const int particles_length = 128;

for (int i = 0; i < particles_length; i++) {
  T0V(VAP_PVS_STATE_FLUSH_REG, 0x00000000);

  const vec3 position = particles[i].position;

  const float consts[] = {
    // 0: local space to clip space transformation matrix
    trans[0][0], trans[0][1], trans[0][2], trans[0][3],
    trans[1][0], trans[1][1], trans[1][2], trans[1][3],
    trans[2][0], trans[2][1], trans[2][2], trans[2][3],
    trans[3][0], trans[3][1], trans[3][2], trans[3][3],

    // 4: dx ("right" change of basis vector)
    local_to_view[0][0], local_to_view[0][1], local_to_view[0][2], 0,

    // 5: dy ("up" change of basis vector)
    local_to_view[1][0], local_to_view[1][1], local_to_view[1][2], 0,

    // 6: particle position, scale
    position.x, position.y, position.z, scale
  };
  // transfer the constants to vertex shader constant memory
  ib_vap_pvs_const_cntl(consts, (sizeof (consts)));

  const int dwords_per_vtx = 2;

  T0V(VAP_VTX_SIZE
      , VAP_VTX_SIZE__DWORDS_PER_VTX(dwords_per_vtx)
      );

  const int vertex_count = 4;
  const vec2 vertices[vertex_count] = {
    {0.0, 0.0f},
    {1.0, 0.0f},
    {1.0, 1.0f},
    {0.0, 1.0f},
  };

  T3(_3D_DRAW_IMMD_2, (1 + vertex_count * dwords_per_vtx) - 1);
  TU( VAP_VF_CNTL__PRIM_TYPE(13) // quad list
    | VAP_VF_CNTL__PRIM_WALK(3)  // embedded/immediate vertex data
    | VAP_VF_CNTL__INDEX_SIZE(0)
    | VAP_VF_CNTL__VTX_REUSE_DIS(0)
    | VAP_VF_CNTL__DUAL_INDEX_MODE(0)
    | VAP_VF_CNTL__USE_ALT_NUM_VERTS(0)
    | VAP_VF_CNTL__NUM_VERTICES(vertex_count)
    );

  // transfer particle quad vertices
  for (int j = 0; j < vertex_count; j++) {
    TF(vertices[i].x);
    TF(vertices[i].y);
  }
}