xref: /dports/graphics/blender/blender-2.91.0/source/blender/blenkernel/intern/particle_system.c

/*
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
 *
 * The Original Code is Copyright (C) 2007 by Janne Karhu.
 * All rights reserved.
 * Adaptive time step
 * Classical SPH
 * Copyright 2011-2012 AutoCRC
 */

/** \file
 * \ingroup bke
 */

#include <stddef.h>

#include <math.h>
#include <stdlib.h>
#include <string.h>

#include "MEM_guardedalloc.h"

#include "DNA_anim_types.h"
#include "DNA_boid_types.h"
#include "DNA_curve_types.h"
#include "DNA_listBase.h"
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "DNA_modifier_types.h"
#include "DNA_object_force_types.h"
#include "DNA_object_types.h"
#include "DNA_particle_types.h"
#include "DNA_scene_types.h"
#include "DNA_texture_types.h"

#include "BLI_blenlib.h"
#include "BLI_edgehash.h"
#include "BLI_kdopbvh.h"
#include "BLI_kdtree.h"
#include "BLI_linklist.h"
#include "BLI_math.h"
#include "BLI_rand.h"
#include "BLI_string_utils.h"
#include "BLI_task.h"
#include "BLI_threads.h"
#include "BLI_utildefines.h"

#include "BKE_animsys.h"
#include "BKE_boids.h"
#include "BKE_collision.h"
#include "BKE_colortools.h"
#include "BKE_effect.h"
#include "BKE_lib_id.h"
#include "BKE_lib_query.h"
#include "BKE_particle.h"

#include "BKE_bvhutils.h"
#include "BKE_cloth.h"
#include "BKE_collection.h"
#include "BKE_lattice.h"
#include "BKE_material.h"
#include "BKE_mesh.h"
#include "BKE_modifier.h"
#include "BKE_object.h"
#include "BKE_pointcache.h"
#include "BKE_scene.h"

#include "DEG_depsgraph.h"
#include "DEG_depsgraph_physics.h"
#include "DEG_depsgraph_query.h"

#include "PIL_time.h"

#include "RE_shader_ext.h"

/* FLUID sim particle import */
#ifdef WITH_FLUID
#  include "DNA_fluid_types.h"
#  include "manta_fluid_API.h"
#endif  // WITH_FLUID

static ThreadRWMutex psys_bvhtree_rwlock = BLI_RWLOCK_INITIALIZER;

/************************************************/
/*          Reacting to system events           */
/************************************************/

static int particles_are_dynamic(ParticleSystem *psys)
{
  if (psys->pointcache->flag & PTCACHE_BAKED) {
    return 0;
  }

  if (psys->part->type == PART_HAIR) {
    return psys->flag & PSYS_HAIR_DYNAMICS;
  }

  return ELEM(psys->part->phystype, PART_PHYS_NEWTON, PART_PHYS_BOIDS, PART_PHYS_FLUID);
}

float psys_get_current_display_percentage(ParticleSystem *psys, const bool use_render_params)
{
  ParticleSettings *part = psys->part;

  if ((use_render_params &&
       !particles_are_dynamic(psys)) ||          /* non-dynamic particles can be rendered fully */
      (part->child_nbr && part->childtype) ||    /* display percentage applies to children */
      (psys->pointcache->flag & PTCACHE_BAKING)) /* baking is always done with full amount */
  {
    return 1.0f;
  }

  return psys->part->disp / 100.0f;
}

static int tot_particles(ParticleSystem *psys, PTCacheID *pid)
{
  if (pid && psys->pointcache->flag & PTCACHE_EXTERNAL) {
    return pid->cache->totpoint;
  }
  if (psys->part->distr == PART_DISTR_GRID && psys->part->from != PART_FROM_VERT) {
    return psys->part->grid_res * psys->part->grid_res * psys->part->grid_res - psys->totunexist;
  }

  return psys->part->totpart - psys->totunexist;
}

void psys_reset(ParticleSystem *psys, int mode)
{
  PARTICLE_P;

  if (ELEM(mode, PSYS_RESET_ALL, PSYS_RESET_DEPSGRAPH)) {
    if (mode == PSYS_RESET_ALL || !(psys->flag & PSYS_EDITED)) {
      /* don't free if not absolutely necessary */
      if (psys->totpart != tot_particles(psys, NULL)) {
        psys_free_particles(psys);
        psys->totpart = 0;
      }

      psys->totkeyed = 0;
      psys->flag &= ~(PSYS_HAIR_DONE | PSYS_KEYED);

      if (psys->edit && psys->free_edit) {
        psys->free_edit(psys->edit);
        psys->edit = NULL;
        psys->free_edit = NULL;
      }
    }
  }
  else if (mode == PSYS_RESET_CACHE_MISS) {
    /* set all particles to be skipped */
    LOOP_PARTICLES
    {
      pa->flag |= PARS_NO_DISP;
    }
  }

  /* reset children */
  if (psys->child) {
    MEM_freeN(psys->child);
    psys->child = NULL;
  }

  psys->totchild = 0;

  /* reset path cache */
  psys_free_path_cache(psys, psys->edit);

  /* reset point cache */
  BKE_ptcache_invalidate(psys->pointcache);

  if (psys->fluid_springs) {
    MEM_freeN(psys->fluid_springs);
    psys->fluid_springs = NULL;
  }

  psys->tot_fluidsprings = psys->alloc_fluidsprings = 0;
}

void psys_unique_name(Object *object, ParticleSystem *psys, const char *defname)
{
  BLI_uniquename(&object->particlesystem,
                 psys,
                 defname,
                 '.',
                 offsetof(ParticleSystem, name),
                 sizeof(psys->name));
}

static void realloc_particles(ParticleSimulationData *sim, int new_totpart)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  ParticleData *newpars = NULL;
  BoidParticle *newboids = NULL;
  PARTICLE_P;
  int totpart, totsaved = 0;

  if (new_totpart < 0) {
    if ((part->distr == PART_DISTR_GRID) && (part->from != PART_FROM_VERT)) {
      totpart = part->grid_res;
      totpart *= totpart * totpart;
    }
    else {
      totpart = part->totpart;
    }
  }
  else {
    totpart = new_totpart;
  }

  if (totpart != psys->totpart) {
    if (psys->edit && psys->free_edit) {
      psys->free_edit(psys->edit);
      psys->edit = NULL;
      psys->free_edit = NULL;
    }

    if (totpart) {
      newpars = MEM_callocN(totpart * sizeof(ParticleData), "particles");
      if (newpars == NULL) {
        return;
      }

      if (psys->part->phystype == PART_PHYS_BOIDS) {
        newboids = MEM_callocN(totpart * sizeof(BoidParticle), "boid particles");

        if (newboids == NULL) {
          /* allocation error! */
          if (newpars) {
            MEM_freeN(newpars);
          }
          return;
        }
      }
    }

    if (psys->particles) {
      totsaved = MIN2(psys->totpart, totpart);
      /*save old pars*/
      if (totsaved) {
        memcpy(newpars, psys->particles, totsaved * sizeof(ParticleData));

        if (psys->particles->boid) {
          memcpy(newboids, psys->particles->boid, totsaved * sizeof(BoidParticle));
        }
      }

      if (psys->particles->keys) {
        MEM_freeN(psys->particles->keys);
      }

      if (psys->particles->boid) {
        MEM_freeN(psys->particles->boid);
      }

      for (p = 0, pa = newpars; p < totsaved; p++, pa++) {
        if (pa->keys) {
          pa->keys = NULL;
          pa->totkey = 0;
        }
      }

      for (p = totsaved, pa = psys->particles + totsaved; p < psys->totpart; p++, pa++) {
        if (pa->hair) {
          MEM_freeN(pa->hair);
        }
      }

      MEM_freeN(psys->particles);
      psys_free_pdd(psys);
    }

    psys->particles = newpars;
    psys->totpart = totpart;

    if (newboids) {
      LOOP_PARTICLES
      {
        pa->boid = newboids++;
      }
    }
  }

  if (psys->child) {
    MEM_freeN(psys->child);
    psys->child = NULL;
    psys->totchild = 0;
  }
}

int psys_get_child_number(Scene *scene, ParticleSystem *psys, const bool use_render_params)
{
  int nbr;

  if (!psys->part->childtype) {
    return 0;
  }

  if (use_render_params) {
    nbr = psys->part->ren_child_nbr;
  }
  else {
    nbr = psys->part->child_nbr;
  }

  return get_render_child_particle_number(&scene->r, nbr, use_render_params);
}

int psys_get_tot_child(Scene *scene, ParticleSystem *psys, const bool use_render_params)
{
  return psys->totpart * psys_get_child_number(scene, psys, use_render_params);
}

/************************************************/
/*          Distribution                        */
/************************************************/

void psys_calc_dmcache(Object *ob, Mesh *mesh_final, Mesh *mesh_original, ParticleSystem *psys)
{
  /* use for building derived mesh mapping info:
   *
   * node: the allocated links - total derived mesh element count
   * nodearray: the array of nodes aligned with the base mesh's elements, so
   *            each original elements can reference its derived elements
   */
  Mesh *me = (Mesh *)ob->data;
  bool use_modifier_stack = psys->part->use_modifier_stack;
  PARTICLE_P;

  /* CACHE LOCATIONS */
  if (!mesh_final->runtime.deformed_only) {
    /* Will use later to speed up subsurf/evaluated mesh. */
    LinkNode *node, *nodedmelem, **nodearray;
    int totdmelem, totelem, i, *origindex, *origindex_poly = NULL;

    if (psys->part->from == PART_FROM_VERT) {
      totdmelem = mesh_final->totvert;

      if (use_modifier_stack) {
        totelem = totdmelem;
        origindex = NULL;
      }
      else {
        totelem = me->totvert;
        origindex = CustomData_get_layer(&mesh_final->vdata, CD_ORIGINDEX);
      }
    }
    else { /* FROM_FACE/FROM_VOLUME */
      totdmelem = mesh_final->totface;

      if (use_modifier_stack) {
        totelem = totdmelem;
        origindex = NULL;
        origindex_poly = NULL;
      }
      else {
        totelem = mesh_original->totface;
        origindex = CustomData_get_layer(&mesh_final->fdata, CD_ORIGINDEX);

        /* for face lookups we need the poly origindex too */
        origindex_poly = CustomData_get_layer(&mesh_final->pdata, CD_ORIGINDEX);
        if (origindex_poly == NULL) {
          origindex = NULL;
        }
      }
    }

    nodedmelem = MEM_callocN(sizeof(LinkNode) * totdmelem, "psys node elems");
    nodearray = MEM_callocN(sizeof(LinkNode *) * totelem, "psys node array");

    for (i = 0, node = nodedmelem; i < totdmelem; i++, node++) {
      int origindex_final;
      node->link = POINTER_FROM_INT(i);

      /* may be vertex or face origindex */
      if (use_modifier_stack) {
        origindex_final = i;
      }
      else {
        origindex_final = origindex ? origindex[i] : ORIGINDEX_NONE;

        /* if we have a poly source, do an index lookup */
        if (origindex_poly && origindex_final != ORIGINDEX_NONE) {
          origindex_final = origindex_poly[origindex_final];
        }
      }

      if (origindex_final != ORIGINDEX_NONE && origindex_final < totelem) {
        if (nodearray[origindex_final]) {
          /* prepend */
          node->next = nodearray[origindex_final];
          nodearray[origindex_final] = node;
        }
        else {
          nodearray[origindex_final] = node;
        }
      }
    }

    /* cache the verts/faces! */
    LOOP_PARTICLES
    {
      if (pa->num < 0) {
        pa->num_dmcache = DMCACHE_NOTFOUND;
        continue;
      }

      if (use_modifier_stack) {
        if (pa->num < totelem) {
          pa->num_dmcache = DMCACHE_ISCHILD;
        }
        else {
          pa->num_dmcache = DMCACHE_NOTFOUND;
        }
      }
      else {
        if (psys->part->from == PART_FROM_VERT) {
          if (pa->num < totelem && nodearray[pa->num]) {
            pa->num_dmcache = POINTER_AS_INT(nodearray[pa->num]->link);
          }
          else {
            pa->num_dmcache = DMCACHE_NOTFOUND;
          }
        }
        else { /* FROM_FACE/FROM_VOLUME */
          pa->num_dmcache = psys_particle_dm_face_lookup(
              mesh_final, mesh_original, pa->num, pa->fuv, nodearray);
        }
      }
    }

    MEM_freeN(nodearray);
    MEM_freeN(nodedmelem);
  }
  else {
    /* TODO PARTICLE, make the following line unnecessary, each function
     * should know to use the num or num_dmcache, set the num_dmcache to
     * an invalid value, just in case */

    LOOP_PARTICLES
    {
      pa->num_dmcache = DMCACHE_NOTFOUND;
    }
  }
}

/* threaded child particle distribution and path caching */
void psys_thread_context_init(ParticleThreadContext *ctx, ParticleSimulationData *sim)
{
  memset(ctx, 0, sizeof(ParticleThreadContext));
  ctx->sim = *sim;
  ctx->mesh = ctx->sim.psmd->mesh_final;
  ctx->ma = BKE_object_material_get(sim->ob, sim->psys->part->omat);
}

void psys_tasks_create(ParticleThreadContext *ctx,
                       int startpart,
                       int endpart,
                       ParticleTask **r_tasks,
                       int *r_numtasks)
{
  ParticleTask *tasks;
  int numtasks = min_ii(BLI_system_thread_count() * 4, endpart - startpart);
  float particles_per_task = (float)(endpart - startpart) / (float)numtasks, p, pnext;
  int i;

  tasks = MEM_callocN(sizeof(ParticleTask) * numtasks, "ParticleThread");
  *r_numtasks = numtasks;
  *r_tasks = tasks;

  p = (float)startpart;
  for (i = 0; i < numtasks; i++, p = pnext) {
    pnext = p + particles_per_task;

    tasks[i].ctx = ctx;
    tasks[i].begin = (int)p;
    tasks[i].end = min_ii((int)pnext, endpart);
  }
}

void psys_tasks_free(ParticleTask *tasks, int numtasks)
{
  int i;

  /* threads */
  for (i = 0; i < numtasks; i++) {
    if (tasks[i].rng) {
      BLI_rng_free(tasks[i].rng);
    }
    if (tasks[i].rng_path) {
      BLI_rng_free(tasks[i].rng_path);
    }
  }

  MEM_freeN(tasks);
}

void psys_thread_context_free(ParticleThreadContext *ctx)
{
  /* path caching */
  if (ctx->vg_length) {
    MEM_freeN(ctx->vg_length);
  }
  if (ctx->vg_clump) {
    MEM_freeN(ctx->vg_clump);
  }
  if (ctx->vg_kink) {
    MEM_freeN(ctx->vg_kink);
  }
  if (ctx->vg_rough1) {
    MEM_freeN(ctx->vg_rough1);
  }
  if (ctx->vg_rough2) {
    MEM_freeN(ctx->vg_rough2);
  }
  if (ctx->vg_roughe) {
    MEM_freeN(ctx->vg_roughe);
  }
  if (ctx->vg_twist) {
    MEM_freeN(ctx->vg_twist);
  }

  if (ctx->sim.psys->lattice_deform_data) {
    BKE_lattice_deform_data_destroy(ctx->sim.psys->lattice_deform_data);
    ctx->sim.psys->lattice_deform_data = NULL;
  }

  /* distribution */
  if (ctx->jit) {
    MEM_freeN(ctx->jit);
  }
  if (ctx->jitoff) {
    MEM_freeN(ctx->jitoff);
  }
  if (ctx->weight) {
    MEM_freeN(ctx->weight);
  }
  if (ctx->index) {
    MEM_freeN(ctx->index);
  }
  if (ctx->seams) {
    MEM_freeN(ctx->seams);
  }
  // if (ctx->vertpart) MEM_freeN(ctx->vertpart);
  BLI_kdtree_3d_free(ctx->tree);

  if (ctx->clumpcurve != NULL) {
    BKE_curvemapping_free(ctx->clumpcurve);
  }
  if (ctx->roughcurve != NULL) {
    BKE_curvemapping_free(ctx->roughcurve);
  }
  if (ctx->twistcurve != NULL) {
    BKE_curvemapping_free(ctx->twistcurve);
  }
}

static void init_particle_texture(ParticleSimulationData *sim, ParticleData *pa, int p)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  ParticleTexture ptex;

  psys_get_texture(sim, pa, &ptex, PAMAP_INIT, 0.f);

  switch (part->type) {
    case PART_EMITTER:
      if (ptex.exist < psys_frand(psys, p + 125)) {
        pa->flag |= PARS_UNEXIST;
      }
      pa->time = part->sta + (part->end - part->sta) * ptex.time;
      break;
    case PART_HAIR:
      if (ptex.exist < psys_frand(psys, p + 125)) {
        pa->flag |= PARS_UNEXIST;
      }
      pa->time = 0.f;
      break;
  }
}

/* set particle parameters that don't change during particle's life */
void init_particle(ParticleSimulationData *sim, ParticleData *pa)
{
  ParticleSettings *part = sim->psys->part;
  float birth_time = (float)(pa - sim->psys->particles) / (float)sim->psys->totpart;

  pa->flag &= ~PARS_UNEXIST;
  pa->time = part->sta + (part->end - part->sta) * birth_time;

  pa->hair_index = 0;
  /* we can't reset to -1 anymore since we've figured out correct index in distribute_particles */
  /* usage other than straight after distribute has to handle this index by itself - jahka*/
  // pa->num_dmcache = DMCACHE_NOTFOUND; /* assume we don't have a derived mesh face */
}

static void initialize_all_particles(ParticleSimulationData *sim)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  /* Grid distributionsets UNEXIST flag, need to take care of
   * it here because later this flag is being reset.
   *
   * We can't do it for any distribution, because it'll then
   * conflict with texture influence, which does not free
   * unexisting particles and only sets flag.
   *
   * It's not so bad, because only grid distribution sets
   * UNEXIST flag.
   */
  const bool emit_from_volume_grid = (part->distr == PART_DISTR_GRID) &&
                                     (!ELEM(part->from, PART_FROM_VERT, PART_FROM_CHILD));
  PARTICLE_P;
  LOOP_PARTICLES
  {
    if (!(emit_from_volume_grid && (pa->flag & PARS_UNEXIST) != 0)) {
      init_particle(sim, pa);
    }
  }
}

static void free_unexisting_particles(ParticleSimulationData *sim)
{
  ParticleSystem *psys = sim->psys;
  PARTICLE_P;

  psys->totunexist = 0;

  LOOP_PARTICLES
  {
    if (pa->flag & PARS_UNEXIST) {
      psys->totunexist++;
    }
  }

  if (psys->totpart && psys->totunexist == psys->totpart) {
    if (psys->particles->boid) {
      MEM_freeN(psys->particles->boid);
    }

    MEM_freeN(psys->particles);
    psys->particles = NULL;
    psys->totpart = psys->totunexist = 0;
  }

  if (psys->totunexist) {
    int newtotpart = psys->totpart - psys->totunexist;
    ParticleData *npa, *newpars;

    npa = newpars = MEM_callocN(newtotpart * sizeof(ParticleData), "particles");

    for (p = 0, pa = psys->particles; p < newtotpart; p++, pa++, npa++) {
      while (pa->flag & PARS_UNEXIST) {
        pa++;
      }

      memcpy(npa, pa, sizeof(ParticleData));
    }

    if (psys->particles->boid) {
      MEM_freeN(psys->particles->boid);
    }
    MEM_freeN(psys->particles);
    psys->particles = newpars;
    psys->totpart -= psys->totunexist;

    if (psys->particles->boid) {
      BoidParticle *newboids = MEM_callocN(psys->totpart * sizeof(BoidParticle), "boid particles");

      LOOP_PARTICLES
      {
        pa->boid = newboids++;
      }
    }
  }
}

static void get_angular_velocity_vector(short avemode, ParticleKey *state, float vec[3])
{
  switch (avemode) {
    case PART_AVE_VELOCITY:
      copy_v3_v3(vec, state->vel);
      break;
    case PART_AVE_HORIZONTAL: {
      float zvec[3];
      zvec[0] = zvec[1] = 0;
      zvec[2] = 1.f;
      cross_v3_v3v3(vec, state->vel, zvec);
      break;
    }
    case PART_AVE_VERTICAL: {
      float zvec[3], temp[3];
      zvec[0] = zvec[1] = 0;
      zvec[2] = 1.f;
      cross_v3_v3v3(temp, state->vel, zvec);
      cross_v3_v3v3(vec, temp, state->vel);
      break;
    }
    case PART_AVE_GLOBAL_X:
      vec[0] = 1.f;
      vec[1] = vec[2] = 0;
      break;
    case PART_AVE_GLOBAL_Y:
      vec[1] = 1.f;
      vec[0] = vec[2] = 0;
      break;
    case PART_AVE_GLOBAL_Z:
      vec[2] = 1.f;
      vec[0] = vec[1] = 0;
      break;
  }
}

void psys_get_birth_coords(
    ParticleSimulationData *sim, ParticleData *pa, ParticleKey *state, float dtime, float cfra)
{
  Object *ob = sim->ob;
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  ParticleTexture ptex;
  float fac, phasefac, nor[3] = {0, 0, 0}, loc[3], vel[3] = {0.0, 0.0, 0.0}, rot[4], q2[4];
  float r_vel[3], r_ave[3], r_rot[4], vec[3], p_vel[3] = {0.0, 0.0, 0.0};
  float x_vec[3] = {1.0, 0.0, 0.0}, utan[3] = {0.0, 1.0, 0.0}, vtan[3] = {0.0, 0.0, 1.0},
        rot_vec[3] = {0.0, 0.0, 0.0};
  float q_phase[4];

  const bool use_boids = ((part->phystype == PART_PHYS_BOIDS) && (pa->boid != NULL));
  const bool use_tangents = ((use_boids == false) &&
                             ((part->tanfac != 0.0f) || (part->rotmode == PART_ROT_NOR_TAN)));

  int p = pa - psys->particles;

  /* get birth location from object       */
  if (use_tangents) {
    psys_particle_on_emitter(sim->psmd,
                             part->from,
                             pa->num,
                             pa->num_dmcache,
                             pa->fuv,
                             pa->foffset,
                             loc,
                             nor,
                             utan,
                             vtan,
                             0);
  }
  else {
    psys_particle_on_emitter(
        sim->psmd, part->from, pa->num, pa->num_dmcache, pa->fuv, pa->foffset, loc, nor, 0, 0, 0);
  }

  /* get possible textural influence */
  psys_get_texture(sim, pa, &ptex, PAMAP_IVEL, cfra);

  /* particles live in global space so    */
  /* let's convert:                       */
  /* -location                            */
  mul_m4_v3(ob->obmat, loc);

  /* -normal                              */
  mul_mat3_m4_v3(ob->obmat, nor);
  normalize_v3(nor);

  /* -tangent                             */
  if (use_tangents) {
#if 0
    float phase = vg_rot ?
                      2.0f *
                          (psys_particle_value_from_verts(sim->psmd->dm, part->from, pa, vg_rot) -
                           0.5f) :
                      0.0f;
#else
    float phase = 0.0f;
#endif
    mul_v3_fl(vtan, -cosf((float)M_PI * (part->tanphase + phase)));
    fac = -sinf((float)M_PI * (part->tanphase + phase));
    madd_v3_v3fl(vtan, utan, fac);

    mul_mat3_m4_v3(ob->obmat, vtan);

    copy_v3_v3(utan, nor);
    mul_v3_fl(utan, dot_v3v3(vtan, nor));
    sub_v3_v3(vtan, utan);

    normalize_v3(vtan);
  }

  /* -velocity (boids need this even if there's no random velocity) */
  if (part->randfac != 0.0f || (part->phystype == PART_PHYS_BOIDS && pa->boid)) {
    r_vel[0] = 2.0f * (psys_frand(psys, p + 10) - 0.5f);
    r_vel[1] = 2.0f * (psys_frand(psys, p + 11) - 0.5f);
    r_vel[2] = 2.0f * (psys_frand(psys, p + 12) - 0.5f);

    mul_mat3_m4_v3(ob->obmat, r_vel);
    normalize_v3(r_vel);
  }

  /* -angular velocity                    */
  if (part->avemode == PART_AVE_RAND) {
    r_ave[0] = 2.0f * (psys_frand(psys, p + 13) - 0.5f);
    r_ave[1] = 2.0f * (psys_frand(psys, p + 14) - 0.5f);
    r_ave[2] = 2.0f * (psys_frand(psys, p + 15) - 0.5f);

    mul_mat3_m4_v3(ob->obmat, r_ave);
    normalize_v3(r_ave);
  }

  /* -rotation                            */
  if (part->randrotfac != 0.0f) {
    r_rot[0] = 2.0f * (psys_frand(psys, p + 16) - 0.5f);
    r_rot[1] = 2.0f * (psys_frand(psys, p + 17) - 0.5f);
    r_rot[2] = 2.0f * (psys_frand(psys, p + 18) - 0.5f);
    r_rot[3] = 2.0f * (psys_frand(psys, p + 19) - 0.5f);
    normalize_qt(r_rot);

    mat4_to_quat(rot, ob->obmat);
    mul_qt_qtqt(r_rot, r_rot, rot);
  }

  if (use_boids) {
    float dvec[3], q[4], mat[3][3];

    copy_v3_v3(state->co, loc);

    /* boids don't get any initial velocity  */
    zero_v3(state->vel);

    /* boids store direction in ave */
    if (fabsf(nor[2]) == 1.0f) {
      sub_v3_v3v3(state->ave, loc, ob->obmat[3]);
      normalize_v3(state->ave);
    }
    else {
      copy_v3_v3(state->ave, nor);
    }

    /* calculate rotation matrix */
    project_v3_v3v3(dvec, r_vel, state->ave);
    sub_v3_v3v3(mat[0], state->ave, dvec);
    normalize_v3(mat[0]);
    negate_v3_v3(mat[2], r_vel);
    normalize_v3(mat[2]);
    cross_v3_v3v3(mat[1], mat[2], mat[0]);

    /* apply rotation */
    mat3_to_quat_is_ok(q, mat);
    copy_qt_qt(state->rot, q);
  }
  else {
    /* conversion done so now we apply new: */
    /* -velocity from:                      */

    /*      *reactions                      */
    if (dtime > 0.f) {
      sub_v3_v3v3(vel, pa->state.vel, pa->prev_state.vel);
    }

    /*      *emitter velocity               */
    if (dtime != 0.f && part->obfac != 0.f) {
      sub_v3_v3v3(vel, loc, state->co);
      mul_v3_fl(vel, part->obfac / dtime);
    }

    /*      *emitter normal                 */
    if (part->normfac != 0.f) {
      madd_v3_v3fl(vel, nor, part->normfac);
    }

    /*      *emitter tangent                */
    if (sim->psmd && part->tanfac != 0.f) {
      madd_v3_v3fl(vel, vtan, part->tanfac);
    }

    /*      *emitter object orientation     */
    if (part->ob_vel[0] != 0.f) {
      normalize_v3_v3(vec, ob->obmat[0]);
      madd_v3_v3fl(vel, vec, part->ob_vel[0]);
    }
    if (part->ob_vel[1] != 0.f) {
      normalize_v3_v3(vec, ob->obmat[1]);
      madd_v3_v3fl(vel, vec, part->ob_vel[1]);
    }
    if (part->ob_vel[2] != 0.f) {
      normalize_v3_v3(vec, ob->obmat[2]);
      madd_v3_v3fl(vel, vec, part->ob_vel[2]);
    }

    /*      *texture                        */
    /* TODO */

    /*      *random                         */
    if (part->randfac != 0.f) {
      madd_v3_v3fl(vel, r_vel, part->randfac);
    }

    /*      *particle                       */
    if (part->partfac != 0.f) {
      madd_v3_v3fl(vel, p_vel, part->partfac);
    }

    mul_v3_v3fl(state->vel, vel, ptex.ivel);

    /* -location from emitter               */
    copy_v3_v3(state->co, loc);

    /* -rotation                            */
    unit_qt(state->rot);

    if (part->rotmode) {
      bool use_global_space;

      /* create vector into which rotation is aligned */
      switch (part->rotmode) {
        case PART_ROT_NOR:
        case PART_ROT_NOR_TAN:
          copy_v3_v3(rot_vec, nor);
          use_global_space = false;
          break;
        case PART_ROT_VEL:
          copy_v3_v3(rot_vec, vel);
          use_global_space = true;
          break;
        case PART_ROT_GLOB_X:
        case PART_ROT_GLOB_Y:
        case PART_ROT_GLOB_Z:
          rot_vec[part->rotmode - PART_ROT_GLOB_X] = 1.0f;
          use_global_space = true;
          break;
        case PART_ROT_OB_X:
        case PART_ROT_OB_Y:
        case PART_ROT_OB_Z:
          copy_v3_v3(rot_vec, ob->obmat[part->rotmode - PART_ROT_OB_X]);
          use_global_space = false;
          break;
        default:
          use_global_space = true;
          break;
      }

      /* create rotation quat */

      if (use_global_space) {
        negate_v3(rot_vec);
        vec_to_quat(q2, rot_vec, OB_POSX, OB_POSZ);

        /* randomize rotation quat */
        if (part->randrotfac != 0.0f) {
          interp_qt_qtqt(rot, q2, r_rot, part->randrotfac);
        }
        else {
          copy_qt_qt(rot, q2);
        }
      }
      else {
        /* calculate rotation in local-space */
        float q_obmat[4];
        float q_imat[4];

        mat4_to_quat(q_obmat, ob->obmat);
        invert_qt_qt_normalized(q_imat, q_obmat);

        if (part->rotmode != PART_ROT_NOR_TAN) {
          float rot_vec_local[3];

          /* rot_vec */
          negate_v3(rot_vec);
          copy_v3_v3(rot_vec_local, rot_vec);
          mul_qt_v3(q_imat, rot_vec_local);
          normalize_v3(rot_vec_local);

          vec_to_quat(q2, rot_vec_local, OB_POSX, OB_POSZ);
        }
        else {
          /* (part->rotmode == PART_ROT_NOR_TAN) */
          float tmat[3][3];

          /* note: utan_local is not taken from 'utan', we calculate from rot_vec/vtan */
          /* note: it looks like rotation phase may be applied twice (once with vtan, again below)
           * however this isn't the case - campbell */
          float *rot_vec_local = tmat[0];
          float *vtan_local = tmat[1];
          float *utan_local = tmat[2];

          /* use tangents */
          BLI_assert(use_tangents == true);

          /* rot_vec */
          copy_v3_v3(rot_vec_local, rot_vec);
          mul_qt_v3(q_imat, rot_vec_local);

          /* vtan_local */
          copy_v3_v3(vtan_local, vtan); /* flips, cant use */
          mul_qt_v3(q_imat, vtan_local);

          /* ensure orthogonal matrix (rot_vec aligned) */
          cross_v3_v3v3(utan_local, vtan_local, rot_vec_local);
          cross_v3_v3v3(vtan_local, utan_local, rot_vec_local);

          /* note: no need to normalize */
          mat3_to_quat(q2, tmat);
        }

        /* randomize rotation quat */
        if (part->randrotfac != 0.0f) {
          mul_qt_qtqt(r_rot, r_rot, q_imat);
          interp_qt_qtqt(rot, q2, r_rot, part->randrotfac);
        }
        else {
          copy_qt_qt(rot, q2);
        }

        mul_qt_qtqt(rot, q_obmat, rot);
      }

      /* rotation phase */
      phasefac = part->phasefac;
      if (part->randphasefac != 0.0f) {
        phasefac += part->randphasefac * psys_frand(psys, p + 20);
      }
      axis_angle_to_quat(q_phase, x_vec, phasefac * (float)M_PI);

      /* combine base rotation & phase */
      mul_qt_qtqt(state->rot, rot, q_phase);
    }

    /* -angular velocity                    */

    zero_v3(state->ave);

    if (part->avemode) {
      if (part->avemode == PART_AVE_RAND) {
        copy_v3_v3(state->ave, r_ave);
      }
      else {
        get_angular_velocity_vector(part->avemode, state, state->ave);
      }

      normalize_v3(state->ave);
      mul_v3_fl(state->ave, part->avefac);
    }
  }
}

/* recursively evaluate emitter parent anim at cfra */
static void evaluate_emitter_anim(struct Depsgraph *depsgraph,
                                  Scene *scene,
                                  Object *ob,
                                  float cfra)
{
  if (ob->parent) {
    evaluate_emitter_anim(depsgraph, scene, ob->parent, cfra);
  }

  BKE_object_where_is_calc_time(depsgraph, scene, ob, cfra);
}

/* sets particle to the emitter surface with initial velocity & rotation */
void reset_particle(ParticleSimulationData *sim, ParticleData *pa, float dtime, float cfra)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part;
  ParticleTexture ptex;
  int p = pa - psys->particles;
  part = psys->part;

  /* get precise emitter matrix if particle is born */
  if (part->type != PART_HAIR && dtime > 0.f && pa->time < cfra && pa->time >= sim->psys->cfra) {
    evaluate_emitter_anim(sim->depsgraph, sim->scene, sim->ob, pa->time);

    psys->flag |= PSYS_OB_ANIM_RESTORE;
  }

  psys_get_birth_coords(sim, pa, &pa->state, dtime, cfra);

  /* Initialize particle settings which depends on texture.
   *
   * We could only do it now because we'll need to know coordinate
   * before sampling the texture.
   */
  init_particle_texture(sim, pa, p);

  if (part->phystype == PART_PHYS_BOIDS && pa->boid) {
    BoidParticle *bpa = pa->boid;

    /* and gravity in r_ve */
    bpa->gravity[0] = bpa->gravity[1] = 0.0f;
    bpa->gravity[2] = -1.0f;
    if ((sim->scene->physics_settings.flag & PHYS_GLOBAL_GRAVITY) &&
        (sim->scene->physics_settings.gravity[2] != 0.0f)) {
      bpa->gravity[2] = sim->scene->physics_settings.gravity[2];
    }

    bpa->data.health = part->boids->health;
    bpa->data.mode = eBoidMode_InAir;
    bpa->data.state_id = ((BoidState *)part->boids->states.first)->id;
    bpa->data.acc[0] = bpa->data.acc[1] = bpa->data.acc[2] = 0.0f;
  }

  if (part->type == PART_HAIR) {
    pa->lifetime = 100.0f;
  }
  else {
    /* initialize the lifetime, in case the texture coordinates
     * are from Particles/Strands, which would cause undefined values
     */
    pa->lifetime = part->lifetime * (1.0f - part->randlife * psys_frand(psys, p + 21));
    pa->dietime = pa->time + pa->lifetime;

    /* get possible textural influence */
    psys_get_texture(sim, pa, &ptex, PAMAP_LIFE, cfra);

    pa->lifetime = part->lifetime * ptex.life;

    if (part->randlife != 0.0f) {
      pa->lifetime *= 1.0f - part->randlife * psys_frand(psys, p + 21);
    }
  }

  pa->dietime = pa->time + pa->lifetime;

  if ((sim->psys->pointcache) && (sim->psys->pointcache->flag & PTCACHE_BAKED) &&
      (sim->psys->pointcache->mem_cache.first)) {
    float dietime = psys_get_dietime_from_cache(sim->psys->pointcache, p);
    pa->dietime = MIN2(pa->dietime, dietime);
  }

  if (pa->time > cfra) {
    pa->alive = PARS_UNBORN;
  }
  else if (pa->dietime <= cfra) {
    pa->alive = PARS_DEAD;
  }
  else {
    pa->alive = PARS_ALIVE;
  }

  pa->state.time = cfra;
}
static void reset_all_particles(ParticleSimulationData *sim, float dtime, float cfra, int from)
{
  ParticleData *pa;
  int p, totpart = sim->psys->totpart;

  for (p = from, pa = sim->psys->particles + from; p < totpart; p++, pa++) {
    reset_particle(sim, pa, dtime, cfra);
  }
}
/************************************************/
/*          Particle targets                    */
/************************************************/
ParticleSystem *psys_get_target_system(Object *ob, ParticleTarget *pt)
{
  ParticleSystem *psys = NULL;

  if (pt->ob == NULL || pt->ob == ob) {
    psys = BLI_findlink(&ob->particlesystem, pt->psys - 1);
  }
  else {
    psys = BLI_findlink(&pt->ob->particlesystem, pt->psys - 1);
  }

  if (psys) {
    pt->flag |= PTARGET_VALID;
  }
  else {
    pt->flag &= ~PTARGET_VALID;
  }

  return psys;
}
/************************************************/
/*          Keyed particles                     */
/************************************************/
/* Counts valid keyed targets */
void psys_count_keyed_targets(ParticleSimulationData *sim)
{
  ParticleSystem *psys = sim->psys, *kpsys;
  ParticleTarget *pt = psys->targets.first;
  int keys_valid = 1;
  psys->totkeyed = 0;

  for (; pt; pt = pt->next) {
    kpsys = psys_get_target_system(sim->ob, pt);

    if (kpsys && kpsys->totpart) {
      psys->totkeyed += keys_valid;
      if (psys->flag & PSYS_KEYED_TIMING && pt->duration != 0.0f) {
        psys->totkeyed += 1;
      }
    }
    else {
      keys_valid = 0;
    }
  }

  psys->totkeyed *= psys->flag & PSYS_KEYED_TIMING ? 1 : psys->part->keyed_loops;
}

static void set_keyed_keys(ParticleSimulationData *sim)
{
  ParticleSystem *psys = sim->psys;
  ParticleSimulationData ksim = {0};
  ParticleTarget *pt;
  PARTICLE_P;
  ParticleKey *key;
  int totpart = psys->totpart, k, totkeys = psys->totkeyed;
  int keyed_flag = 0;

  ksim.depsgraph = sim->depsgraph;
  ksim.scene = sim->scene;

  /* no proper targets so let's clear and bail out */
  if (psys->totkeyed == 0) {
    free_keyed_keys(psys);
    psys->flag &= ~PSYS_KEYED;
    return;
  }

  if (totpart && psys->particles->totkey != totkeys) {
    free_keyed_keys(psys);

    key = MEM_callocN(totpart * totkeys * sizeof(ParticleKey), "Keyed keys");

    LOOP_PARTICLES
    {
      pa->keys = key;
      pa->totkey = totkeys;
      key += totkeys;
    }
  }

  psys->flag &= ~PSYS_KEYED;

  pt = psys->targets.first;
  for (k = 0; k < totkeys; k++) {
    ksim.ob = pt->ob ? pt->ob : sim->ob;
    ksim.psys = BLI_findlink(&ksim.ob->particlesystem, pt->psys - 1);
    keyed_flag = (ksim.psys->flag & PSYS_KEYED);
    ksim.psys->flag &= ~PSYS_KEYED;

    LOOP_PARTICLES
    {
      key = pa->keys + k;
      key->time = -1.0; /* use current time */

      const int p_ksim = (ksim.psys->totpart) ? p % ksim.psys->totpart : 0;
      psys_get_particle_state(&ksim, p_ksim, key, 1);

      if (psys->flag & PSYS_KEYED_TIMING) {
        key->time = pa->time + pt->time;
        if (pt->duration != 0.0f && k + 1 < totkeys) {
          copy_particle_key(key + 1, key, 1);
          (key + 1)->time = pa->time + pt->time + pt->duration;
        }
      }
      else if (totkeys > 1) {
        key->time = pa->time + (float)k / (float)(totkeys - 1) * pa->lifetime;
      }
      else {
        key->time = pa->time;
      }
    }

    if (psys->flag & PSYS_KEYED_TIMING && pt->duration != 0.0f) {
      k++;
    }

    ksim.psys->flag |= keyed_flag;

    pt = (pt->next && pt->next->flag & PTARGET_VALID) ? pt->next : psys->targets.first;
  }

  psys->flag |= PSYS_KEYED;
}

/************************************************/
/*          Point Cache                         */
/************************************************/
void psys_make_temp_pointcache(Object *ob, ParticleSystem *psys)
{
  PointCache *cache = psys->pointcache;

  if (cache->flag & PTCACHE_DISK_CACHE && BLI_listbase_is_empty(&cache->mem_cache)) {
    PTCacheID pid;
    BKE_ptcache_id_from_particles(&pid, ob, psys);
    cache->flag &= ~PTCACHE_DISK_CACHE;
    BKE_ptcache_disk_to_mem(&pid);
    cache->flag |= PTCACHE_DISK_CACHE;
  }
}
static void psys_clear_temp_pointcache(ParticleSystem *psys)
{
  if (psys->pointcache->flag & PTCACHE_DISK_CACHE) {
    BKE_ptcache_free_mem(&psys->pointcache->mem_cache);
  }
}
void psys_get_pointcache_start_end(Scene *scene, ParticleSystem *psys, int *sfra, int *efra)
{
  ParticleSettings *part = psys->part;

  *sfra = max_ii(1, (int)part->sta);
  *efra = min_ii((int)(part->end + part->lifetime + 1.0f), max_ii(scene->r.pefra, scene->r.efra));
}

/************************************************/
/*          Effectors                           */
/************************************************/
static void psys_update_particle_bvhtree(ParticleSystem *psys, float cfra)
{
  if (psys) {
    PARTICLE_P;
    int totpart = 0;
    bool need_rebuild;

    BLI_rw_mutex_lock(&psys_bvhtree_rwlock, THREAD_LOCK_READ);
    need_rebuild = !psys->bvhtree || psys->bvhtree_frame != cfra;
    BLI_rw_mutex_unlock(&psys_bvhtree_rwlock);

    if (need_rebuild) {
      LOOP_SHOWN_PARTICLES
      {
        totpart++;
      }

      BLI_rw_mutex_lock(&psys_bvhtree_rwlock, THREAD_LOCK_WRITE);

      BLI_bvhtree_free(psys->bvhtree);
      psys->bvhtree = BLI_bvhtree_new(totpart, 0.0, 4, 6);

      LOOP_SHOWN_PARTICLES
      {
        if (pa->alive == PARS_ALIVE) {
          if (pa->state.time == cfra) {
            BLI_bvhtree_insert(psys->bvhtree, p, pa->prev_state.co, 1);
          }
          else {
            BLI_bvhtree_insert(psys->bvhtree, p, pa->state.co, 1);
          }
        }
      }
      BLI_bvhtree_balance(psys->bvhtree);

      psys->bvhtree_frame = cfra;

      BLI_rw_mutex_unlock(&psys_bvhtree_rwlock);
    }
  }
}
void psys_update_particle_tree(ParticleSystem *psys, float cfra)
{
  if (psys) {
    PARTICLE_P;
    int totpart = 0;

    if (!psys->tree || psys->tree_frame != cfra) {
      LOOP_SHOWN_PARTICLES
      {
        totpart++;
      }

      BLI_kdtree_3d_free(psys->tree);
      psys->tree = BLI_kdtree_3d_new(psys->totpart);

      LOOP_SHOWN_PARTICLES
      {
        if (pa->alive == PARS_ALIVE) {
          if (pa->state.time == cfra) {
            BLI_kdtree_3d_insert(psys->tree, p, pa->prev_state.co);
          }
          else {
            BLI_kdtree_3d_insert(psys->tree, p, pa->state.co);
          }
        }
      }
      BLI_kdtree_3d_balance(psys->tree);

      psys->tree_frame = cfra;
    }
  }
}

static void psys_update_effectors(ParticleSimulationData *sim)
{
  BKE_effectors_free(sim->psys->effectors);
  sim->psys->effectors = BKE_effectors_create(
      sim->depsgraph, sim->ob, sim->psys, sim->psys->part->effector_weights);
  precalc_guides(sim, sim->psys->effectors);
}

static void integrate_particle(
    ParticleSettings *part,
    ParticleData *pa,
    float dtime,
    float *external_acceleration,
    void (*force_func)(void *forcedata, ParticleKey *state, float *force, float *impulse),
    void *forcedata)
{
#define ZERO_F43 \
  { \
    {0.0f, 0.0f, 0.0f}, {0.0f, 0.0f, 0.0f}, {0.0f, 0.0f, 0.0f}, \
    { \
      0.0f, 0.0f, 0.0f \
    } \
  }

  ParticleKey states[5];
  float force[3], acceleration[3], impulse[3], dx[4][3] = ZERO_F43, dv[4][3] = ZERO_F43, oldpos[3];
  float pa_mass = (part->flag & PART_SIZEMASS ? part->mass * pa->size : part->mass);
  int i, steps = 1;
  int integrator = part->integrator;

#undef ZERO_F43

  copy_v3_v3(oldpos, pa->state.co);

  /* Verlet integration behaves strangely with moving emitters, so do first step with euler. */
  if (pa->prev_state.time < 0.f && integrator == PART_INT_VERLET) {
    integrator = PART_INT_EULER;
  }

  switch (integrator) {
    case PART_INT_EULER:
      steps = 1;
      break;
    case PART_INT_MIDPOINT:
      steps = 2;
      break;
    case PART_INT_RK4:
      steps = 4;
      break;
    case PART_INT_VERLET:
      steps = 1;
      break;
  }

  for (i = 0; i < steps; i++) {
    copy_particle_key(states + i, &pa->state, 1);
  }

  states->time = 0.f;

  for (i = 0; i < steps; i++) {
    zero_v3(force);
    zero_v3(impulse);

    force_func(forcedata, states + i, force, impulse);

    /* force to acceleration*/
    mul_v3_v3fl(acceleration, force, 1.0f / pa_mass);

    if (external_acceleration) {
      add_v3_v3(acceleration, external_acceleration);
    }

    /* calculate next state */
    add_v3_v3(states[i].vel, impulse);

    switch (integrator) {
      case PART_INT_EULER:
        madd_v3_v3v3fl(pa->state.co, states->co, states->vel, dtime);
        madd_v3_v3v3fl(pa->state.vel, states->vel, acceleration, dtime);
        break;
      case PART_INT_MIDPOINT:
        if (i == 0) {
          madd_v3_v3v3fl(states[1].co, states->co, states->vel, dtime * 0.5f);
          madd_v3_v3v3fl(states[1].vel, states->vel, acceleration, dtime * 0.5f);
          states[1].time = dtime * 0.5f;
          /*fra=sim->psys->cfra+0.5f*dfra;*/
        }
        else {
          madd_v3_v3v3fl(pa->state.co, states->co, states[1].vel, dtime);
          madd_v3_v3v3fl(pa->state.vel, states->vel, acceleration, dtime);
        }
        break;
      case PART_INT_RK4:
        switch (i) {
          case 0:
            copy_v3_v3(dx[0], states->vel);
            mul_v3_fl(dx[0], dtime);
            copy_v3_v3(dv[0], acceleration);
            mul_v3_fl(dv[0], dtime);

            madd_v3_v3v3fl(states[1].co, states->co, dx[0], 0.5f);
            madd_v3_v3v3fl(states[1].vel, states->vel, dv[0], 0.5f);
            states[1].time = dtime * 0.5f;
            /*fra=sim->psys->cfra+0.5f*dfra;*/
            break;
          case 1:
            madd_v3_v3v3fl(dx[1], states->vel, dv[0], 0.5f);
            mul_v3_fl(dx[1], dtime);
            copy_v3_v3(dv[1], acceleration);
            mul_v3_fl(dv[1], dtime);

            madd_v3_v3v3fl(states[2].co, states->co, dx[1], 0.5f);
            madd_v3_v3v3fl(states[2].vel, states->vel, dv[1], 0.5f);
            states[2].time = dtime * 0.5f;
            break;
          case 2:
            madd_v3_v3v3fl(dx[2], states->vel, dv[1], 0.5f);
            mul_v3_fl(dx[2], dtime);
            copy_v3_v3(dv[2], acceleration);
            mul_v3_fl(dv[2], dtime);

            add_v3_v3v3(states[3].co, states->co, dx[2]);
            add_v3_v3v3(states[3].vel, states->vel, dv[2]);
            states[3].time = dtime;
            /*fra=cfra;*/
            break;
          case 3:
            add_v3_v3v3(dx[3], states->vel, dv[2]);
            mul_v3_fl(dx[3], dtime);
            copy_v3_v3(dv[3], acceleration);
            mul_v3_fl(dv[3], dtime);

            madd_v3_v3v3fl(pa->state.co, states->co, dx[0], 1.0f / 6.0f);
            madd_v3_v3fl(pa->state.co, dx[1], 1.0f / 3.0f);
            madd_v3_v3fl(pa->state.co, dx[2], 1.0f / 3.0f);
            madd_v3_v3fl(pa->state.co, dx[3], 1.0f / 6.0f);

            madd_v3_v3v3fl(pa->state.vel, states->vel, dv[0], 1.0f / 6.0f);
            madd_v3_v3fl(pa->state.vel, dv[1], 1.0f / 3.0f);
            madd_v3_v3fl(pa->state.vel, dv[2], 1.0f / 3.0f);
            madd_v3_v3fl(pa->state.vel, dv[3], 1.0f / 6.0f);
        }
        break;
      case PART_INT_VERLET: /* Verlet integration */
        madd_v3_v3v3fl(pa->state.vel, pa->prev_state.vel, acceleration, dtime);
        madd_v3_v3v3fl(pa->state.co, pa->prev_state.co, pa->state.vel, dtime);

        sub_v3_v3v3(pa->state.vel, pa->state.co, oldpos);
        mul_v3_fl(pa->state.vel, 1.0f / dtime);
        break;
    }
  }
}

/* -------------------------------------------------------------------- */
/** \name SPH fluid physics
 *
 * In theory, there could be unlimited implementation of SPH simulators
 *
 * This code uses in some parts adapted algorithms
 * from the pseudo code as outlined in the Research paper:
 *
 * Titled: Particle-based Viscoelastic Fluid Simulation.
 * Authors: Simon Clavet, Philippe Beaudoin and Pierre Poulin
 * Website: http://www.iro.umontreal.ca/labs/infographie/papers/Clavet-2005-PVFS/
 *
 * Presented at Siggraph, (2005)
 *
 * \{ */

#define PSYS_FLUID_SPRINGS_INITIAL_SIZE 256
static ParticleSpring *sph_spring_add(ParticleSystem *psys, ParticleSpring *spring)
{
  /* Are more refs required? */
  if (psys->alloc_fluidsprings == 0 || psys->fluid_springs == NULL) {
    psys->alloc_fluidsprings = PSYS_FLUID_SPRINGS_INITIAL_SIZE;
    psys->fluid_springs = (ParticleSpring *)MEM_callocN(
        psys->alloc_fluidsprings * sizeof(ParticleSpring), "Particle Fluid Springs");
  }
  else if (psys->tot_fluidsprings == psys->alloc_fluidsprings) {
    /* Double the number of refs allocated */
    psys->alloc_fluidsprings *= 2;
    psys->fluid_springs = (ParticleSpring *)MEM_reallocN(
        psys->fluid_springs, psys->alloc_fluidsprings * sizeof(ParticleSpring));
  }

  memcpy(psys->fluid_springs + psys->tot_fluidsprings, spring, sizeof(ParticleSpring));
  psys->tot_fluidsprings++;

  return psys->fluid_springs + psys->tot_fluidsprings - 1;
}
static void sph_spring_delete(ParticleSystem *psys, int j)
{
  if (j != psys->tot_fluidsprings - 1) {
    psys->fluid_springs[j] = psys->fluid_springs[psys->tot_fluidsprings - 1];
  }

  psys->tot_fluidsprings--;

  if (psys->tot_fluidsprings < psys->alloc_fluidsprings / 2 &&
      psys->alloc_fluidsprings > PSYS_FLUID_SPRINGS_INITIAL_SIZE) {
    psys->alloc_fluidsprings /= 2;
    psys->fluid_springs = (ParticleSpring *)MEM_reallocN(
        psys->fluid_springs, psys->alloc_fluidsprings * sizeof(ParticleSpring));
  }
}
static void sph_springs_modify(ParticleSystem *psys, float dtime)
{
  SPHFluidSettings *fluid = psys->part->fluid;
  ParticleData *pa1, *pa2;
  ParticleSpring *spring = psys->fluid_springs;

  float h, d, Rij[3], rij, Lij;
  int i;

  float yield_ratio = fluid->yield_ratio;
  float plasticity = fluid->plasticity_constant;
  /* scale things according to dtime */
  float timefix = 25.f * dtime;

  if ((fluid->flag & SPH_VISCOELASTIC_SPRINGS) == 0 || fluid->spring_k == 0.f) {
    return;
  }

  /* Loop through the springs */
  for (i = 0; i < psys->tot_fluidsprings; i++, spring++) {
    pa1 = psys->particles + spring->particle_index[0];
    pa2 = psys->particles + spring->particle_index[1];

    sub_v3_v3v3(Rij, pa2->prev_state.co, pa1->prev_state.co);
    rij = normalize_v3(Rij);

    /* adjust rest length */
    Lij = spring->rest_length;
    d = yield_ratio * timefix * Lij;

    if (rij > Lij + d) {  // Stretch
      spring->rest_length += plasticity * (rij - Lij - d) * timefix;
    }
    else if (rij < Lij - d) {  // Compress
      spring->rest_length -= plasticity * (Lij - d - rij) * timefix;
    }

    h = 4.f * pa1->size;

    if (spring->rest_length > h) {
      spring->delete_flag = 1;
    }
  }

  /* Loop through springs backwaqrds - for efficient delete function */
  for (i = psys->tot_fluidsprings - 1; i >= 0; i--) {
    if (psys->fluid_springs[i].delete_flag) {
      sph_spring_delete(psys, i);
    }
  }
}
static EdgeHash *sph_springhash_build(ParticleSystem *psys)
{
  EdgeHash *springhash = NULL;
  ParticleSpring *spring;
  int i = 0;

  springhash = BLI_edgehash_new_ex(__func__, psys->tot_fluidsprings);

  for (i = 0, spring = psys->fluid_springs; i < psys->tot_fluidsprings; i++, spring++) {
    BLI_edgehash_insert(
        springhash, spring->particle_index[0], spring->particle_index[1], POINTER_FROM_INT(i + 1));
  }

  return springhash;
}

#define SPH_NEIGHBORS 512
typedef struct SPHNeighbor {
  ParticleSystem *psys;
  int index;
} SPHNeighbor;

typedef struct SPHRangeData {
  SPHNeighbor neighbors[SPH_NEIGHBORS];
  int tot_neighbors;

  float *data;

  ParticleSystem *npsys;
  ParticleData *pa;

  float h;
  float mass;
  float massfac;
  int use_size;
} SPHRangeData;

static void sph_evaluate_func(BVHTree *tree,
                              ParticleSystem **psys,
                              const float co[3],
                              SPHRangeData *pfr,
                              float interaction_radius,
                              BVHTree_RangeQuery callback)
{
  int i;

  pfr->tot_neighbors = 0;

  for (i = 0; i < 10 && psys[i]; i++) {
    pfr->npsys = psys[i];
    pfr->massfac = psys[i]->part->mass / pfr->mass;
    pfr->use_size = psys[i]->part->flag & PART_SIZEMASS;

    if (tree) {
      BLI_bvhtree_range_query(tree, co, interaction_radius, callback, pfr);
      break;
    }

    BLI_rw_mutex_lock(&psys_bvhtree_rwlock, THREAD_LOCK_READ);

    BLI_bvhtree_range_query(psys[i]->bvhtree, co, interaction_radius, callback, pfr);

    BLI_rw_mutex_unlock(&psys_bvhtree_rwlock);
  }
}
static void sph_density_accum_cb(void *userdata, int index, const float co[3], float squared_dist)
{
  SPHRangeData *pfr = (SPHRangeData *)userdata;
  ParticleData *npa = pfr->npsys->particles + index;
  float q;
  float dist;

  UNUSED_VARS(co);

  if (npa == pfr->pa || squared_dist < FLT_EPSILON) {
    return;
  }

  /* Ugh! One particle has too many neighbors! If some aren't taken into
   * account, the forces will be biased by the tree search order. This
   * effectively adds energy to the system, and results in a churning motion.
   * But, we have to stop somewhere, and it's not the end of the world.
   * - jahka and z0r
   */
  if (pfr->tot_neighbors >= SPH_NEIGHBORS) {
    return;
  }

  pfr->neighbors[pfr->tot_neighbors].index = index;
  pfr->neighbors[pfr->tot_neighbors].psys = pfr->npsys;
  pfr->tot_neighbors++;

  dist = sqrtf(squared_dist);
  q = (1.f - dist / pfr->h) * pfr->massfac;

  if (pfr->use_size) {
    q *= npa->size;
  }

  pfr->data[0] += q * q;
  pfr->data[1] += q * q * q;
}

/*
 * Find the Courant number for an SPH particle (used for adaptive time step).
 */
static void sph_particle_courant(SPHData *sphdata, SPHRangeData *pfr)
{
  ParticleData *pa, *npa;
  int i;
  float flow[3], offset[3], dist;

  zero_v3(flow);

  dist = 0.0f;
  if (pfr->tot_neighbors > 0) {
    pa = pfr->pa;
    for (i = 0; i < pfr->tot_neighbors; i++) {
      npa = pfr->neighbors[i].psys->particles + pfr->neighbors[i].index;
      sub_v3_v3v3(offset, pa->prev_state.co, npa->prev_state.co);
      dist += len_v3(offset);
      add_v3_v3(flow, npa->prev_state.vel);
    }
    dist += sphdata->psys[0]->part->fluid->radius;  // TODO: remove this? - z0r
    sphdata->element_size = dist / pfr->tot_neighbors;
    mul_v3_v3fl(sphdata->flow, flow, 1.0f / pfr->tot_neighbors);
  }
  else {
    sphdata->element_size = FLT_MAX;
    copy_v3_v3(sphdata->flow, flow);
  }
}
static void sph_force_cb(void *sphdata_v, ParticleKey *state, float *force, float *UNUSED(impulse))
{
  SPHData *sphdata = (SPHData *)sphdata_v;
  ParticleSystem **psys = sphdata->psys;
  ParticleData *pa = sphdata->pa;
  SPHFluidSettings *fluid = psys[0]->part->fluid;
  ParticleSpring *spring = NULL;
  SPHRangeData pfr;
  SPHNeighbor *pfn;
  float *gravity = sphdata->gravity;
  EdgeHash *springhash = sphdata->eh;

  float q, u, rij, dv[3];
  float pressure, near_pressure;

  float visc = fluid->viscosity_omega;
  float stiff_visc = fluid->viscosity_beta *
                     (fluid->flag & SPH_FAC_VISCOSITY ? fluid->viscosity_omega : 1.f);

  float inv_mass = 1.0f / sphdata->mass;
  float spring_constant = fluid->spring_k;

  /* 4.0 seems to be a pretty good value */
  float interaction_radius = fluid->radius *
                             (fluid->flag & SPH_FAC_RADIUS ? 4.0f * pa->size : 1.0f);
  float h = interaction_radius * sphdata->hfac;
  /* 4.77 is an experimentally determined density factor */
  float rest_density = fluid->rest_density * (fluid->flag & SPH_FAC_DENSITY ? 4.77f : 1.f);
  float rest_length = fluid->rest_length *
                      (fluid->flag & SPH_FAC_REST_LENGTH ? 2.588f * pa->size : 1.f);

  float stiffness = fluid->stiffness_k;
  float stiffness_near_fac = fluid->stiffness_knear *
                             (fluid->flag & SPH_FAC_REPULSION ? fluid->stiffness_k : 1.f);

  ParticleData *npa;
  float vec[3];
  float vel[3];
  float co[3];
  float data[2];
  float density, near_density;

  int i, spring_index, index = pa - psys[0]->particles;

  data[0] = data[1] = 0;
  pfr.data = data;
  pfr.h = h;
  pfr.pa = pa;
  pfr.mass = sphdata->mass;

  sph_evaluate_func(NULL, psys, state->co, &pfr, interaction_radius, sph_density_accum_cb);

  density = data[0];
  near_density = data[1];

  pressure = stiffness * (density - rest_density);
  near_pressure = stiffness_near_fac * near_density;

  pfn = pfr.neighbors;
  for (i = 0; i < pfr.tot_neighbors; i++, pfn++) {
    npa = pfn->psys->particles + pfn->index;

    madd_v3_v3v3fl(co, npa->prev_state.co, npa->prev_state.vel, state->time);

    sub_v3_v3v3(vec, co, state->co);
    rij = normalize_v3(vec);

    q = (1.f - rij / h) * pfn->psys->part->mass * inv_mass;

    if (pfn->psys->part->flag & PART_SIZEMASS) {
      q *= npa->size;
    }

    copy_v3_v3(vel, npa->prev_state.vel);

    /* Double Density Relaxation */
    madd_v3_v3fl(force, vec, -(pressure + near_pressure * q) * q);

    /* Viscosity */
    if (visc > 0.f || stiff_visc > 0.f) {
      sub_v3_v3v3(dv, vel, state->vel);
      u = dot_v3v3(vec, dv);

      if (u < 0.f && visc > 0.f) {
        madd_v3_v3fl(force, vec, 0.5f * q * visc * u);
      }

      if (u > 0.f && stiff_visc > 0.f) {
        madd_v3_v3fl(force, vec, 0.5f * q * stiff_visc * u);
      }
    }

    if (spring_constant > 0.f) {
      /* Viscoelastic spring force */
      if (pfn->psys == psys[0] && fluid->flag & SPH_VISCOELASTIC_SPRINGS && springhash) {
        /* BLI_edgehash_lookup appears to be thread-safe. - z0r */
        spring_index = POINTER_AS_INT(BLI_edgehash_lookup(springhash, index, pfn->index));

        if (spring_index) {
          spring = psys[0]->fluid_springs + spring_index - 1;

          madd_v3_v3fl(
              force, vec, -10.f * spring_constant * (1.f - rij / h) * (spring->rest_length - rij));
        }
        else if (fluid->spring_frames == 0 ||
                 (pa->prev_state.time - pa->time) <= fluid->spring_frames) {
          ParticleSpring temp_spring;
          temp_spring.particle_index[0] = index;
          temp_spring.particle_index[1] = pfn->index;
          temp_spring.rest_length = (fluid->flag & SPH_CURRENT_REST_LENGTH) ? rij : rest_length;
          temp_spring.delete_flag = 0;

          BLI_buffer_append(&sphdata->new_springs, ParticleSpring, temp_spring);
        }
      }
      else { /* PART_SPRING_HOOKES - Hooke's spring force */
        madd_v3_v3fl(force, vec, -10.f * spring_constant * (1.f - rij / h) * (rest_length - rij));
      }
    }
  }

  /* Artificial buoyancy force in negative gravity direction  */
  if (fluid->buoyancy > 0.f && gravity) {
    madd_v3_v3fl(force, gravity, fluid->buoyancy * (density - rest_density));
  }

  if (sphdata->pass == 0 && psys[0]->part->time_flag & PART_TIME_AUTOSF) {
    sph_particle_courant(sphdata, &pfr);
  }
  sphdata->pass++;
}

static void sphclassical_density_accum_cb(void *userdata,
                                          int index,
                                          const float co[3],
                                          float UNUSED(squared_dist))
{
  SPHRangeData *pfr = (SPHRangeData *)userdata;
  ParticleData *npa = pfr->npsys->particles + index;
  float q;
  float qfac = 21.0f / (256.f * (float)M_PI);
  float rij, rij_h;
  float vec[3];

  /* Exclude particles that are more than 2h away. Can't use squared_dist here
   * because it is not accurate enough. Use current state, i.e. the output of
   * basic_integrate() - z0r */
  sub_v3_v3v3(vec, npa->state.co, co);
  rij = len_v3(vec);
  rij_h = rij / pfr->h;
  if (rij_h > 2.0f) {
    return;
  }

  /* Smoothing factor. Utilize the Wendland kernel. gnuplot:
   *     q1(x) = (2.0 - x)**4 * ( 1.0 + 2.0 * x)
   *     plot [0:2] q1(x) */
  q = qfac / pow3f(pfr->h) * pow4f(2.0f - rij_h) * (1.0f + 2.0f * rij_h);
  q *= pfr->npsys->part->mass;

  if (pfr->use_size) {
    q *= pfr->pa->size;
  }

  pfr->data[0] += q;
  pfr->data[1] += q / npa->sphdensity;
}

static void sphclassical_neighbor_accum_cb(void *userdata,
                                           int index,
                                           const float co[3],
                                           float UNUSED(squared_dist))
{
  SPHRangeData *pfr = (SPHRangeData *)userdata;
  ParticleData *npa = pfr->npsys->particles + index;
  float rij, rij_h;
  float vec[3];

  if (pfr->tot_neighbors >= SPH_NEIGHBORS) {
    return;
  }

  /* Exclude particles that are more than 2h away. Can't use squared_dist here
   * because it is not accurate enough. Use current state, i.e. the output of
   * basic_integrate() - z0r */
  sub_v3_v3v3(vec, npa->state.co, co);
  rij = len_v3(vec);
  rij_h = rij / pfr->h;
  if (rij_h > 2.0f) {
    return;
  }

  pfr->neighbors[pfr->tot_neighbors].index = index;
  pfr->neighbors[pfr->tot_neighbors].psys = pfr->npsys;
  pfr->tot_neighbors++;
}
static void sphclassical_force_cb(void *sphdata_v,
                                  ParticleKey *state,
                                  float *force,
                                  float *UNUSED(impulse))
{
  SPHData *sphdata = (SPHData *)sphdata_v;
  ParticleSystem **psys = sphdata->psys;
  ParticleData *pa = sphdata->pa;
  SPHFluidSettings *fluid = psys[0]->part->fluid;
  SPHRangeData pfr;
  SPHNeighbor *pfn;
  float *gravity = sphdata->gravity;

  float dq, u, rij, dv[3];
  float pressure, npressure;

  float visc = fluid->viscosity_omega;

  float interaction_radius;
  float h, hinv;
  /* 4.77 is an experimentally determined density factor */
  float rest_density = fluid->rest_density * (fluid->flag & SPH_FAC_DENSITY ? 4.77f : 1.0f);

  /* Use speed of sound squared */
  float stiffness = pow2f(fluid->stiffness_k);

  ParticleData *npa;
  float vec[3];
  float co[3];
  float pressureTerm;

  int i;

  float qfac2 = 42.0f / (256.0f * (float)M_PI);
  float rij_h;

  /* 4.0 here is to be consistent with previous formulation/interface */
  interaction_radius = fluid->radius * (fluid->flag & SPH_FAC_RADIUS ? 4.0f * pa->size : 1.0f);
  h = interaction_radius * sphdata->hfac;
  hinv = 1.0f / h;

  pfr.h = h;
  pfr.pa = pa;

  sph_evaluate_func(
      NULL, psys, state->co, &pfr, interaction_radius, sphclassical_neighbor_accum_cb);
  pressure = stiffness * (pow7f(pa->sphdensity / rest_density) - 1.0f);

  /* multiply by mass so that we return a force, not accel */
  qfac2 *= sphdata->mass / pow3f(pfr.h);

  pfn = pfr.neighbors;
  for (i = 0; i < pfr.tot_neighbors; i++, pfn++) {
    npa = pfn->psys->particles + pfn->index;
    if (npa == pa) {
      /* we do not contribute to ourselves */
      continue;
    }

    /* Find vector to neighbor. Exclude particles that are more than 2h
     * away. Can't use current state here because it may have changed on
     * another thread - so do own mini integration. Unlike basic_integrate,
     * SPH integration depends on neighboring particles. - z0r */
    madd_v3_v3v3fl(co, npa->prev_state.co, npa->prev_state.vel, state->time);
    sub_v3_v3v3(vec, co, state->co);
    rij = normalize_v3(vec);
    rij_h = rij / pfr.h;
    if (rij_h > 2.0f) {
      continue;
    }

    npressure = stiffness * (pow7f(npa->sphdensity / rest_density) - 1.0f);

    /* First derivative of smoothing factor. Utilize the Wendland kernel.
     * gnuplot:
     *     q2(x) = 2.0 * (2.0 - x)**4 - 4.0 * (2.0 - x)**3 * (1.0 + 2.0 * x)
     *     plot [0:2] q2(x)
     * Particles > 2h away are excluded above. */
    dq = qfac2 * (2.0f * pow4f(2.0f - rij_h) - 4.0f * pow3f(2.0f - rij_h) * (1.0f + 2.0f * rij_h));

    if (pfn->psys->part->flag & PART_SIZEMASS) {
      dq *= npa->size;
    }

    pressureTerm = pressure / pow2f(pa->sphdensity) + npressure / pow2f(npa->sphdensity);

    /* Note that 'minus' is removed, because vec = vecBA, not vecAB.
     * This applies to the viscosity calculation below, too. */
    madd_v3_v3fl(force, vec, pressureTerm * dq);

    /* Viscosity */
    if (visc > 0.0f) {
      sub_v3_v3v3(dv, npa->prev_state.vel, pa->prev_state.vel);
      u = dot_v3v3(vec, dv);
      /* Apply parameters */
      u *= -dq * hinv * visc / (0.5f * npa->sphdensity + 0.5f * pa->sphdensity);
      madd_v3_v3fl(force, vec, u);
    }
  }

  /* Artificial buoyancy force in negative gravity direction  */
  if (fluid->buoyancy > 0.f && gravity) {
    madd_v3_v3fl(force, gravity, fluid->buoyancy * (pa->sphdensity - rest_density));
  }

  if (sphdata->pass == 0 && psys[0]->part->time_flag & PART_TIME_AUTOSF) {
    sph_particle_courant(sphdata, &pfr);
  }
  sphdata->pass++;
}

static void sphclassical_calc_dens(ParticleData *pa, float UNUSED(dfra), SPHData *sphdata)
{
  ParticleSystem **psys = sphdata->psys;
  SPHFluidSettings *fluid = psys[0]->part->fluid;
  /* 4.0 seems to be a pretty good value */
  float interaction_radius = fluid->radius *
                             (fluid->flag & SPH_FAC_RADIUS ? 4.0f * psys[0]->part->size : 1.0f);
  SPHRangeData pfr;
  float data[2];

  data[0] = 0;
  data[1] = 0;
  pfr.data = data;
  pfr.h = interaction_radius * sphdata->hfac;
  pfr.pa = pa;
  pfr.mass = sphdata->mass;

  sph_evaluate_func(
      NULL, psys, pa->state.co, &pfr, interaction_radius, sphclassical_density_accum_cb);
  pa->sphdensity = min_ff(max_ff(data[0], fluid->rest_density * 0.9f), fluid->rest_density * 1.1f);
}

void psys_sph_init(ParticleSimulationData *sim, SPHData *sphdata)
{
  ParticleTarget *pt;
  int i;

  BLI_buffer_field_init(&sphdata->new_springs, ParticleSpring);

  /* Add other coupled particle systems. */
  sphdata->psys[0] = sim->psys;
  for (i = 1, pt = sim->psys->targets.first; i < 10; i++, pt = (pt ? pt->next : NULL)) {
    sphdata->psys[i] = pt ? psys_get_target_system(sim->ob, pt) : NULL;
  }

  if (psys_uses_gravity(sim)) {
    sphdata->gravity = sim->scene->physics_settings.gravity;
  }
  else {
    sphdata->gravity = NULL;
  }
  sphdata->eh = sph_springhash_build(sim->psys);

  /* These per-particle values should be overridden later, but just for
   * completeness we give them default values now. */
  sphdata->pa = NULL;
  sphdata->mass = 1.0f;

  if (sim->psys->part->fluid->solver == SPH_SOLVER_DDR) {
    sphdata->force_cb = sph_force_cb;
    sphdata->density_cb = sph_density_accum_cb;
    sphdata->hfac = 1.0f;
  }
  else {
    /* SPH_SOLVER_CLASSICAL */
    sphdata->force_cb = sphclassical_force_cb;
    sphdata->density_cb = sphclassical_density_accum_cb;
    sphdata->hfac = 0.5f;
  }
}

static void psys_sph_flush_springs(SPHData *sphdata)
{
  for (int i = 0; i < sphdata->new_springs.count; i++) {
    /* sph_spring_add is not thread-safe. - z0r */
    sph_spring_add(sphdata->psys[0], &BLI_buffer_at(&sphdata->new_springs, ParticleSpring, i));
  }

  BLI_buffer_field_free(&sphdata->new_springs);
}

void psys_sph_finalize(SPHData *sphdata)
{
  psys_sph_flush_springs(sphdata);

  if (sphdata->eh) {
    BLI_edgehash_free(sphdata->eh, NULL);
    sphdata->eh = NULL;
  }
}

/* Sample the density field at a point in space. */
void psys_sph_density(BVHTree *tree, SPHData *sphdata, float co[3], float vars[2])
{
  ParticleSystem **psys = sphdata->psys;
  SPHFluidSettings *fluid = psys[0]->part->fluid;
  /* 4.0 seems to be a pretty good value */
  float interaction_radius = fluid->radius *
                             (fluid->flag & SPH_FAC_RADIUS ? 4.0f * psys[0]->part->size : 1.0f);
  SPHRangeData pfr;
  float density[2];

  density[0] = density[1] = 0.0f;
  pfr.data = density;
  pfr.h = interaction_radius * sphdata->hfac;
  pfr.mass = sphdata->mass;

  sph_evaluate_func(tree, psys, co, &pfr, interaction_radius, sphdata->density_cb);

  vars[0] = pfr.data[0];
  vars[1] = pfr.data[1];
}

static void sph_integrate(ParticleSimulationData *sim,
                          ParticleData *pa,
                          float dfra,
                          SPHData *sphdata)
{
  ParticleSettings *part = sim->psys->part;
  // float timestep = psys_get_timestep(sim); // UNUSED
  float pa_mass = part->mass * (part->flag & PART_SIZEMASS ? pa->size : 1.f);
  float dtime = dfra * psys_get_timestep(sim);
  // int steps = 1; // UNUSED
  float effector_acceleration[3];

  sphdata->pa = pa;
  sphdata->mass = pa_mass;
  sphdata->pass = 0;
  // sphdata.element_size and sphdata.flow are set in the callback.

  /* restore previous state and treat gravity & effectors as external acceleration*/
  sub_v3_v3v3(effector_acceleration, pa->state.vel, pa->prev_state.vel);
  mul_v3_fl(effector_acceleration, 1.f / dtime);

  copy_particle_key(&pa->state, &pa->prev_state, 0);

  integrate_particle(part, pa, dtime, effector_acceleration, sphdata->force_cb, sphdata);
}

/** \} */

/************************************************/
/*          Basic physics                       */
/************************************************/
typedef struct EfData {
  ParticleTexture ptex;
  ParticleSimulationData *sim;
  ParticleData *pa;
} EfData;
static void basic_force_cb(void *efdata_v, ParticleKey *state, float *force, float *impulse)
{
  EfData *efdata = (EfData *)efdata_v;
  ParticleSimulationData *sim = efdata->sim;
  ParticleSettings *part = sim->psys->part;
  ParticleData *pa = efdata->pa;
  EffectedPoint epoint;
  RNG *rng = sim->rng;

  /* add effectors */
  pd_point_from_particle(efdata->sim, efdata->pa, state, &epoint);
  if (part->type != PART_HAIR || part->effector_weights->flag & EFF_WEIGHT_DO_HAIR) {
    BKE_effectors_apply(sim->psys->effectors,
                        sim->colliders,
                        part->effector_weights,
                        &epoint,
                        force,
                        NULL,
                        impulse);
  }

  mul_v3_fl(force, efdata->ptex.field);
  mul_v3_fl(impulse, efdata->ptex.field);

  /* calculate air-particle interaction */
  if (part->dragfac != 0.0f) {
    madd_v3_v3fl(force, state->vel, -part->dragfac * pa->size * pa->size * len_v3(state->vel));
  }

  /* brownian force */
  if (part->brownfac != 0.0f) {
    force[0] += (BLI_rng_get_float(rng) - 0.5f) * part->brownfac;
    force[1] += (BLI_rng_get_float(rng) - 0.5f) * part->brownfac;
    force[2] += (BLI_rng_get_float(rng) - 0.5f) * part->brownfac;
  }

  if (part->flag & PART_ROT_DYN && epoint.ave) {
    copy_v3_v3(pa->state.ave, epoint.ave);
  }
}
/* gathers all forces that effect particles and calculates a new state for the particle */
static void basic_integrate(ParticleSimulationData *sim, int p, float dfra, float cfra)
{
  ParticleSettings *part = sim->psys->part;
  ParticleData *pa = sim->psys->particles + p;
  ParticleKey tkey;
  float dtime = dfra * psys_get_timestep(sim), time;
  float *gravity = NULL, gr[3];
  EfData efdata;

  psys_get_texture(sim, pa, &efdata.ptex, PAMAP_PHYSICS, cfra);

  efdata.pa = pa;
  efdata.sim = sim;

  /* add global acceleration (gravitation) */
  if (psys_uses_gravity(sim) &&
      /* normal gravity is too strong for hair so it's disabled by default */
      (part->type != PART_HAIR || part->effector_weights->flag & EFF_WEIGHT_DO_HAIR)) {
    zero_v3(gr);
    madd_v3_v3fl(gr,
                 sim->scene->physics_settings.gravity,
                 part->effector_weights->global_gravity * efdata.ptex.gravity);
    gravity = gr;
  }

  /* maintain angular velocity */
  copy_v3_v3(pa->state.ave, pa->prev_state.ave);

  integrate_particle(part, pa, dtime, gravity, basic_force_cb, &efdata);

  /* damp affects final velocity */
  if (part->dampfac != 0.f) {
    mul_v3_fl(pa->state.vel, 1.f - part->dampfac * efdata.ptex.damp * 25.f * dtime);
  }

  // copy_v3_v3(pa->state.ave, states->ave);

  /* finally we do guides */
  time = (cfra - pa->time) / pa->lifetime;
  CLAMP(time, 0.0f, 1.0f);

  copy_v3_v3(tkey.co, pa->state.co);
  copy_v3_v3(tkey.vel, pa->state.vel);
  tkey.time = pa->state.time;

  if (part->type != PART_HAIR) {
    if (do_guides(sim->depsgraph, sim->psys->part, sim->psys->effectors, &tkey, p, time)) {
      copy_v3_v3(pa->state.co, tkey.co);
      /* guides don't produce valid velocity */
      sub_v3_v3v3(pa->state.vel, tkey.co, pa->prev_state.co);
      mul_v3_fl(pa->state.vel, 1.0f / dtime);
      pa->state.time = tkey.time;
    }
  }
}
static void basic_rotate(ParticleSettings *part, ParticleData *pa, float dfra, float timestep)
{
  float rotfac, rot1[4], rot2[4] = {1.0, 0.0, 0.0, 0.0}, dtime = dfra * timestep, extrotfac;

  if ((part->flag & PART_ROTATIONS) == 0) {
    unit_qt(pa->state.rot);
    return;
  }

  if (part->flag & PART_ROT_DYN) {
    extrotfac = len_v3(pa->state.ave);
  }
  else {
    extrotfac = 0.0f;
  }

  if ((part->flag & PART_ROT_DYN) &&
      ELEM(part->avemode, PART_AVE_VELOCITY, PART_AVE_HORIZONTAL, PART_AVE_VERTICAL)) {
    float angle;
    float len1 = len_v3(pa->prev_state.vel);
    float len2 = len_v3(pa->state.vel);
    float vec[3];

    if (len1 == 0.0f || len2 == 0.0f) {
      zero_v3(pa->state.ave);
    }
    else {
      cross_v3_v3v3(pa->state.ave, pa->prev_state.vel, pa->state.vel);
      normalize_v3(pa->state.ave);
      angle = dot_v3v3(pa->prev_state.vel, pa->state.vel) / (len1 * len2);
      mul_v3_fl(pa->state.ave, saacos(angle) / dtime);
    }

    get_angular_velocity_vector(part->avemode, &pa->state, vec);
    axis_angle_to_quat(rot2, vec, dtime * part->avefac);
  }

  rotfac = len_v3(pa->state.ave);
  if (rotfac == 0.0f || (part->flag & PART_ROT_DYN) == 0 || extrotfac == 0.0f) {
    unit_qt(rot1);
  }
  else {
    axis_angle_to_quat(rot1, pa->state.ave, rotfac * dtime);
  }
  mul_qt_qtqt(pa->state.rot, rot1, pa->prev_state.rot);
  mul_qt_qtqt(pa->state.rot, rot2, pa->state.rot);

  /* keep rotation quat in good health */
  normalize_qt(pa->state.rot);
}

/************************************************
 * Collisions
 *
 * The algorithm is roughly:
 *  1. Use a BVH tree to search for faces that a particle may collide with.
 *  2. Use Newton's method to find the exact time at which the collision occurs.
 *     https://en.wikipedia.org/wiki/Newton's_method
 *
 ************************************************/
#define COLLISION_MIN_RADIUS 0.001f
#define COLLISION_MIN_DISTANCE 0.0001f
#define COLLISION_ZERO 0.00001f
#define COLLISION_INIT_STEP 0.00008f
typedef float (*NRDistanceFunc)(float *p, float radius, ParticleCollisionElement *pce, float *nor);
static float nr_signed_distance_to_plane(float *p,
                                         float radius,
                                         ParticleCollisionElement *pce,
                                         float *nor)
{
  float p0[3], e1[3], e2[3], d;

  sub_v3_v3v3(e1, pce->x1, pce->x0);
  sub_v3_v3v3(e2, pce->x2, pce->x0);
  sub_v3_v3v3(p0, p, pce->x0);

  cross_v3_v3v3(nor, e1, e2);
  normalize_v3(nor);

  d = dot_v3v3(p0, nor);

  if (pce->inv_nor == -1) {
    if (d < 0.f) {
      pce->inv_nor = 1;
    }
    else {
      pce->inv_nor = 0;
    }
  }

  if (pce->inv_nor == 1) {
    negate_v3(nor);
    d = -d;
  }

  return d - radius;
}
static float nr_distance_to_edge(float *p,
                                 float radius,
                                 ParticleCollisionElement *pce,
                                 float *UNUSED(nor))
{
  float v0[3], v1[3], v2[3], c[3];

  sub_v3_v3v3(v0, pce->x1, pce->x0);
  sub_v3_v3v3(v1, p, pce->x0);
  sub_v3_v3v3(v2, p, pce->x1);

  cross_v3_v3v3(c, v1, v2);

  return fabsf(len_v3(c) / len_v3(v0)) - radius;
}
static float nr_distance_to_vert(float *p,
                                 float radius,
                                 ParticleCollisionElement *pce,
                                 float *UNUSED(nor))
{
  return len_v3v3(p, pce->x0) - radius;
}
static void collision_interpolate_element(ParticleCollisionElement *pce,
                                          float t,
                                          float fac,
                                          ParticleCollision *col)
{
  /* t is the current time for newton rhapson */
  /* fac is the starting factor for current collision iteration */
  /* The col->fac's are factors for the particle subframe step start
   * and end during collision modifier step. */
  float f = fac + t * (1.f - fac);
  float mul = col->fac1 + f * (col->fac2 - col->fac1);
  if (pce->tot > 0) {
    madd_v3_v3v3fl(pce->x0, pce->x[0], pce->v[0], mul);

    if (pce->tot > 1) {
      madd_v3_v3v3fl(pce->x1, pce->x[1], pce->v[1], mul);

      if (pce->tot > 2) {
        madd_v3_v3v3fl(pce->x2, pce->x[2], pce->v[2], mul);
      }
    }
  }
}
static void collision_point_velocity(ParticleCollisionElement *pce)
{
  float v[3];

  copy_v3_v3(pce->vel, pce->v[0]);

  if (pce->tot > 1) {
    sub_v3_v3v3(v, pce->v[1], pce->v[0]);
    madd_v3_v3fl(pce->vel, v, pce->uv[0]);

    if (pce->tot > 2) {
      sub_v3_v3v3(v, pce->v[2], pce->v[0]);
      madd_v3_v3fl(pce->vel, v, pce->uv[1]);
    }
  }
}
static float collision_point_distance_with_normal(
    float p[3], ParticleCollisionElement *pce, float fac, ParticleCollision *col, float *nor)
{
  if (fac >= 0.f) {
    collision_interpolate_element(pce, 0.f, fac, col);
  }

  switch (pce->tot) {
    case 1: {
      sub_v3_v3v3(nor, p, pce->x0);
      return normalize_v3(nor);
    }
    case 2: {
      float u, e[3], vec[3];
      sub_v3_v3v3(e, pce->x1, pce->x0);
      sub_v3_v3v3(vec, p, pce->x0);
      u = dot_v3v3(vec, e) / dot_v3v3(e, e);

      madd_v3_v3v3fl(nor, vec, e, -u);
      return normalize_v3(nor);
    }
    case 3:
      return nr_signed_distance_to_plane(p, 0.f, pce, nor);
  }
  return 0;
}
static void collision_point_on_surface(
    const float p[3], ParticleCollisionElement *pce, float fac, ParticleCollision *col, float *co)
{
  collision_interpolate_element(pce, 0.f, fac, col);

  switch (pce->tot) {
    case 1: {
      sub_v3_v3v3(co, p, pce->x0);
      normalize_v3(co);
      madd_v3_v3v3fl(co, pce->x0, co, col->radius);
      break;
    }
    case 2: {
      float u, e[3], vec[3], nor[3];
      sub_v3_v3v3(e, pce->x1, pce->x0);
      sub_v3_v3v3(vec, p, pce->x0);
      u = dot_v3v3(vec, e) / dot_v3v3(e, e);

      madd_v3_v3v3fl(nor, vec, e, -u);
      normalize_v3(nor);

      madd_v3_v3v3fl(co, pce->x0, e, pce->uv[0]);
      madd_v3_v3fl(co, nor, col->radius);
      break;
    }
    case 3: {
      float p0[3], e1[3], e2[3], nor[3];

      sub_v3_v3v3(e1, pce->x1, pce->x0);
      sub_v3_v3v3(e2, pce->x2, pce->x0);
      sub_v3_v3v3(p0, p, pce->x0);

      cross_v3_v3v3(nor, e1, e2);
      normalize_v3(nor);

      if (pce->inv_nor == 1) {
        negate_v3(nor);
      }

      madd_v3_v3v3fl(co, pce->x0, nor, col->radius);
      madd_v3_v3fl(co, e1, pce->uv[0]);
      madd_v3_v3fl(co, e2, pce->uv[1]);
      break;
    }
  }
}
/* find first root in range [0-1] starting from 0 */
static float collision_newton_rhapson(ParticleCollision *col,
                                      float radius,
                                      ParticleCollisionElement *pce,
                                      NRDistanceFunc distance_func)
{
  float t0, t1, dt_init, d0, d1, dd, n[3];
  int iter;

  pce->inv_nor = -1;

  if (col->inv_total_time > 0.0f) {
    /* Initial step size should be small, but not too small or floating point
     * precision errors will appear. - z0r */
    dt_init = COLLISION_INIT_STEP * col->inv_total_time;
  }
  else {
    dt_init = 0.001f;
  }

  /* start from the beginning */
  t0 = 0.f;
  collision_interpolate_element(pce, t0, col->f, col);
  d0 = distance_func(col->co1, radius, pce, n);
  t1 = dt_init;
  d1 = 0.f;

  for (iter = 0; iter < 10; iter++) {  //, itersum++) {
    /* get current location */
    collision_interpolate_element(pce, t1, col->f, col);
    interp_v3_v3v3(pce->p, col->co1, col->co2, t1);

    d1 = distance_func(pce->p, radius, pce, n);

    /* particle already inside face, so report collision */
    if (iter == 0 && d0 < 0.f && d0 > -radius) {
      copy_v3_v3(pce->p, col->co1);
      copy_v3_v3(pce->nor, n);
      pce->inside = 1;
      return 0.f;
    }

    /* Zero gradient (no movement relative to element). Can't step from
     * here. */
    if (d1 == d0) {
      /* If first iteration, try from other end where the gradient may be
       * greater. Note: code duplicated below. */
      if (iter == 0) {
        t0 = 1.f;
        collision_interpolate_element(pce, t0, col->f, col);
        d0 = distance_func(col->co2, radius, pce, n);
        t1 = 1.0f - dt_init;
        d1 = 0.f;
        continue;
      }

      return -1.f;
    }

    dd = (t1 - t0) / (d1 - d0);

    t0 = t1;
    d0 = d1;

    t1 -= d1 * dd;

    /* Particle moving away from plane could also mean a strangely rotating
     * face, so check from end. Note: code duplicated above. */
    if (iter == 0 && t1 < 0.f) {
      t0 = 1.f;
      collision_interpolate_element(pce, t0, col->f, col);
      d0 = distance_func(col->co2, radius, pce, n);
      t1 = 1.0f - dt_init;
      d1 = 0.f;
      continue;
    }
    if (iter == 1 && (t1 < -COLLISION_ZERO || t1 > 1.f)) {
      return -1.f;
    }

    if (d1 <= COLLISION_ZERO && d1 >= -COLLISION_ZERO) {
      if (t1 >= -COLLISION_ZERO && t1 <= 1.f) {
        if (distance_func == nr_signed_distance_to_plane) {
          copy_v3_v3(pce->nor, n);
        }

        CLAMP(t1, 0.f, 1.f);

        return t1;
      }

      return -1.f;
    }
  }
  return -1.0;
}
static int collision_sphere_to_tri(ParticleCollision *col,
                                   float radius,
                                   ParticleCollisionElement *pce,
                                   float *t)
{
  ParticleCollisionElement *result = &col->pce;
  float ct, u, v;

  pce->inv_nor = -1;
  pce->inside = 0;

  ct = collision_newton_rhapson(col, radius, pce, nr_signed_distance_to_plane);

  if (ct >= 0.f && ct < *t && (result->inside == 0 || pce->inside == 1)) {
    float e1[3], e2[3], p0[3];
    float e1e1, e1e2, e1p0, e2e2, e2p0, inv;

    sub_v3_v3v3(e1, pce->x1, pce->x0);
    sub_v3_v3v3(e2, pce->x2, pce->x0);
    /* XXX: add radius correction here? */
    sub_v3_v3v3(p0, pce->p, pce->x0);

    e1e1 = dot_v3v3(e1, e1);
    e1e2 = dot_v3v3(e1, e2);
    e1p0 = dot_v3v3(e1, p0);
    e2e2 = dot_v3v3(e2, e2);
    e2p0 = dot_v3v3(e2, p0);

    inv = 1.f / (e1e1 * e2e2 - e1e2 * e1e2);
    u = (e2e2 * e1p0 - e1e2 * e2p0) * inv;
    v = (e1e1 * e2p0 - e1e2 * e1p0) * inv;

    if (u >= 0.f && u <= 1.f && v >= 0.f && u + v <= 1.f) {
      *result = *pce;

      /* normal already calculated in pce */

      result->uv[0] = u;
      result->uv[1] = v;

      *t = ct;
      return 1;
    }
  }
  return 0;
}
static int collision_sphere_to_edges(ParticleCollision *col,
                                     float radius,
                                     ParticleCollisionElement *pce,
                                     float *t)
{
  ParticleCollisionElement edge[3], *cur = NULL, *hit = NULL;
  ParticleCollisionElement *result = &col->pce;

  float ct;
  int i;

  for (i = 0; i < 3; i++) {
    cur = edge + i;
    cur->x[0] = pce->x[i];
    cur->x[1] = pce->x[(i + 1) % 3];
    cur->v[0] = pce->v[i];
    cur->v[1] = pce->v[(i + 1) % 3];
    cur->tot = 2;
    cur->inside = 0;

    ct = collision_newton_rhapson(col, radius, cur, nr_distance_to_edge);

    if (ct >= 0.f && ct < *t) {
      float u, e[3], vec[3];

      sub_v3_v3v3(e, cur->x1, cur->x0);
      sub_v3_v3v3(vec, cur->p, cur->x0);
      u = dot_v3v3(vec, e) / dot_v3v3(e, e);

      if (u < 0.f || u > 1.f) {
        break;
      }

      *result = *cur;

      madd_v3_v3v3fl(result->nor, vec, e, -u);
      normalize_v3(result->nor);

      result->uv[0] = u;

      hit = cur;
      *t = ct;
    }
  }

  return hit != NULL;
}
static int collision_sphere_to_verts(ParticleCollision *col,
                                     float radius,
                                     ParticleCollisionElement *pce,
                                     float *t)
{
  ParticleCollisionElement vert[3], *cur = NULL, *hit = NULL;
  ParticleCollisionElement *result = &col->pce;

  float ct;
  int i;

  for (i = 0; i < 3; i++) {
    cur = vert + i;
    cur->x[0] = pce->x[i];
    cur->v[0] = pce->v[i];
    cur->tot = 1;
    cur->inside = 0;

    ct = collision_newton_rhapson(col, radius, cur, nr_distance_to_vert);

    if (ct >= 0.f && ct < *t) {
      *result = *cur;

      sub_v3_v3v3(result->nor, cur->p, cur->x0);
      normalize_v3(result->nor);

      hit = cur;
      *t = ct;
    }
  }

  return hit != NULL;
}
/* Callback for BVHTree near test */
void BKE_psys_collision_neartest_cb(void *userdata,
                                    int index,
                                    const BVHTreeRay *ray,
                                    BVHTreeRayHit *hit)
{
  ParticleCollision *col = (ParticleCollision *)userdata;
  ParticleCollisionElement pce;
  const MVertTri *vt = &col->md->tri[index];
  MVert *x = col->md->x;
  MVert *v = col->md->current_v;
  float t = hit->dist / col->original_ray_length;
  int collision = 0;

  pce.x[0] = x[vt->tri[0]].co;
  pce.x[1] = x[vt->tri[1]].co;
  pce.x[2] = x[vt->tri[2]].co;

  pce.v[0] = v[vt->tri[0]].co;
  pce.v[1] = v[vt->tri[1]].co;
  pce.v[2] = v[vt->tri[2]].co;

  pce.tot = 3;
  pce.inside = 0;
  pce.index = index;

  collision = collision_sphere_to_tri(col, ray->radius, &pce, &t);
  if (col->pce.inside == 0) {
    collision += collision_sphere_to_edges(col, ray->radius, &pce, &t);
    collision += collision_sphere_to_verts(col, ray->radius, &pce, &t);
  }

  if (collision) {
    hit->dist = col->original_ray_length * t;
    hit->index = index;

    collision_point_velocity(&col->pce);

    col->hit = col->current;
  }
}
static int collision_detect(ParticleData *pa,
                            ParticleCollision *col,
                            BVHTreeRayHit *hit,
                            ListBase *colliders)
{
  const int raycast_flag = BVH_RAYCAST_DEFAULT & ~(BVH_RAYCAST_WATERTIGHT);
  ColliderCache *coll;
  float ray_dir[3];

  if (BLI_listbase_is_empty(colliders)) {
    return 0;
  }

  sub_v3_v3v3(ray_dir, col->co2, col->co1);
  hit->index = -1;
  hit->dist = col->original_ray_length = normalize_v3(ray_dir);
  col->pce.inside = 0;

  /* even if particle is stationary we want to check for moving colliders */
  /* if hit.dist is zero the bvhtree_ray_cast will just ignore everything */
  if (hit->dist == 0.0f) {
    hit->dist = col->original_ray_length = 0.000001f;
  }

  for (coll = colliders->first; coll; coll = coll->next) {
    /* for boids: don't check with current ground object; also skip if permeated */
    bool skip = false;

    for (int i = 0; i < col->skip_count; i++) {
      if (coll->ob == col->skip[i]) {
        skip = true;
        break;
      }
    }

    if (skip) {
      continue;
    }

    /* particles should not collide with emitter at birth */
    if (coll->ob == col->emitter && pa->time < col->cfra && pa->time >= col->old_cfra) {
      continue;
    }

    col->current = coll->ob;
    col->md = coll->collmd;
    col->fac1 = (col->old_cfra - coll->collmd->time_x) /
                (coll->collmd->time_xnew - coll->collmd->time_x);
    col->fac2 = (col->cfra - coll->collmd->time_x) /
                (coll->collmd->time_xnew - coll->collmd->time_x);

    if (col->md && col->md->bvhtree) {
      BLI_bvhtree_ray_cast_ex(col->md->bvhtree,
                              col->co1,
                              ray_dir,
                              col->radius,
                              hit,
                              BKE_psys_collision_neartest_cb,
                              col,
                              raycast_flag);
    }
  }

  return hit->index >= 0;
}
static int collision_response(ParticleSimulationData *sim,
                              ParticleData *pa,
                              ParticleCollision *col,
                              BVHTreeRayHit *hit,
                              int kill,
                              int dynamic_rotation)
{
  ParticleCollisionElement *pce = &col->pce;
  PartDeflect *pd = col->hit->pd;
  RNG *rng = sim->rng;
  /* point of collision */
  float co[3];
  /* location factor of collision between this iteration */
  float x = hit->dist / col->original_ray_length;
  /* time factor of collision between timestep */
  float f = col->f + x * (1.0f - col->f);
  /* time since previous collision (in seconds) */
  float dt1 = (f - col->f) * col->total_time;
  /* time left after collision (in seconds) */
  float dt2 = (1.0f - f) * col->total_time;
  /* did particle pass through the collision surface? */
  int through = (BLI_rng_get_float(rng) < pd->pdef_perm) ? 1 : 0;

  /* calculate exact collision location */
  interp_v3_v3v3(co, col->co1, col->co2, x);

  /* particle dies in collision */
  if (through == 0 && (kill || pd->flag & PDEFLE_KILL_PART)) {
    pa->alive = PARS_DYING;
    pa->dietime = col->old_cfra + (col->cfra - col->old_cfra) * f;

    copy_v3_v3(pa->state.co, co);
    interp_v3_v3v3(pa->state.vel, pa->prev_state.vel, pa->state.vel, f);
    interp_qt_qtqt(pa->state.rot, pa->prev_state.rot, pa->state.rot, f);
    interp_v3_v3v3(pa->state.ave, pa->prev_state.ave, pa->state.ave, f);

    /* particle is dead so we don't need to calculate further */
    return 0;
  }
  /* figure out velocity and other data after collision */

  /* velocity directly before collision to be modified into velocity directly after collision */
  float v0[3];
  /* normal component of v0 */
  float v0_nor[3];
  /* tangential component of v0 */
  float v0_tan[3];
  /* tangential component of collision surface velocity */
  float vc_tan[3];
  float v0_dot, vc_dot;
  float damp = pd->pdef_damp + pd->pdef_rdamp * 2 * (BLI_rng_get_float(rng) - 0.5f);
  float frict = pd->pdef_frict + pd->pdef_rfrict * 2 * (BLI_rng_get_float(rng) - 0.5f);
  float distance, nor[3], dot;

  CLAMP(damp, 0.0f, 1.0f);
  CLAMP(frict, 0.0f, 1.0f);

  /* get exact velocity right before collision */
  madd_v3_v3v3fl(v0, col->ve1, col->acc, dt1);

  /* Convert collider velocity from `1/frame_step` to `1/s` TODO:
   * here we assume 1 frame step for collision modifier. */
  mul_v3_fl(pce->vel, col->inv_timestep);

  /* calculate tangential particle velocity */
  v0_dot = dot_v3v3(pce->nor, v0);
  madd_v3_v3v3fl(v0_tan, v0, pce->nor, -v0_dot);

  /* calculate tangential collider velocity */
  vc_dot = dot_v3v3(pce->nor, pce->vel);
  madd_v3_v3v3fl(vc_tan, pce->vel, pce->nor, -vc_dot);

  /* handle friction effects (tangential and angular velocity) */
  if (frict > 0.0f) {
    /* angular <-> linear velocity */
    if (dynamic_rotation) {
      float vr_tan[3], v1_tan[3], ave[3];

      /* linear velocity of particle surface */
      cross_v3_v3v3(vr_tan, pce->nor, pa->state.ave);
      mul_v3_fl(vr_tan, pa->size);

      /* change to coordinates that move with the collision plane */
      sub_v3_v3v3(v1_tan, v0_tan, vc_tan);

      /* The resulting velocity is a weighted average of particle cm & surface
       * velocity. This weight (related to particle's moment of inertia) could
       * be made a parameter for angular <-> linear conversion.
       */
      madd_v3_v3fl(v1_tan, vr_tan, -0.4);
      mul_v3_fl(v1_tan, 1.0f / 1.4f); /* 1/(1+0.4) */

      /* rolling friction is around 0.01 of sliding friction
       * (could be made a parameter) */
      mul_v3_fl(v1_tan, 1.0f - 0.01f * frict);

      /* surface_velocity is opposite to cm velocity */
      negate_v3_v3(vr_tan, v1_tan);

      /* get back to global coordinates */
      add_v3_v3(v1_tan, vc_tan);

      /* convert to angular velocity*/
      cross_v3_v3v3(ave, vr_tan, pce->nor);
      mul_v3_fl(ave, 1.0f / MAX2(pa->size, 0.001f));

      /* only friction will cause change in linear & angular velocity */
      interp_v3_v3v3(pa->state.ave, pa->state.ave, ave, frict);
      interp_v3_v3v3(v0_tan, v0_tan, v1_tan, frict);
    }
    else {
      /* just basic friction (unphysical due to the friction model used in Blender) */
      interp_v3_v3v3(v0_tan, v0_tan, vc_tan, frict);
    }
  }

  /* Stickiness was possibly added before,
   * so cancel that before calculating new normal velocity.
   * Otherwise particles go flying out of the surface
   * because of high reversed sticky velocity. */
  if (v0_dot < 0.0f) {
    v0_dot += pd->pdef_stickness;
    if (v0_dot > 0.0f) {
      v0_dot = 0.0f;
    }
  }

  /* damping and flipping of velocity around normal */
  v0_dot *= 1.0f - damp;
  vc_dot *= through ? damp : 1.0f;

  /* calculate normal particle velocity */
  /* special case for object hitting the particle from behind */
  if (through == 0 && ((vc_dot > 0.0f && v0_dot > 0.0f && vc_dot > v0_dot) ||
                       (vc_dot < 0.0f && v0_dot < 0.0f && vc_dot < v0_dot))) {
    mul_v3_v3fl(v0_nor, pce->nor, vc_dot);
  }
  else if (v0_dot > 0.f) {
    mul_v3_v3fl(v0_nor, pce->nor, vc_dot + v0_dot);
  }
  else {
    mul_v3_v3fl(v0_nor, pce->nor, vc_dot + (through ? 1.0f : -1.0f) * v0_dot);
  }

  /* combine components together again */
  add_v3_v3v3(v0, v0_nor, v0_tan);

  if (col->boid) {
    /* keep boids above ground */
    BoidParticle *bpa = pa->boid;
    if (bpa->data.mode == eBoidMode_OnLand || co[2] <= col->boid_z) {
      co[2] = col->boid_z;
      v0[2] = 0.0f;
    }
  }

  /* re-apply acceleration to final location and velocity */
  madd_v3_v3v3fl(pa->state.co, co, v0, dt2);
  madd_v3_v3fl(pa->state.co, col->acc, 0.5f * dt2 * dt2);
  madd_v3_v3v3fl(pa->state.vel, v0, col->acc, dt2);

  /* make sure particle stays on the right side of the surface */
  if (!through) {
    distance = collision_point_distance_with_normal(co, pce, -1.f, col, nor);

    if (distance < col->radius + COLLISION_MIN_DISTANCE) {
      madd_v3_v3fl(co, nor, col->radius + COLLISION_MIN_DISTANCE - distance);
    }

    dot = dot_v3v3(nor, v0);
    if (dot < 0.f) {
      madd_v3_v3fl(v0, nor, -dot);
    }

    distance = collision_point_distance_with_normal(pa->state.co, pce, 1.f, col, nor);

    if (distance < col->radius + COLLISION_MIN_DISTANCE) {
      madd_v3_v3fl(pa->state.co, nor, col->radius + COLLISION_MIN_DISTANCE - distance);
    }

    dot = dot_v3v3(nor, pa->state.vel);
    if (dot < 0.f) {
      madd_v3_v3fl(pa->state.vel, nor, -dot);
    }
  }

  /* add stickiness to surface */
  madd_v3_v3fl(pa->state.vel, pce->nor, -pd->pdef_stickness);

  /* set coordinates for next iteration */
  copy_v3_v3(col->co1, co);
  copy_v3_v3(col->co2, pa->state.co);

  copy_v3_v3(col->ve1, v0);
  copy_v3_v3(col->ve2, pa->state.vel);

  col->f = f;

  /* if permeability random roll succeeded, disable collider for this sim step */
  if (through) {
    col->skip[col->skip_count++] = col->hit;
  }

  return 1;
}
static void collision_fail(ParticleData *pa, ParticleCollision *col)
{
  /* final chance to prevent total failure, so stick to the surface and hope for the best */
  collision_point_on_surface(col->co1, &col->pce, 1.f, col, pa->state.co);

  copy_v3_v3(pa->state.vel, col->pce.vel);
  mul_v3_fl(pa->state.vel, col->inv_timestep);

  /* printf("max iterations\n"); */
}

/* Particle - Mesh collision detection and response
 * Features:
 * -friction and damping
 * -angular momentum <-> linear momentum
 * -high accuracy by re-applying particle acceleration after collision
 * -handles moving, rotating and deforming meshes
 * -uses Newton-Rhapson iteration to find the collisions
 * -handles spherical particles and (nearly) point like particles
 */
static void collision_check(ParticleSimulationData *sim, int p, float dfra, float cfra)
{
  ParticleSettings *part = sim->psys->part;
  ParticleData *pa = sim->psys->particles + p;
  ParticleCollision col;
  BVHTreeRayHit hit;
  int collision_count = 0;

  float timestep = psys_get_timestep(sim);

  memset(&col, 0, sizeof(ParticleCollision));

  col.total_time = timestep * dfra;
  col.inv_total_time = 1.0f / col.total_time;
  col.inv_timestep = 1.0f / timestep;

  col.cfra = cfra;
  col.old_cfra = sim->psys->cfra;

  /* get acceleration (from gravity, forcefields etc. to be re-applied in collision response) */
  sub_v3_v3v3(col.acc, pa->state.vel, pa->prev_state.vel);
  mul_v3_fl(col.acc, 1.f / col.total_time);

  /* set values for first iteration */
  copy_v3_v3(col.co1, pa->prev_state.co);
  copy_v3_v3(col.co2, pa->state.co);
  copy_v3_v3(col.ve1, pa->prev_state.vel);
  copy_v3_v3(col.ve2, pa->state.vel);
  col.f = 0.0f;

  col.radius = ((part->flag & PART_SIZE_DEFL) || (part->phystype == PART_PHYS_BOIDS)) ?
                   pa->size :
                   COLLISION_MIN_RADIUS;

  /* override for boids */
  if (part->phystype == PART_PHYS_BOIDS && part->boids->options & BOID_ALLOW_LAND) {
    col.boid = 1;
    col.boid_z = pa->state.co[2];
    col.skip[col.skip_count++] = pa->boid->ground;
  }

  /* 10 iterations to catch multiple collisions */
  while (collision_count < PARTICLE_COLLISION_MAX_COLLISIONS) {
    if (collision_detect(pa, &col, &hit, sim->colliders)) {

      collision_count++;

      if (collision_count == PARTICLE_COLLISION_MAX_COLLISIONS) {
        collision_fail(pa, &col);
      }
      else if (collision_response(
                   sim, pa, &col, &hit, part->flag & PART_DIE_ON_COL, part->flag & PART_ROT_DYN) ==
               0) {
        return;
      }
    }
    else {
      return;
    }
  }
}
/************************************************/
/*          Hair                                */
/************************************************/
/* check if path cache or children need updating and do it if needed */
static void psys_update_path_cache(ParticleSimulationData *sim,
                                   float cfra,
                                   const bool use_render_params)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  ParticleEditSettings *pset = &sim->scene->toolsettings->particle;
  int distr = 0, alloc = 0, skip = 0;

  if ((psys->part->childtype &&
       psys->totchild != psys_get_tot_child(sim->scene, psys, use_render_params)) ||
      psys->recalc & ID_RECALC_PSYS_RESET) {
    alloc = 1;
  }

  if (alloc || psys->recalc & ID_RECALC_PSYS_CHILD ||
      (psys->vgroup[PSYS_VG_DENSITY] && (sim->ob && sim->ob->mode & OB_MODE_WEIGHT_PAINT))) {
    distr = 1;
  }

  if (distr) {
    if (alloc) {
      realloc_particles(sim, sim->psys->totpart);
    }

    if (psys_get_tot_child(sim->scene, psys, use_render_params)) {
      /* don't generate children while computing the hair keys */
      if (!(psys->part->type == PART_HAIR) || (psys->flag & PSYS_HAIR_DONE)) {
        distribute_particles(sim, PART_FROM_CHILD);

        if (part->childtype == PART_CHILD_FACES && part->parents != 0.0f) {
          psys_find_parents(sim, use_render_params);
        }
      }
    }
    else {
      psys_free_children(psys);
    }
  }

  if ((part->type == PART_HAIR || psys->flag & PSYS_KEYED ||
       psys->pointcache->flag & PTCACHE_BAKED) == 0) {
    skip = 1; /* only hair, keyed and baked stuff can have paths */
  }
  else if (part->ren_as != PART_DRAW_PATH &&
           !(part->type == PART_HAIR && ELEM(part->ren_as, PART_DRAW_OB, PART_DRAW_GR))) {
    skip = 1; /* particle visualization must be set as path */
  }
  else if (DEG_get_mode(sim->depsgraph) != DAG_EVAL_RENDER) {
    if (part->draw_as != PART_DRAW_REND) {
      skip = 1; /* draw visualization */
    }
    else if (psys->pointcache->flag & PTCACHE_BAKING) {
      skip = 1; /* no need to cache paths while baking dynamics */
    }
    else if (psys_in_edit_mode(sim->depsgraph, psys)) {
      if ((pset->flag & PE_DRAW_PART) == 0) {
        skip = 1;
      }
      else if (part->childtype == 0 &&
               (psys->flag & PSYS_HAIR_DYNAMICS && psys->pointcache->flag & PTCACHE_BAKED) == 0) {
        skip = 1; /* in edit mode paths are needed for child particles and dynamic hair */
      }
    }
  }

  if (!skip) {
    psys_cache_paths(sim, cfra, use_render_params);

    /* for render, child particle paths are computed on the fly */
    if (part->childtype) {
      if (!psys->totchild) {
        skip = 1;
      }
      else if (psys->part->type == PART_HAIR && (psys->flag & PSYS_HAIR_DONE) == 0) {
        skip = 1;
      }

      if (!skip) {
        psys_cache_child_paths(sim, cfra, 0, use_render_params);
      }
    }
  }
  else if (psys->pathcache) {
    psys_free_path_cache(psys, NULL);
  }
}

static bool psys_hair_use_simulation(ParticleData *pa, float max_length)
{
  /* Minimum segment length relative to average length.
   * Hairs with segments below this length will be excluded from the simulation,
   * because otherwise the solver will become unstable.
   * The hair system should always make sure the hair segments have reasonable length ratios,
   * but this can happen in old files when e.g. cutting hair.
   */
  const float min_length = 0.1f * max_length;

  HairKey *key;
  int k;

  if (pa->totkey < 2) {
    return false;
  }

  for (k = 1, key = pa->hair + 1; k < pa->totkey; k++, key++) {
    float length = len_v3v3(key->co, (key - 1)->co);
    if (length < min_length) {
      return false;
    }
  }

  return true;
}

static MDeformVert *hair_set_pinning(MDeformVert *dvert, float weight)
{
  if (dvert) {
    if (!dvert->totweight) {
      dvert->dw = MEM_callocN(sizeof(MDeformWeight), "deformWeight");
      dvert->totweight = 1;
    }

    dvert->dw->weight = weight;
    dvert++;
  }
  return dvert;
}

static void hair_create_input_mesh(ParticleSimulationData *sim,
                                   int totpoint,
                                   int totedge,
                                   Mesh **r_mesh,
                                   ClothHairData **r_hairdata)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  Mesh *mesh;
  ClothHairData *hairdata;
  MVert *mvert;
  MEdge *medge;
  MDeformVert *dvert;
  HairKey *key;
  PARTICLE_P;
  int k, hair_index;
  float hairmat[4][4];
  float max_length;
  float hair_radius;

  mesh = *r_mesh;
  if (!mesh) {
    *r_mesh = mesh = BKE_mesh_new_nomain(totpoint, totedge, 0, 0, 0);
    CustomData_add_layer(&mesh->vdata, CD_MDEFORMVERT, CD_CALLOC, NULL, mesh->totvert);
    BKE_mesh_update_customdata_pointers(mesh, false);
  }
  mvert = mesh->mvert;
  medge = mesh->medge;
  dvert = mesh->dvert;

  hairdata = *r_hairdata;
  if (!hairdata) {
    *r_hairdata = hairdata = MEM_mallocN(sizeof(ClothHairData) * totpoint, "hair data");
  }

  /* calculate maximum segment length */
  max_length = 0.0f;
  LOOP_PARTICLES
  {
    if (!(pa->flag & PARS_UNEXIST)) {
      for (k = 1, key = pa->hair + 1; k < pa->totkey; k++, key++) {
        float length = len_v3v3(key->co, (key - 1)->co);
        if (max_length < length) {
          max_length = length;
        }
      }
    }
  }

  psys->clmd->sim_parms->vgroup_mass = 1;

  /* XXX placeholder for more flexible future hair settings */
  hair_radius = part->size;

  /* make vgroup for pin roots etc.. */
  hair_index = 1;
  LOOP_PARTICLES
  {
    if (!(pa->flag & PARS_UNEXIST)) {
      float root_mat[4][4];
      float bending_stiffness;
      bool use_hair;

      pa->hair_index = hair_index;
      use_hair = psys_hair_use_simulation(pa, max_length);

      psys_mat_hair_to_object(sim->ob, sim->psmd->mesh_final, psys->part->from, pa, hairmat);
      mul_m4_m4m4(root_mat, sim->ob->obmat, hairmat);
      normalize_m4(root_mat);

      bending_stiffness = CLAMPIS(
          1.0f - part->bending_random * psys_frand(psys, p + 666), 0.0f, 1.0f);

      for (k = 0, key = pa->hair; k < pa->totkey; k++, key++) {
        ClothHairData *hair;
        float *co, *co_next;

        co = key->co;
        co_next = (key + 1)->co;

        /* create fake root before actual root to resist bending */
        if (k == 0) {
          hair = &psys->clmd->hairdata[pa->hair_index - 1];
          copy_v3_v3(hair->loc, root_mat[3]);
          copy_m3_m4(hair->rot, root_mat);

          hair->radius = hair_radius;
          hair->bending_stiffness = bending_stiffness;

          add_v3_v3v3(mvert->co, co, co);
          sub_v3_v3(mvert->co, co_next);
          mul_m4_v3(hairmat, mvert->co);

          medge->v1 = pa->hair_index - 1;
          medge->v2 = pa->hair_index;

          dvert = hair_set_pinning(dvert, 1.0f);

          mvert++;
          medge++;
        }

        /* store root transform in cloth data */
        hair = &psys->clmd->hairdata[pa->hair_index + k];
        copy_v3_v3(hair->loc, root_mat[3]);
        copy_m3_m4(hair->rot, root_mat);

        hair->radius = hair_radius;
        hair->bending_stiffness = bending_stiffness;

        copy_v3_v3(mvert->co, co);
        mul_m4_v3(hairmat, mvert->co);

        if (k) {
          medge->v1 = pa->hair_index + k - 1;
          medge->v2 = pa->hair_index + k;
        }

        /* roots and disabled hairs should be 1.0, the rest can be anything from 0.0 to 1.0 */
        if (use_hair) {
          dvert = hair_set_pinning(dvert, key->weight);
        }
        else {
          dvert = hair_set_pinning(dvert, 1.0f);
        }

        mvert++;
        if (k) {
          medge++;
        }
      }

      hair_index += pa->totkey + 1;
    }
  }
}

static void do_hair_dynamics(ParticleSimulationData *sim)
{
  ParticleSystem *psys = sim->psys;
  PARTICLE_P;
  EffectorWeights *clmd_effweights;
  int totpoint;
  int totedge;
  float(*deformedVerts)[3];
  bool realloc_roots;

  if (!psys->clmd) {
    psys->clmd = (ClothModifierData *)BKE_modifier_new(eModifierType_Cloth);
    psys->clmd->sim_parms->goalspring = 0.0f;
    psys->clmd->sim_parms->flags |= CLOTH_SIMSETTINGS_FLAG_RESIST_SPRING_COMPRESS;
    psys->clmd->coll_parms->flags &= ~CLOTH_COLLSETTINGS_FLAG_SELF;
  }

  /* count simulated points */
  totpoint = 0;
  totedge = 0;
  LOOP_PARTICLES
  {
    if (!(pa->flag & PARS_UNEXIST)) {
      /* "out" dm contains all hairs */
      totedge += pa->totkey;
      totpoint += pa->totkey + 1; /* +1 for virtual root point */
    }
  }

  /* whether hair root info array has to be reallocated */
  realloc_roots = false;
  if (psys->hair_in_mesh) {
    Mesh *mesh = psys->hair_in_mesh;
    if (totpoint != mesh->totvert || totedge != mesh->totedge) {
      BKE_id_free(NULL, mesh);
      psys->hair_in_mesh = NULL;
      realloc_roots = true;
    }
  }

  if (!psys->hair_in_mesh || !psys->clmd->hairdata || realloc_roots) {
    if (psys->clmd->hairdata) {
      MEM_freeN(psys->clmd->hairdata);
      psys->clmd->hairdata = NULL;
    }
  }

  hair_create_input_mesh(sim, totpoint, totedge, &psys->hair_in_mesh, &psys->clmd->hairdata);

  if (psys->hair_out_mesh) {
    BKE_id_free(NULL, psys->hair_out_mesh);
  }

  psys->clmd->point_cache = psys->pointcache;
  /* for hair sim we replace the internal cloth effector weights temporarily
   * to use the particle settings
   */
  clmd_effweights = psys->clmd->sim_parms->effector_weights;
  psys->clmd->sim_parms->effector_weights = psys->part->effector_weights;

  BKE_id_copy_ex(NULL, &psys->hair_in_mesh->id, (ID **)&psys->hair_out_mesh, LIB_ID_COPY_LOCALIZE);
  deformedVerts = BKE_mesh_vert_coords_alloc(psys->hair_out_mesh, NULL);
  clothModifier_do(
      psys->clmd, sim->depsgraph, sim->scene, sim->ob, psys->hair_in_mesh, deformedVerts);
  BKE_mesh_vert_coords_apply(psys->hair_out_mesh, deformedVerts);

  MEM_freeN(deformedVerts);

  /* restore cloth effector weights */
  psys->clmd->sim_parms->effector_weights = clmd_effweights;
}
static void hair_step(ParticleSimulationData *sim, float cfra, const bool use_render_params)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  PARTICLE_P;
  float disp = psys_get_current_display_percentage(psys, use_render_params);

  LOOP_PARTICLES
  {
    pa->size = part->size;
    if (part->randsize > 0.0f) {
      pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
    }

    if (psys_frand(psys, p) > disp) {
      pa->flag |= PARS_NO_DISP;
    }
    else {
      pa->flag &= ~PARS_NO_DISP;
    }
  }

  if (psys->recalc & ID_RECALC_PSYS_RESET) {
    /* need this for changing subsurf levels */
    psys_calc_dmcache(sim->ob, sim->psmd->mesh_final, sim->psmd->mesh_original, psys);

    if (psys->clmd) {
      cloth_free_modifier(psys->clmd);
    }
  }

  /* dynamics with cloth simulation, psys->particles can be NULL with 0 particles T25519. */
  if (psys->part->type == PART_HAIR && psys->flag & PSYS_HAIR_DYNAMICS && psys->particles) {
    do_hair_dynamics(sim);
  }

  /* following lines were removed r29079 but cause bug T22811, see report for details */
  psys_update_effectors(sim);
  psys_update_path_cache(sim, cfra, use_render_params);

  psys->flag |= PSYS_HAIR_UPDATED;
}

static void save_hair(ParticleSimulationData *sim, float UNUSED(cfra))
{
  Object *ob = sim->ob;
  ParticleSystem *psys = sim->psys;
  HairKey *key, *root;
  PARTICLE_P;

  invert_m4_m4(ob->imat, ob->obmat);

  psys->lattice_deform_data = psys_create_lattice_deform_data(sim);

  if (psys->totpart == 0) {
    return;
  }

  /* save new keys for elements if needed */
  LOOP_PARTICLES
  {
    /* first time alloc */
    if (pa->totkey == 0 || pa->hair == NULL) {
      pa->hair = MEM_callocN((psys->part->hair_step + 1) * sizeof(HairKey), "HairKeys");
      pa->totkey = 0;
    }

    key = root = pa->hair;
    key += pa->totkey;

    /* convert from global to geometry space */
    copy_v3_v3(key->co, pa->state.co);
    mul_m4_v3(ob->imat, key->co);

    if (pa->totkey) {
      sub_v3_v3(key->co, root->co);
      psys_vec_rot_to_face(sim->psmd->mesh_final, pa, key->co);
    }

    key->time = pa->state.time;

    key->weight = 1.0f - key->time / 100.0f;

    pa->totkey++;

    /* Root is always in the origin of hair space
     * so we set it to be so after the last key is saved. */
    if (pa->totkey == psys->part->hair_step + 1) {
      zero_v3(root->co);
    }
  }
}

/* Code for an adaptive time step based on the Courant-Friedrichs-Lewy
 * condition. */
static const float MIN_TIMESTEP = 1.0f / 101.0f;
/* Tolerance of 1.5 means the last subframe neither favors growing nor
 * shrinking (e.g if it were 1.3, the last subframe would tend to be too
 * small). */
static const float TIMESTEP_EXPANSION_FACTOR = 0.1f;
static const float TIMESTEP_EXPANSION_TOLERANCE = 1.5f;

/* Calculate the speed of the particle relative to the local scale of the
 * simulation. This should be called once per particle during a simulation
 * step, after the velocity has been updated. element_size defines the scale of
 * the simulation, and is typically the distance to neighboring particles. */
static void update_courant_num(
    ParticleSimulationData *sim, ParticleData *pa, float dtime, SPHData *sphdata, SpinLock *spin)
{
  float relative_vel[3];

  sub_v3_v3v3(relative_vel, pa->prev_state.vel, sphdata->flow);

  const float courant_num = len_v3(relative_vel) * dtime / sphdata->element_size;
  if (sim->courant_num < courant_num) {
    BLI_spin_lock(spin);
    if (sim->courant_num < courant_num) {
      sim->courant_num = courant_num;
    }
    BLI_spin_unlock(spin);
  }
}
static float get_base_time_step(ParticleSettings *part)
{
  return 1.0f / (float)(part->subframes + 1);
}
/* Update time step size to suit current conditions. */
static void update_timestep(ParticleSystem *psys, ParticleSimulationData *sim)
{
  float dt_target;
  if (sim->courant_num == 0.0f) {
    dt_target = 1.0f;
  }
  else {
    dt_target = psys->dt_frac * (psys->part->courant_target / sim->courant_num);
  }

  /* Make sure the time step is reasonable. For some reason, the CLAMP macro
   * doesn't work here. The time step becomes too large. - z0r */
  if (dt_target < MIN_TIMESTEP) {
    dt_target = MIN_TIMESTEP;
  }
  else if (dt_target > get_base_time_step(psys->part)) {
    dt_target = get_base_time_step(psys->part);
  }

  /* Decrease time step instantly, but increase slowly. */
  if (dt_target > psys->dt_frac) {
    psys->dt_frac = interpf(dt_target, psys->dt_frac, TIMESTEP_EXPANSION_FACTOR);
  }
  else {
    psys->dt_frac = dt_target;
  }
}

static float sync_timestep(ParticleSystem *psys, float t_frac)
{
  /* Sync with frame end if it's close. */
  if (t_frac == 1.0f) {
    return psys->dt_frac;
  }
  if (t_frac + (psys->dt_frac * TIMESTEP_EXPANSION_TOLERANCE) >= 1.0f) {
    return 1.0f - t_frac;
  }

  return psys->dt_frac;
}

/************************************************/
/*          System Core                         */
/************************************************/

typedef struct DynamicStepSolverTaskData {
  ParticleSimulationData *sim;

  float cfra;
  float timestep;
  float dtime;

  SpinLock spin;
} DynamicStepSolverTaskData;

static void dynamics_step_sphdata_reduce(const void *__restrict UNUSED(userdata),
                                         void *__restrict UNUSED(join_v),
                                         void *__restrict chunk_v)
{
  SPHData *sphdata = chunk_v;

  psys_sph_flush_springs(sphdata);
}

static void dynamics_step_sph_ddr_task_cb_ex(void *__restrict userdata,
                                             const int p,
                                             const TaskParallelTLS *__restrict tls)
{
  DynamicStepSolverTaskData *data = userdata;
  ParticleSimulationData *sim = data->sim;
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;

  SPHData *sphdata = tls->userdata_chunk;

  ParticleData *pa;

  if ((pa = psys->particles + p)->state.time <= 0.0f) {
    return;
  }

  /* do global forces & effectors */
  basic_integrate(sim, p, pa->state.time, data->cfra);

  /* actual fluids calculations */
  sph_integrate(sim, pa, pa->state.time, sphdata);

  if (sim->colliders) {
    collision_check(sim, p, pa->state.time, data->cfra);
  }

  /* SPH particles are not physical particles, just interpolation
   * particles,  thus rotation has not a direct sense for them */
  basic_rotate(part, pa, pa->state.time, data->timestep);

  if (part->time_flag & PART_TIME_AUTOSF) {
    update_courant_num(sim, pa, data->dtime, sphdata, &data->spin);
  }
}

static void dynamics_step_sph_classical_basic_integrate_task_cb_ex(
    void *__restrict userdata, const int p, const TaskParallelTLS *__restrict UNUSED(tls))
{
  DynamicStepSolverTaskData *data = userdata;
  ParticleSimulationData *sim = data->sim;
  ParticleSystem *psys = sim->psys;

  ParticleData *pa;

  if ((pa = psys->particles + p)->state.time <= 0.0f) {
    return;
  }

  basic_integrate(sim, p, pa->state.time, data->cfra);
}

static void dynamics_step_sph_classical_calc_density_task_cb_ex(
    void *__restrict userdata, const int p, const TaskParallelTLS *__restrict tls)
{
  DynamicStepSolverTaskData *data = userdata;
  ParticleSimulationData *sim = data->sim;
  ParticleSystem *psys = sim->psys;

  SPHData *sphdata = tls->userdata_chunk;

  ParticleData *pa;

  if ((pa = psys->particles + p)->state.time <= 0.0f) {
    return;
  }

  sphclassical_calc_dens(pa, pa->state.time, sphdata);
}

static void dynamics_step_sph_classical_integrate_task_cb_ex(void *__restrict userdata,
                                                             const int p,
                                                             const TaskParallelTLS *__restrict tls)
{
  DynamicStepSolverTaskData *data = userdata;
  ParticleSimulationData *sim = data->sim;
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;

  SPHData *sphdata = tls->userdata_chunk;

  ParticleData *pa;

  if ((pa = psys->particles + p)->state.time <= 0.0f) {
    return;
  }

  /* actual fluids calculations */
  sph_integrate(sim, pa, pa->state.time, sphdata);

  if (sim->colliders) {
    collision_check(sim, p, pa->state.time, data->cfra);
  }

  /* SPH particles are not physical particles, just interpolation
   * particles,  thus rotation has not a direct sense for them */
  basic_rotate(part, pa, pa->state.time, data->timestep);

  if (part->time_flag & PART_TIME_AUTOSF) {
    update_courant_num(sim, pa, data->dtime, sphdata, &data->spin);
  }
}

/* unbaked particles are calculated dynamically */
static void dynamics_step(ParticleSimulationData *sim, float cfra)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  BoidBrainData bbd;
  ParticleTexture ptex;
  PARTICLE_P;
  float timestep;
  /* frame & time changes */
  float dfra, dtime;
  float birthtime, dietime;

  /* where have we gone in time since last time */
  dfra = cfra - psys->cfra;

  timestep = psys_get_timestep(sim);
  dtime = dfra * timestep;

  if (dfra < 0.0f) {
    LOOP_EXISTING_PARTICLES
    {
      psys_get_texture(sim, pa, &ptex, PAMAP_SIZE, cfra);
      pa->size = part->size * ptex.size;
      if (part->randsize > 0.0f) {
        pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
      }

      reset_particle(sim, pa, dtime, cfra);
    }
    return;
  }

  /* for now do both, boids us 'rng' */
  sim->rng = BLI_rng_new_srandom(31415926 + (int)cfra + psys->seed);

  psys_update_effectors(sim);

  if (part->type != PART_HAIR) {
    sim->colliders = BKE_collider_cache_create(sim->depsgraph, sim->ob, part->collision_group);
  }

  /* initialize physics type specific stuff */
  switch (part->phystype) {
    case PART_PHYS_BOIDS: {
      ParticleTarget *pt = psys->targets.first;
      bbd.sim = sim;
      bbd.part = part;
      bbd.cfra = cfra;
      bbd.dfra = dfra;
      bbd.timestep = timestep;
      bbd.rng = sim->rng;

      psys_update_particle_tree(psys, cfra);

      boids_precalc_rules(part, cfra);

      for (; pt; pt = pt->next) {
        ParticleSystem *psys_target = psys_get_target_system(sim->ob, pt);
        if (psys_target && psys_target != psys) {
          psys_update_particle_tree(psys_target, cfra);
        }
      }
      break;
    }
    case PART_PHYS_FLUID: {
      ParticleTarget *pt = psys->targets.first;
      psys_update_particle_bvhtree(psys, cfra);

      for (; pt;
           pt = pt->next) { /* Updating others systems particle tree for fluid-fluid interaction */
        if (pt->ob) {
          psys_update_particle_bvhtree(BLI_findlink(&pt->ob->particlesystem, pt->psys - 1), cfra);
        }
      }
      break;
    }
  }
  /* initialize all particles for dynamics */
  LOOP_SHOWN_PARTICLES
  {
    copy_particle_key(&pa->prev_state, &pa->state, 1);

    psys_get_texture(sim, pa, &ptex, PAMAP_SIZE, cfra);

    pa->size = part->size * ptex.size;
    if (part->randsize > 0.0f) {
      pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
    }

    birthtime = pa->time;
    dietime = pa->dietime;

    /* store this, so we can do multiple loops over particles */
    pa->state.time = dfra;

    if (dietime <= cfra && psys->cfra < dietime) {
      /* particle dies some time between this and last step */
      pa->state.time = dietime - ((birthtime > psys->cfra) ? birthtime : psys->cfra);
      pa->alive = PARS_DYING;
    }
    else if (birthtime <= cfra && birthtime >= psys->cfra) {
      /* particle is born some time between this and last step*/
      reset_particle(sim, pa, dfra * timestep, cfra);
      pa->alive = PARS_ALIVE;
      pa->state.time = cfra - birthtime;
    }
    else if (dietime < cfra) {
      /* nothing to be done when particle is dead */
    }

    /* only reset unborn particles if they're shown or if the particle is born soon*/
    if (pa->alive == PARS_UNBORN &&
        (part->flag & PART_UNBORN || (cfra + psys->pointcache->step > pa->time))) {
      reset_particle(sim, pa, dtime, cfra);
    }
    else if (part->phystype == PART_PHYS_NO) {
      reset_particle(sim, pa, dtime, cfra);
    }

    if (ELEM(pa->alive, PARS_ALIVE, PARS_DYING) == 0 ||
        (pa->flag & (PARS_UNEXIST | PARS_NO_DISP))) {
      pa->state.time = -1.f;
    }
  }

  switch (part->phystype) {
    case PART_PHYS_NEWTON: {
      LOOP_DYNAMIC_PARTICLES
      {
        /* do global forces & effectors */
        basic_integrate(sim, p, pa->state.time, cfra);

        /* deflection */
        if (sim->colliders) {
          collision_check(sim, p, pa->state.time, cfra);
        }

        /* rotations */
        basic_rotate(part, pa, pa->state.time, timestep);
      }
      break;
    }
    case PART_PHYS_BOIDS: {
      LOOP_DYNAMIC_PARTICLES
      {
        bbd.goal_ob = NULL;

        boid_brain(&bbd, p, pa);

        if (pa->alive != PARS_DYING) {
          boid_body(&bbd, pa);

          /* deflection */
          if (sim->colliders) {
            collision_check(sim, p, pa->state.time, cfra);
          }
        }
      }
      break;
    }
    case PART_PHYS_FLUID: {
      SPHData sphdata;
      psys_sph_init(sim, &sphdata);

      DynamicStepSolverTaskData task_data = {
          .sim = sim,
          .cfra = cfra,
          .timestep = timestep,
          .dtime = dtime,
      };

      BLI_spin_init(&task_data.spin);

      if (part->fluid->solver == SPH_SOLVER_DDR) {
        /* Apply SPH forces using double-density relaxation algorithm
         * (Clavat et. al.) */

        TaskParallelSettings settings;
        BLI_parallel_range_settings_defaults(&settings);
        settings.use_threading = (psys->totpart > 100);
        settings.userdata_chunk = &sphdata;
        settings.userdata_chunk_size = sizeof(sphdata);
        settings.func_reduce = dynamics_step_sphdata_reduce;
        BLI_task_parallel_range(
            0, psys->totpart, &task_data, dynamics_step_sph_ddr_task_cb_ex, &settings);

        sph_springs_modify(psys, timestep);
      }
      else {
        /* SPH_SOLVER_CLASSICAL */
        /* Apply SPH forces using classical algorithm (due to Gingold
         * and Monaghan). Note that, unlike double-density relaxation,
         * this algorithm is separated into distinct loops. */

        {
          TaskParallelSettings settings;
          BLI_parallel_range_settings_defaults(&settings);
          settings.use_threading = (psys->totpart > 100);
          BLI_task_parallel_range(0,
                                  psys->totpart,
                                  &task_data,
                                  dynamics_step_sph_classical_basic_integrate_task_cb_ex,
                                  &settings);
        }

        /* calculate summation density */
        /* Note that we could avoid copying sphdata for each thread here (it's only read here),
         * but doubt this would gain us anything except confusion... */
        {
          TaskParallelSettings settings;
          BLI_parallel_range_settings_defaults(&settings);
          settings.use_threading = (psys->totpart > 100);
          settings.userdata_chunk = &sphdata;
          settings.userdata_chunk_size = sizeof(sphdata);
          settings.func_reduce = dynamics_step_sphdata_reduce;
          BLI_task_parallel_range(0,
                                  psys->totpart,
                                  &task_data,
                                  dynamics_step_sph_classical_calc_density_task_cb_ex,
                                  &settings);
        }

        /* do global forces & effectors */
        {
          TaskParallelSettings settings;
          BLI_parallel_range_settings_defaults(&settings);
          settings.use_threading = (psys->totpart > 100);
          settings.userdata_chunk = &sphdata;
          settings.userdata_chunk_size = sizeof(sphdata);
          settings.func_reduce = dynamics_step_sphdata_reduce;
          BLI_task_parallel_range(0,
                                  psys->totpart,
                                  &task_data,
                                  dynamics_step_sph_classical_integrate_task_cb_ex,
                                  &settings);
        }
      }

      BLI_spin_end(&task_data.spin);

      psys_sph_finalize(&sphdata);
      break;
    }
  }

  /* finalize particle state and time after dynamics */
  LOOP_DYNAMIC_PARTICLES
  {
    if (pa->alive == PARS_DYING) {
      pa->alive = PARS_DEAD;
      pa->state.time = pa->dietime;
    }
    else {
      pa->state.time = cfra;
    }
  }

  BKE_collider_cache_free(&sim->colliders);
  BLI_rng_free(sim->rng);
  sim->rng = NULL;
}

static void update_children(ParticleSimulationData *sim, const bool use_render_params)
{
  if ((sim->psys->part->type == PART_HAIR) && (sim->psys->flag & PSYS_HAIR_DONE) == 0) {
    /* don't generate children while growing hair - waste of time */
    psys_free_children(sim->psys);
  }
  else if (sim->psys->part->childtype) {
    if (sim->psys->totchild != psys_get_tot_child(sim->scene, sim->psys, use_render_params)) {
      distribute_particles(sim, PART_FROM_CHILD);
    }
    else {
      /* Children are up to date, nothing to do. */
    }
  }
  else {
    psys_free_children(sim->psys);
  }
}
/* updates cached particles' alive & other flags etc..*/
static void cached_step(ParticleSimulationData *sim, float cfra, const bool use_render_params)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  ParticleTexture ptex;
  PARTICLE_P;
  float disp, dietime;

  psys_update_effectors(sim);

  disp = psys_get_current_display_percentage(psys, use_render_params);

  LOOP_PARTICLES
  {
    psys_get_texture(sim, pa, &ptex, PAMAP_SIZE, cfra);
    pa->size = part->size * ptex.size;
    if (part->randsize > 0.0f) {
      pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
    }

    psys->lattice_deform_data = psys_create_lattice_deform_data(sim);

    dietime = pa->dietime;

    /* update alive status and push events */
    if (pa->time > cfra) {
      pa->alive = PARS_UNBORN;
      if (part->flag & PART_UNBORN && (psys->pointcache->flag & PTCACHE_EXTERNAL) == 0) {
        reset_particle(sim, pa, 0.0f, cfra);
      }
    }
    else if (dietime <= cfra) {
      pa->alive = PARS_DEAD;
    }
    else {
      pa->alive = PARS_ALIVE;
    }

    if (psys->lattice_deform_data) {
      BKE_lattice_deform_data_destroy(psys->lattice_deform_data);
      psys->lattice_deform_data = NULL;
    }

    if (psys_frand(psys, p) > disp) {
      pa->flag |= PARS_NO_DISP;
    }
    else {
      pa->flag &= ~PARS_NO_DISP;
    }
  }
}

static bool particles_has_flip(short parttype)
{
  return (parttype == PART_FLUID_FLIP);
}

static bool particles_has_tracer(short parttype)
{
  return (parttype == PART_FLUID_TRACER);
}

static bool particles_has_spray(short parttype)
{
  return ((parttype == PART_FLUID_SPRAY) || (parttype == PART_FLUID_SPRAYFOAM) ||
          (parttype == PART_FLUID_SPRAYFOAMBUBBLE));
}

static bool particles_has_bubble(short parttype)
{
  return ((parttype == PART_FLUID_BUBBLE) || (parttype == PART_FLUID_FOAMBUBBLE) ||
          (parttype == PART_FLUID_SPRAYFOAMBUBBLE));
}

static bool particles_has_foam(short parttype)
{
  return ((parttype == PART_FLUID_FOAM) || (parttype == PART_FLUID_SPRAYFOAM) ||
          (parttype == PART_FLUID_SPRAYFOAMBUBBLE));
}

static void particles_fluid_step(ParticleSimulationData *sim,
                                 int cfra,
                                 const bool use_render_params)
{
  ParticleSystem *psys = sim->psys;
  if (psys->particles) {
    MEM_freeN(psys->particles);
    psys->particles = 0;
    psys->totpart = 0;
  }

#ifndef WITH_FLUID
  UNUSED_VARS(use_render_params, cfra);
#else
  {
    Object *ob = sim->ob;
    FluidModifierData *fmd = (FluidModifierData *)BKE_modifiers_findby_type(ob,
                                                                            eModifierType_Fluid);

    if (fmd && fmd->domain && fmd->domain->fluid) {
      FluidDomainSettings *fds = fmd->domain;

      ParticleSettings *part = psys->part;
      ParticleData *pa = NULL;

      int p, totpart = 0, tottypepart = 0;
      int flagActivePart, activeParts = 0;
      float posX, posY, posZ, velX, velY, velZ;
      float resX, resY, resZ;
      int upres = 1;
      char debugStrBuffer[256];
      float tmp[3] = {0}, tmp2[3] = {0};

      /* Helper variables for scaling. */
      float min[3], max[3], size[3], cell_size_scaled[3], max_size;

      /* Sanity check: parts also enabled in fluid domain? */
      if ((particles_has_flip(part->type) &&
           (fds->particle_type & FLUID_DOMAIN_PARTICLE_FLIP) == 0) ||
          (particles_has_spray(part->type) &&
           (fds->particle_type & FLUID_DOMAIN_PARTICLE_SPRAY) == 0) ||
          (particles_has_bubble(part->type) &&
           (fds->particle_type & FLUID_DOMAIN_PARTICLE_BUBBLE) == 0) ||
          (particles_has_foam(part->type) &&
           (fds->particle_type & FLUID_DOMAIN_PARTICLE_FOAM) == 0) ||
          (particles_has_tracer(part->type) &&
           (fds->particle_type & FLUID_DOMAIN_PARTICLE_TRACER) == 0)) {
        BLI_snprintf(debugStrBuffer,
                     sizeof(debugStrBuffer),
                     "particles_fluid_step::error - found particle system that is not enabled in "
                     "fluid domain\n");
        return;
      }

      /* Count particle amount. tottypepart is only important for snd particles. */
      if (part->type == PART_FLUID_FLIP) {
        tottypepart = totpart = manta_liquid_get_num_flip_particles(fds->fluid);
      }
      if (particles_has_spray(part->type) || particles_has_bubble(part->type) ||
          particles_has_foam(part->type) || particles_has_tracer(part->type)) {
        totpart = manta_liquid_get_num_snd_particles(fds->fluid);

        /* tottypepart is the amount of particles of a snd particle type. */
        for (p = 0; p < totpart; p++) {
          flagActivePart = manta_liquid_get_snd_particle_flag_at(fds->fluid, p);
          if (particles_has_spray(part->type) && (flagActivePart & PARTICLE_TYPE_SPRAY)) {
            tottypepart++;
          }
          if (particles_has_bubble(part->type) && (flagActivePart & PARTICLE_TYPE_BUBBLE)) {
            tottypepart++;
          }
          if (particles_has_foam(part->type) && (flagActivePart & PARTICLE_TYPE_FOAM)) {
            tottypepart++;
          }
          if (particles_has_tracer(part->type) && (flagActivePart & PARTICLE_TYPE_TRACER)) {
            tottypepart++;
          }
        }
      }
      /* Sanity check: no particles present. */
      if (!totpart || !tottypepart) {
        return;
      }

      /* How many particles to display? */
      tottypepart = (use_render_params) ? tottypepart : (part->disp * tottypepart) / 100;

      part->totpart = tottypepart;
      part->sta = part->end = 1.0f;
      part->lifetime = sim->scene->r.efra + 1;

      /* Allocate particles. */
      realloc_particles(sim, part->totpart);

      /* Set some randomness when choosing which particles to display. */
      sim->rng = BLI_rng_new_srandom(31415926 + (int)cfra + psys->seed);
      double r, dispProb = (double)part->disp / 100.0;

      /* Loop over *all* particles. Will break out of loop before tottypepart amount exceeded. */
      for (p = 0, pa = psys->particles; p < totpart; p++) {

        /* Apply some randomness and determine which particles to skip. */
        r = BLI_rng_get_double(sim->rng);
        if (r > dispProb) {
          continue;
        }

        /* flag, res, upres, pos, vel for FLIP and snd particles have different getters. */
        if (part->type == PART_FLUID_FLIP) {
          flagActivePart = manta_liquid_get_flip_particle_flag_at(fds->fluid, p);

          resX = (float)manta_get_res_x(fds->fluid);
          resY = (float)manta_get_res_y(fds->fluid);
          resZ = (float)manta_get_res_z(fds->fluid);

          upres = 1;

          posX = manta_liquid_get_flip_particle_position_x_at(fds->fluid, p);
          posY = manta_liquid_get_flip_particle_position_y_at(fds->fluid, p);
          posZ = manta_liquid_get_flip_particle_position_z_at(fds->fluid, p);

          velX = manta_liquid_get_flip_particle_velocity_x_at(fds->fluid, p);
          velY = manta_liquid_get_flip_particle_velocity_y_at(fds->fluid, p);
          velZ = manta_liquid_get_flip_particle_velocity_z_at(fds->fluid, p);
        }
        else if (particles_has_spray(part->type) || particles_has_bubble(part->type) ||
                 particles_has_foam(part->type) || particles_has_tracer(part->type)) {
          flagActivePart = manta_liquid_get_snd_particle_flag_at(fds->fluid, p);

          resX = (float)manta_liquid_get_particle_res_x(fds->fluid);
          resY = (float)manta_liquid_get_particle_res_y(fds->fluid);
          resZ = (float)manta_liquid_get_particle_res_z(fds->fluid);

          upres = manta_liquid_get_particle_upres(fds->fluid);

          posX = manta_liquid_get_snd_particle_position_x_at(fds->fluid, p);
          posY = manta_liquid_get_snd_particle_position_y_at(fds->fluid, p);
          posZ = manta_liquid_get_snd_particle_position_z_at(fds->fluid, p);

          velX = manta_liquid_get_snd_particle_velocity_x_at(fds->fluid, p);
          velY = manta_liquid_get_snd_particle_velocity_y_at(fds->fluid, p);
          velZ = manta_liquid_get_snd_particle_velocity_z_at(fds->fluid, p);
        }
        else {
          BLI_snprintf(debugStrBuffer,
                       sizeof(debugStrBuffer),
                       "particles_fluid_step::error - unknown particle system type\n");
          return;
        }
#  if 0
        /* Debugging: Print type of particle system and current particles. */
        printf("system type is %d and particle type is %d\n", part->type, flagActivePart);
#  endif

        /* Type of particle must match current particle system type
         * (only important for snd particles). */
        if ((flagActivePart & PARTICLE_TYPE_SPRAY) && !particles_has_spray(part->type)) {
          continue;
        }
        if ((flagActivePart & PARTICLE_TYPE_BUBBLE) && !particles_has_bubble(part->type)) {
          continue;
        }
        if ((flagActivePart & PARTICLE_TYPE_FOAM) && !particles_has_foam(part->type)) {
          continue;
        }
        if ((flagActivePart & PARTICLE_TYPE_TRACER) && !particles_has_tracer(part->type)) {
          continue;
        }
#  if 0
        /* Debugging: Print type of particle system and current particles. */
        printf("system type is %d and particle type is %d\n", part->type, flagActivePart);
#  endif
        /* Particle system has allocated 'tottypepart' particles - so break early before exceeded.
         */
        if (activeParts >= tottypepart) {
          break;
        }

        /* Only show active particles, i.e. filter out dead particles that just Mantaflow needs.
         * Mantaflow convention: PARTICLE_TYPE_DELETE == inactive particle. */
        if ((flagActivePart & PARTICLE_TYPE_DELETE) == 0) {
          activeParts++;

          /* Use particle system settings for particle size. */
          pa->size = part->size;
          if (part->randsize > 0.0f) {
            pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
          }

          /* Get size (dimension) but considering scaling */
          copy_v3_v3(cell_size_scaled, fds->cell_size);
          mul_v3_v3(cell_size_scaled, ob->scale);
          madd_v3fl_v3fl_v3fl_v3i(min, fds->p0, cell_size_scaled, fds->res_min);
          madd_v3fl_v3fl_v3fl_v3i(max, fds->p0, cell_size_scaled, fds->res_max);
          sub_v3_v3v3(size, max, min);

          /* Biggest dimension will be used for up-scaling. */
          max_size = MAX3(size[0] / (float)upres, size[1] / (float)upres, size[2] / (float)upres);

          /* Set particle position. */
          const float posParticle[3] = {posX, posY, posZ};
          copy_v3_v3(pa->state.co, posParticle);

          /* Normalize to unit cube around 0. */
          float resDomain[3] = {resX, resY, resZ};
          mul_v3_fl(resDomain, 0.5f);
          sub_v3_v3(pa->state.co, resDomain);
          mul_v3_fl(pa->state.co, fds->dx);

          /* Match domain dimension / size. */
          float scaleAbs[3] = {
              1. / fabsf(ob->scale[0]), 1. / fabsf(ob->scale[1]), 1. / fabsf(ob->scale[2])};
          mul_v3_fl(scaleAbs, max_size);
          mul_v3_v3(pa->state.co, scaleAbs);

          /* Match domain scale. */
          mul_m4_v3(ob->obmat, pa->state.co);

          /* Add origin offset to particle position. */
          zero_v3(tmp);
          zero_v3(tmp2);
          sub_v3_v3v3(tmp2, fds->p1, fds->p0);
          mul_v3_fl(tmp2, 0.5f);
          add_v3_v3v3(tmp, tmp, fds->p1);
          sub_v3_v3(tmp, tmp2);
          mul_v3_v3(tmp, ob->scale);
          add_v3_v3(pa->state.co, tmp);
#  if 0
          /* Debugging: Print particle coordinates. */
          printf("pa->state.co[0]: %f, pa->state.co[1]: %f, pa->state.co[2]: %f\n",
          pa->state.co[0], pa->state.co[1], pa->state.co[2]);
#  endif
          /* Set particle velocity. */
          const float velParticle[3] = {velX, velY, velZ};
          copy_v3_v3(pa->state.vel, velParticle);
          mul_v3_fl(pa->state.vel, fds->dx);
#  if 0
          /* Debugging: Print particle velocity. */
          printf("pa->state.vel[0]: %f, pa->state.vel[1]: %f, pa->state.vel[2]: %f\n",
          pa->state.vel[0], pa->state.vel[1], pa->state.vel[2]);
#  endif
          /* Set default angular velocity and particle rotation. */
          zero_v3(pa->state.ave);
          unit_qt(pa->state.rot);

          pa->time = 1.f;
          pa->dietime = sim->scene->r.efra + 1;
          pa->lifetime = sim->scene->r.efra;
          pa->alive = PARS_ALIVE;

          /* Increasing particle settings pointer only for active particles. */
          pa++;
        }
      }
#  if 0
      /* Debugging: Print number of active particles. */
      printf("active parts: %d\n", activeParts);
#  endif
      totpart = psys->totpart = part->totpart = activeParts;

      BLI_rng_free(sim->rng);
      sim->rng = NULL;

    } /* Fluid sim particles done. */
  }
#endif /* WITH_FLUID */
}

static int emit_particles(ParticleSimulationData *sim, PTCacheID *pid, float UNUSED(cfra))
{
  ParticleSystem *psys = sim->psys;
  int oldtotpart = psys->totpart;
  int totpart = tot_particles(psys, pid);

  if (totpart != oldtotpart) {
    realloc_particles(sim, totpart);
  }

  return totpart - oldtotpart;
}

/**
 * Calculates the next state for all particles of the system.
 * In particles code most 'cfra - ending' are frames,
 * 'time - ending' are 'cfra * timestep' (seconds).
 *
 * 1. Emit particles
 * 2. Check cache (if used) and return if frame is cached
 * 3. Do dynamics
 * 4. Save to cache
 */
static void system_step(ParticleSimulationData *sim, float cfra, const bool use_render_params)
{
  ParticleSystem *psys = sim->psys;
  ParticleSettings *part = psys->part;
  PointCache *cache = psys->pointcache;
  PTCacheID ptcacheid, *pid = NULL;
  PARTICLE_P;
  float disp, cache_cfra = cfra; /*, *vg_vel= 0, *vg_tan= 0, *vg_rot= 0, *vg_size= 0; */
  int startframe = 0, endframe = 100, oldtotpart = 0;

  /* cache shouldn't be used for hair or "continue physics" */
  if (part->type != PART_HAIR) {
    psys_clear_temp_pointcache(psys);

    /* set suitable cache range automatically */
    if ((cache->flag & (PTCACHE_BAKING | PTCACHE_BAKED)) == 0) {
      psys_get_pointcache_start_end(sim->scene, psys, &cache->startframe, &cache->endframe);
    }

    pid = &ptcacheid;
    BKE_ptcache_id_from_particles(pid, sim->ob, psys);

    BKE_ptcache_id_time(pid, sim->scene, 0.0f, &startframe, &endframe, NULL);

    /* clear everything on start frame, or when psys needs full reset! */
    if ((cfra == startframe) || (psys->recalc & ID_RECALC_PSYS_RESET)) {
      BKE_ptcache_id_reset(sim->scene, pid, PTCACHE_RESET_OUTDATED);
      BKE_ptcache_validate(cache, startframe);
      cache->flag &= ~PTCACHE_REDO_NEEDED;
    }

    CLAMP(cache_cfra, startframe, endframe);
  }

  /* 1. emit particles and redo particles if needed */
  oldtotpart = psys->totpart;
  if (emit_particles(sim, pid, cfra) || psys->recalc & ID_RECALC_PSYS_RESET) {
    distribute_particles(sim, part->from);
    initialize_all_particles(sim);
    /* reset only just created particles (on startframe all particles are recreated) */
    reset_all_particles(sim, 0.0, cfra, oldtotpart);
    free_unexisting_particles(sim);

    if (psys->fluid_springs) {
      MEM_freeN(psys->fluid_springs);
      psys->fluid_springs = NULL;
    }

    psys->tot_fluidsprings = psys->alloc_fluidsprings = 0;

    /* flag for possible explode modifiers after this system */
    sim->psmd->flag |= eParticleSystemFlag_Pars;

    BKE_ptcache_id_clear(pid, PTCACHE_CLEAR_AFTER, cfra);
  }

  /* 2. try to read from the cache */
  if (pid) {
    int cache_result = BKE_ptcache_read(pid, cache_cfra, true);

    if (ELEM(cache_result, PTCACHE_READ_EXACT, PTCACHE_READ_INTERPOLATED)) {
      cached_step(sim, cfra, use_render_params);
      update_children(sim, use_render_params);
      psys_update_path_cache(sim, cfra, use_render_params);

      BKE_ptcache_validate(cache, (int)cache_cfra);

      if (cache_result == PTCACHE_READ_INTERPOLATED && cache->flag & PTCACHE_REDO_NEEDED) {
        BKE_ptcache_write(pid, (int)cache_cfra);
      }

      return;
    }
    /* Cache is supposed to be baked, but no data was found so bail out */
    if (cache->flag & PTCACHE_BAKED) {
      psys_reset(psys, PSYS_RESET_CACHE_MISS);
      return;
    }
    if (cache_result == PTCACHE_READ_OLD) {
      psys->cfra = (float)cache->simframe;
      cached_step(sim, psys->cfra, use_render_params);
    }

    /* if on second frame, write cache for first frame */
    if (psys->cfra == startframe && (cache->flag & PTCACHE_OUTDATED || cache->last_exact == 0)) {
      BKE_ptcache_write(pid, startframe);
    }
  }
  else {
    BKE_ptcache_invalidate(cache);
  }

  /* 3. do dynamics */
  /* set particles to be not calculated TODO: can't work with pointcache */
  disp = psys_get_current_display_percentage(psys, use_render_params);

  LOOP_PARTICLES
  {
    if (psys_frand(psys, p) > disp) {
      pa->flag |= PARS_NO_DISP;
    }
    else {
      pa->flag &= ~PARS_NO_DISP;
    }
  }

  if (psys->totpart) {
    int dframe, totframesback = 0;
    float t_frac, dt_frac;

    /* handle negative frame start at the first frame by doing
     * all the steps before the first frame */
    if ((int)cfra == startframe && part->sta < startframe) {
      totframesback = (startframe - (int)part->sta);
    }

    if (!(part->time_flag & PART_TIME_AUTOSF)) {
      /* Constant time step */
      psys->dt_frac = get_base_time_step(part);
    }
    else if ((int)cfra == startframe) {
      /* Variable time step; initialize to sub-frames. */
      psys->dt_frac = get_base_time_step(part);
    }
    else if (psys->dt_frac < MIN_TIMESTEP) {
      /* Variable time step; subsequent frames */
      psys->dt_frac = MIN_TIMESTEP;
    }

    for (dframe = -totframesback; dframe <= 0; dframe++) {
      /* simulate each subframe */
      dt_frac = psys->dt_frac;
      for (t_frac = dt_frac; t_frac <= 1.0f; t_frac += dt_frac) {
        sim->courant_num = 0.0f;
        dynamics_step(sim, cfra + dframe + t_frac - 1.f);
        psys->cfra = cfra + dframe + t_frac - 1.f;

        if (part->time_flag & PART_TIME_AUTOSF) {
          update_timestep(psys, sim);
        }
        /* Even without AUTOSF dt_frac may not add up to 1.0 due to float precision. */
        dt_frac = sync_timestep(psys, t_frac);
      }
    }
  }

  /* 4. only write cache starting from second frame */
  if (pid) {
    BKE_ptcache_validate(cache, (int)cache_cfra);
    if ((int)cache_cfra != startframe) {
      BKE_ptcache_write(pid, (int)cache_cfra);
    }
  }

  update_children(sim, use_render_params);

  /* cleanup */
  if (psys->lattice_deform_data) {
    BKE_lattice_deform_data_destroy(psys->lattice_deform_data);
    psys->lattice_deform_data = NULL;
  }
}

/* system type has changed so set sensible defaults and clear non applicable flags */
void psys_changed_type(Object *ob, ParticleSystem *psys)
{
  ParticleSettings *part = psys->part;
  PTCacheID pid;

  BKE_ptcache_id_from_particles(&pid, ob, psys);

  if (part->phystype != PART_PHYS_KEYED) {
    psys->flag &= ~PSYS_KEYED;
  }

  if (part->type == PART_HAIR) {
    if (ELEM(part->ren_as, PART_DRAW_NOT, PART_DRAW_PATH, PART_DRAW_OB, PART_DRAW_GR) == 0) {
      part->ren_as = PART_DRAW_PATH;
    }

    if (part->distr == PART_DISTR_GRID) {
      part->distr = PART_DISTR_JIT;
    }

    if (ELEM(part->draw_as, PART_DRAW_NOT, PART_DRAW_REND, PART_DRAW_PATH) == 0) {
      part->draw_as = PART_DRAW_REND;
    }

    CLAMP(part->path_start, 0.0f, 100.0f);
    CLAMP(part->path_end, 0.0f, 100.0f);

    BKE_ptcache_id_clear(&pid, PTCACHE_CLEAR_ALL, 0);
  }
  else {
    free_hair(ob, psys, 1);

    CLAMP(part->path_start, 0.0f, MAX2(100.0f, part->end + part->lifetime));
    CLAMP(part->path_end, 0.0f, MAX2(100.0f, part->end + part->lifetime));
  }

  psys_reset(psys, PSYS_RESET_ALL);
}
void psys_check_boid_data(ParticleSystem *psys)
{
  BoidParticle *bpa;
  PARTICLE_P;

  pa = psys->particles;

  if (!pa) {
    return;
  }

  if (psys->part && psys->part->phystype == PART_PHYS_BOIDS) {
    if (!pa->boid) {
      bpa = MEM_callocN(psys->totpart * sizeof(BoidParticle), "Boid Data");

      LOOP_PARTICLES
      {
        pa->boid = bpa++;
      }
    }
  }
  else if (pa->boid) {
    MEM_freeN(pa->boid);
    LOOP_PARTICLES
    {
      pa->boid = NULL;
    }
  }
}

void BKE_particlesettings_fluid_default_settings(ParticleSettings *part)
{
  SPHFluidSettings *fluid = part->fluid;

  fluid->spring_k = 0.f;
  fluid->plasticity_constant = 0.1f;
  fluid->yield_ratio = 0.1f;
  fluid->rest_length = 1.f;
  fluid->viscosity_omega = 2.f;
  fluid->viscosity_beta = 0.1f;
  fluid->stiffness_k = 1.f;
  fluid->stiffness_knear = 1.f;
  fluid->rest_density = 1.f;
  fluid->buoyancy = 0.f;
  fluid->radius = 1.f;
  fluid->flag |= SPH_FAC_REPULSION | SPH_FAC_DENSITY | SPH_FAC_RADIUS | SPH_FAC_VISCOSITY |
                 SPH_FAC_REST_LENGTH;
}

static void psys_prepare_physics(ParticleSimulationData *sim)
{
  ParticleSettings *part = sim->psys->part;

  if (ELEM(part->phystype, PART_PHYS_NO, PART_PHYS_KEYED)) {
    PTCacheID pid;
    BKE_ptcache_id_from_particles(&pid, sim->ob, sim->psys);
    BKE_ptcache_id_clear(&pid, PTCACHE_CLEAR_ALL, 0);
  }
  else {
    free_keyed_keys(sim->psys);
    sim->psys->flag &= ~PSYS_KEYED;
  }

  /* RNA Update must ensure this is true. */
  if (part->phystype == PART_PHYS_BOIDS) {
    BLI_assert(part->boids != NULL);
  }
  else if (part->phystype == PART_PHYS_FLUID) {
    BLI_assert(part->fluid != NULL);
  }

  psys_check_boid_data(sim->psys);
}
static int hair_needs_recalc(ParticleSystem *psys)
{
  if (!(psys->flag & PSYS_EDITED) && (!psys->edit || !psys->edit->edited) &&
      ((psys->flag & PSYS_HAIR_DONE) == 0 || psys->recalc & ID_RECALC_PSYS_RESET ||
       (psys->part->flag & PART_HAIR_REGROW && !psys->edit))) {
    return 1;
  }

  return 0;
}

static ParticleSettings *particle_settings_localize(ParticleSettings *particle_settings)
{
  ParticleSettings *particle_settings_local = (ParticleSettings *)BKE_id_copy_ex(
      NULL, (ID *)&particle_settings->id, NULL, LIB_ID_COPY_LOCALIZE);
  return particle_settings_local;
}

static void particle_settings_free_local(ParticleSettings *particle_settings)
{
  BKE_libblock_free_datablock(&particle_settings->id, 0);
  BKE_libblock_free_data(&particle_settings->id, false);
  MEM_freeN(particle_settings);
}

/* main particle update call, checks that things are ok on the large scale and
 * then advances in to actual particle calculations depending on particle type */
void particle_system_update(struct Depsgraph *depsgraph,
                            Scene *scene,
                            Object *ob,
                            ParticleSystem *psys,
                            const bool use_render_params)
{
  ParticleSimulationData sim = {0};
  ParticleSettings *part = psys->part;
  ParticleSystem *psys_orig = psys_orig_get(psys);
  float cfra;
  ParticleSystemModifierData *psmd = psys_get_modifier(ob, psys);

  /* drawdata is outdated after ANY change */
  if (psys->pdd) {
    psys->pdd->flag &= ~PARTICLE_DRAW_DATA_UPDATED;
  }

  if (!psys_check_enabled(ob, psys, use_render_params)) {
    return;
  }

  cfra = DEG_get_ctime(depsgraph);

  sim.depsgraph = depsgraph;
  sim.scene = scene;
  sim.ob = ob;
  sim.psys = psys;
  sim.psmd = psmd;

  /* system was already updated from modifier stack */
  if (sim.psmd->flag & eParticleSystemFlag_psys_updated) {
    sim.psmd->flag &= ~eParticleSystemFlag_psys_updated;
    /* make sure it really was updated to cfra */
    if (psys->cfra == cfra) {
      return;
    }
  }

  if (!sim.psmd->mesh_final) {
    return;
  }

  if (part->from != PART_FROM_VERT) {
    BKE_mesh_tessface_ensure(sim.psmd->mesh_final);
  }

  /* to verify if we need to restore object afterwards */
  psys->flag &= ~PSYS_OB_ANIM_RESTORE;

  if (psys->recalc & ID_RECALC_PSYS_RESET) {
    psys->totunexist = 0;
  }

  /* setup necessary physics type dependent additional data if it doesn't yet exist */
  psys_prepare_physics(&sim);

  if (part->type == PART_HAIR) {
    /* nothing to do so bail out early */
    if (psys->totpart == 0 && part->totpart == 0) {
      psys_free_path_cache(psys, NULL);
      free_hair(ob, psys, 0);
      psys->flag |= PSYS_HAIR_DONE;
    }
    /* (re-)create hair */
    else if (hair_needs_recalc(psys)) {
      float hcfra = 0.0f;
      int i, recalc = psys->recalc;

      free_hair(ob, psys, 0);

      if (psys_orig->edit && psys_orig->free_edit) {
        psys_orig->free_edit(psys_orig->edit);
        psys_orig->edit = NULL;
        psys_orig->free_edit = NULL;
      }

      /* first step is negative so particles get killed and reset */
      psys->cfra = 1.0f;

      ParticleSettings *part_local = part;
      if ((part->flag & PART_HAIR_REGROW) == 0) {
        part_local = particle_settings_localize(part);
        psys->part = part_local;
      }

      for (i = 0; i <= part->hair_step; i++) {
        hcfra = 100.0f * (float)i / (float)psys->part->hair_step;
        if ((part->flag & PART_HAIR_REGROW) == 0) {
          const AnimationEvalContext anim_eval_context = BKE_animsys_eval_context_construct(
              depsgraph, hcfra);
          BKE_animsys_evaluate_animdata(
              &part_local->id, part_local->adt, &anim_eval_context, ADT_RECALC_ANIM, false);
        }
        system_step(&sim, hcfra, use_render_params);
        psys->cfra = hcfra;
        psys->recalc = 0;
        save_hair(&sim, hcfra);
      }

      if (part_local != part) {
        particle_settings_free_local(part_local);
        psys->part = part;
      }

      psys->flag |= PSYS_HAIR_DONE;
      psys->recalc = recalc;
    }
    else if (psys->flag & PSYS_EDITED) {
      psys->flag |= PSYS_HAIR_DONE;
    }

    if (psys->flag & PSYS_HAIR_DONE) {
      hair_step(&sim, cfra, use_render_params);
    }
  }
  else if (particles_has_flip(part->type) || particles_has_spray(part->type) ||
           particles_has_bubble(part->type) || particles_has_foam(part->type) ||
           particles_has_tracer(part->type)) {
    particles_fluid_step(&sim, (int)cfra, use_render_params);
  }
  else {
    switch (part->phystype) {
      case PART_PHYS_NO:
      case PART_PHYS_KEYED: {
        PARTICLE_P;
        float disp = psys_get_current_display_percentage(psys, use_render_params);
        bool free_unexisting = false;

        /* Particles without dynamics haven't been reset yet because they don't use pointcache */
        if (psys->recalc & ID_RECALC_PSYS_RESET) {
          psys_reset(psys, PSYS_RESET_ALL);
        }

        if (emit_particles(&sim, NULL, cfra) || (psys->recalc & ID_RECALC_PSYS_RESET)) {
          free_keyed_keys(psys);
          distribute_particles(&sim, part->from);
          initialize_all_particles(&sim);
          free_unexisting = true;

          /* flag for possible explode modifiers after this system */
          sim.psmd->flag |= eParticleSystemFlag_Pars;
        }

        LOOP_EXISTING_PARTICLES
        {
          pa->size = part->size;
          if (part->randsize > 0.0f) {
            pa->size *= 1.0f - part->randsize * psys_frand(psys, p + 1);
          }

          reset_particle(&sim, pa, 0.0, cfra);

          if (psys_frand(psys, p) > disp) {
            pa->flag |= PARS_NO_DISP;
          }
          else {
            pa->flag &= ~PARS_NO_DISP;
          }
        }

        /* free unexisting after resetting particles */
        if (free_unexisting) {
          free_unexisting_particles(&sim);
        }

        if (part->phystype == PART_PHYS_KEYED) {
          psys_count_keyed_targets(&sim);
          set_keyed_keys(&sim);
          psys_update_path_cache(&sim, (int)cfra, use_render_params);
        }
        break;
      }
      default: {
        /* the main dynamic particle system step */
        system_step(&sim, cfra, use_render_params);
        break;
      }
    }
  }

  /* make sure emitter is left at correct time (particle emission can change this) */
  if (psys->flag & PSYS_OB_ANIM_RESTORE) {
    evaluate_emitter_anim(depsgraph, scene, ob, cfra);
    psys->flag &= ~PSYS_OB_ANIM_RESTORE;
  }

  if (psys_orig->edit) {
    psys_orig->edit->flags |= PT_CACHE_EDIT_UPDATE_PARTICLE_FROM_EVAL;
  }

  psys->cfra = cfra;
  psys->recalc = 0;

  if (DEG_is_active(depsgraph)) {
    if (psys_orig != psys) {
      if (psys_orig->edit != NULL && psys_orig->edit->psys == psys_orig) {
        psys_orig->edit->psys_eval = psys;
        psys_orig->edit->psmd_eval = psmd;
      }
      psys_orig->flag = (psys->flag & ~PSYS_SHARED_CACHES);
      psys_orig->cfra = psys->cfra;
      psys_orig->recalc = psys->recalc;
      psys_orig->part->totpart = part->totpart;
    }
  }

  /* Save matrix for duplicators,
   * at rendertime the actual dupliobject's matrix is used so don't update! */
  invert_m4_m4(psys->imat, ob->obmat);

  BKE_particle_batch_cache_dirty_tag(psys, BKE_PARTICLE_BATCH_DIRTY_ALL);
}

/* ID looper */

void BKE_particlesystem_id_loop(ParticleSystem *psys, ParticleSystemIDFunc func, void *userdata)
{
  ParticleTarget *pt;

  func(psys, (ID **)&psys->part, userdata, IDWALK_CB_USER | IDWALK_CB_NEVER_NULL);
  func(psys, (ID **)&psys->target_ob, userdata, IDWALK_CB_NOP);
  func(psys, (ID **)&psys->parent, userdata, IDWALK_CB_NOP);

  for (pt = psys->targets.first; pt; pt = pt->next) {
    func(psys, (ID **)&pt->ob, userdata, IDWALK_CB_NOP);
  }

  /* Even though psys->part should never be NULL, this can happen as an exception during deletion.
   * See ID_REMAP_SKIP/FORCE/FLAG_NEVER_NULL_USAGE in BKE_library_remap. */
  if (psys->part && psys->part->phystype == PART_PHYS_BOIDS) {
    ParticleData *pa;
    int p;

    for (p = 0, pa = psys->particles; p < psys->totpart; p++, pa++) {
      func(psys, (ID **)&pa->boid->ground, userdata, IDWALK_CB_NOP);
    }
  }
}

void BKE_particlesystem_reset_all(struct Object *object)
{
  for (ModifierData *md = object->modifiers.first; md != NULL; md = md->next) {
    if (md->type != eModifierType_ParticleSystem) {
      continue;
    }
    ParticleSystemModifierData *psmd = (ParticleSystemModifierData *)md;
    ParticleSystem *psys = psmd->psys;
    psys->recalc |= ID_RECALC_PSYS_RESET;
  }
}

/* **** Depsgraph evaluation **** */

void BKE_particle_settings_eval_reset(struct Depsgraph *depsgraph,
                                      ParticleSettings *particle_settings)
{
  DEG_debug_print_eval(depsgraph, __func__, particle_settings->id.name, particle_settings);
  particle_settings->id.recalc |= ID_RECALC_PSYS_RESET;
}

void BKE_particle_system_eval_init(struct Depsgraph *depsgraph, Object *object)
{
  DEG_debug_print_eval(depsgraph, __func__, object->id.name, object);
  for (ParticleSystem *psys = object->particlesystem.first; psys != NULL; psys = psys->next) {
    psys->recalc |= (psys->part->id.recalc & ID_RECALC_PSYS_ALL);
  }
}