/*
* Default kernel arguments
* arg0:
* input, __read_only image2d_t
* arg1:
* output, __write_only image2d_t
* suppose cl can get width/height pixels from
* get_image_width/get_image_height
*/
XCamReturn
CLImageKernel::pre_execute (SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output)
{
XCamReturn ret = XCAM_RETURN_NO_ERROR;
SmartPtr<CLContext> context = get_context ();
#define XCAM_CL_MAX_ARGS 256
CLArgument args[XCAM_CL_MAX_ARGS];
uint32_t arg_count = XCAM_CL_MAX_ARGS;
CLWorkSize work_size;
ret = prepare_arguments (input, output, args, arg_count, work_size);
XCAM_ASSERT (arg_count);
for (uint32_t i = 0; i < arg_count; ++i) {
ret = set_argument (i, args[i].arg_adress, args[i].arg_size);
XCAM_FAIL_RETURN (
WARNING,
ret == XCAM_RETURN_NO_ERROR,
ret,
"cl image kernel(%s) set argc(%d) failed", get_kernel_name (), i);
}
XCAM_ASSERT (work_size.global[0]);
ret = set_work_size (work_size.dim, work_size.global, work_size.local);
XCAM_FAIL_RETURN (
WARNING,
ret == XCAM_RETURN_NO_ERROR,
ret,
"cl image kernel(%s) set work size failed", get_kernel_name ());
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLDemosaicImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & video_info = output->get_video_info ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
arg_count = 2;
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = video_info.width / 2;
work_size.global[1] = video_info.height / 2;
work_size.local[0] = 4;
work_size.local[1] = 4;
return XCAM_RETURN_NO_ERROR;
}
void hardware::code::Real::get_work_sizes(const cl_kernel kernel, size_t* ls, size_t* gs, cl_uint* num_groups) const
{
Opencl_Module::get_work_sizes(kernel, ls, gs, num_groups);
// Query specific sizes for kernels if needed
std::string kernelname = get_kernel_name(kernel);
if (kernelname.compare("get_elem_vector") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("set_elem_vector") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("real_ratio") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("real_product") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("real_sum") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("real_subtraction") == 0) {
*ls = 1;
*gs = 1;
*num_groups = 1;
}
if (kernelname.compare("update_alpha_cgm") == 0) {
*ls = 16;
*gs = 16;
*num_groups = 1;
}
if (kernelname.compare("update_beta_cgm") == 0) {
*ls = 16;
*gs = 16;
*num_groups = 1;
}
if (kernelname.compare("update_zeta_cgm") == 0) {
*ls = 16;
*gs = 16;
*num_groups = 1;
}
}
XCamReturn
CLRgbPipeImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & video_info = input->get_video_info ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
if (_image_in_list.size () < 4) {
while (_image_in_list.size () < 4) {
_image_in_list.push_back (_image_in);
}
} else {
_image_in_list.pop_front ();
_image_in_list.push_back (_image_in);
}
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
args[0].arg_adress = &_image_out->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_tnr_config;
args[1].arg_size = sizeof (CLRgbPipeTnrConfig);
uint8_t index = 0;
for (std::list<SmartPtr<CLImage>>::iterator it = _image_in_list.begin (); it != _image_in_list.end (); it++) {
args[2 + index].arg_adress = &(*it)->get_mem_id ();
args[2 + index].arg_size = sizeof (cl_mem);
index++;
}
arg_count = 2 + index;
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = XCAM_ALIGN_UP(video_info.width, 16);
work_size.global[1] = XCAM_ALIGN_UP(video_info.height, 16);
work_size.local[0] = 8;
work_size.local[1] = 4;
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLKernel::set_arguments (const CLArgList &args, const CLWorkSize &work_size)
{
XCamReturn ret = XCAM_RETURN_NO_ERROR;
uint32_t i_count = 0;
XCAM_FAIL_RETURN (
ERROR, _arg_list.empty (), XCAM_RETURN_ERROR_PARAM,
"cl image kernel(%s) arguments was already set, can NOT be set twice", get_kernel_name ());
for (CLArgList::const_iterator iter = args.begin (); iter != args.end (); ++iter, ++i_count) {
const SmartPtr<CLArgument> &arg = *iter;
XCAM_FAIL_RETURN (
WARNING, arg.ptr (),
XCAM_RETURN_ERROR_PARAM, "cl image kernel(%s) argc(%d) is NULL", get_kernel_name (), i_count);
void *adress = NULL;
uint32_t size = 0;
arg->get_value (adress, size);
ret = set_argument (i_count, adress, size);
XCAM_FAIL_RETURN (
WARNING, ret == XCAM_RETURN_NO_ERROR,
ret, "cl image kernel(%s) set argc(%d) failed", get_kernel_name (), i_count);
}
ret = set_work_size (work_size);
XCAM_FAIL_RETURN (
WARNING, ret == XCAM_RETURN_NO_ERROR, ret,
"cl image kernel(%s) set worksize(global:%dx%dx%d, local:%dx%dx%d) failed",
XCAM_STR(get_kernel_name ()),
(int)work_size.global[0], (int)work_size.global[1], (int)work_size.global[2],
(int)work_size.local[0], (int)work_size.local[1], (int)work_size.local[2]);
_arg_list = args;
return ret;
}
XCamReturn
CLDemoImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & video_info = input->get_video_info ();
CLImageDesc image_info;
uint32_t channel_bits = XCAM_ALIGN_UP (video_info.color_bits, 8);
image_info.format.image_channel_order = CL_R;
if (channel_bits == 8)
image_info.format.image_channel_data_type = CL_UNORM_INT8;
else if (channel_bits == 16)
image_info.format.image_channel_data_type = CL_UNORM_INT16;
image_info.width = video_info.strides[0] / CLImage::calculate_pixel_bytes (image_info.format);
image_info.height = (video_info.size / video_info.strides[0]) / 4 * 4;
image_info.row_pitch = video_info.strides[0];
_image_in = new CLVaImage (context, input, image_info);
_image_out = new CLVaImage (context, output, image_info);
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
arg_count = 2;
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = image_info.width;
work_size.global[1] = image_info.height;
work_size.local[0] = 8;
work_size.local[1] = 4;
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLGammaImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
_gamma_table_buffer = new CLBuffer(
context, sizeof(float)*XCAM_GAMMA_TABLE_SIZE,
CL_MEM_READ_WRITE | CL_MEM_USE_HOST_PTR , &_gamma_table);
//CL_MEM_READ_ONLY
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid () && _gamma_table_buffer->is_valid());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid () && _gamma_table_buffer->is_valid(),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
args[2].arg_adress = &_gamma_table_buffer->get_mem_id();
args[2].arg_size = sizeof (cl_mem);
arg_count = 3;
const CLImageDesc out_info = _image_out->get_image_desc ();
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = out_info.width/4;
work_size.global[1] = out_info.height/2;
work_size.local[0] = 4;
work_size.local[1] = 4;
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLBnrImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
args[2].arg_adress = &_bnr_config.bnr_gain;
args[2].arg_size = sizeof (cl_float);
args[3].arg_adress = &_bnr_config.direction;
args[3].arg_size = sizeof (cl_float);
arg_count = 4;
const CLImageDesc out_info = _image_out->get_image_desc ();
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = out_info.width;
work_size.global[1] = out_info.height;
work_size.local[0] = 8;
work_size.local[1] = 4;
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLTnrImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & video_info = input->get_video_info ();
_image_in = new CLVaImage (context, input);
if (CL_TNR_TYPE_RGB == _type) {
if (_image_in_list.size () < _frame_count) {
while (_image_in_list.size () < _frame_count) {
_image_in_list.push_back (_image_in);
}
} else {
_image_in_list.pop_front ();
_image_in_list.push_back (_image_in);
}
}
_image_out = new CLVaImage (context, output);
if (CL_TNR_TYPE_YUV == _type) {
if (!_image_out_prev.ptr ()) {
_image_out_prev = _image_in;
}
}
_vertical_offset = video_info.aligned_height;
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.local[0] = 8;
work_size.local[1] = 4;
if (CL_TNR_TYPE_YUV == _type) {
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out_prev->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
args[2].arg_adress = &_image_out->get_mem_id ();
args[2].arg_size = sizeof (cl_mem);
args[3].arg_adress = &_vertical_offset;
args[3].arg_size = sizeof (_vertical_offset);
args[4].arg_adress = &_gain;
args[4].arg_size = sizeof (_gain);
args[5].arg_adress = &_thr_Y;
args[5].arg_size = sizeof (_thr_Y);
args[6].arg_adress = &_thr_C;
args[6].arg_size = sizeof (_thr_C);
work_size.global[0] = video_info.width / 2;
work_size.global[1] = video_info.height / 2;
arg_count = 7;
}
else if (CL_TNR_TYPE_RGB == _type) {
const CLImageDesc out_info = _image_out->get_image_desc ();
work_size.global[0] = out_info.width;
work_size.global[1] = out_info.height;
args[0].arg_adress = &_image_out->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_thr_Y;
args[1].arg_size = sizeof (_thr_Y);
args[2].arg_adress = &_frame_count;
args[2].arg_size = sizeof (_frame_count);
uint8_t index = 0;
for (std::list<SmartPtr<CLImage>>::iterator it = _image_in_list.begin (); it != _image_in_list.end (); it++) {
args[3 + index].arg_adress = &(*it)->get_mem_id ();
args[3 + index].arg_size = sizeof (cl_mem);
index++;
}
arg_count = 3 + index;
}
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLWaveletDenoiseImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & video_info_in = input->get_video_info ();
const VideoBufferInfo & video_info_out = output->get_video_info ();
_input_image = new CLVaBuffer (context, input);
_reconstruct_image = new CLVaBuffer (context, output);
_details_image = _handler->get_details_image ();
_approx_image = _handler->get_approx_image ();
_decomposition_levels = WAVELET_DECOMPOSITION_LEVELS;
_soft_threshold = _handler->get_denoise_config ().threshold[0];
_hard_threshold = _handler->get_denoise_config ().threshold[1];
_input_y_offset = video_info_in.offsets[0] / 4;
_output_y_offset = video_info_out.offsets[0] / 4;
_input_uv_offset = video_info_in.aligned_height;
_output_uv_offset = video_info_out.aligned_height;
XCAM_ASSERT (_input_image->is_valid () && _reconstruct_image->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_input_image->is_valid () && _reconstruct_image->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
//set args;
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.local[0] = 8;
work_size.local[1] = 4;
if (_current_layer % 2) {
args[0].arg_adress = &_input_image->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_approx_image->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
} else {
args[0].arg_adress = &_approx_image->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_input_image->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
}
args[2].arg_adress = &_details_image->get_mem_id ();
args[2].arg_size = sizeof (cl_mem);
args[3].arg_adress = &_reconstruct_image->get_mem_id ();
args[3].arg_size = sizeof (cl_mem);
args[4].arg_adress = &_input_y_offset;
args[4].arg_size = sizeof (_input_y_offset);
args[5].arg_adress = &_output_y_offset;
args[5].arg_size = sizeof (_output_y_offset);
args[6].arg_adress = &_input_uv_offset;
args[6].arg_size = sizeof (_input_uv_offset);
args[7].arg_adress = &_output_uv_offset;
args[7].arg_size = sizeof (_output_uv_offset);
args[8].arg_adress = &_current_layer;
args[8].arg_size = sizeof (_current_layer);
args[9].arg_adress = &_decomposition_levels;
args[9].arg_size = sizeof (_decomposition_levels);
args[10].arg_adress = &_hard_threshold;
args[10].arg_size = sizeof (_hard_threshold);
args[11].arg_adress = &_soft_threshold;
args[11].arg_size = sizeof (_soft_threshold);
#if WAVELET_DENOISE_UV
work_size.global[0] = video_info_in.width / 16;
work_size.global[1] = video_info_in.height / 2;
#else
work_size.global[0] = video_info_in.width / 16;
work_size.global[1] = video_info_in.height;
#endif
arg_count = 12;
return XCAM_RETURN_NO_ERROR;
}
int install_from_cwd(Options *op)
{
Package *p;
CommandList *c;
int ret;
int ran_pre_install_hook = FALSE;
HookScriptStatus res;
static const char* edit_your_xf86config =
"Please update your XF86Config or xorg.conf file as "
"appropriate; see the file /usr/share/doc/"
"NVIDIA_GLX-1.0/README.txt for details.";
/*
* validate the manifest file in the cwd, and process it, building
* a Package struct
*/
if ((p = parse_manifest(op)) == NULL) goto failed;
if (!op->x_files_packaged) {
edit_your_xf86config = "";
}
ui_set_title(op, "%s (%s)", p->description, p->version);
/*
* warn the user if "legacy" GPUs are installed in this system
* and if no supported GPU is found, at all.
*/
check_for_nvidia_graphics_devices(op, p);
/* check that we are not running any X server */
if (!check_for_running_x(op)) goto failed;
/* make sure the kernel module is unloaded */
if (!check_for_unloaded_kernel_module(op)) goto failed;
/* ask the user to accept the license */
if (!get_license_acceptance(op)) goto exit_install;
ui_log(op, "Installing NVIDIA driver version %s.", p->version);
/*
* determine the current NVIDIA version (if any); ask the user if
* they really want to overwrite the existing installation
*/
if (!check_for_existing_driver(op, p)) goto exit_install;
/*
* check to see if an alternate method of installation is already installed
* or is available, but not installed; ask the user if they really want to
* install anyway despite the presence/availability of an alternate install.
*/
if (!check_for_alternate_install(op)) goto exit_install;
/* run the distro preinstall hook */
res = run_distro_hook(op, "pre-install");
if (res == HOOK_SCRIPT_FAIL) {
if (ui_multiple_choice(op, CONTINUE_ABORT_CHOICES,
NUM_CONTINUE_ABORT_CHOICES,
CONTINUE_CHOICE, /* Default choice */
"The distribution-provided pre-install "
"script failed! Are you sure you want "
"to continue?") == ABORT_CHOICE) {
goto failed;
}
} else if (res == HOOK_SCRIPT_SUCCESS) {
if (ui_multiple_choice(op, CONTINUE_ABORT_CHOICES,
NUM_CONTINUE_ABORT_CHOICES,
CONTINUE_CHOICE, /* Default choice */
"The distribution-provided pre-install script "
"completed successfully. If this is the first "
"time you have run the installer, this script "
"may have helped disable Nouveau, but a reboot "
"may be required first. "
"Would you like to continue, or would you "
"prefer to abort installation to reboot the "
"system?") == ABORT_CHOICE) {
goto exit_install;
}
ran_pre_install_hook = TRUE;
}
/* fail if the nouveau driver is currently in use */
if (!check_for_nouveau(op)) goto failed;
/* ask if we should install the UVM kernel module */
should_install_uvm(op, p);
/* attempt to build the kernel modules for the target kernel */
if (!op->no_kernel_module) {
if (!install_kernel_modules(op, p)) {
goto failed;
}
} else {
ui_warn(op, "You specified the '--no-kernel-module' command line "
"option, nvidia-installer will not install a kernel "
"module as part of this driver installation, and it will "
"not remove existing NVIDIA kernel modules not part of "
"an earlier NVIDIA driver installation. Please ensure "
"that an NVIDIA kernel module matching this driver version "
"is installed seperately.");
/* no_kernel_module should imply no DKMS */
if (op->dkms) {
ui_warn(op, "You have specified both the '--no-kernel-module' "
"and the '--dkms' command line options. The '--dkms' "
"option will be ignored.");
op->dkms = FALSE;
}
}
/*
* if we are only installing the kernel module, then remove
* everything else from the package; otherwise do some
* OpenGL-specific stuff
*/
if (op->kernel_module_only) {
remove_non_kernel_module_files_from_package(op, p);
} else {
/* ask for the XFree86 and OpenGL installation prefixes. */
if (!get_prefixes(op)) goto failed;
/* ask if we should install the OpenGL header files */
should_install_opengl_headers(op, p);
/*
* select the appropriate TLS class, modifying the package as
* necessary.
*/
select_tls_class(op, p);
/*
* if the package contains any libGL.la or .desktop files,
* process them (perform some search and replacing so
* that they reflect the correct installation path, etc)
* and add them to the package list (files to be installed).
*/
process_libGL_la_files(op, p);
process_dot_desktop_files(op, p);
#if defined(NV_X86_64)
/*
* ask if we should install the 32bit compatibility files on
* this machine.
*/
should_install_compat32_files(op, p);
#endif /* NV_X86_64 */
}
if (op->no_opengl_files) {
remove_opengl_files_from_package(op, p);
}
/*
* now that we have the installation prefixes, build the
* destination for each file to be installed
*/
if (!set_destinations(op, p)) goto failed;
/*
* if we are installing OpenGL libraries, ensure that a symlink gets
* installed to /usr/lib/libGL.so.1. add_libgl_abi_symlink() sets its own
* destination, so it must be called after set_destinations().
*/
if (!op->kernel_module_only && !op->no_opengl_files) {
add_libgl_abi_symlink(op, p);
}
/*
* uninstall the existing driver; this needs to be done before
* building the command list.
*
* XXX if we uninstall now, then build the command list, and
* then ask the user if they really want to execute the
* command list, if the user decides not to execute the
* command list, they'll be left with no driver installed.
*/
if (!op->kernel_module_only) {
if (!run_existing_uninstaller(op)) goto failed;
}
if (!check_libglvnd_files(op, p)) {
goto failed;
}
/* build a list of operations to execute to do the install */
if ((c = build_command_list(op, p)) == NULL) goto failed;
/* call the ui to get approval for the list of commands */
if (!ui_approve_command_list(op, c, "%s", p->description)) {
goto exit_install;
}
/* initialize the backup log file */
if (!op->kernel_module_only) {
if (!init_backup(op, p)) goto failed;
}
/* execute the command list */
if (!do_install(op, p, c)) goto failed;
/* Register, build, and install the module with DKMS, if requested */
if (op->dkms && !dkms_install_module(op, p->version, get_kernel_name(op)))
goto failed;
/*
* Leave the RM loaded in case an X server with OutputClass-based driver
* matching is being used.
*/
if (!op->no_kernel_module || op->dkms) {
if (!load_kernel_module(op, p->kernel_modules[0].module_name)) {
goto failed;
}
}
/* run the distro postinstall script */
run_distro_hook(op, "post-install");
/*
* check that everything is installed properly (post-install
* sanity check)
*/
check_installed_files_from_package(op, p);
if (!check_runtime_configuration(op, p)) goto failed;
/* done */
if (op->kernel_module_only || op->no_nvidia_xconfig_question) {
ui_message(op, "Installation of the kernel module for the %s "
"(version %s) is now complete.",
p->description, p->version);
} else {
/* ask the user if they would like to run nvidia-xconfig */
const char *msg = "Would you like to run the nvidia-xconfig utility "
"to automatically update your X configuration file "
"so that the NVIDIA X driver will be used when you "
"restart X? Any pre-existing X configuration "
"file will be backed up.";
ret = run_nvidia_xconfig(op, FALSE, msg, op->run_nvidia_xconfig);
if (ret) {
ui_message(op, "Your X configuration file has been successfully "
"updated. Installation of the %s (version: %s) is now "
"complete.", p->description, p->version);
} else {
ui_message(op, "Installation of the %s (version: %s) is now "
"complete. %s", p->description,
p->version, edit_your_xf86config);
}
}
free_package(p);
return TRUE;
failed:
/*
* something bad happened during installation; print an error
* message and return FALSE
*/
if (op->logging) {
ui_error(op, "Installation has failed. Please see the file '%s' "
"for details. You may find suggestions on fixing "
"installation problems in the README available on the "
"Linux driver download page at www.nvidia.com.",
op->log_file_name);
} else {
ui_error(op, "Installation has failed. You may find suggestions "
"on fixing installation problems in the README available "
"on the Linux driver download page at www.nvidia.com.");
}
if (ran_pre_install_hook)
run_distro_hook(op, "failed-install");
/* fall through into exit_install... */
exit_install:
/*
* we are exiting installation; this can happen for reasons that
* do not merit the error message (e.g., the user declined the
* license agreement)
*/
free_package(p);
return FALSE;
} /* install_from_cwd() */
XCamReturn
CLNewTonemappingImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
const VideoBufferInfo & in_video_info = input->get_video_info ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
_image_height = in_video_info.aligned_height;
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
SmartPtr<X3aStats> stats = input->find_3a_stats ();
XCAM_ASSERT (stats.ptr ());
XCam3AStats *stats_ptr = stats->get_stats ();
XCAM_ASSERT (stats_ptr);
int stats_totalnum = stats_ptr->info.width * stats_ptr->info.height;
int hist_bin_count = 1 << stats_ptr->info.bit_depth;
int y_max = 0;
float y_avg = 0.0f;
for(int i = hist_bin_count - 1; i >= 0; i--)
{
if(stats_ptr->hist_y[i] > 0)
{
y_max = i;
break;
}
}
for(int i = 0; i < hist_bin_count - 1; i++)
{
y_avg += i * stats_ptr->hist_y[i];
}
y_max = y_max + 1;
y_avg = y_avg / stats_totalnum;
int* hist_log = new int[hist_bin_count];
int* sort_y = new int[stats_totalnum + 1];
float* map_index_leq = new float[hist_bin_count];
int* map_index_log = new int[hist_bin_count];
for(int i = 0; i < hist_bin_count; i++)
{
hist_log[i] = 0;
}
int thres = hist_bin_count * (53 * hist_bin_count / 256) / (2 * y_avg + 1);
int y_max0 = (y_max > thres) ? thres : y_max;
int y_max1 = (y_max - thres) > 0 ? (y_max - thres) : 0;
if(y_max < thres)
{
y_max0 = hist_bin_count - 1;
}
float t0 = 0.01f * y_max0;
float t1 = 0.01f * y_max1;
float max0_log = log(y_max0 + t0);
float max1_log = log(y_max1 + t1 + 0.01f);
float t0_log = log(t0);
float t1_log = log(t1 + 0.01f);
float factor0;
if(y_max < thres)
factor0 = (hist_bin_count - 1) / (max0_log - t0_log + 0.01f);
else
factor0 = y_max0 / (max0_log - t0_log + 0.01f);
float factor1 = y_max1 / (max1_log - t1_log + 0.01f);
if(y_max < thres)
{
for(int i = 0; i < y_max; i++)
{
int index = (int)((log(i + t0) - t0_log) * factor0 + 0.5f);
hist_log[index] += stats_ptr->hist_y[i];
map_index_log[i] = index;
}
}
else
{
for(int i = 0; i < y_max0; i++)
{
int index = (int)((log(i + t0) - t0_log) * factor0 + 0.5f);
hist_log[index] += stats_ptr->hist_y[i];
map_index_log[i] = index;
}
for(int i = y_max0; i < y_max; i++)
{
int r = y_max - i;
int index = (int)((log(r + t1) - t1_log) * factor1 + 0.5f);
index = y_max - index;
hist_log[index] += stats_ptr->hist_y[i];
map_index_log[i] = index;
}
}
for(int i = y_max; i < hist_bin_count; i++)
{
map_index_log[i] = map_index_log[y_max - 1];
}
int sort_index = 1;
for(int i = 0; i < hist_bin_count; i++)
{
for(int l = 0; l < hist_log[i]; l++)
{
sort_y[sort_index] = i;
sort_index++;
}
}
sort_y[0] = 0;
for(int i = 1; i < hist_bin_count; i++)
{
hist_log[i] += hist_log[i - 1];
}
int map_leq_index[256];
Haleq(sort_y, hist_log, map_leq_index, 0, sort_y[stats_totalnum], 0, 0, 255);
map_leq_index[255] = hist_bin_count;
map_leq_index[0] = 0;
for(int i = 1; i < 256; i++)
{
if(map_leq_index[i] < map_leq_index[i - 1])
map_leq_index[i] = map_leq_index[i - 1];
}
int k;
for(int i = 0; i < 255; i++)
{
for(k = map_leq_index[i]; k < map_leq_index[i + 1]; k++)
{
map_index_leq[k] = (float)i;
}
}
for(int i = 0; i < hist_bin_count; i++)
{
_map_hist[i] = map_index_leq[map_index_log[i]] / 255.0f;
}
delete[] hist_log;
delete[] map_index_leq;
delete[] map_index_log;
delete[] sort_y;
_map_hist_buffer = new CLBuffer(
context, sizeof(float) * hist_bin_count,
CL_MEM_READ_WRITE | CL_MEM_USE_HOST_PTR, &_map_hist);
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
args[2].arg_adress = &_map_hist_buffer->get_mem_id ();
args[2].arg_size = sizeof (cl_mem);
args[3].arg_adress = &_image_height;
args[3].arg_size = sizeof (int);
arg_count = 4;
const CLImageDesc out_info = _image_out->get_image_desc ();
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = out_info.width;
work_size.global[1] = out_info.height / 4;
work_size.local[0] = 8;
work_size.local[1] = 8;
return XCAM_RETURN_NO_ERROR;
}
XCamReturn
CLTonemappingImageKernel::prepare_arguments (
SmartPtr<DrmBoBuffer> &input, SmartPtr<DrmBoBuffer> &output,
CLArgument args[], uint32_t &arg_count,
CLWorkSize &work_size)
{
SmartPtr<CLContext> context = get_context ();
_image_in = new CLVaImage (context, input);
_image_out = new CLVaImage (context, output);
const VideoBufferInfo & in_video_info = input->get_video_info ();
_image_height = in_video_info.aligned_height;
XCAM_ASSERT (_image_in->is_valid () && _image_out->is_valid ());
XCAM_FAIL_RETURN (
WARNING,
_image_in->is_valid () && _image_out->is_valid (),
XCAM_RETURN_ERROR_MEM,
"cl image kernel(%s) in/out memory not available", get_kernel_name ());
SmartPtr<X3aStats> stats = input->find_3a_stats ();
XCAM_FAIL_RETURN (
ERROR,
stats.ptr (),
XCAM_RETURN_ERROR_MEM,
"CLTonemappingImageKernel find_3a_stats failed");
XCam3AStats *stats_ptr = stats->get_stats ();
XCAM_ASSERT (stats_ptr);
int pixel_totalnum = stats_ptr->info.aligned_width * stats_ptr->info.aligned_height;
int pixel_num = 0;
int hist_bin_count = 1 << stats_ptr->info.bit_depth;
int64_t cumulative_value = 0;
int saturated_thresh = pixel_totalnum * 0.003f;
int percent_90_thresh = pixel_totalnum * 0.1f;
int medium_thresh = pixel_totalnum * 0.5f;
float y_saturated = 0;
float y_percent_90 = 0;
float y_average = 0;
float y_medium = 0;
for (int i = (hist_bin_count - 1); i >= 0; i--)
{
pixel_num += stats_ptr->hist_y[i];
if ((y_saturated == 0) && (pixel_num >= saturated_thresh))
{
y_saturated = i;
}
if ((y_percent_90 == 0) && (pixel_num >= percent_90_thresh))
{
y_percent_90 = i;
}
if ((y_medium == 0) && (pixel_num >= medium_thresh))
{
y_medium = i;
}
cumulative_value += i * stats_ptr->hist_y[i];
}
y_average = cumulative_value / pixel_totalnum;
if (y_saturated < (hist_bin_count - 1)) {
y_saturated = y_saturated + 1;
}
_y_target = (hist_bin_count / y_saturated) * (1.5 * y_medium + 0.5 * y_average) / 2;
if (_y_target < 4) {
_y_target = 4;
}
if ((_y_target > y_saturated) || (y_saturated < 4)) {
_y_target = y_saturated / 4;
}
_y_max = hist_bin_count * (2 * y_saturated + _y_target) / y_saturated - y_saturated - _y_target;
_y_target = _y_target / pow(2, stats_ptr->info.bit_depth - 8);
_y_max = _y_max / pow(2, stats_ptr->info.bit_depth - 8);
//set args;
args[0].arg_adress = &_image_in->get_mem_id ();
args[0].arg_size = sizeof (cl_mem);
args[1].arg_adress = &_image_out->get_mem_id ();
args[1].arg_size = sizeof (cl_mem);
args[2].arg_adress = &_y_max;
args[2].arg_size = sizeof (float);
args[3].arg_adress = &_y_target;
args[3].arg_size = sizeof (float);
args[4].arg_adress = &_image_height;
args[4].arg_size = sizeof (int);
arg_count = 5;
const CLImageDesc out_info = _image_out->get_image_desc ();
work_size.dim = XCAM_DEFAULT_IMAGE_DIM;
work_size.global[0] = out_info.width;
work_size.global[1] = out_info.height / 4;
work_size.local[0] = 8;
work_size.local[1] = 8;
return XCAM_RETURN_NO_ERROR;
}
