runner_doiact.h

/*******************************************************************************
 * This file is part of SWIFT.
 * Copyright (c) 2012 Pedro Gonnet (pedro.gonnet@durham.ac.uk)
 *               2016 Matthieu Schaller (matthieu.schaller@durham.ac.uk)
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published
 * by the Free Software Foundation, either version 3 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 Lesser General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 *
 ******************************************************************************/

/* Before including this file, define FUNCTION, which is the
   name of the interaction function. This creates the interaction functions
   runner_dopair_FUNCTION, runner_dopair_FUNCTION_naive, runner_doself_FUNCTION,
   and runner_dosub_FUNCTION calling the pairwise interaction function
   runner_iact_FUNCTION. */

#define PASTE(x, y) x##_##y

#define _DOPAIR1(f) PASTE(runner_dopair1, f)
#define DOPAIR1 _DOPAIR1(FUNCTION)

#define _DOPAIR2(f) PASTE(runner_dopair2, f)
#define DOPAIR2 _DOPAIR2(FUNCTION)

#define _DOPAIR_SUBSET(f) PASTE(runner_dopair_subset, f)
#define DOPAIR_SUBSET _DOPAIR_SUBSET(FUNCTION)

#define _DOPAIR_SUBSET_NAIVE(f) PASTE(runner_dopair_subset_naive, f)
#define DOPAIR_SUBSET_NAIVE _DOPAIR_SUBSET_NAIVE(FUNCTION)

#define _DOPAIR_NAIVE(f) PASTE(runner_dopair_naive, f)
#define DOPAIR_NAIVE _DOPAIR_NAIVE(FUNCTION)

#define _DOSELF_NAIVE(f) PASTE(runner_doself_naive, f)
#define DOSELF_NAIVE _DOSELF_NAIVE(FUNCTION)

#define _DOSELF1(f) PASTE(runner_doself1, f)
#define DOSELF1 _DOSELF1(FUNCTION)

#define _DOSELF2(f) PASTE(runner_doself2, f)
#define DOSELF2 _DOSELF2(FUNCTION)

#define _DOSELF_SUBSET(f) PASTE(runner_doself_subset, f)
#define DOSELF_SUBSET _DOSELF_SUBSET(FUNCTION)

#define _DOSUB_SELF1(f) PASTE(runner_dosub_self1, f)
#define DOSUB_SELF1 _DOSUB_SELF1(FUNCTION)

#define _DOSUB_PAIR1(f) PASTE(runner_dosub_pair1, f)
#define DOSUB_PAIR1 _DOSUB_PAIR1(FUNCTION)

#define _DOSUB_SELF2(f) PASTE(runner_dosub_self2, f)
#define DOSUB_SELF2 _DOSUB_SELF2(FUNCTION)

#define _DOSUB_PAIR2(f) PASTE(runner_dosub_pair2, f)
#define DOSUB_PAIR2 _DOSUB_PAIR2(FUNCTION)

#define _DOSUB_SUBSET(f) PASTE(runner_dosub_subset, f)
#define DOSUB_SUBSET _DOSUB_SUBSET(FUNCTION)

#define _IACT_NONSYM(f) PASTE(runner_iact_nonsym, f)
#define IACT_NONSYM _IACT_NONSYM(FUNCTION)

#define _IACT(f) PASTE(runner_iact, f)
#define IACT _IACT(FUNCTION)

#define _IACT_NONSYM_VEC(f) PASTE(runner_iact_nonsym_vec, f)
#define IACT_NONSYM_VEC _IACT_NONSYM_VEC(FUNCTION)

#define _IACT_VEC(f) PASTE(runner_iact_vec, f)
#define IACT_VEC _IACT_VEC(FUNCTION)

#define _TIMER_DOSELF(f) PASTE(timer_doself, f)
#define TIMER_DOSELF _TIMER_DOSELF(FUNCTION)

#define _TIMER_DOPAIR(f) PASTE(timer_dopair, f)
#define TIMER_DOPAIR _TIMER_DOPAIR(FUNCTION)

#define _TIMER_DOSUB_SELF(f) PASTE(timer_dosub_self, f)
#define TIMER_DOSUB_SELF _TIMER_DOSUB_SELF(FUNCTION)

#define _TIMER_DOSUB_PAIR(f) PASTE(timer_dosub_pair, f)
#define TIMER_DOSUB_PAIR _TIMER_DOSUB_PAIR(FUNCTION)

#define _TIMER_DOSELF_SUBSET(f) PASTE(timer_doself_subset, f)
#define TIMER_DOSELF_SUBSET _TIMER_DOSELF_SUBSET(FUNCTION)

#define _TIMER_DOPAIR_SUBSET(f) PASTE(timer_dopair_subset, f)
#define TIMER_DOPAIR_SUBSET _TIMER_DOPAIR_SUBSET(FUNCTION)

/**
 * @brief Compute the interactions between a cell pair.
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param cj The second #cell.
 */
void DOPAIR_NAIVE(struct runner *r, struct cell *restrict ci,
                  struct cell *restrict cj) {

  const struct engine *e = r->e;

  error("Don't use in actual runs ! Slow code !");

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif
  TIMER_TIC;

  /* Anything to do here? */
  if (!cell_is_active(ci, e) && !cell_is_active(cj, e)) return;

  const int count_i = ci->count;
  const int count_j = cj->count;
  struct part *restrict parts_i = ci->parts;
  struct part *restrict parts_j = cj->parts;

  /* Get the relative distance between the pairs, wrapping. */
  double shift[3] = {0.0, 0.0, 0.0};
  for (int k = 0; k < 3; k++) {
    if (cj->loc[k] - ci->loc[k] < -e->s->dim[k] / 2)
      shift[k] = e->s->dim[k];
    else if (cj->loc[k] - ci->loc[k] > e->s->dim[k] / 2)
      shift[k] = -e->s->dim[k];
  }

  /* Loop over the parts in ci. */
  for (int pid = 0; pid < count_i; pid++) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts_i[pid];
    const float hi = pi->h;

    double pix[3];
    for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];
    const float hig2 = hi * hi * kernel_gamma2;

    /* Loop over the parts in cj. */
    for (int pjd = 0; pjd < count_j; pjd++) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pj = &parts_j[pjd];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pix[k] - pj->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 < hig2 || r2 < pj->h * pj->h * kernel_gamma2) {

#ifndef WITH_VECTORIZATION

        IACT(r2, dx, hi, pj->h, pi, pj);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hi;
        hjq[icount] = pj->h;
        piq[icount] = pi;
        pjq[icount] = pj;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(TIMER_DOPAIR);
}

void DOSELF_NAIVE(struct runner *r, struct cell *restrict c) {

  const struct engine *e = r->e;

  error("Don't use in actual runs ! Slow code !");

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif

  TIMER_TIC;

  /* Anything to do here? */
  if (!cell_is_active(c, e)) return;

  const int count = c->count;
  struct part *restrict parts = c->parts;

  /* Loop over the parts in ci. */
  for (int pid = 0; pid < count; pid++) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts[pid];
    const double pix[3] = {pi->x[0], pi->x[1], pi->x[2]};
    const float hi = pi->h;
    const float hig2 = hi * hi * kernel_gamma2;

    /* Loop over the parts in cj. */
    for (int pjd = pid + 1; pjd < count; pjd++) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pj = &parts[pjd];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pix[k] - pj->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 < hig2 || r2 < pj->h * pj->h * kernel_gamma2) {

#ifndef WITH_VECTORIZATION

        IACT(r2, dx, hi, pj->h, pi, pj);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hi;
        hjq[icount] = pj->h;
        piq[icount] = pi;
        pjq[icount] = pj;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(TIMER_DOSELF);
}

/**
 * @brief Compute the interactions between a cell pair, but only for the
 *      given indices in ci.
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param parts_i The #part to interact with @c cj.
 * @param ind The list of indices of particles in @c ci to interact with.
 * @param count The number of particles in @c ind.
 * @param cj The second #cell.
 */
void DOPAIR_SUBSET_NAIVE(struct runner *r, struct cell *restrict ci,
                         struct part *restrict parts_i, int *restrict ind,
                         int count, struct cell *restrict cj) {

  struct engine *e = r->e;

  error("Don't use in actual runs ! Slow code !");

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif

  TIMER_TIC;

  const int count_j = cj->count;
  struct part *restrict parts_j = cj->parts;

  /* Get the relative distance between the pairs, wrapping. */
  double shift[3] = {0.0, 0.0, 0.0};
  for (int k = 0; k < 3; k++) {
    if (cj->loc[k] - ci->loc[k] < -e->s->dim[k] / 2)
      shift[k] = e->s->dim[k];
    else if (cj->loc[k] - ci->loc[k] > e->s->dim[k] / 2)
      shift[k] = -e->s->dim[k];
  }

  /* Loop over the parts_i. */
  for (int pid = 0; pid < count; pid++) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts_i[ind[pid]];
    double pix[3];
    for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];
    const float hi = pi->h;
    const float hig2 = hi * hi * kernel_gamma2;

    /* Loop over the parts in cj. */
    for (int pjd = 0; pjd < count_j; pjd++) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pj = &parts_j[pjd];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pix[k] - pj->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 < hig2) {

#ifndef WITH_VECTORIZATION

        IACT_NONSYM(r2, dx, hi, pj->h, pi, pj);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hi;
        hjq[icount] = pj->h;
        piq[icount] = pi;
        pjq[icount] = pj;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(timer_dopair_subset);
}

/**
 * @brief Compute the interactions between a cell pair, but only for the
 *      given indices in ci.
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param parts_i The #part to interact with @c cj.
 * @param ind The list of indices of particles in @c ci to interact with.
 * @param count The number of particles in @c ind.
 * @param cj The second #cell.
 */
void DOPAIR_SUBSET(struct runner *r, struct cell *restrict ci,
                   struct part *restrict parts_i, int *restrict ind, int count,
                   struct cell *restrict cj) {

  struct engine *e = r->e;

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif

  TIMER_TIC;

  const int count_j = cj->count;
  struct part *restrict parts_j = cj->parts;

  /* Get the relative distance between the pairs, wrapping. */
  double shift[3] = {0.0, 0.0, 0.0};
  for (int k = 0; k < 3; k++) {
    if (cj->loc[k] - ci->loc[k] < -e->s->dim[k] / 2)
      shift[k] = e->s->dim[k];
    else if (cj->loc[k] - ci->loc[k] > e->s->dim[k] / 2)
      shift[k] = -e->s->dim[k];
  }

  /* Get the sorting index. */
  int sid = 0;
  for (int k = 0; k < 3; k++)
    sid = 3 * sid + ((cj->loc[k] - ci->loc[k] + shift[k] < 0)
                         ? 0
                         : (cj->loc[k] - ci->loc[k] + shift[k] > 0) ? 2 : 1);

  /* Switch the cells around? */
  const int flipped = runner_flip[sid];
  sid = sortlistID[sid];

  /* Have the cells been sorted? */
  if (!(cj->sorted & (1 << sid))) error("Trying to interact unsorted cells.");

  /* Pick-out the sorted lists. */
  const struct entry *restrict sort_j = &cj->sort[sid * (cj->count + 1)];
  const float dxj = cj->dx_max;

  /* Parts are on the left? */
  if (!flipped) {

    /* Loop over the parts_i. */
    for (int pid = 0; pid < count; pid++) {

      /* Get a hold of the ith part in ci. */
      struct part *restrict pi = &parts_i[ind[pid]];
      double pix[3];
      for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];

      const float hi = pi->h;
      const float hig2 = hi * hi * kernel_gamma2;
      const float di = hi * kernel_gamma + dxj + pix[0] * runner_shift[sid][0] +
                       pix[1] * runner_shift[sid][1] +
                       pix[2] * runner_shift[sid][2];

      /* Loop over the parts in cj. */
      for (int pjd = 0; pjd < count_j && sort_j[pjd].d < di; pjd++) {

        /* Get a pointer to the jth particle. */
        struct part *restrict pj = &parts_j[sort_j[pjd].i];

        /* Compute the pairwise distance. */
        float r2 = 0.0f;
        float dx[3];
        for (int k = 0; k < 3; k++) {
          dx[k] = pix[k] - pj->x[k];
          r2 += dx[k] * dx[k];
        }

        /* Hit or miss? */
        if (r2 < hig2) {

#ifndef WITH_VECTORIZATION

          IACT_NONSYM(r2, dx, hi, pj->h, pi, pj);

#else

          /* Add this interaction to the queue. */
          r2q[icount] = r2;
          dxq[3 * icount + 0] = dx[0];
          dxq[3 * icount + 1] = dx[1];
          dxq[3 * icount + 2] = dx[2];
          hiq[icount] = hi;
          hjq[icount] = pj->h;
          piq[icount] = pi;
          pjq[icount] = pj;
          icount += 1;

          /* Flush? */
          if (icount == VEC_SIZE) {
            IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
            icount = 0;
          }

#endif
        }

      } /* loop over the parts in cj. */

    } /* loop over the parts in ci. */

  }

  /* Parts are on the right. */
  else {

    /* Loop over the parts_i. */
    for (int pid = 0; pid < count; pid++) {

      /* Get a hold of the ith part in ci. */
      struct part *restrict pi = &parts_i[ind[pid]];
      double pix[3];
      for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];
      const float hi = pi->h;
      const float hig2 = hi * hi * kernel_gamma2;
      const float di =
          -hi * kernel_gamma - dxj + pix[0] * runner_shift[sid][0] +
          pix[1] * runner_shift[sid][1] + pix[2] * runner_shift[sid][2];

      /* Loop over the parts in cj. */
      for (int pjd = count_j - 1; pjd >= 0 && di < sort_j[pjd].d; pjd--) {

        /* Get a pointer to the jth particle. */
        struct part *restrict pj = &parts_j[sort_j[pjd].i];

        /* Compute the pairwise distance. */
        float r2 = 0.0f;
        float dx[3];
        for (int k = 0; k < 3; k++) {
          dx[k] = pix[k] - pj->x[k];
          r2 += dx[k] * dx[k];
        }

        /* Hit or miss? */
        if (r2 < hig2) {

#ifndef WITH_VECTORIZATION

          IACT_NONSYM(r2, dx, hi, pj->h, pi, pj);

#else

          /* Add this interaction to the queue. */
          r2q[icount] = r2;
          dxq[3 * icount + 0] = dx[0];
          dxq[3 * icount + 1] = dx[1];
          dxq[3 * icount + 2] = dx[2];
          hiq[icount] = hi;
          hjq[icount] = pj->h;
          piq[icount] = pi;
          pjq[icount] = pj;
          icount += 1;

          /* Flush? */
          if (icount == VEC_SIZE) {
            IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
            icount = 0;
          }

#endif
        }

      } /* loop over the parts in cj. */

    } /* loop over the parts in ci. */
  }

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(timer_dopair_subset);
}

/**
 * @brief Compute the interactions between a cell pair, but only for the
 *      given indices in ci.
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param parts The #part to interact.
 * @param ind The list of indices of particles in @c ci to interact with.
 * @param count The number of particles in @c ind.
 */
void DOSELF_SUBSET(struct runner *r, struct cell *restrict ci,
                   struct part *restrict parts, int *restrict ind, int count) {

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif

  TIMER_TIC;

  const int count_i = ci->count;
  struct part *restrict parts_j = ci->parts;

  /* Loop over the parts in ci. */
  for (int pid = 0; pid < count; pid++) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts[ind[pid]];
    const double pix[3] = {pi->x[0], pi->x[1], pi->x[2]};
    const float hi = pi->h;
    const float hig2 = hi * hi * kernel_gamma2;

    /* Loop over the parts in cj. */
    for (int pjd = 0; pjd < count_i; pjd++) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pj = &parts_j[pjd];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pix[k] - pj->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 > 0.0f && r2 < hig2) {

#ifndef WITH_VECTORIZATION

        IACT_NONSYM(r2, dx, hi, pj->h, pi, pj);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hi;
        hjq[icount] = pj->h;
        piq[icount] = pi;
        pjq[icount] = pj;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(timer_dopair_subset);
}

/**
 * @brief Compute the interactions between a cell pair (non-symmetric).
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param cj The second #cell.
 */
void DOPAIR1(struct runner *r, struct cell *ci, struct cell *cj) {

  const struct engine *restrict e = r->e;

#ifdef WITH_VECTORIZATION
  int icount = 0;
  float r2q[VEC_SIZE] __attribute__((aligned(16)));
  float hiq[VEC_SIZE] __attribute__((aligned(16)));
  float hjq[VEC_SIZE] __attribute__((aligned(16)));
  float dxq[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq[VEC_SIZE], *pjq[VEC_SIZE];
#endif

  TIMER_TIC;

  /* Anything to do here? */
  if (!cell_is_active(ci, e) && !cell_is_active(cj, e)) return;

#ifdef SWIFT_DEBUG_CHECKS
  cell_is_drifted(ci, e);
  cell_is_drifted(cj, e);
#endif

  /* Get the sort ID. */
  double shift[3] = {0.0, 0.0, 0.0};
  const int sid = space_getsid(e->s, &ci, &cj, shift);

  /* Have the cells been sorted? */
  if (!(ci->sorted & (1 << sid)) || !(cj->sorted & (1 << sid)))
    error("Trying to interact unsorted cells.");

  /* Get the cutoff shift. */
  double rshift = 0.0;
  for (int k = 0; k < 3; k++) rshift += shift[k] * runner_shift[sid][k];

  /* Pick-out the sorted lists. */
  const struct entry *restrict sort_i = &ci->sort[sid * (ci->count + 1)];
  const struct entry *restrict sort_j = &cj->sort[sid * (cj->count + 1)];

  /* Get some other useful values. */
  const double hi_max = ci->h_max * kernel_gamma - rshift;
  const double hj_max = cj->h_max * kernel_gamma;
  const int count_i = ci->count;
  const int count_j = cj->count;
  struct part *restrict parts_i = ci->parts;
  struct part *restrict parts_j = cj->parts;
  const double di_max = sort_i[count_i - 1].d - rshift;
  const double dj_min = sort_j[0].d;
  const float dx_max = (ci->dx_max + cj->dx_max);

  /* Loop over the parts in ci. */
  for (int pid = count_i - 1;
       pid >= 0 && sort_i[pid].d + hi_max + dx_max > dj_min; pid--) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts_i[sort_i[pid].i];
    if (!part_is_active(pi, e)) continue;
    const float hi = pi->h;
    const double di = sort_i[pid].d + hi * kernel_gamma + dx_max - rshift;
    if (di < dj_min) continue;

    double pix[3];
    for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];
    const float hig2 = hi * hi * kernel_gamma2;

    /* Loop over the parts in cj. */
    for (int pjd = 0; pjd < count_j && sort_j[pjd].d < di; pjd++) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pj = &parts_j[sort_j[pjd].i];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pix[k] - pj->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 < hig2) {

#ifndef WITH_VECTORIZATION

        IACT_NONSYM(r2, dx, hi, pj->h, pi, pj);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hi;
        hjq[icount] = pj->h;
        piq[icount] = pi;
        pjq[icount] = pj;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

  /* Loop over the parts in cj. */
  for (int pjd = 0; pjd < count_j && sort_j[pjd].d - hj_max - dx_max < di_max;
       pjd++) {

    /* Get a hold of the jth part in cj. */
    struct part *restrict pj = &parts_j[sort_j[pjd].i];
    if (!part_is_active(pj, e)) continue;
    const float hj = pj->h;
    const double dj = sort_j[pjd].d - hj * kernel_gamma - dx_max - rshift;
    if (dj > di_max) continue;

    double pjx[3];
    for (int k = 0; k < 3; k++) pjx[k] = pj->x[k] + shift[k];
    const float hjg2 = hj * hj * kernel_gamma2;

    /* Loop over the parts in ci. */
    for (int pid = count_i - 1; pid >= 0 && sort_i[pid].d > dj; pid--) {

      /* Get a pointer to the jth particle. */
      struct part *restrict pi = &parts_i[sort_i[pid].i];

      /* Compute the pairwise distance. */
      float r2 = 0.0f;
      float dx[3];
      for (int k = 0; k < 3; k++) {
        dx[k] = pjx[k] - pi->x[k];
        r2 += dx[k] * dx[k];
      }

      /* Hit or miss? */
      if (r2 < hjg2) {

#ifndef WITH_VECTORIZATION

        IACT_NONSYM(r2, dx, hj, pi->h, pj, pi);

#else

        /* Add this interaction to the queue. */
        r2q[icount] = r2;
        dxq[3 * icount + 0] = dx[0];
        dxq[3 * icount + 1] = dx[1];
        dxq[3 * icount + 2] = dx[2];
        hiq[icount] = hj;
        hjq[icount] = pi->h;
        piq[icount] = pj;
        pjq[icount] = pi;
        icount += 1;

        /* Flush? */
        if (icount == VEC_SIZE) {
          IACT_NONSYM_VEC(r2q, dxq, hiq, hjq, piq, pjq);
          icount = 0;
        }

#endif
      }

    } /* loop over the parts in cj. */

  } /* loop over the parts in ci. */

#ifdef WITH_VECTORIZATION
  /* Pick up any leftovers. */
  if (icount > 0)
    for (int k = 0; k < icount; k++)
      IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif

  TIMER_TOC(TIMER_DOPAIR);
}

/**
 * @brief Compute the interactions between a cell pair (symmetric)
 *
 * @param r The #runner.
 * @param ci The first #cell.
 * @param cj The second #cell.
 */
void DOPAIR2(struct runner *r, struct cell *ci, struct cell *cj) {

  struct engine *restrict e = r->e;

#ifdef WITH_VECTORIZATION
  int icount1 = 0;
  float r2q1[VEC_SIZE] __attribute__((aligned(16)));
  float hiq1[VEC_SIZE] __attribute__((aligned(16)));
  float hjq1[VEC_SIZE] __attribute__((aligned(16)));
  float dxq1[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq1[VEC_SIZE], *pjq1[VEC_SIZE];
  int icount2 = 0;
  float r2q2[VEC_SIZE] __attribute__((aligned(16)));
  float hiq2[VEC_SIZE] __attribute__((aligned(16)));
  float hjq2[VEC_SIZE] __attribute__((aligned(16)));
  float dxq2[3 * VEC_SIZE] __attribute__((aligned(16)));
  struct part *piq2[VEC_SIZE], *pjq2[VEC_SIZE];
#endif

  TIMER_TIC;

  /* Anything to do here? */
  if (!cell_is_active(ci, e) && !cell_is_active(cj, e)) return;

#ifdef SWIFT_DEBUG_CHECKS
  cell_is_drifted(ci, e);
  cell_is_drifted(cj, e);
#endif

  /* Get the shift ID. */
  double shift[3] = {0.0, 0.0, 0.0};
  const int sid = space_getsid(e->s, &ci, &cj, shift);

  /* Have the cells been sorted? */
  if (!(ci->sorted & (1 << sid)) || !(cj->sorted & (1 << sid)))
    error("Trying to interact unsorted cells.");

  /* Get the cutoff shift. */
  double rshift = 0.0;
  for (int k = 0; k < 3; k++) rshift += shift[k] * runner_shift[sid][k];

  /* Pick-out the sorted lists. */
  struct entry *restrict sort_i = &ci->sort[sid * (ci->count + 1)];
  struct entry *restrict sort_j = &cj->sort[sid * (cj->count + 1)];

  /* Get some other useful values. */
  const double hi_max = ci->h_max * kernel_gamma - rshift;
  const double hj_max = cj->h_max * kernel_gamma;
  const int count_i = ci->count;
  const int count_j = cj->count;
  struct part *restrict parts_i = ci->parts;
  struct part *restrict parts_j = cj->parts;
  const double di_max = sort_i[count_i - 1].d - rshift;
  const double dj_min = sort_j[0].d;
  const double dx_max = (ci->dx_max + cj->dx_max);

  /* Collect the number of parts left and right below dt. */
  int countdt_i = 0, countdt_j = 0;
  struct entry *restrict sortdt_i = NULL, *restrict sortdt_j = NULL;
  if (cell_is_all_active(ci, e)) {
    sortdt_i = sort_i;
    countdt_i = count_i;
  } else if (cell_is_active(ci, e)) {
    if (posix_memalign((void *)&sortdt_i, VEC_SIZE * sizeof(float),
                       sizeof(struct entry) * count_i) != 0)
      error("Failed to allocate dt sortlists.");
    for (int k = 0; k < count_i; k++)
      if (part_is_active(&parts_i[sort_i[k].i], e)) {
        sortdt_i[countdt_i] = sort_i[k];
        countdt_i += 1;
      }
  }
  if (cell_is_all_active(cj, e)) {
    sortdt_j = sort_j;
    countdt_j = count_j;
  } else if (cell_is_active(cj, e)) {
    if (posix_memalign((void *)&sortdt_j, VEC_SIZE * sizeof(float),
                       sizeof(struct entry) * count_j) != 0)
      error("Failed to allocate dt sortlists.");
    for (int k = 0; k < count_j; k++)
      if (part_is_active(&parts_j[sort_j[k].i], e)) {
        sortdt_j[countdt_j] = sort_j[k];
        countdt_j += 1;
      }
  }

  /* Loop over the parts in ci. */
  for (int pid = count_i - 1;
       pid >= 0 && sort_i[pid].d + hi_max + dx_max > dj_min; pid--) {

    /* Get a hold of the ith part in ci. */
    struct part *restrict pi = &parts_i[sort_i[pid].i];
    const float hi = pi->h;
    const double di = sort_i[pid].d + hi * kernel_gamma + dx_max - rshift;
    if (di < dj_min) continue;

    double pix[3];
    for (int k = 0; k < 3; k++) pix[k] = pi->x[k] - shift[k];
    const float hig2 = hi * hi * kernel_gamma2;

    /* Look at valid dt parts only? */
    if (!part_is_active(pi, e)) {

      /* Loop over the parts in cj within dt. */