// Copyright 2009-2021 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#include "../common/tutorial/tutorial_device.h"

namespace embree {

const int numPhi = 20;
const int numTheta = 2*numPhi;
const int numSpheres = 64;

/* state of the lazy geometry */
enum LazyState
{
  LAZY_INVALID = 0,   // the geometry is not yet created
  LAZY_CREATE = 1,    // one thread is creating the geometry
  LAZY_COMMIT = 2,    // possible multiple threads are committing the geometry
  LAZY_VALID = 3      // the geometry is created
};

/* representation for our lazy geometry */
struct LazyGeometry
{
  ALIGNED_STRUCT_(16)
  RTCGeometry geometry;
  LazyState state;
  RTCScene object;
  int userID;
  Vec3fa center;
  float radius;
};

LazyGeometry* g_objects[numSpheres];

void instanceBoundsFunc(const struct RTCBoundsFunctionArguments* args)
{
  const LazyGeometry* instance = (const LazyGeometry*) args->geometryUserPtr;
  RTCBounds* bounds_o = args->bounds_o;
  Vec3fa lower = instance->center-Vec3fa(instance->radius);
  Vec3fa upper = instance->center+Vec3fa(instance->radius);
  bounds_o->lower_x = lower.x;
  bounds_o->lower_y = lower.y;
  bounds_o->lower_z = lower.z;
  bounds_o->upper_x = upper.x;
  bounds_o->upper_y = upper.y;
  bounds_o->upper_z = upper.z;
}

unsigned int createTriangulatedSphere (RTCScene scene, const Vec3fa& p, float r)
{
  /* create triangle mesh */
  RTCGeometry geom = rtcNewGeometry (g_device, RTC_GEOMETRY_TYPE_TRIANGLE);

  /* map triangle and vertex buffers */
  Vertex* vertices = (Vertex*) rtcSetNewGeometryBuffer(geom, RTC_BUFFER_TYPE_VERTEX, 0, RTC_FORMAT_FLOAT3, sizeof(Vertex), numTheta*(numPhi+1));
  Triangle* triangles = (Triangle*) rtcSetNewGeometryBuffer(geom, RTC_BUFFER_TYPE_INDEX, 0, RTC_FORMAT_UINT3, sizeof(Triangle), 2*numTheta*(numPhi-1));

  /* create sphere */
  int tri = 0;
  const float rcpNumTheta = rcp((float)numTheta);
  const float rcpNumPhi   = rcp((float)numPhi);
  for (int phi=0; phi<=numPhi; phi++)
  {
    for (int theta=0; theta<numTheta; theta++)
    {
      const float phif   = phi*float(pi)*rcpNumPhi;
      const float thetaf = theta*2.0f*float(pi)*rcpNumTheta;

      Vertex& v = vertices[phi*numTheta+theta];
      v.x = p.x + r*sin(phif)*sin(thetaf);
      v.y = p.y + r*cos(phif);
      v.z = p.z + r*sin(phif)*cos(thetaf);
    }
    if (phi == 0) continue;

    for (int theta=1; theta<=numTheta; theta++)
    {
      int p00 = (phi-1)*numTheta+theta-1;
      int p01 = (phi-1)*numTheta+theta%numTheta;
      int p10 = phi*numTheta+theta-1;
      int p11 = phi*numTheta+theta%numTheta;

      if (phi > 1) {
        triangles[tri].v0 = p10;
        triangles[tri].v1 = p01;
        triangles[tri].v2 = p00;
        tri++;
      }

      if (phi < numPhi) {
        triangles[tri].v0 = p11;
        triangles[tri].v1 = p01;
        triangles[tri].v2 = p10;
        tri++;
      }
    }
  }

  rtcCommitGeometry(geom);
  unsigned int geomID = rtcAttachGeometry(scene,geom);
  rtcReleaseGeometry(geom);
  return geomID;
}

void lazyCreate(LazyGeometry* instance)
{
  const bool join_commit_supported = rtcGetDeviceProperty(g_device,RTC_DEVICE_PROPERTY_JOIN_COMMIT_SUPPORTED);
  const bool parallel_commit_supported = rtcGetDeviceProperty(g_device,RTC_DEVICE_PROPERTY_PARALLEL_COMMIT_SUPPORTED);
    
  /* one thread will switch the object from the LAZY_INVALID state to the LAZY_CREATE state */
  if (atomic_cmpxchg((int32_t*)&instance->state,LAZY_INVALID,LAZY_CREATE) == 0)
  {
    /* create the geometry */
    //printf("creating sphere %i (lazy)\n",instance->userID);
    instance->object = rtcNewScene(g_device);
    createTriangulatedSphere(instance->object,instance->center,instance->radius);

    /* when no parallel commit mode at all is supported let only a single thread build */
    if (!join_commit_supported && !parallel_commit_supported)
      rtcCommitScene(instance->object);

    /* now switch to the LAZY_COMMIT state */
    __memory_barrier();
    instance->state = LAZY_COMMIT;
  }
  else
  {
    /* wait until the geometry got created */
    while (atomic_cmpxchg((int32_t*)&instance->state,10,11) < LAZY_COMMIT) {
      // instead of actively spinning here, best use a condition to let the thread sleep, or let it help in the creation stage
    }
  }

  /* if we support rtcCommit to get called from multiple threads, then this should be preferred for performance reasons */
  if (parallel_commit_supported)
    rtcCommitScene(instance->object);
  
  /* otherwise we could fallback to rtcJoinCommit scene, which has lower performance */
  else if (join_commit_supported)
    rtcJoinCommitScene(instance->object);

  /* switch to LAZY_VALID state */
  atomic_cmpxchg((int32_t*)&instance->state,LAZY_COMMIT,LAZY_VALID);
}

void eagerCreate(LazyGeometry* instance)
{
  //printf("creating sphere %i (eager)\n",instance->userID);
  instance->object = rtcNewScene(g_device);
  createTriangulatedSphere(instance->object,instance->center,instance->radius);
  rtcCommitScene(instance->object);
  instance->state = LAZY_VALID;
}

void instanceIntersectFuncN(const RTCIntersectFunctionNArguments* args)
{
  const int* valid = args->valid;
  void* ptr  = args->geometryUserPtr;
  RTCRayQueryContext* context = args->context;
  RTCRayHitN* rays = (RTCRayHitN*)args->rayhit;
  assert(args->N == 1);
  LazyGeometry* instance = (LazyGeometry*)ptr;

  if (!valid[0])
    return;
  
  Ray *ray = (Ray *)rays;
  
  /* create the object if it is not yet created */
  if (instance->state != LAZY_VALID)
    lazyCreate(instance);
  
  /* trace ray inside object */
  const unsigned int geomID = ray->geomID;
  ray->geomID = RTC_INVALID_GEOMETRY_ID;
  RTCIntersectArguments iargs;
  rtcInitIntersectArguments(&iargs);
  iargs.context = context;
  rtcIntersect1(instance->object,RTCRayHit_(*ray),&iargs);
  if (ray->geomID == RTC_INVALID_GEOMETRY_ID) ray->geomID = geomID;
  else ray->instID[0] = instance->userID;
}

void instanceOccludedFuncN(const RTCOccludedFunctionNArguments* args)
{
  const int* valid = args->valid;
  void* ptr  = args->geometryUserPtr;
  RTCRayQueryContext* context = args->context;
  RTCRayHitN* rays = (RTCRayHitN*)args->ray;
  assert(args->N == 1);
  LazyGeometry* instance = (LazyGeometry*)ptr;

  if (!valid[0])
    return;
  
  Ray *ray = (Ray *)rays;
  /* create the object if it is not yet created */
  if (instance->state != LAZY_VALID)
    lazyCreate(instance);
  
  /* trace ray inside object */
  RTCOccludedArguments sargs;
  rtcInitOccludedArguments(&sargs);
  sargs.context = context;
  rtcOccluded1(instance->object,RTCRay_(*ray),&sargs);
}

LazyGeometry* createLazyObject (RTCScene scene, int userID, const Vec3fa& center, const float radius)
{
  LazyGeometry* instance = (LazyGeometry*) alignedUSMMalloc(sizeof(LazyGeometry),16);
  instance->state = LAZY_INVALID;
  instance->object = nullptr;
  instance->userID = userID;
  instance->center = center;
  instance->radius = radius;
  instance->geometry = rtcNewGeometry(g_device, RTC_GEOMETRY_TYPE_USER);
  rtcSetGeometryUserPrimitiveCount(instance->geometry,1);
  rtcSetGeometryUserData(instance->geometry,instance);
  rtcSetGeometryBoundsFunction(instance->geometry,instanceBoundsFunc,nullptr);
  rtcSetGeometryIntersectFunction(instance->geometry,instanceIntersectFuncN);
  rtcSetGeometryOccludedFunction (instance->geometry,instanceOccludedFuncN);
  rtcCommitGeometry(instance->geometry);
  rtcAttachGeometry(scene,instance->geometry);
  rtcReleaseGeometry(instance->geometry);

  /* if we do not support the join mode then Embree also does not
   * support lazy build */
  if (!rtcGetDeviceProperty(g_device,RTC_DEVICE_PROPERTY_JOIN_COMMIT_SUPPORTED))
    eagerCreate(instance);

  return instance;
}

/* creates a ground plane */
unsigned int createGroundPlane (RTCScene scene)
{
  /* create a triangulated plane with 2 triangles and 4 vertices */
  RTCGeometry geom = rtcNewGeometry (g_device, RTC_GEOMETRY_TYPE_TRIANGLE);

  /* set vertices */
  Vertex* vertices = (Vertex*) rtcSetNewGeometryBuffer(geom, RTC_BUFFER_TYPE_VERTEX, 0, RTC_FORMAT_FLOAT3, sizeof(Vertex), 4);
  vertices[0].x = -10; vertices[0].y = -2; vertices[0].z = -10;
  vertices[1].x = -10; vertices[1].y = -2; vertices[1].z = +10;
  vertices[2].x = +10; vertices[2].y = -2; vertices[2].z = -10;
  vertices[3].x = +10; vertices[3].y = -2; vertices[3].z = +10;

  /* set triangles */
  Triangle* triangles = (Triangle*) rtcSetNewGeometryBuffer(geom, RTC_BUFFER_TYPE_INDEX, 0, RTC_FORMAT_UINT3, sizeof(Triangle), 2);
  triangles[0].v0 = 0; triangles[0].v1 = 1; triangles[0].v2 = 2;
  triangles[1].v0 = 1; triangles[1].v1 = 3; triangles[1].v2 = 2;

  rtcCommitGeometry(geom);
  unsigned int geomID = rtcAttachGeometry(scene,geom);
  rtcReleaseGeometry(geom);
  return geomID;
}

/* scene data */
RTCScene g_scene  = nullptr;

/* called by the C++ code for initialization */
extern "C" void device_init (char* cfg)
{
  /* create scene */
  g_scene = rtcNewScene(g_device);
  
  /* instantiate geometry */
  createGroundPlane(g_scene);
  for (int i=0; i<numSpheres; i++) {
    float a = 2.0f*float(M_PI)*(float)i/(float)numSpheres;
    g_objects[i] = createLazyObject(g_scene,i,10.0f*Vec3fa(cosf(a),0,sinf(a)),1);
  }
  rtcCommitScene (g_scene);
}

/* task that renders a single screen tile */
Vec3fa renderPixelStandard(float x, float y, const ISPCCamera& camera, RayStats& stats)
{
  /* initialize ray */
  Ray ray(Vec3fa(camera.xfm.p), Vec3fa(normalize(x*camera.xfm.l.vx + y*camera.xfm.l.vy + camera.xfm.l.vz)), 0.0f, inf, 0.0f, -1, RTC_INVALID_GEOMETRY_ID, RTC_INVALID_GEOMETRY_ID);
  ray.instID[0] = 4;

  /* intersect ray with scene */
  rtcIntersect1(g_scene,RTCRayHit_(ray));
  RayStats_addRay(stats);

  /* shade pixels */
  Vec3fa color = Vec3fa(0.0f);
  if (ray.geomID != RTC_INVALID_GEOMETRY_ID)
  {
    Vec3fa diffuse = Vec3fa(1.0f);
    color = color + diffuse*0.5;
    Vec3fa lightDir = normalize(Vec3fa(-1,-1,-1));

    /* initialize shadow ray */
    Ray shadow(ray.org + ray.tfar*ray.dir, neg(lightDir), 0.001f, inf);

    /* trace shadow ray */
    rtcOccluded1(g_scene,RTCRay_(shadow));
    RayStats_addShadowRay(stats);

    /* add light contribution */
    if (shadow.tfar >= 0.0f)
      color = color + diffuse*clamp(-dot(lightDir,normalize(ray.Ng)),0.0f,1.0f);
  }
  return color;
}

/* renders a single screen tile */
void renderTileStandard(int taskIndex,
                        int threadIndex,
                        int* pixels,
                        const unsigned int width,
                        const unsigned int height,
                        const float time,
                        const ISPCCamera& camera,
                        const int numTilesX,
                        const int numTilesY)
{
  const unsigned int tileY = taskIndex / numTilesX;
  const unsigned int tileX = taskIndex - tileY * numTilesX;
  const unsigned int x0 = tileX * TILE_SIZE_X;
  const unsigned int x1 = min(x0+TILE_SIZE_X,width);
  const unsigned int y0 = tileY * TILE_SIZE_Y;
  const unsigned int y1 = min(y0+TILE_SIZE_Y,height);

  for (unsigned int y=y0; y<y1; y++) for (unsigned int x=x0; x<x1; x++)
  {
    /* calculate pixel color */
    Vec3fa color = renderPixelStandard((float)x,(float)y,camera,g_stats[threadIndex]);

    /* write color to framebuffer */
    unsigned int r = (unsigned int) (255.0f * clamp(color.x,0.0f,1.0f));
    unsigned int g = (unsigned int) (255.0f * clamp(color.y,0.0f,1.0f));
    unsigned int b = (unsigned int) (255.0f * clamp(color.z,0.0f,1.0f));
    pixels[y*width+x] = (b << 16) + (g << 8) + r;
  }
}

/* task that renders a single screen tile */
void renderTileTask (int taskIndex, int threadIndex, int* pixels,
                         const unsigned int width,
                         const unsigned int height,
                         const float time,
                         const ISPCCamera& camera,
                         const int numTilesX,
                         const int numTilesY)
{
  renderTileStandard(taskIndex,threadIndex,pixels,width,height,time,camera,numTilesX,numTilesY);
}

extern "C" void renderFrameStandard (int* pixels,
                          const unsigned int width,
                          const unsigned int height,
                          const float time,
                          const ISPCCamera& camera)
{
  /* render all pixels */
  const int numTilesX = (width +TILE_SIZE_X-1)/TILE_SIZE_X;
  const int numTilesY = (height+TILE_SIZE_Y-1)/TILE_SIZE_Y;
  parallel_for(size_t(0),size_t(numTilesX*numTilesY),[&](const range<size_t>& range) {
    const int threadIndex = (int)TaskScheduler::threadIndex();
    for (size_t i=range.begin(); i<range.end(); i++)
      renderTileTask((int)i,threadIndex,pixels,width,height,time,camera,numTilesX,numTilesY);
  }); 
}

/* called by the C++ code to render */
extern "C" void device_render (int* pixels,
                           const unsigned int width,
                           const unsigned int height,
                           const float time,
                           const ISPCCamera& camera)
{
}

/* called by the C++ code for cleanup */
extern "C" void device_cleanup ()
{
  for (int i=0; i<numSpheres; i++) {
    if (g_objects[i]->object) rtcReleaseScene(g_objects[i]->object);
    delete g_objects[i];
  }
  rtcReleaseScene (g_scene); g_scene = nullptr;
}

} // namespace embree