device_cuda.cpp

/*
 * Copyright 2011-2013 Blender Foundation
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <climits>
#include <limits.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "device/device.h"
#include "device/device_denoising.h"
#include "device/device_intern.h"
#include "device/device_split_kernel.h"

#include "render/buffers.h"

#include "kernel/filter/filter_defines.h"

#ifdef WITH_CUDA_DYNLOAD
#  include "cuew.h"
#else
#  include "util/util_opengl.h"
#  include <cuda.h>
#  include <cudaGL.h>
#endif
#include "util/util_debug.h"
#include "util/util_foreach.h"
#include "util/util_logging.h"
#include "util/util_map.h"
#include "util/util_md5.h"
#include "util/util_opengl.h"
#include "util/util_path.h"
#include "util/util_string.h"
#include "util/util_system.h"
#include "util/util_types.h"
#include "util/util_time.h"
#include "util/util_windows.h"

#include "kernel/split/kernel_split_data_types.h"

CCL_NAMESPACE_BEGIN

#ifndef WITH_CUDA_DYNLOAD

/* Transparently implement some functions, so majority of the file does not need
 * to worry about difference between dynamically loaded and linked CUDA at all.
 */

namespace {

const char *cuewErrorString(CUresult result)
{
  /* We can only give error code here without major code duplication, that
   * should be enough since dynamic loading is only being disabled by folks
   * who knows what they're doing anyway.
   *
   * NOTE: Avoid call from several threads.
   */
  static string error;
  error = string_printf("%d", result);
  return error.c_str();
}

const char *cuewCompilerPath()
{
  return CYCLES_CUDA_NVCC_EXECUTABLE;
}

int cuewCompilerVersion()
{
  return (CUDA_VERSION / 100) + (CUDA_VERSION % 100 / 10);
}

} /* namespace */
#endif /* WITH_CUDA_DYNLOAD */

class CUDADevice;

class CUDASplitKernel : public DeviceSplitKernel {
  CUDADevice *device;

 public:
  explicit CUDASplitKernel(CUDADevice *device);

  virtual uint64_t state_buffer_size(device_memory &kg, device_memory &data, size_t num_threads);

  virtual bool enqueue_split_kernel_data_init(const KernelDimensions &dim,
                                              RenderTile &rtile,
                                              int num_global_elements,
                                              device_memory &kernel_globals,
                                              device_memory &kernel_data_,
                                              device_memory &split_data,
                                              device_memory &ray_state,
                                              device_memory &queue_index,
                                              device_memory &use_queues_flag,
                                              device_memory &work_pool_wgs);

  virtual SplitKernelFunction *get_split_kernel_function(const string &kernel_name,
                                                         const DeviceRequestedFeatures &);
  virtual int2 split_kernel_local_size();
  virtual int2 split_kernel_global_size(device_memory &kg, device_memory &data, DeviceTask *task);
};

/* Utility to push/pop CUDA context. */
class CUDAContextScope {
 public:
  CUDAContextScope(CUDADevice *device);
  ~CUDAContextScope();

 private:
  CUDADevice *device;
};

class CUDADevice : public Device {
 public:
  DedicatedTaskPool task_pool;
  CUdevice cuDevice;
  CUcontext cuContext;
  CUmodule cuModule, cuFilterModule;
  size_t device_texture_headroom;
  size_t device_working_headroom;
  bool move_texture_to_host;
  size_t map_host_used;
  size_t map_host_limit;
  int can_map_host;
  int cuDevId;
  int cuDevArchitecture;
  bool first_error;
  CUDASplitKernel *split_kernel;

  struct CUDAMem {
    CUDAMem() : texobject(0), array(0), use_mapped_host(false)
    {
    }

    CUtexObject texobject;
    CUarray array;

    /* If true, a mapped host memory in shared_pointer is being used. */
    bool use_mapped_host;
  };
  typedef map<device_memory *, CUDAMem> CUDAMemMap;
  CUDAMemMap cuda_mem_map;

  struct PixelMem {
    GLuint cuPBO;
    CUgraphicsResource cuPBOresource;
    GLuint cuTexId;
    int w, h;
  };
  map<device_ptr, PixelMem> pixel_mem_map;

  /* Bindless Textures */
  device_vector<TextureInfo> texture_info;
  bool need_texture_info;

  CUdeviceptr cuda_device_ptr(device_ptr mem)
  {
    return (CUdeviceptr)mem;
  }

  static bool have_precompiled_kernels()
  {
    string cubins_path = path_get("lib");
    return path_exists(cubins_path);
  }

  virtual bool show_samples() const
  {
    /* The CUDADevice only processes one tile at a time, so showing samples is fine. */
    return true;
  }

  virtual BVHLayoutMask get_bvh_layout_mask() const
  {
    return BVH_LAYOUT_BVH2;
  }

  /*#ifdef NDEBUG
#define cuda_abort()
#else
#define cuda_abort() abort()
#endif*/
  void cuda_error_documentation()
  {
    if (first_error) {
      fprintf(stderr,
              "\nRefer to the Cycles GPU rendering documentation for possible solutions:\n");
      fprintf(stderr,
              "https://docs.blender.org/manual/en/latest/render/cycles/gpu_rendering.html\n\n");
      first_error = false;
    }
  }

#define cuda_assert(stmt) \
  { \
    CUresult result = stmt; \
\
    if (result != CUDA_SUCCESS) { \
      string message = string_printf( \
          "CUDA error: %s in %s, line %d", cuewErrorString(result), #stmt, __LINE__); \
      if (error_msg == "") \
        error_msg = message; \
      fprintf(stderr, "%s\n", message.c_str()); \
      /*cuda_abort();*/ \
      cuda_error_documentation(); \
    } \
  } \
  (void)0

  bool cuda_error_(CUresult result, const string &stmt)
  {
    if (result == CUDA_SUCCESS)
      return false;

    string message = string_printf("CUDA error at %s: %s", stmt.c_str(), cuewErrorString(result));
    if (error_msg == "")
      error_msg = message;
    fprintf(stderr, "%s\n", message.c_str());
    cuda_error_documentation();
    return true;
  }

#define cuda_error(stmt) cuda_error_(stmt, #stmt)

  void cuda_error_message(const string &message)
  {
    if (error_msg == "")
      error_msg = message;
    fprintf(stderr, "%s\n", message.c_str());
    cuda_error_documentation();
  }

  CUDADevice(DeviceInfo &info, Stats &stats, Profiler &profiler, bool background_)
      : Device(info, stats, profiler, background_),
        texture_info(this, "__texture_info", MEM_TEXTURE)
  {
    first_error = true;
    background = background_;

    cuDevId = info.num;
    cuDevice = 0;
    cuContext = 0;

    cuModule = 0;
    cuFilterModule = 0;

    split_kernel = NULL;

    need_texture_info = false;

    device_texture_headroom = 0;
    device_working_headroom = 0;
    move_texture_to_host = false;
    map_host_limit = 0;
    map_host_used = 0;
    can_map_host = 0;

    /* Intialize CUDA. */
    if (cuda_error(cuInit(0)))
      return;

    /* Setup device and context. */
    if (cuda_error(cuDeviceGet(&cuDevice, cuDevId)))
      return;

    /* CU_CTX_MAP_HOST for mapping host memory when out of device memory.
     * CU_CTX_LMEM_RESIZE_TO_MAX for reserving local memory ahead of render,
     * so we can predict which memory to map to host. */
    cuda_assert(
        cuDeviceGetAttribute(&can_map_host, CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY, cuDevice));

    unsigned int ctx_flags = CU_CTX_LMEM_RESIZE_TO_MAX;
    if (can_map_host) {
      ctx_flags |= CU_CTX_MAP_HOST;
      init_host_memory();
    }

    /* Create context. */
    CUresult result;

    if (background) {
      result = cuCtxCreate(&cuContext, ctx_flags, cuDevice);
    }
    else {
      result = cuGLCtxCreate(&cuContext, ctx_flags, cuDevice);

      if (result != CUDA_SUCCESS) {
        result = cuCtxCreate(&cuContext, ctx_flags, cuDevice);
        background = true;
      }
    }

    if (cuda_error_(result, "cuCtxCreate"))
      return;

    int major, minor;
    cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, cuDevId);
    cuDeviceGetAttribute(&minor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, cuDevId);
    cuDevArchitecture = major * 100 + minor * 10;

    /* Pop context set by cuCtxCreate. */
    cuCtxPopCurrent(NULL);
  }

  ~CUDADevice()
  {
    task_pool.stop();

    delete split_kernel;

    texture_info.free();

    cuda_assert(cuCtxDestroy(cuContext));
  }

  bool support_device(const DeviceRequestedFeatures & /*requested_features*/)
  {
    int major, minor;
    cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, cuDevId);
    cuDeviceGetAttribute(&minor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, cuDevId);

    /* We only support sm_30 and above */
    if (major < 3) {
      cuda_error_message(string_printf(
          "CUDA device supported only with compute capability 3.0 or up, found %d.%d.",
          major,
          minor));
      return false;
    }

    return true;
  }

  bool use_adaptive_compilation()
  {
    return DebugFlags().cuda.adaptive_compile;
  }

  bool use_split_kernel()
  {
    return DebugFlags().cuda.split_kernel;
  }

  /* Common NVCC flags which stays the same regardless of shading model,
   * kernel sources md5 and only depends on compiler or compilation settings.
   */
  string compile_kernel_get_common_cflags(const DeviceRequestedFeatures &requested_features,
                                          bool filter = false,
                                          bool split = false)
  {
    const int machine = system_cpu_bits();
    const string source_path = path_get("source");
    const string include_path = source_path;
    string cflags = string_printf(
        "-m%d "
        "--ptxas-options=\"-v\" "
        "--use_fast_math "
        "-DNVCC "
        "-I\"%s\"",
        machine,
        include_path.c_str());
    if (!filter && use_adaptive_compilation()) {
      cflags += " " + requested_features.get_build_options();
    }
    const char *extra_cflags = getenv("CYCLES_CUDA_EXTRA_CFLAGS");
    if (extra_cflags) {
      cflags += string(" ") + string(extra_cflags);
    }
#ifdef WITH_CYCLES_DEBUG
    cflags += " -D__KERNEL_DEBUG__";
#endif

    if (split) {
      cflags += " -D__SPLIT__";
    }

    return cflags;
  }

  bool compile_check_compiler()
  {
    const char *nvcc = cuewCompilerPath();
    if (nvcc == NULL) {
      cuda_error_message(
          "CUDA nvcc compiler not found. "
          "Install CUDA toolkit in default location.");
      return false;
    }
    const int cuda_version = cuewCompilerVersion();
    VLOG(1) << "Found nvcc " << nvcc << ", CUDA version " << cuda_version << ".";
    const int major = cuda_version / 10, minor = cuda_version % 10;
    if (cuda_version == 0) {
      cuda_error_message("CUDA nvcc compiler version could not be parsed.");
      return false;
    }
    if (cuda_version < 80) {
      printf(
          "Unsupported CUDA version %d.%d detected, "
          "you need CUDA 8.0 or newer.\n",
          major,
          minor);
      return false;
    }
    else if (cuda_version != 101) {
      printf(
          "CUDA version %d.%d detected, build may succeed but only "
          "CUDA 10.1 is officially supported.\n",
          major,
          minor);
    }
    return true;
  }

  string compile_kernel(const DeviceRequestedFeatures &requested_features,
                        bool filter = false,
                        bool split = false)
  {
    const char *name, *source;
    if (filter) {
      name = "filter";
      source = "filter.cu";
    }
    else if (split) {
      name = "kernel_split";
      source = "kernel_split.cu";
    }
    else {
      name = "kernel";
      source = "kernel.cu";
    }
    /* Compute cubin name. */
    int major, minor;
    cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, cuDevId);
    cuDeviceGetAttribute(&minor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, cuDevId);

    /* Attempt to use kernel provided with Blender. */
    if (!use_adaptive_compilation()) {
      const string cubin = path_get(string_printf("lib/%s_sm_%d%d.cubin", name, major, minor));
      VLOG(1) << "Testing for pre-compiled kernel " << cubin << ".";
      if (path_exists(cubin)) {
        VLOG(1) << "Using precompiled kernel.";
        return cubin;
      }
      const string ptx = path_get(string_printf("lib/%s_compute_%d%d.ptx", name, major, minor));
      VLOG(1) << "Testing for pre-compiled kernel " << ptx << ".";
      if (path_exists(ptx)) {
        VLOG(1) << "Using precompiled kernel.";
        return ptx;
      }
    }

    const string common_cflags = compile_kernel_get_common_cflags(
        requested_features, filter, split);

    /* Try to use locally compiled kernel. */
    const string source_path = path_get("source");
    const string kernel_md5 = path_files_md5_hash(source_path);

    /* We include cflags into md5 so changing cuda toolkit or changing other
     * compiler command line arguments makes sure cubin gets re-built.
     */
    const string cubin_md5 = util_md5_string(kernel_md5 + common_cflags);

    const string cubin_file = string_printf(
        "cycles_%s_sm%d%d_%s.cubin", name, major, minor, cubin_md5.c_str());
    const string cubin = path_cache_get(path_join("kernels", cubin_file));
    VLOG(1) << "Testing for locally compiled kernel " << cubin << ".";
    if (path_exists(cubin)) {
      VLOG(1) << "Using locally compiled kernel.";
      return cubin;
    }

#ifdef _WIN32
    if (have_precompiled_kernels()) {
      if (major < 3) {
        cuda_error_message(
            string_printf("CUDA device requires compute capability 3.0 or up, "
                          "found %d.%d. Your GPU is not supported.",
                          major,
                          minor));
      }
      else {
        cuda_error_message(
            string_printf("CUDA binary kernel for this graphics card compute "
                          "capability (%d.%d) not found.",
                          major,
                          minor));
      }
      return "";
    }
#endif

    /* Compile. */
    if (!compile_check_compiler()) {
      return "";
    }
    const char *nvcc = cuewCompilerPath();
    const string kernel = path_join(path_join(source_path, "kernel"),
                                    path_join("kernels", path_join("cuda", source)));
    double starttime = time_dt();
    printf("Compiling CUDA kernel ...\n");

    path_create_directories(cubin);

    string command = string_printf(
        "\"%s\" "
        "-arch=sm_%d%d "
        "--cubin \"%s\" "
        "-o \"%s\" "
        "%s ",
        nvcc,
        major,
        minor,
        kernel.c_str(),
        cubin.c_str(),
        common_cflags.c_str());

    printf("%s\n", command.c_str());

    if (system(command.c_str()) == -1) {
      cuda_error_message(
          "Failed to execute compilation command, "
          "see console for details.");
      return "";
    }

    /* Verify if compilation succeeded */
    if (!path_exists(cubin)) {
      cuda_error_message(
          "CUDA kernel compilation failed, "
          "see console for details.");
      return "";
    }

    printf("Kernel compilation finished in %.2lfs.\n", time_dt() - starttime);

    return cubin;
  }

  bool load_kernels(const DeviceRequestedFeatures &requested_features)
  {
    /* TODO(sergey): Support kernels re-load for CUDA devices.
     *
     * Currently re-loading kernel will invalidate memory pointers,
     * causing problems in cuCtxSynchronize.
     */
    if (cuFilterModule && cuModule) {
      VLOG(1) << "Skipping kernel reload, not currently supported.";
      return true;
    }

    /* check if cuda init succeeded */
    if (cuContext == 0)
      return false;

    /* check if GPU is supported */
    if (!support_device(requested_features))
      return false;

    /* get kernel */
    string cubin = compile_kernel(requested_features, false, use_split_kernel());
    if (cubin == "")
      return false;

    string filter_cubin = compile_kernel(requested_features, true, false);
    if (filter_cubin == "")
      return false;

    /* open module */
    CUDAContextScope scope(this);

    string cubin_data;
    CUresult result;

    if (path_read_text(cubin, cubin_data))
      result = cuModuleLoadData(&cuModule, cubin_data.c_str());
    else
      result = CUDA_ERROR_FILE_NOT_FOUND;

    if (cuda_error_(result, "cuModuleLoad"))
      cuda_error_message(string_printf("Failed loading CUDA kernel %s.", cubin.c_str()));

    if (path_read_text(filter_cubin, cubin_data))
      result = cuModuleLoadData(&cuFilterModule, cubin_data.c_str());
    else
      result = CUDA_ERROR_FILE_NOT_FOUND;

    if (cuda_error_(result, "cuModuleLoad"))
      cuda_error_message(string_printf("Failed loading CUDA kernel %s.", filter_cubin.c_str()));

    if (result == CUDA_SUCCESS) {
      reserve_local_memory(requested_features);
    }

    return (result == CUDA_SUCCESS);
  }

  void reserve_local_memory(const DeviceRequestedFeatures &requested_features)
  {
    if (use_split_kernel()) {
      /* Split kernel mostly uses global memory and adaptive compilation,
       * difficult to predict how much is needed currently. */
      return;
    }

    /* Together with CU_CTX_LMEM_RESIZE_TO_MAX, this reserves local memory
     * needed for kernel launches, so that we can reliably figure out when
     * to allocate scene data in mapped host memory. */
    CUDAContextScope scope(this);

    size_t total = 0, free_before = 0, free_after = 0;
    cuMemGetInfo(&free_before, &total);

    /* Get kernel function. */
    CUfunction cuPathTrace;

    if (requested_features.use_integrator_branched) {
      cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_branched_path_trace"));
    }
    else {
      cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_path_trace"));
    }

    cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1));

    int min_blocks, num_threads_per_block;
    cuda_assert(cuOccupancyMaxPotentialBlockSize(
        &min_blocks, &num_threads_per_block, cuPathTrace, NULL, 0, 0));

    /* Launch kernel, using just 1 block appears sufficient to reserve
     * memory for all multiprocessors. It would be good to do this in
     * parallel for the multi GPU case still to make it faster. */
    CUdeviceptr d_work_tiles = 0;
    uint total_work_size = 0;

    void *args[] = {&d_work_tiles, &total_work_size};

    cuda_assert(cuLaunchKernel(cuPathTrace, 1, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));

    cuda_assert(cuCtxSynchronize());

    cuMemGetInfo(&free_after, &total);
    VLOG(1) << "Local memory reserved " << string_human_readable_number(free_before - free_after)
            << " bytes. (" << string_human_readable_size(free_before - free_after) << ")";

#if 0
    /* For testing mapped host memory, fill up device memory. */
    const size_t keep_mb = 1024;

    while (free_after > keep_mb * 1024 * 1024LL) {
      CUdeviceptr tmp;
      cuda_assert(cuMemAlloc(&tmp, 10 * 1024 * 1024LL));
      cuMemGetInfo(&free_after, &total);
    }
#endif
  }

  void init_host_memory()
  {
    /* Limit amount of host mapped memory, because allocating too much can
     * cause system instability. Leave at least half or 4 GB of system
     * memory free, whichever is smaller. */
    size_t default_limit = 4 * 1024 * 1024 * 1024LL;
    size_t system_ram = system_physical_ram();

    if (system_ram > 0) {
      if (system_ram / 2 > default_limit) {
        map_host_limit = system_ram - default_limit;
      }
      else {
        map_host_limit = system_ram / 2;
      }
    }
    else {
      VLOG(1) << "Mapped host memory disabled, failed to get system RAM";
      map_host_limit = 0;
    }

    /* Amount of device memory to keep is free after texture memory
     * and working memory allocations respectively. We set the working
     * memory limit headroom lower so that some space is left after all
     * texture memory allocations. */
    device_working_headroom = 32 * 1024 * 1024LL;   // 32MB
    device_texture_headroom = 128 * 1024 * 1024LL;  // 128MB

    VLOG(1) << "Mapped host memory limit set to " << string_human_readable_number(map_host_limit)
            << " bytes. (" << string_human_readable_size(map_host_limit) << ")";
  }

  void load_texture_info()
  {
    if (need_texture_info) {
      texture_info.copy_to_device();
      need_texture_info = false;
    }
  }

  void move_textures_to_host(size_t size, bool for_texture)
  {
    /* Signal to reallocate textures in host memory only. */
    move_texture_to_host = true;

    while (size > 0) {
      /* Find suitable memory allocation to move. */
      device_memory *max_mem = NULL;
      size_t max_size = 0;
      bool max_is_image = false;

      foreach (CUDAMemMap::value_type &pair, cuda_mem_map) {
        device_memory &mem = *pair.first;
        CUDAMem *cmem = &pair.second;

        bool is_texture = (mem.type == MEM_TEXTURE) && (&mem != &texture_info);
        bool is_image = is_texture && (mem.data_height > 1);

        /* Can't move this type of memory. */
        if (!is_texture || cmem->array) {
          continue;
        }

        /* Already in host memory. */
        if (cmem->use_mapped_host) {
          continue;
        }

        /* For other textures, only move image textures. */
        if (for_texture && !is_image) {
          continue;
        }

        /* Try to move largest allocation, prefer moving images. */
        if (is_image > max_is_image || (is_image == max_is_image && mem.device_size > max_size)) {
          max_is_image = is_image;
          max_size = mem.device_size;
          max_mem = &mem;
        }
      }

      /* Move to host memory. This part is mutex protected since
       * multiple CUDA devices could be moving the memory. The
       * first one will do it, and the rest will adopt the pointer. */
      if (max_mem) {
        VLOG(1) << "Move memory from device to host: " << max_mem->name;

        static thread_mutex move_mutex;
        thread_scoped_lock lock(move_mutex);

        /* Preserve the original device pointer, in case of multi device
         * we can't change it because the pointer mapping would break. */
        device_ptr prev_pointer = max_mem->device_pointer;
        size_t prev_size = max_mem->device_size;

        tex_free(*max_mem);
        tex_alloc(*max_mem);
        size = (max_size >= size) ? 0 : size - max_size;

        max_mem->device_pointer = prev_pointer;
        max_mem->device_size = prev_size;
      }
      else {
        break;
      }
    }

    /* Update texture info array with new pointers. */
    load_texture_info();

    move_texture_to_host = false;
  }

  CUDAMem *generic_alloc(device_memory &mem, size_t pitch_padding = 0)
  {
    CUDAContextScope scope(this);

    CUdeviceptr device_pointer = 0;
    size_t size = mem.memory_size() + pitch_padding;

    CUresult mem_alloc_result = CUDA_ERROR_OUT_OF_MEMORY;
    const char *status = "";

    /* First try allocating in device memory, respecting headroom. We make
     * an exception for texture info. It is small and frequently accessed,
     * so treat it as working memory.
     *
     * If there is not enough room for working memory, we will try to move
     * textures to host memory, assuming the performance impact would have
     * been worse for working memory. */
    bool is_texture = (mem.type == MEM_TEXTURE) && (&mem != &texture_info);
    bool is_image = is_texture && (mem.data_height > 1);

    size_t headroom = (is_texture) ? device_texture_headroom : device_working_headroom;

    size_t total = 0, free = 0;
    cuMemGetInfo(&free, &total);

    /* Move textures to host memory if needed. */
    if (!move_texture_to_host && !is_image && (size + headroom) >= free && can_map_host) {
      move_textures_to_host(size + headroom - free, is_texture);
      cuMemGetInfo(&free, &total);
    }

    /* Allocate in device memory. */
    if (!move_texture_to_host && (size + headroom) < free) {
      mem_alloc_result = cuMemAlloc(&device_pointer, size);
      if (mem_alloc_result == CUDA_SUCCESS) {
        status = " in device memory";
      }
    }

    /* Fall back to mapped host memory if needed and possible. */

    void *shared_pointer = 0;

    if (mem_alloc_result != CUDA_SUCCESS && can_map_host) {
      if (mem.shared_pointer) {
        /* Another device already allocated host memory. */
        mem_alloc_result = CUDA_SUCCESS;
        shared_pointer = mem.shared_pointer;
      }
      else if (map_host_used + size < map_host_limit) {
        /* Allocate host memory ourselves. */
        mem_alloc_result = cuMemHostAlloc(
            &shared_pointer, size, CU_MEMHOSTALLOC_DEVICEMAP | CU_MEMHOSTALLOC_WRITECOMBINED);

        assert((mem_alloc_result == CUDA_SUCCESS && shared_pointer != 0) ||
               (mem_alloc_result != CUDA_SUCCESS && shared_pointer == 0));
      }

      if (mem_alloc_result == CUDA_SUCCESS) {
        cuda_assert(cuMemHostGetDevicePointer_v2(&device_pointer, shared_pointer, 0));
        map_host_used += size;
        status = " in host memory";
      }
      else {
        status = " failed, out of host memory";
      }
    }

    if (mem_alloc_result != CUDA_SUCCESS) {
      status = " failed, out of device and host memory";
      cuda_assert(mem_alloc_result);
    }

    if (mem.name) {
      VLOG(1) << "Buffer allocate: " << mem.name << ", "
              << string_human_readable_number(mem.memory_size()) << " bytes. ("
              << string_human_readable_size(mem.memory_size()) << ")" << status;
    }

    mem.device_pointer = (device_ptr)device_pointer;
    mem.device_size = size;
    stats.mem_alloc(size);

    if (!mem.device_pointer) {
      return NULL;
    }

    /* Insert into map of allocations. */
    CUDAMem *cmem = &cuda_mem_map[&mem];
    if (shared_pointer != 0) {
      /* Replace host pointer with our host allocation. Only works if
       * CUDA memory layout is the same and has no pitch padding. Also
       * does not work if we move textures to host during a render,
       * since other devices might be using the memory. */

      if (!move_texture_to_host && pitch_padding == 0 && mem.host_pointer &&
          mem.host_pointer != shared_pointer) {
        memcpy(shared_pointer, mem.host_pointer, size);

        /* A Call to device_memory::host_free() should be preceded by
         * a call to device_memory::device_free() for host memory
         * allocated by a device to be handled properly. Two exceptions
         * are here and a call in OptiXDevice::generic_alloc(), where
         * the current host memory can be assumed to be allocated by
         * device_memory::host_alloc(), not by a device */

        mem.host_free();
        mem.host_pointer = shared_pointer;
      }
      mem.shared_pointer = shared_pointer;
      mem.shared_counter++;
      cmem->use_mapped_host = true;
    }
    else {
      cmem->use_mapped_host = false;
    }

    return cmem;
  }

  void generic_copy_to(device_memory &mem)
  {
    if (mem.host_pointer && mem.device_pointer) {
      CUDAContextScope scope(this);

      /* If use_mapped_host of mem is false, the current device only
       * uses device memory allocated by cuMemAlloc regardless of
       * mem.host_pointer and mem.shared_pointer, and should copy
       * data from mem.host_pointer. */

      if (cuda_mem_map[&mem].use_mapped_host == false || mem.host_pointer != mem.shared_pointer) {
        cuda_assert(cuMemcpyHtoD(
            cuda_device_ptr(mem.device_pointer), mem.host_pointer, mem.memory_size()));
      }
    }
  }

  void generic_free(device_memory &mem)
  {
    if (mem.device_pointer) {
      CUDAContextScope scope(this);
      const CUDAMem &cmem = cuda_mem_map[&mem];

      /* If cmem.use_mapped_host is true, reference counting is used
       * to safely free a mapped host memory. */

      if (cmem.use_mapped_host) {
        assert(mem.shared_pointer);
        if (mem.shared_pointer) {
          assert(mem.shared_counter > 0);
          if (--mem.shared_counter == 0) {
            if (mem.host_pointer == mem.shared_pointer) {
              mem.host_pointer = 0;
            }
            cuMemFreeHost(mem.shared_pointer);
            mem.shared_pointer = 0;
          }
        }
        map_host_used -= mem.device_size;
      }
      else {
        /* Free device memory. */
        cuMemFree(mem.device_pointer);
      }

      stats.mem_free(mem.device_size);
      mem.device_pointer = 0;
      mem.device_size = 0;

      cuda_mem_map.erase(cuda_mem_map.find(&mem));
    }
  }

  void mem_alloc(device_memory &mem)
  {
    if (mem.type == MEM_PIXELS && !background) {
      pixels_alloc(mem);
    }
    else if (mem.type == MEM_TEXTURE) {
      assert(!"mem_alloc not supported for textures.");
    }
    else {
      generic_alloc(mem);
    }
  }

  void mem_copy_to(device_memory &mem)
  {
    if (mem.type == MEM_PIXELS) {
      assert(!"mem_copy_to not supported for pixels.");
    }
    else if (mem.type == MEM_TEXTURE) {
      tex_free(mem);
      tex_alloc(mem);
    }
    else {
      if (!mem.device_pointer) {
        generic_alloc(mem);
      }

      generic_copy_to(mem);
    }
  }

  void mem_copy_from(device_memory &mem, int y, int w, int h, int elem)
  {
    if (mem.type == MEM_PIXELS && !background) {
      pixels_copy_from(mem, y, w, h);
    }
    else if (mem.type == MEM_TEXTURE) {
      assert(!"mem_copy_from not supported for textures.");
    }
    else {
      CUDAContextScope scope(this);
      size_t offset = elem * y * w;
      size_t size = elem * w * h;

      if (mem.host_pointer && mem.device_pointer) {
        cuda_assert(cuMemcpyDtoH(
            (uchar *)mem.host_pointer + offset, (CUdeviceptr)(mem.device_pointer + offset), size));
      }
      else if (mem.host_pointer) {
        memset((char *)mem.host_pointer + offset, 0, size);
      }
    }
  }

  void mem_zero(device_memory &mem)
  {
    if (!mem.device_pointer) {
      mem_alloc(mem);
    }

    if (mem.host_pointer) {
      memset(mem.host_pointer, 0, mem.memory_size());
    }

    /* If use_mapped_host of mem is false, mem.device_pointer currently
     * refers to device memory regardless of mem.host_pointer and
     * mem.shared_pointer. */

    if (mem.device_pointer &&
        (cuda_mem_map[&mem].use_mapped_host == false || mem.host_pointer != mem.shared_pointer)) {
      CUDAContextScope scope(this);
      cuda_assert(cuMemsetD8(cuda_device_ptr(mem.device_pointer), 0, mem.memory_size()));
    }
  }

  void mem_free(device_memory &mem)
  {
    if (mem.type == MEM_PIXELS && !background) {
      pixels_free(mem);
    }
    else if (mem.type == MEM_TEXTURE) {
      tex_free(mem);
    }
    else {
      generic_free(mem);
    }
  }

  virtual device_ptr mem_alloc_sub_ptr(device_memory &mem, int offset, int /*size*/)
  {
    return (device_ptr)(((char *)mem.device_pointer) + mem.memory_elements_size(offset));
  }

  void const_copy_to(const char *name, void *host, size_t size)
  {
    CUDAContextScope scope(this);
    CUdeviceptr mem;
    size_t bytes;

    cuda_assert(cuModuleGetGlobal(&mem, &bytes, cuModule, name));
    // assert(bytes == size);
    cuda_assert(cuMemcpyHtoD(mem, host, size));
  }

  void tex_alloc(device_memory &mem)
  {
    CUDAContextScope scope(this);

    /* General variables for both architectures */
    string bind_name = mem.name;
    size_t dsize = datatype_size(mem.data_type);
    size_t size = mem.memory_size();

    CUaddress_mode address_mode = CU_TR_ADDRESS_MODE_WRAP;
    switch (mem.extension) {
      case EXTENSION_REPEAT:
        address_mode = CU_TR_ADDRESS_MODE_WRAP;
        break;
      case EXTENSION_EXTEND:
        address_mode = CU_TR_ADDRESS_MODE_CLAMP;
        break;
      case EXTENSION_CLIP:
        address_mode = CU_TR_ADDRESS_MODE_BORDER;
        break;
      default:
        assert(0);
        break;
    }

    CUfilter_mode filter_mode;
    if (mem.interpolation == INTERPOLATION_CLOSEST) {
      filter_mode = CU_TR_FILTER_MODE_POINT;
    }
    else {
      filter_mode = CU_TR_FILTER_MODE_LINEAR;
    }

    /* Data Storage */
    if (mem.interpolation == INTERPOLATION_NONE) {
      generic_alloc(mem);
      generic_copy_to(mem);

      CUdeviceptr cumem;
      size_t cubytes;

      cuda_assert(cuModuleGetGlobal(&cumem, &cubytes, cuModule, bind_name.c_str()));

      if (cubytes == 8) {
        /* 64 bit device pointer */
        uint64_t ptr = mem.device_pointer;
        cuda_assert(cuMemcpyHtoD(cumem, (void *)&ptr, cubytes));
      }
      else {
        /* 32 bit device pointer */
        uint32_t ptr = (uint32_t)mem.device_pointer;
        cuda_assert(cuMemcpyHtoD(cumem, (void *)&ptr, cubytes));
      }
      return;
    }

    /* Image Texture Storage */
    CUarray_format_enum format;
    switch (mem.data_type) {
      case TYPE_UCHAR:
        format = CU_AD_FORMAT_UNSIGNED_INT8;
        break;
      case TYPE_UINT16:
        format = CU_AD_FORMAT_UNSIGNED_INT16;
        break;
      case TYPE_UINT:
        format = CU_AD_FORMAT_UNSIGNED_INT32;
        break;
      case TYPE_INT:
        format = CU_AD_FORMAT_SIGNED_INT32;
        break;
      case TYPE_FLOAT:
        format = CU_AD_FORMAT_FLOAT;
        break;
      case TYPE_HALF:
        format = CU_AD_FORMAT_HALF;
        break;
      default:
        assert(0);
        return;
    }

    CUDAMem *cmem = NULL;
    CUarray array_3d = NULL;
    size_t src_pitch = mem.data_width * dsize * mem.data_elements;
    size_t dst_pitch = src_pitch;

    if (mem.data_depth > 1) {
      /* 3D texture using array, there is no API for linear memory. */
      CUDA_ARRAY3D_DESCRIPTOR desc;

      desc.Width = mem.data_width;
      desc.Height = mem.data_height;
      desc.Depth = mem.data_depth;
      desc.Format = format;
      desc.NumChannels = mem.data_elements;
      desc.Flags = 0;

      VLOG(1) << "Array 3D allocate: " << mem.name << ", "
              << string_human_readable_number(mem.memory_size()) << " bytes. ("
              << string_human_readable_size(mem.memory_size()) << ")";

      cuda_assert(cuArray3DCreate(&array_3d, &desc));

      if (!array_3d) {
        return;
      }

      CUDA_MEMCPY3D param;
      memset(&param, 0, sizeof(param));
      param.dstMemoryType = CU_MEMORYTYPE_ARRAY;
      param.dstArray = array_3d;
      param.srcMemoryType = CU_MEMORYTYPE_HOST;
      param.srcHost = mem.host_pointer;
      param.srcPitch = src_pitch;
      param.WidthInBytes = param.srcPitch;
      param.Height = mem.data_height;
      param.Depth = mem.data_depth;

      cuda_assert(cuMemcpy3D(&param));

      mem.device_pointer = (device_ptr)array_3d;
      mem.device_size = size;
      stats.mem_alloc(size);

      cmem = &cuda_mem_map[&mem];
      cmem->texobject = 0;
      cmem->array = array_3d;
    }
    else if (mem.data_height > 0) {
      /* 2D texture, using pitch aligned linear memory. */
      int alignment = 0;
      cuda_assert(
          cuDeviceGetAttribute(&alignment, CU_DEVICE_ATTRIBUTE_TEXTURE_PITCH_ALIGNMENT, cuDevice));
      dst_pitch = align_up(src_pitch, alignment);
      size_t dst_size = dst_pitch * mem.data_height;

      cmem = generic_alloc(mem, dst_size - mem.memory_size());
      if (!cmem) {
        return;
      }

      CUDA_MEMCPY2D param;
      memset(&param, 0, sizeof(param));
      param.dstMemoryType = CU_MEMORYTYPE_DEVICE;
      param.dstDevice = mem.device_pointer;
      param.dstPitch = dst_pitch;
      param.srcMemoryType = CU_MEMORYTYPE_HOST;
      param.srcHost = mem.host_pointer;
      param.srcPitch = src_pitch;
      param.WidthInBytes = param.srcPitch;
      param.Height = mem.data_height;

      cuda_assert(cuMemcpy2DUnaligned(&param));
    }
    else {
      /* 1D texture, using linear memory. */
      cmem = generic_alloc(mem);
      if (!cmem) {
        return;
      }

      cuda_assert(cuMemcpyHtoD(mem.device_pointer, mem.host_pointer, size));
    }

    /* Kepler+, bindless textures. */
    int flat_slot = 0;
    if (string_startswith(mem.name, "__tex_image")) {
      int pos = string(mem.name).rfind("_");
      flat_slot = atoi(mem.name + pos + 1);
    }
    else {
      assert(0);
    }

    CUDA_RESOURCE_DESC resDesc;
    memset(&resDesc, 0, sizeof(resDesc));

    if (array_3d) {
      resDesc.resType = CU_RESOURCE_TYPE_ARRAY;
      resDesc.res.array.hArray = array_3d;
      resDesc.flags = 0;
    }
    else if (mem.data_height > 0) {
      resDesc.resType = CU_RESOURCE_TYPE_PITCH2D;
      resDesc.res.pitch2D.devPtr = mem.device_pointer;
      resDesc.res.pitch2D.format = format;
      resDesc.res.pitch2D.numChannels = mem.data_elements;
      resDesc.res.pitch2D.height = mem.data_height;
      resDesc.res.pitch2D.width = mem.data_width;
      resDesc.res.pitch2D.pitchInBytes = dst_pitch;
    }
    else {
      resDesc.resType = CU_RESOURCE_TYPE_LINEAR;
      resDesc.res.linear.devPtr = mem.device_pointer;
      resDesc.res.linear.format = format;
      resDesc.res.linear.numChannels = mem.data_elements;
      resDesc.res.linear.sizeInBytes = mem.device_size;
    }

    CUDA_TEXTURE_DESC texDesc;
    memset(&texDesc, 0, sizeof(texDesc));
    texDesc.addressMode[0] = address_mode;
    texDesc.addressMode[1] = address_mode;
    texDesc.addressMode[2] = address_mode;
    texDesc.filterMode = filter_mode;
    texDesc.flags = CU_TRSF_NORMALIZED_COORDINATES;

    cuda_assert(cuTexObjectCreate(&cmem->texobject, &resDesc, &texDesc, NULL));

    /* Resize once */
    if (flat_slot >= texture_info.size()) {
      /* Allocate some slots in advance, to reduce amount
       * of re-allocations. */
      texture_info.resize(flat_slot + 128);
    }

    /* Set Mapping and tag that we need to (re-)upload to device */
    TextureInfo &info = texture_info[flat_slot];
    info.data = (uint64_t)cmem->texobject;
    info.cl_buffer = 0;
    info.interpolation = mem.interpolation;
    info.extension = mem.extension;
    info.width = mem.data_width;
    info.height = mem.data_height;
    info.depth = mem.data_depth;
    need_texture_info = true;
  }

  void tex_free(device_memory &mem)
  {
    if (mem.device_pointer) {
      CUDAContextScope scope(this);
      const CUDAMem &cmem = cuda_mem_map[&mem];

      if (cmem.texobject) {
        /* Free bindless texture. */
        cuTexObjectDestroy(cmem.texobject);
      }

      if (cmem.array) {
        /* Free array. */
        cuArrayDestroy(cmem.array);
        stats.mem_free(mem.device_size);
        mem.device_pointer = 0;
        mem.device_size = 0;

        cuda_mem_map.erase(cuda_mem_map.find(&mem));
      }
      else {
        generic_free(mem);
      }
    }
  }

#define CUDA_GET_BLOCKSIZE(func, w, h) \
  int threads_per_block; \
  cuda_assert( \
      cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, func)); \
  int threads = (int)sqrt((float)threads_per_block); \
  int xblocks = ((w) + threads - 1) / threads; \
  int yblocks = ((h) + threads - 1) / threads;

#define CUDA_LAUNCH_KERNEL(func, args) \
  cuda_assert(cuLaunchKernel(func, xblocks, yblocks, 1, threads, threads, 1, 0, 0, args, 0));

/* Similar as above, but for 1-dimensional blocks. */
#define CUDA_GET_BLOCKSIZE_1D(func, w, h) \
  int threads_per_block; \
  cuda_assert( \
      cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, func)); \
  int xblocks = ((w) + threads_per_block - 1) / threads_per_block; \
  int yblocks = h;

#define CUDA_LAUNCH_KERNEL_1D(func, args) \
  cuda_assert(cuLaunchKernel(func, xblocks, yblocks, 1, threads_per_block, 1, 1, 0, 0, args, 0));

  bool denoising_non_local_means(device_ptr image_ptr,
                                 device_ptr guide_ptr,
                                 device_ptr variance_ptr,
                                 device_ptr out_ptr,
                                 DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    int stride = task->buffer.stride;
    int w = task->buffer.width;
    int h = task->buffer.h;
    int r = task->nlm_state.r;
    int f = task->nlm_state.f;
    float a = task->nlm_state.a;
    float k_2 = task->nlm_state.k_2;

    int pass_stride = task->buffer.pass_stride;
    int num_shifts = (2 * r + 1) * (2 * r + 1);
    int channel_offset = task->nlm_state.is_color ? task->buffer.pass_stride : 0;
    int frame_offset = 0;

    if (have_error())
      return false;

    CUdeviceptr difference = cuda_device_ptr(task->buffer.temporary_mem.device_pointer);
    CUdeviceptr blurDifference = difference + sizeof(float) * pass_stride * num_shifts;
    CUdeviceptr weightAccum = difference + 2 * sizeof(float) * pass_stride * num_shifts;
    CUdeviceptr scale_ptr = 0;

    cuda_assert(cuMemsetD8(weightAccum, 0, sizeof(float) * pass_stride));
    cuda_assert(cuMemsetD8(out_ptr, 0, sizeof(float) * pass_stride));

    {
      CUfunction cuNLMCalcDifference, cuNLMBlur, cuNLMCalcWeight, cuNLMUpdateOutput;
      cuda_assert(cuModuleGetFunction(
          &cuNLMCalcDifference, cuFilterModule, "kernel_cuda_filter_nlm_calc_difference"));
      cuda_assert(cuModuleGetFunction(&cuNLMBlur, cuFilterModule, "kernel_cuda_filter_nlm_blur"));
      cuda_assert(cuModuleGetFunction(
          &cuNLMCalcWeight, cuFilterModule, "kernel_cuda_filter_nlm_calc_weight"));
      cuda_assert(cuModuleGetFunction(
          &cuNLMUpdateOutput, cuFilterModule, "kernel_cuda_filter_nlm_update_output"));

      cuda_assert(cuFuncSetCacheConfig(cuNLMCalcDifference, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(cuFuncSetCacheConfig(cuNLMBlur, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(cuFuncSetCacheConfig(cuNLMCalcWeight, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(cuFuncSetCacheConfig(cuNLMUpdateOutput, CU_FUNC_CACHE_PREFER_L1));

      CUDA_GET_BLOCKSIZE_1D(cuNLMCalcDifference, w * h, num_shifts);

      void *calc_difference_args[] = {&guide_ptr,
                                      &variance_ptr,
                                      &scale_ptr,
                                      &difference,
                                      &w,
                                      &h,
                                      &stride,
                                      &pass_stride,
                                      &r,
                                      &channel_offset,
                                      &frame_offset,
                                      &a,
                                      &k_2};
      void *blur_args[] = {&difference, &blurDifference, &w, &h, &stride, &pass_stride, &r, &f};
      void *calc_weight_args[] = {
          &blurDifference, &difference, &w, &h, &stride, &pass_stride, &r, &f};
      void *update_output_args[] = {&blurDifference,
                                    &image_ptr,
                                    &out_ptr,
                                    &weightAccum,
                                    &w,
                                    &h,
                                    &stride,
                                    &pass_stride,
                                    &channel_offset,
                                    &r,
                                    &f};

      CUDA_LAUNCH_KERNEL_1D(cuNLMCalcDifference, calc_difference_args);
      CUDA_LAUNCH_KERNEL_1D(cuNLMBlur, blur_args);
      CUDA_LAUNCH_KERNEL_1D(cuNLMCalcWeight, calc_weight_args);
      CUDA_LAUNCH_KERNEL_1D(cuNLMBlur, blur_args);
      CUDA_LAUNCH_KERNEL_1D(cuNLMUpdateOutput, update_output_args);
    }

    {
      CUfunction cuNLMNormalize;
      cuda_assert(cuModuleGetFunction(
          &cuNLMNormalize, cuFilterModule, "kernel_cuda_filter_nlm_normalize"));
      cuda_assert(cuFuncSetCacheConfig(cuNLMNormalize, CU_FUNC_CACHE_PREFER_L1));
      void *normalize_args[] = {&out_ptr, &weightAccum, &w, &h, &stride};
      CUDA_GET_BLOCKSIZE(cuNLMNormalize, w, h);
      CUDA_LAUNCH_KERNEL(cuNLMNormalize, normalize_args);
      cuda_assert(cuCtxSynchronize());
    }

    return !have_error();
  }

  bool denoising_construct_transform(DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterConstructTransform;
    cuda_assert(cuModuleGetFunction(
        &cuFilterConstructTransform, cuFilterModule, "kernel_cuda_filter_construct_transform"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterConstructTransform, CU_FUNC_CACHE_PREFER_SHARED));
    CUDA_GET_BLOCKSIZE(cuFilterConstructTransform, task->storage.w, task->storage.h);

    void *args[] = {&task->buffer.mem.device_pointer,
                    &task->tile_info_mem.device_pointer,
                    &task->storage.transform.device_pointer,
                    &task->storage.rank.device_pointer,
                    &task->filter_area,
                    &task->rect,
                    &task->radius,
                    &task->pca_threshold,
                    &task->buffer.pass_stride,
                    &task->buffer.frame_stride,
                    &task->buffer.use_time};
    CUDA_LAUNCH_KERNEL(cuFilterConstructTransform, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_accumulate(device_ptr color_ptr,
                            device_ptr color_variance_ptr,
                            device_ptr scale_ptr,
                            int frame,
                            DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    int r = task->radius;
    int f = 4;
    float a = 1.0f;
    float k_2 = task->nlm_k_2;

    int w = task->reconstruction_state.source_w;
    int h = task->reconstruction_state.source_h;
    int stride = task->buffer.stride;
    int frame_offset = frame * task->buffer.frame_stride;
    int t = task->tile_info->frames[frame];

    int pass_stride = task->buffer.pass_stride;
    int num_shifts = (2 * r + 1) * (2 * r + 1);

    if (have_error())
      return false;

    CUdeviceptr difference = cuda_device_ptr(task->buffer.temporary_mem.device_pointer);
    CUdeviceptr blurDifference = difference + sizeof(float) * pass_stride * num_shifts;

    CUfunction cuNLMCalcDifference, cuNLMBlur, cuNLMCalcWeight, cuNLMConstructGramian;
    cuda_assert(cuModuleGetFunction(
        &cuNLMCalcDifference, cuFilterModule, "kernel_cuda_filter_nlm_calc_difference"));
    cuda_assert(cuModuleGetFunction(&cuNLMBlur, cuFilterModule, "kernel_cuda_filter_nlm_blur"));
    cuda_assert(cuModuleGetFunction(
        &cuNLMCalcWeight, cuFilterModule, "kernel_cuda_filter_nlm_calc_weight"));
    cuda_assert(cuModuleGetFunction(
        &cuNLMConstructGramian, cuFilterModule, "kernel_cuda_filter_nlm_construct_gramian"));

    cuda_assert(cuFuncSetCacheConfig(cuNLMCalcDifference, CU_FUNC_CACHE_PREFER_L1));
    cuda_assert(cuFuncSetCacheConfig(cuNLMBlur, CU_FUNC_CACHE_PREFER_L1));
    cuda_assert(cuFuncSetCacheConfig(cuNLMCalcWeight, CU_FUNC_CACHE_PREFER_L1));
    cuda_assert(cuFuncSetCacheConfig(cuNLMConstructGramian, CU_FUNC_CACHE_PREFER_SHARED));

    CUDA_GET_BLOCKSIZE_1D(cuNLMCalcDifference,
                          task->reconstruction_state.source_w *
                              task->reconstruction_state.source_h,
                          num_shifts);

    void *calc_difference_args[] = {&color_ptr,
                                    &color_variance_ptr,
                                    &scale_ptr,
                                    &difference,
                                    &w,
                                    &h,
                                    &stride,
                                    &pass_stride,
                                    &r,
                                    &pass_stride,
                                    &frame_offset,
                                    &a,
                                    &k_2};
    void *blur_args[] = {&difference, &blurDifference, &w, &h, &stride, &pass_stride, &r, &f};
    void *calc_weight_args[] = {
        &blurDifference, &difference, &w, &h, &stride, &pass_stride, &r, &f};
    void *construct_gramian_args[] = {&t,
                                      &blurDifference,
                                      &task->buffer.mem.device_pointer,
                                      &task->storage.transform.device_pointer,
                                      &task->storage.rank.device_pointer,
                                      &task->storage.XtWX.device_pointer,
                                      &task->storage.XtWY.device_pointer,
                                      &task->reconstruction_state.filter_window,
                                      &w,
                                      &h,
                                      &stride,
                                      &pass_stride,
                                      &r,
                                      &f,
                                      &frame_offset,
                                      &task->buffer.use_time};

    CUDA_LAUNCH_KERNEL_1D(cuNLMCalcDifference, calc_difference_args);
    CUDA_LAUNCH_KERNEL_1D(cuNLMBlur, blur_args);
    CUDA_LAUNCH_KERNEL_1D(cuNLMCalcWeight, calc_weight_args);
    CUDA_LAUNCH_KERNEL_1D(cuNLMBlur, blur_args);
    CUDA_LAUNCH_KERNEL_1D(cuNLMConstructGramian, construct_gramian_args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_solve(device_ptr output_ptr, DenoisingTask *task)
  {
    CUfunction cuFinalize;
    cuda_assert(cuModuleGetFunction(&cuFinalize, cuFilterModule, "kernel_cuda_filter_finalize"));
    cuda_assert(cuFuncSetCacheConfig(cuFinalize, CU_FUNC_CACHE_PREFER_L1));
    void *finalize_args[] = {&output_ptr,
                             &task->storage.rank.device_pointer,
                             &task->storage.XtWX.device_pointer,
                             &task->storage.XtWY.device_pointer,
                             &task->filter_area,
                             &task->reconstruction_state.buffer_params.x,
                             &task->render_buffer.samples};
    CUDA_GET_BLOCKSIZE(
        cuFinalize, task->reconstruction_state.source_w, task->reconstruction_state.source_h);
    CUDA_LAUNCH_KERNEL(cuFinalize, finalize_args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_combine_halves(device_ptr a_ptr,
                                device_ptr b_ptr,
                                device_ptr mean_ptr,
                                device_ptr variance_ptr,
                                int r,
                                int4 rect,
                                DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterCombineHalves;
    cuda_assert(cuModuleGetFunction(
        &cuFilterCombineHalves, cuFilterModule, "kernel_cuda_filter_combine_halves"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterCombineHalves, CU_FUNC_CACHE_PREFER_L1));
    CUDA_GET_BLOCKSIZE(
        cuFilterCombineHalves, task->rect.z - task->rect.x, task->rect.w - task->rect.y);

    void *args[] = {&mean_ptr, &variance_ptr, &a_ptr, &b_ptr, &rect, &r};
    CUDA_LAUNCH_KERNEL(cuFilterCombineHalves, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_divide_shadow(device_ptr a_ptr,
                               device_ptr b_ptr,
                               device_ptr sample_variance_ptr,
                               device_ptr sv_variance_ptr,
                               device_ptr buffer_variance_ptr,
                               DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterDivideShadow;
    cuda_assert(cuModuleGetFunction(
        &cuFilterDivideShadow, cuFilterModule, "kernel_cuda_filter_divide_shadow"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterDivideShadow, CU_FUNC_CACHE_PREFER_L1));
    CUDA_GET_BLOCKSIZE(
        cuFilterDivideShadow, task->rect.z - task->rect.x, task->rect.w - task->rect.y);

    void *args[] = {&task->render_buffer.samples,
                    &task->tile_info_mem.device_pointer,
                    &a_ptr,
                    &b_ptr,
                    &sample_variance_ptr,
                    &sv_variance_ptr,
                    &buffer_variance_ptr,
                    &task->rect,
                    &task->render_buffer.pass_stride,
                    &task->render_buffer.offset};
    CUDA_LAUNCH_KERNEL(cuFilterDivideShadow, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_get_feature(int mean_offset,
                             int variance_offset,
                             device_ptr mean_ptr,
                             device_ptr variance_ptr,
                             float scale,
                             DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterGetFeature;
    cuda_assert(cuModuleGetFunction(
        &cuFilterGetFeature, cuFilterModule, "kernel_cuda_filter_get_feature"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterGetFeature, CU_FUNC_CACHE_PREFER_L1));
    CUDA_GET_BLOCKSIZE(
        cuFilterGetFeature, task->rect.z - task->rect.x, task->rect.w - task->rect.y);

    void *args[] = {&task->render_buffer.samples,
                    &task->tile_info_mem.device_pointer,
                    &mean_offset,
                    &variance_offset,
                    &mean_ptr,
                    &variance_ptr,
                    &scale,
                    &task->rect,
                    &task->render_buffer.pass_stride,
                    &task->render_buffer.offset};
    CUDA_LAUNCH_KERNEL(cuFilterGetFeature, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_write_feature(int out_offset,
                               device_ptr from_ptr,
                               device_ptr buffer_ptr,
                               DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterWriteFeature;
    cuda_assert(cuModuleGetFunction(
        &cuFilterWriteFeature, cuFilterModule, "kernel_cuda_filter_write_feature"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterWriteFeature, CU_FUNC_CACHE_PREFER_L1));
    CUDA_GET_BLOCKSIZE(cuFilterWriteFeature, task->filter_area.z, task->filter_area.w);

    void *args[] = {&task->render_buffer.samples,
                    &task->reconstruction_state.buffer_params,
                    &task->filter_area,
                    &from_ptr,
                    &buffer_ptr,
                    &out_offset,
                    &task->rect};
    CUDA_LAUNCH_KERNEL(cuFilterWriteFeature, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  bool denoising_detect_outliers(device_ptr image_ptr,
                                 device_ptr variance_ptr,
                                 device_ptr depth_ptr,
                                 device_ptr output_ptr,
                                 DenoisingTask *task)
  {
    if (have_error())
      return false;

    CUDAContextScope scope(this);

    CUfunction cuFilterDetectOutliers;
    cuda_assert(cuModuleGetFunction(
        &cuFilterDetectOutliers, cuFilterModule, "kernel_cuda_filter_detect_outliers"));
    cuda_assert(cuFuncSetCacheConfig(cuFilterDetectOutliers, CU_FUNC_CACHE_PREFER_L1));
    CUDA_GET_BLOCKSIZE(
        cuFilterDetectOutliers, task->rect.z - task->rect.x, task->rect.w - task->rect.y);

    void *args[] = {&image_ptr,
                    &variance_ptr,
                    &depth_ptr,
                    &output_ptr,
                    &task->rect,
                    &task->buffer.pass_stride};

    CUDA_LAUNCH_KERNEL(cuFilterDetectOutliers, args);
    cuda_assert(cuCtxSynchronize());

    return !have_error();
  }

  void denoise(RenderTile &rtile, DenoisingTask &denoising)
  {
    denoising.functions.construct_transform = function_bind(
        &CUDADevice::denoising_construct_transform, this, &denoising);
    denoising.functions.accumulate = function_bind(
        &CUDADevice::denoising_accumulate, this, _1, _2, _3, _4, &denoising);
    denoising.functions.solve = function_bind(&CUDADevice::denoising_solve, this, _1, &denoising);
    denoising.functions.divide_shadow = function_bind(
        &CUDADevice::denoising_divide_shadow, this, _1, _2, _3, _4, _5, &denoising);
    denoising.functions.non_local_means = function_bind(
        &CUDADevice::denoising_non_local_means, this, _1, _2, _3, _4, &denoising);
    denoising.functions.combine_halves = function_bind(
        &CUDADevice::denoising_combine_halves, this, _1, _2, _3, _4, _5, _6, &denoising);
    denoising.functions.get_feature = function_bind(
        &CUDADevice::denoising_get_feature, this, _1, _2, _3, _4, _5, &denoising);
    denoising.functions.write_feature = function_bind(
        &CUDADevice::denoising_write_feature, this, _1, _2, _3, &denoising);
    denoising.functions.detect_outliers = function_bind(
        &CUDADevice::denoising_detect_outliers, this, _1, _2, _3, _4, &denoising);

    denoising.filter_area = make_int4(rtile.x, rtile.y, rtile.w, rtile.h);
    denoising.render_buffer.samples = rtile.sample;
    denoising.buffer.gpu_temporary_mem = true;

    denoising.run_denoising(&rtile);
  }

  void path_trace(DeviceTask &task, RenderTile &rtile, device_vector<WorkTile> &work_tiles)
  {
    scoped_timer timer(&rtile.buffers->render_time);

    if (have_error())
      return;

    CUDAContextScope scope(this);
    CUfunction cuPathTrace;

    /* Get kernel function. */
    if (task.integrator_branched) {
      cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_branched_path_trace"));
    }
    else {
      cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_path_trace"));
    }

    if (have_error()) {
      return;
    }

    cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1));

    /* Kernels for adaptive sampling. */
    CUfunction cuAdaptiveStopping, cuAdaptiveFilterX, cuAdaptiveFilterY, cuAdaptiveScaleSamples;
    if (task.integrator_adaptive) {
      cuda_assert(
          cuModuleGetFunction(&cuAdaptiveStopping, cuModule, "kernel_cuda_adaptive_stopping"));
      cuda_assert(cuFuncSetCacheConfig(cuAdaptiveStopping, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(
          cuModuleGetFunction(&cuAdaptiveFilterX, cuModule, "kernel_cuda_adaptive_filter_x"));
      cuda_assert(cuFuncSetCacheConfig(cuAdaptiveFilterX, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(
          cuModuleGetFunction(&cuAdaptiveFilterY, cuModule, "kernel_cuda_adaptive_filter_y"));
      cuda_assert(cuFuncSetCacheConfig(cuAdaptiveFilterY, CU_FUNC_CACHE_PREFER_L1));
      cuda_assert(cuModuleGetFunction(
          &cuAdaptiveScaleSamples, cuModule, "kernel_cuda_adaptive_scale_samples"));
      cuda_assert(cuFuncSetCacheConfig(cuAdaptiveScaleSamples, CU_FUNC_CACHE_PREFER_L1));
    }

    /* Allocate work tile. */
    work_tiles.alloc(1);

    WorkTile *wtile = work_tiles.data();
    wtile->x = rtile.x;
    wtile->y = rtile.y;
    wtile->w = rtile.w;
    wtile->h = rtile.h;
    wtile->offset = rtile.offset;
    wtile->stride = rtile.stride;
    wtile->buffer = (float *)cuda_device_ptr(rtile.buffer);

    /* Prepare work size. More step samples render faster, but for now we
     * remain conservative for GPUs connected to a display to avoid driver
     * timeouts and display freezing. */
    int min_blocks, num_threads_per_block;
    cuda_assert(cuOccupancyMaxPotentialBlockSize(
        &min_blocks, &num_threads_per_block, cuPathTrace, NULL, 0, 0));
    if (!info.display_device) {
      min_blocks *= 8;
    }

    uint step_samples = divide_up(min_blocks * num_threads_per_block, wtile->w * wtile->h);

    if (task.integrator_adaptive) {
      /* Force to either 1, 2 or multiple of 4 samples per kernel invocation. */
      if (step_samples == 3) {
        step_samples = 2;
      }
      else if (step_samples > 4) {
        step_samples &= 0xfffffffc;
      }
    }

    /* Render all samples. */
    int start_sample = rtile.start_sample;
    int end_sample = rtile.start_sample + rtile.num_samples;

    for (int sample = start_sample; sample < end_sample; sample += step_samples) {
      /* Setup and copy work tile to device. */
      wtile->start_sample = sample;
      wtile->num_samples = min(step_samples, end_sample - sample);
      work_tiles.copy_to_device();

      CUdeviceptr d_work_tiles = cuda_device_ptr(work_tiles.device_pointer);
      uint total_work_size = wtile->w * wtile->h * wtile->num_samples;
      uint num_blocks = divide_up(total_work_size, num_threads_per_block);

      /* Launch kernel. */
      void *args[] = {&d_work_tiles, &total_work_size};

      cuda_assert(cuLaunchKernel(
          cuPathTrace, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));

      uint filter_sample = sample + wtile->num_samples - 1;
      /* Run the adaptive sampling kernels when we're at a multiple of 4 samples.
       * These are a series of tiny kernels because there is no grid synchronisation
       * from within a kernel, so multiple kernel launches it is. */
      if (task.integrator_adaptive && (filter_sample & 0x3) == 3) {
        total_work_size = wtile->h * wtile->w;
        void *args2[] = {&d_work_tiles, &filter_sample, &total_work_size};
        num_blocks = divide_up(total_work_size, num_threads_per_block);
        cuda_assert(cuLaunchKernel(
            cuAdaptiveStopping, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args2, 0));
        total_work_size = wtile->h;
        num_blocks = divide_up(total_work_size, num_threads_per_block);
        cuda_assert(cuLaunchKernel(
            cuAdaptiveFilterX, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args2, 0));
        total_work_size = wtile->w;
        num_blocks = divide_up(total_work_size, num_threads_per_block);
        cuda_assert(cuLaunchKernel(
            cuAdaptiveFilterY, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args2, 0));
      }

      cuda_assert(cuCtxSynchronize());

      /* Update progress. */
      rtile.sample = sample + wtile->num_samples;
      task.update_progress(&rtile, rtile.w * rtile.h * wtile->num_samples);

      if (task.get_cancel()) {
        if (task.need_finish_queue == false)
          break;
      }
    }

    if (task.integrator_adaptive) {
      CUdeviceptr d_work_tiles = cuda_device_ptr(work_tiles.device_pointer);
      uint total_work_size = wtile->h * wtile->w;
      void *args[] = {&d_work_tiles, &rtile.start_sample, &rtile.sample, &total_work_size};
      uint num_blocks = divide_up(total_work_size, num_threads_per_block);
      cuda_assert(cuLaunchKernel(
          cuAdaptiveScaleSamples, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));
      cuda_assert(cuCtxSynchronize());
      task.update_progress(&rtile, rtile.w * rtile.h * wtile->num_samples);
    }
  }

  void film_convert(DeviceTask &task,
                    device_ptr buffer,
                    device_ptr rgba_byte,
                    device_ptr rgba_half)
  {
    if (have_error())
      return;

    CUDAContextScope scope(this);

    CUfunction cuFilmConvert;
    CUdeviceptr d_rgba = map_pixels((rgba_byte) ? rgba_byte : rgba_half);
    CUdeviceptr d_buffer = cuda_device_ptr(buffer);

    /* get kernel function */
    if (rgba_half) {
      cuda_assert(
          cuModuleGetFunction(&cuFilmConvert, cuModule, "kernel_cuda_convert_to_half_float"));
    }
    else {
      cuda_assert(cuModuleGetFunction(&cuFilmConvert, cuModule, "kernel_cuda_convert_to_byte"));
    }

    float sample_scale = 1.0f / (task.sample + 1);

    /* pass in parameters */
    void *args[] = {&d_rgba,
                    &d_buffer,
                    &sample_scale,
                    &task.x,
                    &task.y,
                    &task.w,
                    &task.h,
                    &task.offset,
                    &task.stride};

    /* launch kernel */
    int threads_per_block;
    cuda_assert(cuFuncGetAttribute(
        &threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, cuFilmConvert));

    int xthreads = (int)sqrt(threads_per_block);
    int ythreads = (int)sqrt(threads_per_block);
    int xblocks = (task.w + xthreads - 1) / xthreads;
    int yblocks = (task.h + ythreads - 1) / ythreads;

    cuda_assert(cuFuncSetCacheConfig(cuFilmConvert, CU_FUNC_CACHE_PREFER_L1));

    cuda_assert(cuLaunchKernel(cuFilmConvert,
                               xblocks,
                               yblocks,
                               1, /* blocks */
                               xthreads,
                               ythreads,
                               1, /* threads */
                               0,
                               0,
                               args,
                               0));

    unmap_pixels((rgba_byte) ? rgba_byte : rgba_half);

    cuda_assert(cuCtxSynchronize());
  }

  void shader(DeviceTask &task)
  {
    if (have_error())
      return;

    CUDAContextScope scope(this);

    CUfunction cuShader;
    CUdeviceptr d_input = cuda_device_ptr(task.shader_input);
    CUdeviceptr d_output = cuda_device_ptr(task.shader_output);

    /* get kernel function */
    if (task.shader_eval_type >= SHADER_EVAL_BAKE) {
      cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_bake"));
    }
    else if (task.shader_eval_type == SHADER_EVAL_DISPLACE) {
      cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_displace"));
    }
    else {
      cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_background"));
    }

    /* do tasks in smaller chunks, so we can cancel it */
    const int shader_chunk_size = 65536;
    const int start = task.shader_x;
    const int end = task.shader_x + task.shader_w;
    int offset = task.offset;

    bool canceled = false;
    for (int sample = 0; sample < task.num_samples && !canceled; sample++) {
      for (int shader_x = start; shader_x < end; shader_x += shader_chunk_size) {
        int shader_w = min(shader_chunk_size, end - shader_x);

        /* pass in parameters */
        void *args[8];
        int arg = 0;
        args[arg++] = &d_input;
        args[arg++] = &d_output;
        args[arg++] = &task.shader_eval_type;
        if (task.shader_eval_type >= SHADER_EVAL_BAKE) {
          args[arg++] = &task.shader_filter;
        }
        args[arg++] = &shader_x;
        args[arg++] = &shader_w;
        args[arg++] = &offset;
        args[arg++] = &sample;

        /* launch kernel */
        int threads_per_block;
        cuda_assert(cuFuncGetAttribute(
            &threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, cuShader));

        int xblocks = (shader_w + threads_per_block - 1) / threads_per_block;

        cuda_assert(cuFuncSetCacheConfig(cuShader, CU_FUNC_CACHE_PREFER_L1));
        cuda_assert(cuLaunchKernel(cuShader,
                                   xblocks,
                                   1,
                                   1, /* blocks */
                                   threads_per_block,
                                   1,
                                   1, /* threads */
                                   0,
                                   0,
                                   args,
                                   0));

        cuda_assert(cuCtxSynchronize());

        if (task.get_cancel()) {
          canceled = true;
          break;
        }
      }

      task.update_progress(NULL);
    }
  }

  CUdeviceptr map_pixels(device_ptr mem)
  {
    if (!background) {
      PixelMem pmem = pixel_mem_map[mem];
      CUdeviceptr buffer;

      size_t bytes;
      cuda_assert(cuGraphicsMapResources(1, &pmem.cuPBOresource, 0));
      cuda_assert(cuGraphicsResourceGetMappedPointer(&buffer, &bytes, pmem.cuPBOresource));

      return buffer;
    }

    return cuda_device_ptr(mem);
  }

  void unmap_pixels(device_ptr mem)
  {
    if (!background) {
      PixelMem pmem = pixel_mem_map[mem];

      cuda_assert(cuGraphicsUnmapResources(1, &pmem.cuPBOresource, 0));
    }
  }

  void pixels_alloc(device_memory &mem)
  {
    PixelMem pmem;

    pmem.w = mem.data_width;
    pmem.h = mem.data_height;

    CUDAContextScope scope(this);

    glGenBuffers(1, &pmem.cuPBO);
    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pmem.cuPBO);
    if (mem.data_type == TYPE_HALF)
      glBufferData(
          GL_PIXEL_UNPACK_BUFFER, pmem.w * pmem.h * sizeof(GLhalf) * 4, NULL, GL_DYNAMIC_DRAW);
    else
      glBufferData(
          GL_PIXEL_UNPACK_BUFFER, pmem.w * pmem.h * sizeof(uint8_t) * 4, NULL, GL_DYNAMIC_DRAW);

    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);

    glActiveTexture(GL_TEXTURE0);
    glGenTextures(1, &pmem.cuTexId);
    glBindTexture(GL_TEXTURE_2D, pmem.cuTexId);
    if (mem.data_type == TYPE_HALF)
      glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA16F, pmem.w, pmem.h, 0, GL_RGBA, GL_HALF_FLOAT, NULL);
    else
      glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, pmem.w, pmem.h, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
    glBindTexture(GL_TEXTURE_2D, 0);

    CUresult result = cuGraphicsGLRegisterBuffer(
        &pmem.cuPBOresource, pmem.cuPBO, CU_GRAPHICS_MAP_RESOURCE_FLAGS_NONE);

    if (result == CUDA_SUCCESS) {
      mem.device_pointer = pmem.cuTexId;
      pixel_mem_map[mem.device_pointer] = pmem;

      mem.device_size = mem.memory_size();
      stats.mem_alloc(mem.device_size);

      return;
    }
    else {
      /* failed to register buffer, fallback to no interop */
      glDeleteBuffers(1, &pmem.cuPBO);
      glDeleteTextures(1, &pmem.cuTexId);

      background = true;
    }
  }

  void pixels_copy_from(device_memory &mem, int y, int w, int h)
  {
    PixelMem pmem = pixel_mem_map[mem.device_pointer];

    CUDAContextScope scope(this);

    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pmem.cuPBO);
    uchar *pixels = (uchar *)glMapBuffer(GL_PIXEL_UNPACK_BUFFER, GL_READ_ONLY);
    size_t offset = sizeof(uchar) * 4 * y * w;
    memcpy((uchar *)mem.host_pointer + offset, pixels + offset, sizeof(uchar) * 4 * w * h);
    glUnmapBuffer(GL_PIXEL_UNPACK_BUFFER);
    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
  }

  void pixels_free(device_memory &mem)
  {
    if (mem.device_pointer) {
      PixelMem pmem = pixel_mem_map[mem.device_pointer];

      CUDAContextScope scope(this);

      cuda_assert(cuGraphicsUnregisterResource(pmem.cuPBOresource));
      glDeleteBuffers(1, &pmem.cuPBO);
      glDeleteTextures(1, &pmem.cuTexId);

      pixel_mem_map.erase(pixel_mem_map.find(mem.device_pointer));
      mem.device_pointer = 0;

      stats.mem_free(mem.device_size);
      mem.device_size = 0;
    }
  }

  void draw_pixels(device_memory &mem,
                   int y,
                   int w,
                   int h,
                   int width,
                   int height,
                   int dx,
                   int dy,
                   int dw,
                   int dh,
                   bool transparent,
                   const DeviceDrawParams &draw_params)
  {
    assert(mem.type == MEM_PIXELS);

    if (!background) {
      const bool use_fallback_shader = (draw_params.bind_display_space_shader_cb == NULL);
      PixelMem pmem = pixel_mem_map[mem.device_pointer];
      float *vpointer;

      CUDAContextScope scope(this);

      /* for multi devices, this assumes the inefficient method that we allocate
       * all pixels on the device even though we only render to a subset */
      size_t offset = 4 * y * w;

      if (mem.data_type == TYPE_HALF)
        offset *= sizeof(GLhalf);
      else
        offset *= sizeof(uint8_t);

      glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pmem.cuPBO);
      glActiveTexture(GL_TEXTURE0);
      glBindTexture(GL_TEXTURE_2D, pmem.cuTexId);
      if (mem.data_type == TYPE_HALF) {
        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, w, h, GL_RGBA, GL_HALF_FLOAT, (void *)offset);
      }
      else {
        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, (void *)offset);
      }
      glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);

      if (transparent) {
        glEnable(GL_BLEND);
        glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
      }

      GLint shader_program;
      if (use_fallback_shader) {
        if (!bind_fallback_display_space_shader(dw, dh)) {
          return;
        }
        shader_program = fallback_shader_program;
      }
      else {
        draw_params.bind_display_space_shader_cb();
        glGetIntegerv(GL_CURRENT_PROGRAM, &shader_program);
      }

      if (!vertex_buffer) {
        glGenBuffers(1, &vertex_buffer);
      }

      glBindBuffer(GL_ARRAY_BUFFER, vertex_buffer);
      /* invalidate old contents -
       * avoids stalling if buffer is still waiting in queue to be rendered */
      glBufferData(GL_ARRAY_BUFFER, 16 * sizeof(float), NULL, GL_STREAM_DRAW);

      vpointer = (float *)glMapBuffer(GL_ARRAY_BUFFER, GL_WRITE_ONLY);

      if (vpointer) {
        /* texture coordinate - vertex pair */
        vpointer[0] = 0.0f;
        vpointer[1] = 0.0f;
        vpointer[2] = dx;
        vpointer[3] = dy;

        vpointer[4] = (float)w / (float)pmem.w;
        vpointer[5] = 0.0f;
        vpointer[6] = (float)width + dx;
        vpointer[7] = dy;

        vpointer[8] = (float)w / (float)pmem.w;
        vpointer[9] = (float)h / (float)pmem.h;
        vpointer[10] = (float)width + dx;
        vpointer[11] = (float)height + dy;

        vpointer[12] = 0.0f;
        vpointer[13] = (float)h / (float)pmem.h;
        vpointer[14] = dx;
        vpointer[15] = (float)height + dy;

        glUnmapBuffer(GL_ARRAY_BUFFER);
      }

      GLuint vertex_array_object;
      GLuint position_attribute, texcoord_attribute;

      glGenVertexArrays(1, &vertex_array_object);
      glBindVertexArray(vertex_array_object);

      texcoord_attribute = glGetAttribLocation(shader_program, "texCoord");
      position_attribute = glGetAttribLocation(shader_program, "pos");

      glEnableVertexAttribArray(texcoord_attribute);
      glEnableVertexAttribArray(position_attribute);

      glVertexAttribPointer(
          texcoord_attribute, 2, GL_FLOAT, GL_FALSE, 4 * sizeof(float), (const GLvoid *)0);
      glVertexAttribPointer(position_attribute,
                            2,
                            GL_FLOAT,
                            GL_FALSE,
                            4 * sizeof(float),
                            (const GLvoid *)(sizeof(float) * 2));

      glDrawArrays(GL_TRIANGLE_FAN, 0, 4);

      if (use_fallback_shader) {
        glUseProgram(0);
      }
      else {
        draw_params.unbind_display_space_shader_cb();
      }

      if (transparent) {
        glDisable(GL_BLEND);
      }

      glBindTexture(GL_TEXTURE_2D, 0);

      return;
    }

    Device::draw_pixels(mem, y, w, h, width, height, dx, dy, dw, dh, transparent, draw_params);
  }

  void thread_run(DeviceTask *task)
  {
    CUDAContextScope scope(this);

    if (task->type == DeviceTask::RENDER) {
      DeviceRequestedFeatures requested_features;
      if (use_split_kernel()) {
        if (split_kernel == NULL) {
          split_kernel = new CUDASplitKernel(this);
          split_kernel->load_kernels(requested_features);
        }
      }

      device_vector<WorkTile> work_tiles(this, "work_tiles", MEM_READ_ONLY);

      /* keep rendering tiles until done */
      RenderTile tile;
      DenoisingTask denoising(this, *task);

      while (task->acquire_tile(this, tile)) {
        if (tile.task == RenderTile::PATH_TRACE) {
          if (use_split_kernel()) {
            device_only_memory<uchar> void_buffer(this, "void_buffer");
            split_kernel->path_trace(task, tile, void_buffer, void_buffer);
          }
          else {
            path_trace(*task, tile, work_tiles);
          }
        }
        else if (tile.task == RenderTile::DENOISE) {
          tile.sample = tile.start_sample + tile.num_samples;

          denoise(tile, denoising);

          task->update_progress(&tile, tile.w * tile.h);
        }

        task->release_tile(tile);

        if (task->get_cancel()) {
          if (task->need_finish_queue == false)
            break;
        }
      }

      work_tiles.free();
    }
    else if (task->type == DeviceTask::SHADER) {
      shader(*task);

      cuda_assert(cuCtxSynchronize());
    }
  }

  class CUDADeviceTask : public DeviceTask {
   public:
    CUDADeviceTask(CUDADevice *device, DeviceTask &task) : DeviceTask(task)
    {
      run = function_bind(&CUDADevice::thread_run, device, this);
    }
  };

  void task_add(DeviceTask &task)
  {
    CUDAContextScope scope(this);

    /* Load texture info. */
    load_texture_info();

    /* Synchronize all memory copies before executing task. */
    cuda_assert(cuCtxSynchronize());

    if (task.type == DeviceTask::FILM_CONVERT) {
      /* must be done in main thread due to opengl access */
      film_convert(task, task.buffer, task.rgba_byte, task.rgba_half);
    }
    else {
      task_pool.push(new CUDADeviceTask(this, task));
    }
  }

  void task_wait()
  {
    task_pool.wait();
  }

  void task_cancel()
  {
    task_pool.cancel();
  }

  friend class CUDASplitKernelFunction;
  friend class CUDASplitKernel;
  friend class CUDAContextScope;
};

/* redefine the cuda_assert macro so it can be used outside of the CUDADevice class
 * now that the definition of that class is complete
 */
#undef cuda_assert
#define cuda_assert(stmt) \
  { \
    CUresult result = stmt; \
\
    if (result != CUDA_SUCCESS) { \
      string message = string_printf("CUDA error: %s in %s", cuewErrorString(result), #stmt); \
      if (device->error_msg == "") \
        device->error_msg = message; \
      fprintf(stderr, "%s\n", message.c_str()); \
      /*cuda_abort();*/ \
      device->cuda_error_documentation(); \
    } \
  } \
  (void)0

/* CUDA context scope. */

CUDAContextScope::CUDAContextScope(CUDADevice *device) : device(device)
{
  cuda_assert(cuCtxPushCurrent(device->cuContext));
}

CUDAContextScope::~CUDAContextScope()
{
  cuda_assert(cuCtxPopCurrent(NULL));
}

/* split kernel */

class CUDASplitKernelFunction : public SplitKernelFunction {
  CUDADevice *device;
  CUfunction func;

 public:
  CUDASplitKernelFunction(CUDADevice *device, CUfunction func) : device(device), func(func)
  {
  }

  /* enqueue the kernel, returns false if there is an error */
  bool enqueue(const KernelDimensions &dim, device_memory & /*kg*/, device_memory & /*data*/)
  {
    return enqueue(dim, NULL);
  }

  /* enqueue the kernel, returns false if there is an error */
  bool enqueue(const KernelDimensions &dim, void *args[])
  {
    if (device->have_error())
      return false;

    CUDAContextScope scope(device);

    /* we ignore dim.local_size for now, as this is faster */
    int threads_per_block;
    cuda_assert(
        cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, func));

    int xblocks = (dim.global_size[0] * dim.global_size[1] + threads_per_block - 1) /
                  threads_per_block;

    cuda_assert(cuFuncSetCacheConfig(func, CU_FUNC_CACHE_PREFER_L1));

    cuda_assert(cuLaunchKernel(func,
                               xblocks,
                               1,
                               1, /* blocks */
                               threads_per_block,
                               1,
                               1, /* threads */
                               0,
                               0,
                               args,
                               0));

    return !device->have_error();
  }
};

CUDASplitKernel::CUDASplitKernel(CUDADevice *device) : DeviceSplitKernel(device), device(device)
{
}

uint64_t CUDASplitKernel::state_buffer_size(device_memory & /*kg*/,
                                            device_memory & /*data*/,
                                            size_t num_threads)
{
  CUDAContextScope scope(device);

  device_vector<uint64_t> size_buffer(device, "size_buffer", MEM_READ_WRITE);
  size_buffer.alloc(1);
  size_buffer.zero_to_device();

  uint threads = num_threads;
  CUdeviceptr d_size = device->cuda_device_ptr(size_buffer.device_pointer);

  struct args_t {
    uint *num_threads;
    CUdeviceptr *size;
  };

  args_t args = {&threads, &d_size};

  CUfunction state_buffer_size;
  cuda_assert(
      cuModuleGetFunction(&state_buffer_size, device->cuModule, "kernel_cuda_state_buffer_size"));

  cuda_assert(cuLaunchKernel(state_buffer_size, 1, 1, 1, 1, 1, 1, 0, 0, (void **)&args, 0));

  size_buffer.copy_from_device(0, 1, 1);
  size_t size = size_buffer[0];
  size_buffer.free();

  return size;
}

bool CUDASplitKernel::enqueue_split_kernel_data_init(const KernelDimensions &dim,
                                                     RenderTile &rtile,
                                                     int num_global_elements,
                                                     device_memory & /*kernel_globals*/,
                                                     device_memory & /*kernel_data*/,
                                                     device_memory &split_data,
                                                     device_memory &ray_state,
                                                     device_memory &queue_index,
                                                     device_memory &use_queues_flag,
                                                     device_memory &work_pool_wgs)
{
  CUDAContextScope scope(device);

  CUdeviceptr d_split_data = device->cuda_device_ptr(split_data.device_pointer);
  CUdeviceptr d_ray_state = device->cuda_device_ptr(ray_state.device_pointer);
  CUdeviceptr d_queue_index = device->cuda_device_ptr(queue_index.device_pointer);
  CUdeviceptr d_use_queues_flag = device->cuda_device_ptr(use_queues_flag.device_pointer);
  CUdeviceptr d_work_pool_wgs = device->cuda_device_ptr(work_pool_wgs.device_pointer);

  CUdeviceptr d_buffer = device->cuda_device_ptr(rtile.buffer);

  int end_sample = rtile.start_sample + rtile.num_samples;
  int queue_size = dim.global_size[0] * dim.global_size[1];

  struct args_t {
    CUdeviceptr *split_data_buffer;
    int *num_elements;
    CUdeviceptr *ray_state;
    int *start_sample;
    int *end_sample;
    int *sx;
    int *sy;
    int *sw;
    int *sh;
    int *offset;
    int *stride;
    CUdeviceptr *queue_index;
    int *queuesize;
    CUdeviceptr *use_queues_flag;
    CUdeviceptr *work_pool_wgs;
    int *num_samples;
    CUdeviceptr *buffer;
  };

  args_t args = {&d_split_data,
                 &num_global_elements,
                 &d_ray_state,
                 &rtile.start_sample,
                 &end_sample,
                 &rtile.x,
                 &rtile.y,
                 &rtile.w,
                 &rtile.h,
                 &rtile.offset,
                 &rtile.stride,
                 &d_queue_index,
                 &queue_size,
                 &d_use_queues_flag,
                 &d_work_pool_wgs,
                 &rtile.num_samples,
                 &d_buffer};

  CUfunction data_init;
  cuda_assert(
      cuModuleGetFunction(&data_init, device->cuModule, "kernel_cuda_path_trace_data_init"));
  if (device->have_error()) {
    return false;
  }

  CUDASplitKernelFunction(device, data_init).enqueue(dim, (void **)&args);

  return !device->have_error();
}

SplitKernelFunction *CUDASplitKernel::get_split_kernel_function(const string &kernel_name,
                                                                const DeviceRequestedFeatures &)
{
  CUDAContextScope scope(device);
  CUfunction func;

  cuda_assert(
      cuModuleGetFunction(&func, device->cuModule, (string("kernel_cuda_") + kernel_name).data()));
  if (device->have_error()) {
    device->cuda_error_message(
        string_printf("kernel \"kernel_cuda_%s\" not found in module", kernel_name.data()));
    return NULL;
  }

  return new CUDASplitKernelFunction(device, func);
}

int2 CUDASplitKernel::split_kernel_local_size()
{
  return make_int2(32, 1);
}

int2 CUDASplitKernel::split_kernel_global_size(device_memory &kg,
                                               device_memory &data,
                                               DeviceTask * /*task*/)
{
  CUDAContextScope scope(device);
  size_t free;
  size_t total;

  cuda_assert(cuMemGetInfo(&free, &total));

  VLOG(1) << "Maximum device allocation size: " << string_human_readable_number(free)
          << " bytes. (" << string_human_readable_size(free) << ").";

  size_t num_elements = max_elements_for_max_buffer_size(kg, data, free / 2);
  size_t side = round_down((int)sqrt(num_elements), 32);
  int2 global_size = make_int2(side, round_down(num_elements / side, 16));
  VLOG(1) << "Global size: " << global_size << ".";
  return global_size;
}

bool device_cuda_init()
{
#ifdef WITH_CUDA_DYNLOAD
  static bool initialized = false;
  static bool result = false;

  if (initialized)
    return result;

  initialized = true;
  int cuew_result = cuewInit(CUEW_INIT_CUDA);
  if (cuew_result == CUEW_SUCCESS) {
    VLOG(1) << "CUEW initialization succeeded";
    if (CUDADevice::have_precompiled_kernels()) {
      VLOG(1) << "Found precompiled kernels";
      result = true;
    }
#  ifndef _WIN32
    else if (cuewCompilerPath() != NULL) {
      VLOG(1) << "Found CUDA compiler " << cuewCompilerPath();
      result = true;
    }
    else {
      VLOG(1) << "Neither precompiled kernels nor CUDA compiler was found,"
              << " unable to use CUDA";
    }
#  endif
  }
  else {
    VLOG(1) << "CUEW initialization failed: "
            << ((cuew_result == CUEW_ERROR_ATEXIT_FAILED) ? "Error setting up atexit() handler" :
                                                            "Error opening the library");
  }

  return result;
#else  /* WITH_CUDA_DYNLOAD */
  return true;
#endif /* WITH_CUDA_DYNLOAD */
}

Device *device_cuda_create(DeviceInfo &info, Stats &stats, Profiler &profiler, bool background)
{
  return new CUDADevice(info, stats, profiler, background);
}

static CUresult device_cuda_safe_init()
{
#ifdef _WIN32
  __try {
    return cuInit(0);
  }
  __except (EXCEPTION_EXECUTE_HANDLER) {
    /* Ignore crashes inside the CUDA driver and hope we can
     * survive even with corrupted CUDA installs. */
    fprintf(stderr, "Cycles CUDA: driver crashed, continuing without CUDA.\n");
  }

  return CUDA_ERROR_NO_DEVICE;
#else
  return cuInit(0);
#endif
}

void device_cuda_info(vector<DeviceInfo> &devices)
{
  CUresult result = device_cuda_safe_init();
  if (result != CUDA_SUCCESS) {
    if (result != CUDA_ERROR_NO_DEVICE)
      fprintf(stderr, "CUDA cuInit: %s\n", cuewErrorString(result));
    return;
  }

  int count = 0;
  result = cuDeviceGetCount(&count);
  if (result != CUDA_SUCCESS) {
    fprintf(stderr, "CUDA cuDeviceGetCount: %s\n", cuewErrorString(result));
    return;
  }

  vector<DeviceInfo> display_devices;

  for (int num = 0; num < count; num++) {
    char name[256];

    result = cuDeviceGetName(name, 256, num);
    if (result != CUDA_SUCCESS) {
      fprintf(stderr, "CUDA cuDeviceGetName: %s\n", cuewErrorString(result));
      continue;
    }

    int major;
    cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, num);
    if (major < 3) {
      VLOG(1) << "Ignoring device \"" << name << "\", this graphics card is no longer supported.";
      continue;
    }

    DeviceInfo info;

    info.type = DEVICE_CUDA;
    info.description = string(name);
    info.num = num;

    info.has_half_images = (major >= 3);
    info.has_volume_decoupled = false;

    int pci_location[3] = {0, 0, 0};
    cuDeviceGetAttribute(&pci_location[0], CU_DEVICE_ATTRIBUTE_PCI_DOMAIN_ID, num);
    cuDeviceGetAttribute(&pci_location[1], CU_DEVICE_ATTRIBUTE_PCI_BUS_ID, num);
    cuDeviceGetAttribute(&pci_location[2], CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID, num);
    info.id = string_printf("CUDA_%s_%04x:%02x:%02x",
                            name,
                            (unsigned int)pci_location[0],
                            (unsigned int)pci_location[1],
                            (unsigned int)pci_location[2]);

    /* If device has a kernel timeout and no compute preemption, we assume
     * it is connected to a display and will freeze the display while doing
     * computations. */
    int timeout_attr = 0, preempt_attr = 0;
    cuDeviceGetAttribute(&timeout_attr, CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT, num);
    cuDeviceGetAttribute(&preempt_attr, CU_DEVICE_ATTRIBUTE_COMPUTE_PREEMPTION_SUPPORTED, num);

    /* The CUDA driver reports compute preemption as not being available on
     * Windows 10 even when it is, due to an issue in application profiles.
     * Detect case where we expect it to be available and override. */
    if (preempt_attr == 0 && (major >= 6) && system_windows_version_at_least(10, 17134)) {
      VLOG(1) << "Assuming device has compute preemption on Windows 10.";
      preempt_attr = 1;
    }

    if (timeout_attr && !preempt_attr) {
      VLOG(1) << "Device is recognized as display.";
      info.description += " (Display)";
      info.display_device = true;
      display_devices.push_back(info);
    }
    else {
      VLOG(1) << "Device has compute preemption or is not used for display.";
      devices.push_back(info);
    }
    VLOG(1) << "Added device \"" << name << "\" with id \"" << info.id << "\".";
  }

  if (!display_devices.empty())
    devices.insert(devices.end(), display_devices.begin(), display_devices.end());
}

string device_cuda_capabilities()
{
  CUresult result = device_cuda_safe_init();
  if (result != CUDA_SUCCESS) {
    if (result != CUDA_ERROR_NO_DEVICE) {
      return string("Error initializing CUDA: ") + cuewErrorString(result);
    }
    return "No CUDA device found\n";
  }

  int count;
  result = cuDeviceGetCount(&count);
  if (result != CUDA_SUCCESS) {
    return string("Error getting devices: ") + cuewErrorString(result);
  }

  string capabilities = "";
  for (int num = 0; num < count; num++) {
    char name[256];
    if (cuDeviceGetName(name, 256, num) != CUDA_SUCCESS) {
      continue;
    }
    capabilities += string("\t") + name + "\n";
    int value;
#define GET_ATTR(attr) \
  { \
    if (cuDeviceGetAttribute(&value, CU_DEVICE_ATTRIBUTE_##attr, num) == CUDA_SUCCESS) { \
      capabilities += string_printf("\t\tCU_DEVICE_ATTRIBUTE_" #attr "\t\t\t%d\n", value); \
    } \
  } \
  (void)0
    /* TODO(sergey): Strip all attributes which are not useful for us
     * or does not depend on the driver.
     */
    GET_ATTR(MAX_THREADS_PER_BLOCK);
    GET_ATTR(MAX_BLOCK_DIM_X);
    GET_ATTR(MAX_BLOCK_DIM_Y);
    GET_ATTR(MAX_BLOCK_DIM_Z);
    GET_ATTR(MAX_GRID_DIM_X);
    GET_ATTR(MAX_GRID_DIM_Y);
    GET_ATTR(MAX_GRID_DIM_Z);
    GET_ATTR(MAX_SHARED_MEMORY_PER_BLOCK);
    GET_ATTR(SHARED_MEMORY_PER_BLOCK);
    GET_ATTR(TOTAL_CONSTANT_MEMORY);
    GET_ATTR(WARP_SIZE);
    GET_ATTR(MAX_PITCH);
    GET_ATTR(MAX_REGISTERS_PER_BLOCK);
    GET_ATTR(REGISTERS_PER_BLOCK);
    GET_ATTR(CLOCK_RATE);
    GET_ATTR(TEXTURE_ALIGNMENT);
    GET_ATTR(GPU_OVERLAP);
    GET_ATTR(MULTIPROCESSOR_COUNT);
    GET_ATTR(KERNEL_EXEC_TIMEOUT);
    GET_ATTR(INTEGRATED);
    GET_ATTR(CAN_MAP_HOST_MEMORY);
    GET_ATTR(COMPUTE_MODE);
    GET_ATTR(MAXIMUM_TEXTURE1D_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE3D_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE3D_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE3D_DEPTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_LAYERED_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE2D_LAYERED_LAYERS);
    GET_ATTR(MAXIMUM_TEXTURE2D_ARRAY_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_ARRAY_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE2D_ARRAY_NUMSLICES);
    GET_ATTR(SURFACE_ALIGNMENT);
    GET_ATTR(CONCURRENT_KERNELS);
    GET_ATTR(ECC_ENABLED);
    GET_ATTR(TCC_DRIVER);
    GET_ATTR(MEMORY_CLOCK_RATE);
    GET_ATTR(GLOBAL_MEMORY_BUS_WIDTH);
    GET_ATTR(L2_CACHE_SIZE);
    GET_ATTR(MAX_THREADS_PER_MULTIPROCESSOR);
    GET_ATTR(ASYNC_ENGINE_COUNT);
    GET_ATTR(UNIFIED_ADDRESSING);
    GET_ATTR(MAXIMUM_TEXTURE1D_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE1D_LAYERED_LAYERS);
    GET_ATTR(CAN_TEX2D_GATHER);
    GET_ATTR(MAXIMUM_TEXTURE2D_GATHER_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_GATHER_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE3D_WIDTH_ALTERNATE);
    GET_ATTR(MAXIMUM_TEXTURE3D_HEIGHT_ALTERNATE);
    GET_ATTR(MAXIMUM_TEXTURE3D_DEPTH_ALTERNATE);
    GET_ATTR(TEXTURE_PITCH_ALIGNMENT);
    GET_ATTR(MAXIMUM_TEXTURECUBEMAP_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURECUBEMAP_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURECUBEMAP_LAYERED_LAYERS);
    GET_ATTR(MAXIMUM_SURFACE1D_WIDTH);
    GET_ATTR(MAXIMUM_SURFACE2D_WIDTH);
    GET_ATTR(MAXIMUM_SURFACE2D_HEIGHT);
    GET_ATTR(MAXIMUM_SURFACE3D_WIDTH);
    GET_ATTR(MAXIMUM_SURFACE3D_HEIGHT);
    GET_ATTR(MAXIMUM_SURFACE3D_DEPTH);
    GET_ATTR(MAXIMUM_SURFACE1D_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_SURFACE1D_LAYERED_LAYERS);
    GET_ATTR(MAXIMUM_SURFACE2D_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_SURFACE2D_LAYERED_HEIGHT);
    GET_ATTR(MAXIMUM_SURFACE2D_LAYERED_LAYERS);
    GET_ATTR(MAXIMUM_SURFACECUBEMAP_WIDTH);
    GET_ATTR(MAXIMUM_SURFACECUBEMAP_LAYERED_WIDTH);
    GET_ATTR(MAXIMUM_SURFACECUBEMAP_LAYERED_LAYERS);
    GET_ATTR(MAXIMUM_TEXTURE1D_LINEAR_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_LINEAR_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_LINEAR_HEIGHT);
    GET_ATTR(MAXIMUM_TEXTURE2D_LINEAR_PITCH);
    GET_ATTR(MAXIMUM_TEXTURE2D_MIPMAPPED_WIDTH);
    GET_ATTR(MAXIMUM_TEXTURE2D_MIPMAPPED_HEIGHT);
    GET_ATTR(COMPUTE_CAPABILITY_MAJOR);
    GET_ATTR(COMPUTE_CAPABILITY_MINOR);
    GET_ATTR(MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH);
    GET_ATTR(STREAM_PRIORITIES_SUPPORTED);
    GET_ATTR(GLOBAL_L1_CACHE_SUPPORTED);
    GET_ATTR(LOCAL_L1_CACHE_SUPPORTED);
    GET_ATTR(MAX_SHARED_MEMORY_PER_MULTIPROCESSOR);
    GET_ATTR(MAX_REGISTERS_PER_MULTIPROCESSOR);
    GET_ATTR(MANAGED_MEMORY);
    GET_ATTR(MULTI_GPU_BOARD);
    GET_ATTR(MULTI_GPU_BOARD_GROUP_ID);
#undef GET_ATTR
    capabilities += "\n";
  }

  return capabilities;
}

CCL_NAMESPACE_END