); end_resume(); Engines which allocate a large far memory chunk of their own might directly set resume_info, resume_len, calc_status to avoid doubling transient memory needs by using these routines. StandardFractal, calcmand, solidguess, and bound_trace_main are a related set of engines for escape-time fractals. They use a common worklist structure for save/resume. Fractals using these must specify calcmand or StandardFractal as the engine in fractalspecificinfo. Other engines don't get btm nor ssg, don't get off-axis symmetry nor panning (the worklist stuff), and are on their own for save/resume. */ #ifndef XFRACT int put_resume(int len, ...) #else int put_resume(va_alist) va_dcl #endif { va_list arg_marker; /* variable arg list */ BYTE *source_ptr; #ifdef XFRACT int len; #endif if (resume_info == 0) return(-1); #ifndef XFRACT va_start(arg_marker,len); #else va_start(arg_marker); len = va_arg(arg_marker,int); #endif while (len) { source_ptr = (BYTE *)va_arg(arg_marker,char *); /* far_memcpy(resume_info+resume_len,source_ptr,len); */ MoveToMemory(source_ptr,(U16)1,(long)len,resume_len,resume_info); resume_len += len; len = va_arg(arg_marker,int); } return(0); } int alloc_resume(int alloclen, int version) { /* WARNING! if alloclen > 4096B, problems may occur with GIF save/restore */ if (resume_info != 0) /* free the prior area if there is one */ MemoryRelease(resume_info); if ((resume_info = MemoryAlloc((U16)sizeof(alloclen), (long)alloclen, FARMEM)) == 0) { static FCODE msg[] = {"\ Warning - insufficient free memory to save status.\n\ You will not be able to resume calculating this image."}; stopmsg(0,msg); calc_status = 3; return(-1); } resume_len = 0; put_resume(sizeof(version),&version,0); calc_status = 2; return(0); } #ifndef XFRACT int get_resume(int len, ...) #else int get_resume(va_alist) va_dcl #endif { va_list arg_marker; /* variable arg list */ BYTE *dest_ptr; #ifdef XFRACT int len; #endif if (resume_info == 0) return(-1); #ifndef XFRACT va_start(arg_marker,len); #else va_start(arg_marker); len = va_arg(arg_marker,int); #endif while (len) { dest_ptr = (BYTE *)va_arg(arg_marker,char *); /* far_memcpy(dest_ptr,resume_info+resume_offset,len); */ MoveFromMemory(dest_ptr,(U16)1,(long)len,resume_offset,resume_info); resume_offset += len; len = va_arg(arg_marker,int); } return(0); } int start_resume(void) { int version; if (resume_info == 0) return(-1); resume_offset = 0; get_resume(sizeof(version),&version,0); return(version); } void end_resume(void) { if (resume_info != 0) /* free the prior area if there is one */ { MemoryRelease(resume_info); resume_info = 0; } } /* Showing orbit requires converting real co-ords to screen co-ords. Define: Xs == xxmax-xx3rd Ys == yy3rd-yymax W == xdots-1 D == ydots-1 We know that: realx == dx0[col] + dx1[row] realy == dy0[row] + dy1[col] dx0[col] == (col/width) * Xs + xxmin dx1[row] == row * delxx dy0[row] == (row/D) * Ys + yymax dy1[col] == col * (0-delyy) so: realx == (col/W) * Xs + xxmin + row * delxx realy == (row/D) * Ys + yymax + col * (0-delyy) and therefore: row == (realx-xxmin - (col/W)*Xs) / Xv (1) col == (realy-yymax - (row/D)*Ys) / Yv (2) substitute (2) into (1) and solve for row: row == ((realx-xxmin)*(0-delyy2)*W*D - (realy-yymax)*Xs*D) / ((0-delyy2)*W*delxx2*D-Ys*Xs) */ /* sleep N * a tenth of a millisecond */ void sleepms_old(long ms) { static long scalems = 0L; int savehelpmode,savetabmode; struct timebx t1,t2; #define SLEEPINIT 250 /* milliseconds for calibration */ savetabmode = tabmode; savehelpmode = helpmode; tabmode = 0; helpmode = -1; if(scalems==0L) /* calibrate */ { /* selects a value of scalems that makes the units 10000 per sec independent of CPU speed */ int i,elapsed; scalems = 1L; if(keypressed()) /* check at start, hope to get start of timeslice */ goto sleepexit; /* calibrate, assume slow computer first */ showtempmsg("Calibrating timer"); do { scalems *= 2; ftimex(&t2); do { /* wait for the start of a new tick */ ftimex(&t1); } while (t2.time == t1.time && t2.millitm == t1.millitm); sleepms_old(10L * SLEEPINIT); /* about 1/4 sec */ ftimex(&t2); if(keypressed()) { scalems = 0L; cleartempmsg(); goto sleepexit; } } while ((elapsed = (int)(t2.time-t1.time)*1000 + t2.millitm-t1.millitm) < SLEEPINIT); /* once more to see if faster (eg multi-tasking) */ do { /* wait for the start of a new tick */ ftimex(&t1); } while (t2.time == t1.time && t2.millitm == t1.millitm); sleepms_old(10L * SLEEPINIT); ftimex(&t2); if ((i = (int)(t2.time-t1.time)*1000 + t2.millitm-t1.millitm) < elapsed) elapsed = (i == 0) ? 1 : i; scalems = (long)((float)SLEEPINIT/(float)(elapsed) * scalems); cleartempmsg(); } if(ms > 10L * SLEEPINIT) { /* using ftime is probably more accurate */ ms /= 10; ftimex(&t1); for(;;) { if(keypressed()) break; ftimex(&t2); if ((long)((t2.time-t1.time)*1000 + t2.millitm-t1.millitm) >= ms) break; } } else if(!keypressed()) { ms *= scalems; while(ms-- >= 0); } sleepexit: tabmode = savetabmode; helpmode = savehelpmode; } static void sleepms_new(long ms) { uclock_t next_time; uclock_t now = usec_clock(); next_time = now + ms*100; while ((now = usec_clock()) < next_time) if(keypressed()) break; } void sleepms(long ms) { if(debugflag == 4020) sleepms_old(ms); else sleepms_new(ms); } /* * wait until wait_time microseconds from the * last call has elapsed. */ #define MAX_INDEX 2 static uclock_t next_time[MAX_INDEX]; void wait_until(int index, uclock_t wait_time) { if(debugflag == 4020) sleepms_old(wait_time); else { uclock_t now; while ( (now = usec_clock()) < next_time[index]) if(keypressed()) break; next_time[index] = now + wait_time*100; /* wait until this time next call */ } } void reset_clock(void) { int i; restart_uclock(); for(i=0;i= 1500) return; i = (int)(dy * plotmx1 - dx * plotmx2); i += sxoffs; if (i < 0 || i >= sxdots) return; j = (int)(dx * plotmy1 - dy * plotmy2); j += syoffs; if (j < 0 || j >= sydots) return; save_sxoffs = sxoffs; save_syoffs = syoffs; sxoffs = syoffs = 0; /* save orbit value */ if(color == -1) { *(save_orbit + orbit_ptr++) = i; *(save_orbit + orbit_ptr++) = j; *(save_orbit + orbit_ptr++) = c = getcolor(i,j); putcolor(i,j,c^orbit_color); } else putcolor(i,j,color); sxoffs = save_sxoffs; syoffs = save_syoffs; if(debugflag == 4030) { if((soundflag&7) == 2) /* sound = x */ w_snd((int)(i*1000/xdots+basehertz)); else if((soundflag&7) > 2) /* sound = y or z */ w_snd((int)(j*1000/ydots+basehertz)); else if(orbit_delay > 0) { wait_until(0,orbit_delay); } } else { if((soundflag&7) == 2) /* sound = x */ w_snd((int)(i+basehertz)); else if((soundflag&7) == 3) /* sound = y */ w_snd((int)(j+basehertz)); else if((soundflag&7) == 4) /* sound = z */ w_snd((int)(i+j+basehertz)); else if(orbit_delay > 0) { wait_until(0,orbit_delay); } } /* placing sleepms here delays each dot */ } void plot_orbit(double real,double imag,int color) { plotdorbit(real-xxmin,imag-yymax,color); } void scrub_orbit(void) { int i,j,c; int save_sxoffs,save_syoffs; mute(); save_sxoffs = sxoffs; save_syoffs = syoffs; sxoffs = syoffs = 0; while(orbit_ptr > 0) { c = *(save_orbit + --orbit_ptr); j = *(save_orbit + --orbit_ptr); i = *(save_orbit + --orbit_ptr); putcolor(i,j,c); } sxoffs = save_sxoffs; syoffs = save_syoffs; } int add_worklist(int xfrom, int xto, int xbegin, int yfrom, int yto, int ybegin, int pass, int sym) { if (num_worklist >= MAXCALCWORK) return(-1); worklist[num_worklist].xxstart = xfrom; worklist[num_worklist].xxstop = xto; worklist[num_worklist].xxbegin = xbegin; worklist[num_worklist].yystart = yfrom; worklist[num_worklist].yystop = yto; worklist[num_worklist].yybegin = ybegin; worklist[num_worklist].pass = pass; worklist[num_worklist].sym = sym; ++num_worklist; tidy_worklist(); return(0); } static int _fastcall combine_worklist(void) /* look for 2 entries which can freely merge */ { int i,j; for (i=0; i= N_ATTR) /* space for more attractors ? */ return; /* Bad luck - no room left ! */ savper = periodicitycheck; savmaxit = maxit; periodicitycheck = 0; old.x = real; /* prepare for f.p orbit calc */ old.y = imag; tempsqrx = sqr(old.x); tempsqry = sqr(old.y); if (maxit < 500) /* we're going to try at least this hard */ maxit = 500; coloriter = 0; overflow = 0; while (++coloriter < maxit) if(curfractalspecific->orbitcalc() || overflow) break; if (coloriter >= maxit) /* if orbit stays in the lake */ { result = new; for (i=0;i<10;i++) { overflow = 0; if(!curfractalspecific->orbitcalc() && !overflow) /* if it stays in the lake */ { /* and doesn't move far, probably */ if(fabs(result.x-new.x) < closenuff && fabs(result.y-new.y) < closenuff) { attr[attractors] = new; attrperiod[attractors] = i+1; attractors++; /* another attractor - coloured lakes ! */ break; } } else { break; } } } if(attractors==0) periodicitycheck = savper; maxit = savmaxit; } #define maxyblk 7 /* must match calcfrac.c */ #define maxxblk 202 /* must match calcfrac.c */ int ssg_blocksize(void) /* used by solidguessing and by zoom panning */ { int blocksize,i; /* blocksize 4 if <300 rows, 8 if 300-599, 16 if 600-1199, 32 if >=1200 */ blocksize=4; i=300; while(i<=ydots) { blocksize+=blocksize; i+=i; } /* increase blocksize if prefix array not big enough */ while(blocksize*(maxxblk-2)