diff --git a/Dockerfile.openpilot_base b/Dockerfile.openpilot_base index 8a6041299..44d8d95e9 100644 --- a/Dockerfile.openpilot_base +++ b/Dockerfile.openpilot_base @@ -18,6 +18,43 @@ RUN /tmp/tools/install_ubuntu_dependencies.sh && \ cd /usr/lib/gcc/arm-none-eabi/* && \ rm -rf arm/ thumb/nofp thumb/v6* thumb/v8* thumb/v7+fp thumb/v7-r+fp.sp +# Add OpenCL +RUN apt-get update && apt-get install -y --no-install-recommends \ + apt-utils \ + alien \ + unzip \ + tar \ + curl \ + xz-utils \ + dbus \ + gcc-arm-none-eabi \ + tmux \ + vim \ + libx11-6 \ + wget \ + && rm -rf /var/lib/apt/lists/* + +RUN mkdir -p /tmp/opencl-driver-intel && \ + cd /tmp/opencl-driver-intel && \ + wget https://github.com/intel/llvm/releases/download/2024-WW14/oclcpuexp-2024.17.3.0.09_rel.tar.gz && \ + wget https://github.com/oneapi-src/oneTBB/releases/download/v2021.12.0/oneapi-tbb-2021.12.0-lin.tgz && \ + mkdir -p /opt/intel/oclcpuexp_2024.17.3.0.09_rel && \ + cd /opt/intel/oclcpuexp_2024.17.3.0.09_rel && \ + tar -zxvf /tmp/opencl-driver-intel/oclcpuexp-2024.17.3.0.09_rel.tar.gz && \ + mkdir -p /etc/OpenCL/vendors && \ + echo /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64/libintelocl.so > /etc/OpenCL/vendors/intel_expcpu.icd && \ + cd /opt/intel && \ + tar -zxvf /tmp/opencl-driver-intel/oneapi-tbb-2021.12.0-lin.tgz && \ + ln -s /opt/intel/oneapi-tbb-2021.12.0/lib/intel64/gcc4.8/libtbb.so /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64 && \ + ln -s /opt/intel/oneapi-tbb-2021.12.0/lib/intel64/gcc4.8/libtbbmalloc.so /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64 && \ + ln -s /opt/intel/oneapi-tbb-2021.12.0/lib/intel64/gcc4.8/libtbb.so.12 /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64 && \ + ln -s /opt/intel/oneapi-tbb-2021.12.0/lib/intel64/gcc4.8/libtbbmalloc.so.2 /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64 && \ + mkdir -p /etc/ld.so.conf.d && \ + echo /opt/intel/oclcpuexp_2024.17.3.0.09_rel/x64 > /etc/ld.so.conf.d/libintelopenclexp.conf && \ + ldconfig -f /etc/ld.so.conf.d/libintelopenclexp.conf && \ + cd / && \ + rm -rf /tmp/opencl-driver-intel + ENV NVIDIA_VISIBLE_DEVICES=all ENV NVIDIA_DRIVER_CAPABILITIES=graphics,utility,compute ENV QTWEBENGINE_DISABLE_SANDBOX=1 diff --git a/SConstruct b/SConstruct index 4f04be624..ca5b7b6cb 100644 --- a/SConstruct +++ b/SConstruct @@ -94,6 +94,7 @@ env = Environment( # Arch-specific flags and paths if arch == "larch64": + env.Append(CPPPATH=["#third_party/opencl/include"]) env.Append(LIBPATH=[ "/usr/local/lib", "/system/vendor/lib64", diff --git a/common/SConscript b/common/SConscript index 15a0e5eff..1c68cf05c 100644 --- a/common/SConscript +++ b/common/SConscript @@ -5,6 +5,7 @@ common_libs = [ 'swaglog.cc', 'util.cc', 'ratekeeper.cc', + 'clutil.cc', ] _common = env.Library('common', common_libs, LIBS="json11") diff --git a/common/clutil.cc b/common/clutil.cc new file mode 100644 index 000000000..f8381a7e0 --- /dev/null +++ b/common/clutil.cc @@ -0,0 +1,98 @@ +#include "common/clutil.h" + +#include +#include +#include + +#include "common/util.h" +#include "common/swaglog.h" + +namespace { // helper functions + +template +std::string get_info(Func get_info_func, Id id, Name param_name) { + size_t size = 0; + CL_CHECK(get_info_func(id, param_name, 0, NULL, &size)); + std::string info(size, '\0'); + CL_CHECK(get_info_func(id, param_name, size, info.data(), NULL)); + return info; +} +inline std::string get_platform_info(cl_platform_id id, cl_platform_info name) { return get_info(&clGetPlatformInfo, id, name); } +inline std::string get_device_info(cl_device_id id, cl_device_info name) { return get_info(&clGetDeviceInfo, id, name); } + +void cl_print_info(cl_platform_id platform, cl_device_id device) { + size_t work_group_size = 0; + cl_device_type device_type = 0; + clGetDeviceInfo(device, CL_DEVICE_MAX_WORK_GROUP_SIZE, sizeof(work_group_size), &work_group_size, NULL); + clGetDeviceInfo(device, CL_DEVICE_TYPE, sizeof(device_type), &device_type, NULL); + const char *type_str = "Other..."; + switch (device_type) { + case CL_DEVICE_TYPE_CPU: type_str ="CL_DEVICE_TYPE_CPU"; break; + case CL_DEVICE_TYPE_GPU: type_str = "CL_DEVICE_TYPE_GPU"; break; + case CL_DEVICE_TYPE_ACCELERATOR: type_str = "CL_DEVICE_TYPE_ACCELERATOR"; break; + } + + LOGD("vendor: %s", get_platform_info(platform, CL_PLATFORM_VENDOR).c_str()); + LOGD("platform version: %s", get_platform_info(platform, CL_PLATFORM_VERSION).c_str()); + LOGD("profile: %s", get_platform_info(platform, CL_PLATFORM_PROFILE).c_str()); + LOGD("extensions: %s", get_platform_info(platform, CL_PLATFORM_EXTENSIONS).c_str()); + LOGD("name: %s", get_device_info(device, CL_DEVICE_NAME).c_str()); + LOGD("device version: %s", get_device_info(device, CL_DEVICE_VERSION).c_str()); + LOGD("max work group size: %zu", work_group_size); + LOGD("type = %d, %s", (int)device_type, type_str); +} + +void cl_print_build_errors(cl_program program, cl_device_id device) { + cl_build_status status; + clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_STATUS, sizeof(status), &status, NULL); + size_t log_size; + clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size); + std::string log(log_size, '\0'); + clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, log_size, &log[0], NULL); + + LOGE("build failed; status=%d, log: %s", status, log.c_str()); +} + +} // namespace + +cl_device_id cl_get_device_id(cl_device_type device_type) { + cl_uint num_platforms = 0; + CL_CHECK(clGetPlatformIDs(0, NULL, &num_platforms)); + std::unique_ptr platform_ids = std::make_unique(num_platforms); + CL_CHECK(clGetPlatformIDs(num_platforms, &platform_ids[0], NULL)); + + for (size_t i = 0; i < num_platforms; ++i) { + LOGD("platform[%zu] CL_PLATFORM_NAME: %s", i, get_platform_info(platform_ids[i], CL_PLATFORM_NAME).c_str()); + + // Get first device + if (cl_device_id device_id = NULL; clGetDeviceIDs(platform_ids[i], device_type, 1, &device_id, NULL) == 0 && device_id) { + cl_print_info(platform_ids[i], device_id); + return device_id; + } + } + LOGE("No valid openCL platform found"); + assert(0); + return nullptr; +} + +cl_context cl_create_context(cl_device_id device_id) { + return CL_CHECK_ERR(clCreateContext(NULL, 1, &device_id, NULL, NULL, &err)); +} + +void cl_release_context(cl_context context) { + clReleaseContext(context); +} + +cl_program cl_program_from_file(cl_context ctx, cl_device_id device_id, const char* path, const char* args) { + return cl_program_from_source(ctx, device_id, util::read_file(path), args); +} + +cl_program cl_program_from_source(cl_context ctx, cl_device_id device_id, const std::string& src, const char* args) { + const char *csrc = src.c_str(); + cl_program prg = CL_CHECK_ERR(clCreateProgramWithSource(ctx, 1, &csrc, NULL, &err)); + if (int err = clBuildProgram(prg, 1, &device_id, args, NULL, NULL); err != 0) { + cl_print_build_errors(prg, device_id); + assert(0); + } + return prg; +} diff --git a/common/clutil.h b/common/clutil.h new file mode 100644 index 000000000..b364e79d4 --- /dev/null +++ b/common/clutil.h @@ -0,0 +1,28 @@ +#pragma once + +#ifdef __APPLE__ +#include +#else +#include +#endif + +#include + +#define CL_CHECK(_expr) \ + do { \ + assert(CL_SUCCESS == (_expr)); \ + } while (0) + +#define CL_CHECK_ERR(_expr) \ + ({ \ + cl_int err = CL_INVALID_VALUE; \ + __typeof__(_expr) _ret = _expr; \ + assert(_ret&& err == CL_SUCCESS); \ + _ret; \ + }) + +cl_device_id cl_get_device_id(cl_device_type device_type); +cl_context cl_create_context(cl_device_id device_id); +void cl_release_context(cl_context context); +cl_program cl_program_from_source(cl_context ctx, cl_device_id device_id, const std::string& src, const char* args = nullptr); +cl_program cl_program_from_file(cl_context ctx, cl_device_id device_id, const char* path, const char* args); diff --git a/common/mat.h b/common/mat.h new file mode 100644 index 000000000..8e10d6197 --- /dev/null +++ b/common/mat.h @@ -0,0 +1,85 @@ +#pragma once + +typedef struct vec3 { + float v[3]; +} vec3; + +typedef struct vec4 { + float v[4]; +} vec4; + +typedef struct mat3 { + float v[3*3]; +} mat3; + +typedef struct mat4 { + float v[4*4]; +} mat4; + +static inline mat3 matmul3(const mat3 &a, const mat3 &b) { + mat3 ret = {{0.0}}; + for (int r=0; r<3; r++) { + for (int c=0; c<3; c++) { + float v = 0.0; + for (int k=0; k<3; k++) { + v += a.v[r*3+k] * b.v[k*3+c]; + } + ret.v[r*3+c] = v; + } + } + return ret; +} + +static inline vec3 matvecmul3(const mat3 &a, const vec3 &b) { + vec3 ret = {{0.0}}; + for (int r=0; r<3; r++) { + for (int c=0; c<3; c++) { + ret.v[r] += a.v[r*3+c] * b.v[c]; + } + } + return ret; +} + +static inline mat4 matmul(const mat4 &a, const mat4 &b) { + mat4 ret = {{0.0}}; + for (int r=0; r<4; r++) { + for (int c=0; c<4; c++) { + float v = 0.0; + for (int k=0; k<4; k++) { + v += a.v[r*4+k] * b.v[k*4+c]; + } + ret.v[r*4+c] = v; + } + } + return ret; +} + +static inline vec4 matvecmul(const mat4 &a, const vec4 &b) { + vec4 ret = {{0.0}}; + for (int r=0; r<4; r++) { + for (int c=0; c<4; c++) { + ret.v[r] += a.v[r*4+c] * b.v[c]; + } + } + return ret; +} + +// scales the input and output space of a transformation matrix +// that assumes pixel-center origin. +static inline mat3 transform_scale_buffer(const mat3 &in, float s) { + // in_pt = ( transform(out_pt/s + 0.5) - 0.5) * s + + mat3 transform_out = {{ + 1.0f/s, 0.0f, 0.5f, + 0.0f, 1.0f/s, 0.5f, + 0.0f, 0.0f, 1.0f, + }}; + + mat3 transform_in = {{ + s, 0.0f, -0.5f*s, + 0.0f, s, -0.5f*s, + 0.0f, 0.0f, 1.0f, + }}; + + return matmul3(transform_in, matmul3(in, transform_out)); +} diff --git a/msgq_repo b/msgq_repo index f9ebdca88..20f249385 160000 --- a/msgq_repo +++ b/msgq_repo @@ -1 +1 @@ -Subproject commit f9ebdca885cfe25295d09de9fd57023a10758de5 +Subproject commit 20f2493855ef32339b80f0ad76b3cb82210dc474 diff --git a/selfdrive/modeld/SConscript b/selfdrive/modeld/SConscript index 72d3f95c4..91f359744 100644 --- a/selfdrive/modeld/SConscript +++ b/selfdrive/modeld/SConscript @@ -1,10 +1,35 @@ import os import glob -Import('env', 'arch') +Import('env', 'envCython', 'arch', 'cereal', 'messaging', 'common', 'visionipc') lenv = env.Clone() +lenvCython = envCython.Clone() -tinygrad_files = ["#"+x for x in glob.glob(env.Dir("#tinygrad_repo").relpath + "/**", recursive=True, root_dir=env.Dir("#").abspath) if 'pycache' not in x] +libs = [cereal, messaging, visionipc, common, 'capnp', 'kj', 'pthread'] +frameworks = [] + +common_src = [ + "models/commonmodel.cc", + "transforms/loadyuv.cc", + "transforms/transform.cc", +] + +# OpenCL is a framework on Mac +if arch == "Darwin": + frameworks += ['OpenCL'] +else: + libs += ['OpenCL'] + +# Set path definitions +for pathdef, fn in {'TRANSFORM': 'transforms/transform.cl', 'LOADYUV': 'transforms/loadyuv.cl'}.items(): + for xenv in (lenv, lenvCython): + xenv['CXXFLAGS'].append(f'-D{pathdef}_PATH=\\"{File(fn).abspath}\\"') + +# Compile cython +cython_libs = envCython["LIBS"] + libs +commonmodel_lib = lenv.Library('commonmodel', common_src) +lenvCython.Program('models/commonmodel_pyx.so', 'models/commonmodel_pyx.pyx', LIBS=[commonmodel_lib, *cython_libs], FRAMEWORKS=frameworks) +tinygrad_files = sorted(["#"+x for x in glob.glob(env.Dir("#tinygrad_repo").relpath + "/**", recursive=True, root_dir=env.Dir("#").abspath) if 'pycache' not in x]) # Get model metadata for model_name in ['driving_vision', 'driving_policy', 'dmonitoring_model']: @@ -13,30 +38,22 @@ for model_name in ['driving_vision', 'driving_policy', 'dmonitoring_model']: cmd = f'python3 {Dir("#selfdrive/modeld").abspath}/get_model_metadata.py {fn}.onnx' lenv.Command(fn + "_metadata.pkl", [fn + ".onnx"] + tinygrad_files + script_files, cmd) -# compile warp -tg_flags = { - 'larch64': 'DEV=QCOM FLOAT16=1 NOLOCALS=1 JIT_BATCH_SIZE=0', - 'Darwin': f'DEV=CPU HOME={os.path.expanduser("~")}', # tinygrad calls brew which needs a $HOME in the env -}.get(arch, 'DEV=CPU CPU_LLVM=1') -image_flag = { - 'larch64': 'IMAGE=2', -}.get(arch, 'IMAGE=0') -script_files = [File(Dir("#selfdrive/modeld").File("compile_warp.py").abspath)] -cmd = f'{tg_flags} python3 {Dir("#selfdrive/modeld").abspath}/compile_warp.py ' -lenv.Command(fn + "warp_tinygrad.pkl", tinygrad_files + script_files, cmd) - def tg_compile(flags, model_name): pythonpath_string = 'PYTHONPATH="${PYTHONPATH}:' + env.Dir("#tinygrad_repo").abspath + '"' fn = File(f"models/{model_name}").abspath return lenv.Command( fn + "_tinygrad.pkl", [fn + ".onnx"] + tinygrad_files, - f'{pythonpath_string} {flags} {image_flag} python3 {Dir("#tinygrad_repo").abspath}/examples/openpilot/compile3.py {fn}.onnx {fn}_tinygrad.pkl' + f'{pythonpath_string} {flags} python3 {Dir("#tinygrad_repo").abspath}/examples/openpilot/compile3.py {fn}.onnx {fn}_tinygrad.pkl' ) # Compile small models for model_name in ['driving_vision', 'driving_policy', 'dmonitoring_model']: - tg_compile(tg_flags, model_name) + flags = { + 'larch64': 'DEV=QCOM FLOAT16=1 NOLOCALS=1 IMAGE=2 JIT_BATCH_SIZE=0', + 'Darwin': f'DEV=CPU HOME={os.path.expanduser("~")}', # tinygrad calls brew which needs a $HOME in the env + }.get(arch, 'DEV=CPU CPU_LLVM=1') + tg_compile(flags, model_name) # Compile BIG model if USB GPU is available if "USBGPU" in os.environ: diff --git a/selfdrive/modeld/compile_warp.py b/selfdrive/modeld/compile_warp.py deleted file mode 100755 index 5adb60e62..000000000 --- a/selfdrive/modeld/compile_warp.py +++ /dev/null @@ -1,206 +0,0 @@ -#!/usr/bin/env python3 -import time -import pickle -import numpy as np -from pathlib import Path -from tinygrad.tensor import Tensor -from tinygrad.helpers import Context -from tinygrad.device import Device -from tinygrad.engine.jit import TinyJit - -from openpilot.system.camerad.cameras.nv12_info import get_nv12_info -from openpilot.common.transformations.model import MEDMODEL_INPUT_SIZE, DM_INPUT_SIZE -from openpilot.common.transformations.camera import _ar_ox_fisheye, _os_fisheye - -MODELS_DIR = Path(__file__).parent / 'models' - -CAMERA_CONFIGS = [ - (_ar_ox_fisheye.width, _ar_ox_fisheye.height), # tici: 1928x1208 - (_os_fisheye.width, _os_fisheye.height), # mici: 1344x760 -] - -UV_SCALE_MATRIX = np.array([[0.5, 0, 0], [0, 0.5, 0], [0, 0, 1]], dtype=np.float32) -UV_SCALE_MATRIX_INV = np.linalg.inv(UV_SCALE_MATRIX) - -IMG_BUFFER_SHAPE = (30, MEDMODEL_INPUT_SIZE[1] // 2, MEDMODEL_INPUT_SIZE[0] // 2) - - -def warp_pkl_path(w, h): - return MODELS_DIR / f'warp_{w}x{h}_tinygrad.pkl' - - -def dm_warp_pkl_path(w, h): - return MODELS_DIR / f'dm_warp_{w}x{h}_tinygrad.pkl' - - -def warp_perspective_tinygrad(src_flat, M_inv, dst_shape, src_shape, stride_pad, ratio): - w_dst, h_dst = dst_shape - h_src, w_src = src_shape - - x = Tensor.arange(w_dst).reshape(1, w_dst).expand(h_dst, w_dst) - y = Tensor.arange(h_dst).reshape(h_dst, 1).expand(h_dst, w_dst) - ones = Tensor.ones_like(x) - dst_coords = x.reshape(1, -1).cat(y.reshape(1, -1)).cat(ones.reshape(1, -1)) - - src_coords = M_inv @ dst_coords - src_coords = src_coords / src_coords[2:3, :] - - x_nn_clipped = Tensor.round(src_coords[0]).clip(0, w_src - 1).cast('int') - y_nn_clipped = Tensor.round(src_coords[1]).clip(0, h_src - 1).cast('int') - idx = y_nn_clipped * w_src + (y_nn_clipped * ratio).cast('int') * stride_pad + x_nn_clipped - - sampled = src_flat[idx] - return sampled - - -def frames_to_tensor(frames, model_w, model_h): - H = (frames.shape[0] * 2) // 3 - W = frames.shape[1] - in_img1 = Tensor.cat(frames[0:H:2, 0::2], - frames[1:H:2, 0::2], - frames[0:H:2, 1::2], - frames[1:H:2, 1::2], - frames[H:H+H//4].reshape((H//2, W//2)), - frames[H+H//4:H+H//2].reshape((H//2, W//2)), dim=0).reshape((6, H//2, W//2)) - return in_img1 - - -def make_frame_prepare(cam_w, cam_h, model_w, model_h): - stride, y_height, _, _ = get_nv12_info(cam_w, cam_h) - uv_offset = stride * y_height - stride_pad = stride - cam_w - - def frame_prepare_tinygrad(input_frame, M_inv): - tg_scale = Tensor(UV_SCALE_MATRIX) - M_inv_uv = tg_scale @ M_inv @ Tensor(UV_SCALE_MATRIX_INV) - with Context(SPLIT_REDUCEOP=0): - y = warp_perspective_tinygrad(input_frame[:cam_h*stride], - M_inv, (model_w, model_h), - (cam_h, cam_w), stride_pad, 1).realize() - u = warp_perspective_tinygrad(input_frame[uv_offset:uv_offset + (cam_h//4)*stride], - M_inv_uv, (model_w//2, model_h//2), - (cam_h//2, cam_w//2), stride_pad, 0.5).realize() - v = warp_perspective_tinygrad(input_frame[uv_offset + (cam_h//4)*stride:uv_offset + (cam_h//2)*stride], - M_inv_uv, (model_w//2, model_h//2), - (cam_h//2, cam_w//2), stride_pad, 0.5).realize() - yuv = y.cat(u).cat(v).reshape((model_h * 3 // 2, model_w)) - tensor = frames_to_tensor(yuv, model_w, model_h) - return tensor - return frame_prepare_tinygrad - - -def make_update_img_input(frame_prepare, model_w, model_h): - def update_img_input_tinygrad(tensor, frame, M_inv): - M_inv = M_inv.to(Device.DEFAULT) - new_img = frame_prepare(frame, M_inv) - full_buffer = tensor[6:].cat(new_img, dim=0).contiguous() - return full_buffer, Tensor.cat(full_buffer[:6], full_buffer[-6:], dim=0).contiguous().reshape(1, 12, model_h//2, model_w//2) - return update_img_input_tinygrad - - -def make_update_both_imgs(frame_prepare, model_w, model_h): - update_img = make_update_img_input(frame_prepare, model_w, model_h) - - def update_both_imgs_tinygrad(calib_img_buffer, new_img, M_inv, - calib_big_img_buffer, new_big_img, M_inv_big): - calib_img_buffer, calib_img_pair = update_img(calib_img_buffer, new_img, M_inv) - calib_big_img_buffer, calib_big_img_pair = update_img(calib_big_img_buffer, new_big_img, M_inv_big) - return calib_img_buffer, calib_img_pair, calib_big_img_buffer, calib_big_img_pair - return update_both_imgs_tinygrad - - -def make_warp_dm(cam_w, cam_h, dm_w, dm_h): - stride, y_height, _, _ = get_nv12_info(cam_w, cam_h) - stride_pad = stride - cam_w - - def warp_dm(input_frame, M_inv): - M_inv = M_inv.to(Device.DEFAULT) - with Context(SPLIT_REDUCEOP=0): - result = warp_perspective_tinygrad(input_frame[:cam_h*stride], M_inv, (dm_w, dm_h), (cam_h, cam_w), stride_pad, 1).reshape(-1, dm_h * dm_w) - return result - return warp_dm - - -def compile_modeld_warp(cam_w, cam_h): - model_w, model_h = MEDMODEL_INPUT_SIZE - _, _, _, yuv_size = get_nv12_info(cam_w, cam_h) - - print(f"Compiling modeld warp for {cam_w}x{cam_h}...") - - frame_prepare = make_frame_prepare(cam_w, cam_h, model_w, model_h) - update_both_imgs = make_update_both_imgs(frame_prepare, model_w, model_h) - update_img_jit = TinyJit(update_both_imgs, prune=True) - - full_buffer = Tensor.zeros(IMG_BUFFER_SHAPE, dtype='uint8').contiguous().realize() - big_full_buffer = Tensor.zeros(IMG_BUFFER_SHAPE, dtype='uint8').contiguous().realize() - full_buffer_np = np.zeros(IMG_BUFFER_SHAPE, dtype=np.uint8) - big_full_buffer_np = np.zeros(IMG_BUFFER_SHAPE, dtype=np.uint8) - - for i in range(10): - new_frame_np = (32 * np.random.randn(yuv_size).astype(np.float32) + 128).clip(0, 255).astype(np.uint8) - img_inputs = [full_buffer, - Tensor.from_blob(new_frame_np.ctypes.data, (yuv_size,), dtype='uint8').realize(), - Tensor(Tensor.randn(3, 3).mul(8).realize().numpy(), device='NPY')] - new_big_frame_np = (32 * np.random.randn(yuv_size).astype(np.float32) + 128).clip(0, 255).astype(np.uint8) - big_img_inputs = [big_full_buffer, - Tensor.from_blob(new_big_frame_np.ctypes.data, (yuv_size,), dtype='uint8').realize(), - Tensor(Tensor.randn(3, 3).mul(8).realize().numpy(), device='NPY')] - inputs = img_inputs + big_img_inputs - Device.default.synchronize() - - inputs_np = [x.numpy() for x in inputs] - inputs_np[0] = full_buffer_np - inputs_np[3] = big_full_buffer_np - - st = time.perf_counter() - out = update_img_jit(*inputs) - full_buffer = out[0].contiguous().realize().clone() - big_full_buffer = out[2].contiguous().realize().clone() - mt = time.perf_counter() - Device.default.synchronize() - et = time.perf_counter() - print(f" [{i+1}/10] enqueue {(mt-st)*1e3:6.2f} ms -- total {(et-st)*1e3:6.2f} ms") - - pkl_path = warp_pkl_path(cam_w, cam_h) - with open(pkl_path, "wb") as f: - pickle.dump(update_img_jit, f) - print(f" Saved to {pkl_path}") - - jit = pickle.load(open(pkl_path, "rb")) - jit(*inputs) - - -def compile_dm_warp(cam_w, cam_h): - dm_w, dm_h = DM_INPUT_SIZE - _, _, _, yuv_size = get_nv12_info(cam_w, cam_h) - - print(f"Compiling DM warp for {cam_w}x{cam_h}...") - - warp_dm = make_warp_dm(cam_w, cam_h, dm_w, dm_h) - warp_dm_jit = TinyJit(warp_dm, prune=True) - - for i in range(10): - inputs = [Tensor.from_blob((32 * Tensor.randn(yuv_size,) + 128).cast(dtype='uint8').realize().numpy().ctypes.data, (yuv_size,), dtype='uint8'), - Tensor(Tensor.randn(3, 3).mul(8).realize().numpy(), device='NPY')] - Device.default.synchronize() - st = time.perf_counter() - warp_dm_jit(*inputs) - mt = time.perf_counter() - Device.default.synchronize() - et = time.perf_counter() - print(f" [{i+1}/10] enqueue {(mt-st)*1e3:6.2f} ms -- total {(et-st)*1e3:6.2f} ms") - - pkl_path = dm_warp_pkl_path(cam_w, cam_h) - with open(pkl_path, "wb") as f: - pickle.dump(warp_dm_jit, f) - print(f" Saved to {pkl_path}") - - -def run_and_save_pickle(): - for cam_w, cam_h in CAMERA_CONFIGS: - compile_modeld_warp(cam_w, cam_h) - compile_dm_warp(cam_w, cam_h) - - -if __name__ == "__main__": - run_and_save_pickle() diff --git a/selfdrive/modeld/dmonitoringmodeld.py b/selfdrive/modeld/dmonitoringmodeld.py index 7d4df713c..fca762c69 100755 --- a/selfdrive/modeld/dmonitoringmodeld.py +++ b/selfdrive/modeld/dmonitoringmodeld.py @@ -3,6 +3,7 @@ import os from openpilot.system.hardware import TICI os.environ['DEV'] = 'QCOM' if TICI else 'CPU' from tinygrad.tensor import Tensor +from tinygrad.dtype import dtypes import time import pickle import numpy as np @@ -15,34 +16,32 @@ from openpilot.common.swaglog import cloudlog from openpilot.common.realtime import config_realtime_process from openpilot.common.transformations.model import dmonitoringmodel_intrinsics from openpilot.common.transformations.camera import _ar_ox_fisheye, _os_fisheye -from openpilot.system.camerad.cameras.nv12_info import get_nv12_info +from openpilot.selfdrive.modeld.models.commonmodel_pyx import CLContext, MonitoringModelFrame from openpilot.selfdrive.modeld.parse_model_outputs import sigmoid, safe_exp +from openpilot.selfdrive.modeld.runners.tinygrad_helpers import qcom_tensor_from_opencl_address PROCESS_NAME = "selfdrive.modeld.dmonitoringmodeld" SEND_RAW_PRED = os.getenv('SEND_RAW_PRED') MODEL_PKL_PATH = Path(__file__).parent / 'models/dmonitoring_model_tinygrad.pkl' METADATA_PATH = Path(__file__).parent / 'models/dmonitoring_model_metadata.pkl' -MODELS_DIR = Path(__file__).parent / 'models' + class ModelState: inputs: dict[str, np.ndarray] output: np.ndarray - def __init__(self): + def __init__(self, cl_ctx): with open(METADATA_PATH, 'rb') as f: model_metadata = pickle.load(f) self.input_shapes = model_metadata['input_shapes'] self.output_slices = model_metadata['output_slices'] + self.frame = MonitoringModelFrame(cl_ctx) self.numpy_inputs = { 'calib': np.zeros(self.input_shapes['calib'], dtype=np.float32), } - self.warp_inputs_np = {'transform': np.zeros((3,3), dtype=np.float32)} - self.warp_inputs = {k: Tensor(v, device='NPY') for k,v in self.warp_inputs_np.items()} - self.frame_buf_params = None self.tensor_inputs = {k: Tensor(v, device='NPY').realize() for k,v in self.numpy_inputs.items()} - self.image_warp = None with open(MODEL_PKL_PATH, "rb") as f: self.model_run = pickle.load(f) @@ -51,15 +50,14 @@ class ModelState: t1 = time.perf_counter() - if self.image_warp is None: - self.frame_buf_params = get_nv12_info(buf.width, buf.height) - warp_path = MODELS_DIR / f'dm_warp_{buf.width}x{buf.height}_tinygrad.pkl' - with open(warp_path, "rb") as f: - self.image_warp = pickle.load(f) - self.warp_inputs['frame'] = Tensor.from_blob(buf.data.ctypes.data, (self.frame_buf_params[3],), dtype='uint8').realize() + input_img_cl = self.frame.prepare(buf, transform.flatten()) + if TICI: + # The imgs tensors are backed by opencl memory, only need init once + if 'input_img' not in self.tensor_inputs: + self.tensor_inputs['input_img'] = qcom_tensor_from_opencl_address(input_img_cl.mem_address, self.input_shapes['input_img'], dtype=dtypes.uint8) + else: + self.tensor_inputs['input_img'] = Tensor(self.frame.buffer_from_cl(input_img_cl).reshape(self.input_shapes['input_img']), dtype=dtypes.uint8).realize() - self.warp_inputs_np['transform'][:] = transform[:] - self.tensor_inputs['input_img'] = self.image_warp(self.warp_inputs['frame'], self.warp_inputs['transform']).realize() output = self.model_run(**self.tensor_inputs).contiguous().realize().uop.base.buffer.numpy() @@ -109,11 +107,12 @@ def get_driverstate_packet(model_output, frame_id: int, location_ts: int, exec_t def main(): config_realtime_process(7, 5) - model = ModelState() + cl_context = CLContext() + model = ModelState(cl_context) cloudlog.warning("models loaded, dmonitoringmodeld starting") cloudlog.warning("connecting to driver stream") - vipc_client = VisionIpcClient("camerad", VisionStreamType.VISION_STREAM_DRIVER, True) + vipc_client = VisionIpcClient("camerad", VisionStreamType.VISION_STREAM_DRIVER, True, cl_context) while not vipc_client.connect(False): time.sleep(0.1) assert vipc_client.is_connected() diff --git a/selfdrive/modeld/modeld.py b/selfdrive/modeld/modeld.py index 7fae36510..1f347dc32 100755 --- a/selfdrive/modeld/modeld.py +++ b/selfdrive/modeld/modeld.py @@ -7,6 +7,7 @@ if USBGPU: os.environ['DEV'] = 'AMD' os.environ['AMD_IFACE'] = 'USB' from tinygrad.tensor import Tensor +from tinygrad.dtype import dtypes import time import pickle import numpy as np @@ -21,13 +22,14 @@ from openpilot.common.params import Params from openpilot.common.filter_simple import FirstOrderFilter from openpilot.common.realtime import config_realtime_process, DT_MDL from openpilot.common.transformations.camera import DEVICE_CAMERAS -from openpilot.system.camerad.cameras.nv12_info import get_nv12_info from openpilot.common.transformations.model import get_warp_matrix from openpilot.selfdrive.controls.lib.desire_helper import DesireHelper from openpilot.selfdrive.controls.lib.drive_helpers import get_accel_from_plan, smooth_value, get_curvature_from_plan from openpilot.selfdrive.modeld.parse_model_outputs import Parser from openpilot.selfdrive.modeld.fill_model_msg import fill_model_msg, fill_pose_msg, PublishState from openpilot.selfdrive.modeld.constants import ModelConstants, Plan +from openpilot.selfdrive.modeld.models.commonmodel_pyx import DrivingModelFrame, CLContext +from openpilot.selfdrive.modeld.runners.tinygrad_helpers import qcom_tensor_from_opencl_address PROCESS_NAME = "selfdrive.modeld.modeld" @@ -37,15 +39,11 @@ VISION_PKL_PATH = Path(__file__).parent / 'models/driving_vision_tinygrad.pkl' POLICY_PKL_PATH = Path(__file__).parent / 'models/driving_policy_tinygrad.pkl' VISION_METADATA_PATH = Path(__file__).parent / 'models/driving_vision_metadata.pkl' POLICY_METADATA_PATH = Path(__file__).parent / 'models/driving_policy_metadata.pkl' -MODELS_DIR = Path(__file__).parent / 'models' LAT_SMOOTH_SECONDS = 0.0 LONG_SMOOTH_SECONDS = 0.3 MIN_LAT_CONTROL_SPEED = 0.3 -IMG_QUEUE_SHAPE = (6*(ModelConstants.MODEL_RUN_FREQ//ModelConstants.MODEL_CONTEXT_FREQ + 1), 128, 256) -assert IMG_QUEUE_SHAPE[0] == 30 - def get_action_from_model(model_output: dict[str, np.ndarray], prev_action: log.ModelDataV2.Action, lat_action_t: float, long_action_t: float, v_ego: float) -> log.ModelDataV2.Action: @@ -138,11 +136,12 @@ class InputQueues: return out class ModelState: + frames: dict[str, DrivingModelFrame] inputs: dict[str, np.ndarray] output: np.ndarray prev_desire: np.ndarray # for tracking the rising edge of the pulse - def __init__(self): + def __init__(self, context: CLContext): with open(VISION_METADATA_PATH, 'rb') as f: vision_metadata = pickle.load(f) self.vision_input_shapes = vision_metadata['input_shapes'] @@ -156,6 +155,7 @@ class ModelState: self.policy_output_slices = policy_metadata['output_slices'] policy_output_size = policy_metadata['output_shapes']['outputs'][1] + self.frames = {name: DrivingModelFrame(context, ModelConstants.MODEL_RUN_FREQ//ModelConstants.MODEL_CONTEXT_FREQ) for name in self.vision_input_names} self.prev_desire = np.zeros(ModelConstants.DESIRE_LEN, dtype=np.float32) # policy inputs @@ -165,17 +165,12 @@ class ModelState: self.full_input_queues.update_dtypes_and_shapes({k: self.numpy_inputs[k].dtype}, {k: self.numpy_inputs[k].shape}) self.full_input_queues.reset() - self.img_queues = {'img': Tensor.zeros(IMG_QUEUE_SHAPE, dtype='uint8').contiguous().realize(), - 'big_img': Tensor.zeros(IMG_QUEUE_SHAPE, dtype='uint8').contiguous().realize(),} - self.full_frames : dict[str, Tensor] = {} - self.transforms_np = {k: np.zeros((3,3), dtype=np.float32) for k in self.img_queues} - self.transforms = {k: Tensor(v, device='NPY').realize() for k, v in self.transforms_np.items()} + # img buffers are managed in openCL transform code + self.vision_inputs: dict[str, Tensor] = {} self.vision_output = np.zeros(vision_output_size, dtype=np.float32) self.policy_inputs = {k: Tensor(v, device='NPY').realize() for k,v in self.numpy_inputs.items()} self.policy_output = np.zeros(policy_output_size, dtype=np.float32) self.parser = Parser() - self.frame_buf_params : dict[str, tuple[int, int, int, int]] = {} - self.update_imgs = None with open(VISION_PKL_PATH, "rb") as f: self.vision_run = pickle.load(f) @@ -193,28 +188,23 @@ class ModelState: inputs['desire_pulse'][0] = 0 new_desire = np.where(inputs['desire_pulse'] - self.prev_desire > .99, inputs['desire_pulse'], 0) self.prev_desire[:] = inputs['desire_pulse'] - if self.update_imgs is None: - for key in bufs.keys(): - w, h = bufs[key].width, bufs[key].height - self.frame_buf_params[key] = get_nv12_info(w, h) - warp_path = MODELS_DIR / f'warp_{w}x{h}_tinygrad.pkl' - with open(warp_path, "rb") as f: - self.update_imgs = pickle.load(f) + imgs_cl = {name: self.frames[name].prepare(bufs[name], transforms[name].flatten()) for name in self.vision_input_names} - for key in bufs.keys(): - self.full_frames[key] = Tensor.from_blob(bufs[key].data.ctypes.data, (self.frame_buf_params[key][3],), dtype='uint8').realize() - self.transforms_np[key][:,:] = transforms[key][:,:] - - out = self.update_imgs(self.img_queues['img'], self.full_frames['img'], self.transforms['img'], - self.img_queues['big_img'], self.full_frames['big_img'], self.transforms['big_img']) - self.img_queues['img'], self.img_queues['big_img'], = out[0].realize(), out[2].realize() - vision_inputs = {'img': out[1], 'big_img': out[3]} + if TICI and not USBGPU: + # The imgs tensors are backed by opencl memory, only need init once + for key in imgs_cl: + if key not in self.vision_inputs: + self.vision_inputs[key] = qcom_tensor_from_opencl_address(imgs_cl[key].mem_address, self.vision_input_shapes[key], dtype=dtypes.uint8) + else: + for key in imgs_cl: + frame_input = self.frames[key].buffer_from_cl(imgs_cl[key]).reshape(self.vision_input_shapes[key]) + self.vision_inputs[key] = Tensor(frame_input, dtype=dtypes.uint8).realize() if prepare_only: return None - self.vision_output = self.vision_run(**vision_inputs).contiguous().realize().uop.base.buffer.numpy() + self.vision_output = self.vision_run(**self.vision_inputs).contiguous().realize().uop.base.buffer.numpy() vision_outputs_dict = self.parser.parse_vision_outputs(self.slice_outputs(self.vision_output, self.vision_output_slices)) self.full_input_queues.enqueue({'features_buffer': vision_outputs_dict['hidden_state'], 'desire_pulse': new_desire}) @@ -224,6 +214,7 @@ class ModelState: self.policy_output = self.policy_run(**self.policy_inputs).contiguous().realize().uop.base.buffer.numpy() policy_outputs_dict = self.parser.parse_policy_outputs(self.slice_outputs(self.policy_output, self.policy_output_slices)) + combined_outputs_dict = {**vision_outputs_dict, **policy_outputs_dict} if SEND_RAW_PRED: combined_outputs_dict['raw_pred'] = np.concatenate([self.vision_output.copy(), self.policy_output.copy()]) @@ -240,8 +231,10 @@ def main(demo=False): config_realtime_process(7, 54) st = time.monotonic() - cloudlog.warning("loading model") - model = ModelState() + cloudlog.warning("setting up CL context") + cl_context = CLContext() + cloudlog.warning("CL context ready; loading model") + model = ModelState(cl_context) cloudlog.warning(f"models loaded in {time.monotonic() - st:.1f}s, modeld starting") # visionipc clients @@ -254,8 +247,8 @@ def main(demo=False): time.sleep(.1) vipc_client_main_stream = VisionStreamType.VISION_STREAM_WIDE_ROAD if main_wide_camera else VisionStreamType.VISION_STREAM_ROAD - vipc_client_main = VisionIpcClient("camerad", vipc_client_main_stream, True) - vipc_client_extra = VisionIpcClient("camerad", VisionStreamType.VISION_STREAM_WIDE_ROAD, False) + vipc_client_main = VisionIpcClient("camerad", vipc_client_main_stream, True, cl_context) + vipc_client_extra = VisionIpcClient("camerad", VisionStreamType.VISION_STREAM_WIDE_ROAD, False, cl_context) cloudlog.warning(f"vision stream set up, main_wide_camera: {main_wide_camera}, use_extra_client: {use_extra_client}") while not vipc_client_main.connect(False): diff --git a/selfdrive/modeld/models/commonmodel.cc b/selfdrive/modeld/models/commonmodel.cc new file mode 100644 index 000000000..d3341e76e --- /dev/null +++ b/selfdrive/modeld/models/commonmodel.cc @@ -0,0 +1,64 @@ +#include "selfdrive/modeld/models/commonmodel.h" + +#include +#include + +#include "common/clutil.h" + +DrivingModelFrame::DrivingModelFrame(cl_device_id device_id, cl_context context, int _temporal_skip) : ModelFrame(device_id, context) { + input_frames = std::make_unique(buf_size); + temporal_skip = _temporal_skip; + input_frames_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, buf_size, NULL, &err)); + img_buffer_20hz_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, (temporal_skip+1)*frame_size_bytes, NULL, &err)); + region.origin = temporal_skip * frame_size_bytes; + region.size = frame_size_bytes; + last_img_cl = CL_CHECK_ERR(clCreateSubBuffer(img_buffer_20hz_cl, CL_MEM_READ_WRITE, CL_BUFFER_CREATE_TYPE_REGION, ®ion, &err)); + + loadyuv_init(&loadyuv, context, device_id, MODEL_WIDTH, MODEL_HEIGHT); + init_transform(device_id, context, MODEL_WIDTH, MODEL_HEIGHT); +} + +cl_mem* DrivingModelFrame::prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection) { + run_transform(yuv_cl, MODEL_WIDTH, MODEL_HEIGHT, frame_width, frame_height, frame_stride, frame_uv_offset, projection); + + for (int i = 0; i < temporal_skip; i++) { + CL_CHECK(clEnqueueCopyBuffer(q, img_buffer_20hz_cl, img_buffer_20hz_cl, (i+1)*frame_size_bytes, i*frame_size_bytes, frame_size_bytes, 0, nullptr, nullptr)); + } + loadyuv_queue(&loadyuv, q, y_cl, u_cl, v_cl, last_img_cl); + + copy_queue(&loadyuv, q, img_buffer_20hz_cl, input_frames_cl, 0, 0, frame_size_bytes); + copy_queue(&loadyuv, q, last_img_cl, input_frames_cl, 0, frame_size_bytes, frame_size_bytes); + + // NOTE: Since thneed is using a different command queue, this clFinish is needed to ensure the image is ready. + clFinish(q); + return &input_frames_cl; +} + +DrivingModelFrame::~DrivingModelFrame() { + deinit_transform(); + loadyuv_destroy(&loadyuv); + CL_CHECK(clReleaseMemObject(input_frames_cl)); + CL_CHECK(clReleaseMemObject(img_buffer_20hz_cl)); + CL_CHECK(clReleaseMemObject(last_img_cl)); + CL_CHECK(clReleaseCommandQueue(q)); +} + + +MonitoringModelFrame::MonitoringModelFrame(cl_device_id device_id, cl_context context) : ModelFrame(device_id, context) { + input_frames = std::make_unique(buf_size); + input_frame_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, buf_size, NULL, &err)); + + init_transform(device_id, context, MODEL_WIDTH, MODEL_HEIGHT); +} + +cl_mem* MonitoringModelFrame::prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection) { + run_transform(yuv_cl, MODEL_WIDTH, MODEL_HEIGHT, frame_width, frame_height, frame_stride, frame_uv_offset, projection); + clFinish(q); + return &y_cl; +} + +MonitoringModelFrame::~MonitoringModelFrame() { + deinit_transform(); + CL_CHECK(clReleaseMemObject(input_frame_cl)); + CL_CHECK(clReleaseCommandQueue(q)); +} diff --git a/selfdrive/modeld/models/commonmodel.h b/selfdrive/modeld/models/commonmodel.h new file mode 100644 index 000000000..176d7eb6d --- /dev/null +++ b/selfdrive/modeld/models/commonmodel.h @@ -0,0 +1,97 @@ +#pragma once + +#include +#include +#include + +#include + +#define CL_USE_DEPRECATED_OPENCL_1_2_APIS +#ifdef __APPLE__ +#include +#else +#include +#endif + +#include "common/mat.h" +#include "selfdrive/modeld/transforms/loadyuv.h" +#include "selfdrive/modeld/transforms/transform.h" + +class ModelFrame { +public: + ModelFrame(cl_device_id device_id, cl_context context) { + q = CL_CHECK_ERR(clCreateCommandQueue(context, device_id, 0, &err)); + } + virtual ~ModelFrame() {} + virtual cl_mem* prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection) { return NULL; } + uint8_t* buffer_from_cl(cl_mem *in_frames, int buffer_size) { + CL_CHECK(clEnqueueReadBuffer(q, *in_frames, CL_TRUE, 0, buffer_size, input_frames.get(), 0, nullptr, nullptr)); + clFinish(q); + return &input_frames[0]; + } + + int MODEL_WIDTH; + int MODEL_HEIGHT; + int MODEL_FRAME_SIZE; + int buf_size; + +protected: + cl_mem y_cl, u_cl, v_cl; + Transform transform; + cl_command_queue q; + std::unique_ptr input_frames; + + void init_transform(cl_device_id device_id, cl_context context, int model_width, int model_height) { + y_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, model_width * model_height, NULL, &err)); + u_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, (model_width / 2) * (model_height / 2), NULL, &err)); + v_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, (model_width / 2) * (model_height / 2), NULL, &err)); + transform_init(&transform, context, device_id); + } + + void deinit_transform() { + transform_destroy(&transform); + CL_CHECK(clReleaseMemObject(v_cl)); + CL_CHECK(clReleaseMemObject(u_cl)); + CL_CHECK(clReleaseMemObject(y_cl)); + } + + void run_transform(cl_mem yuv_cl, int model_width, int model_height, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection) { + transform_queue(&transform, q, + yuv_cl, frame_width, frame_height, frame_stride, frame_uv_offset, + y_cl, u_cl, v_cl, model_width, model_height, projection); + } +}; + +class DrivingModelFrame : public ModelFrame { +public: + DrivingModelFrame(cl_device_id device_id, cl_context context, int _temporal_skip); + ~DrivingModelFrame(); + cl_mem* prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection); + + const int MODEL_WIDTH = 512; + const int MODEL_HEIGHT = 256; + const int MODEL_FRAME_SIZE = MODEL_WIDTH * MODEL_HEIGHT * 3 / 2; + const int buf_size = MODEL_FRAME_SIZE * 2; // 2 frames are temporal_skip frames apart + const size_t frame_size_bytes = MODEL_FRAME_SIZE * sizeof(uint8_t); + +private: + LoadYUVState loadyuv; + cl_mem img_buffer_20hz_cl, last_img_cl, input_frames_cl; + cl_buffer_region region; + int temporal_skip; +}; + +class MonitoringModelFrame : public ModelFrame { +public: + MonitoringModelFrame(cl_device_id device_id, cl_context context); + ~MonitoringModelFrame(); + cl_mem* prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3& projection); + + const int MODEL_WIDTH = 1440; + const int MODEL_HEIGHT = 960; + const int MODEL_FRAME_SIZE = MODEL_WIDTH * MODEL_HEIGHT; + const int buf_size = MODEL_FRAME_SIZE; + +private: + cl_mem input_frame_cl; +}; diff --git a/selfdrive/modeld/models/commonmodel.pxd b/selfdrive/modeld/models/commonmodel.pxd new file mode 100644 index 000000000..4ac64d917 --- /dev/null +++ b/selfdrive/modeld/models/commonmodel.pxd @@ -0,0 +1,27 @@ +# distutils: language = c++ + +from msgq.visionipc.visionipc cimport cl_device_id, cl_context, cl_mem + +cdef extern from "common/mat.h": + cdef struct mat3: + float v[9] + +cdef extern from "common/clutil.h": + cdef unsigned long CL_DEVICE_TYPE_DEFAULT + cl_device_id cl_get_device_id(unsigned long) + cl_context cl_create_context(cl_device_id) + void cl_release_context(cl_context) + +cdef extern from "selfdrive/modeld/models/commonmodel.h": + cppclass ModelFrame: + int buf_size + unsigned char * buffer_from_cl(cl_mem*, int); + cl_mem * prepare(cl_mem, int, int, int, int, mat3) + + cppclass DrivingModelFrame: + int buf_size + DrivingModelFrame(cl_device_id, cl_context, int) + + cppclass MonitoringModelFrame: + int buf_size + MonitoringModelFrame(cl_device_id, cl_context) diff --git a/selfdrive/modeld/models/commonmodel_pyx.pxd b/selfdrive/modeld/models/commonmodel_pyx.pxd new file mode 100644 index 000000000..0bb798625 --- /dev/null +++ b/selfdrive/modeld/models/commonmodel_pyx.pxd @@ -0,0 +1,13 @@ +# distutils: language = c++ + +from msgq.visionipc.visionipc cimport cl_mem +from msgq.visionipc.visionipc_pyx cimport CLContext as BaseCLContext + +cdef class CLContext(BaseCLContext): + pass + +cdef class CLMem: + cdef cl_mem * mem + + @staticmethod + cdef create(void*) diff --git a/selfdrive/modeld/models/commonmodel_pyx.pyx b/selfdrive/modeld/models/commonmodel_pyx.pyx new file mode 100644 index 000000000..5b7d11bc7 --- /dev/null +++ b/selfdrive/modeld/models/commonmodel_pyx.pyx @@ -0,0 +1,74 @@ +# distutils: language = c++ +# cython: c_string_encoding=ascii, language_level=3 + +import numpy as np +cimport numpy as cnp +from libc.string cimport memcpy +from libc.stdint cimport uintptr_t + +from msgq.visionipc.visionipc cimport cl_mem +from msgq.visionipc.visionipc_pyx cimport VisionBuf, CLContext as BaseCLContext +from .commonmodel cimport CL_DEVICE_TYPE_DEFAULT, cl_get_device_id, cl_create_context, cl_release_context +from .commonmodel cimport mat3, ModelFrame as cppModelFrame, DrivingModelFrame as cppDrivingModelFrame, MonitoringModelFrame as cppMonitoringModelFrame + + +cdef class CLContext(BaseCLContext): + def __cinit__(self): + self.device_id = cl_get_device_id(CL_DEVICE_TYPE_DEFAULT) + self.context = cl_create_context(self.device_id) + + def __dealloc__(self): + if self.context: + cl_release_context(self.context) + +cdef class CLMem: + @staticmethod + cdef create(void * cmem): + mem = CLMem() + mem.mem = cmem + return mem + + @property + def mem_address(self): + return (self.mem) + +def cl_from_visionbuf(VisionBuf buf): + return CLMem.create(&buf.buf.buf_cl) + + +cdef class ModelFrame: + cdef cppModelFrame * frame + cdef int buf_size + + def __dealloc__(self): + del self.frame + + def prepare(self, VisionBuf buf, float[:] projection): + cdef mat3 cprojection + memcpy(cprojection.v, &projection[0], 9*sizeof(float)) + cdef cl_mem * data + data = self.frame.prepare(buf.buf.buf_cl, buf.width, buf.height, buf.stride, buf.uv_offset, cprojection) + return CLMem.create(data) + + def buffer_from_cl(self, CLMem in_frames): + cdef unsigned char * data2 + data2 = self.frame.buffer_from_cl(in_frames.mem, self.buf_size) + return np.asarray( data2) + + +cdef class DrivingModelFrame(ModelFrame): + cdef cppDrivingModelFrame * _frame + + def __cinit__(self, CLContext context, int temporal_skip): + self._frame = new cppDrivingModelFrame(context.device_id, context.context, temporal_skip) + self.frame = (self._frame) + self.buf_size = self._frame.buf_size + +cdef class MonitoringModelFrame(ModelFrame): + cdef cppMonitoringModelFrame * _frame + + def __cinit__(self, CLContext context): + self._frame = new cppMonitoringModelFrame(context.device_id, context.context) + self.frame = (self._frame) + self.buf_size = self._frame.buf_size + diff --git a/selfdrive/modeld/runners/tinygrad_helpers.py b/selfdrive/modeld/runners/tinygrad_helpers.py new file mode 100644 index 000000000..776381341 --- /dev/null +++ b/selfdrive/modeld/runners/tinygrad_helpers.py @@ -0,0 +1,8 @@ + +from tinygrad.tensor import Tensor +from tinygrad.helpers import to_mv + +def qcom_tensor_from_opencl_address(opencl_address, shape, dtype): + cl_buf_desc_ptr = to_mv(opencl_address, 8).cast('Q')[0] + rawbuf_ptr = to_mv(cl_buf_desc_ptr, 0x100).cast('Q')[20] # offset 0xA0 is a raw gpu pointer. + return Tensor.from_blob(rawbuf_ptr, shape, dtype=dtype, device='QCOM') diff --git a/selfdrive/modeld/transforms/loadyuv.cc b/selfdrive/modeld/transforms/loadyuv.cc new file mode 100644 index 000000000..c93f5cd03 --- /dev/null +++ b/selfdrive/modeld/transforms/loadyuv.cc @@ -0,0 +1,76 @@ +#include "selfdrive/modeld/transforms/loadyuv.h" + +#include +#include +#include + +void loadyuv_init(LoadYUVState* s, cl_context ctx, cl_device_id device_id, int width, int height) { + memset(s, 0, sizeof(*s)); + + s->width = width; + s->height = height; + + char args[1024]; + snprintf(args, sizeof(args), + "-cl-fast-relaxed-math -cl-denorms-are-zero " + "-DTRANSFORMED_WIDTH=%d -DTRANSFORMED_HEIGHT=%d", + width, height); + cl_program prg = cl_program_from_file(ctx, device_id, LOADYUV_PATH, args); + + s->loadys_krnl = CL_CHECK_ERR(clCreateKernel(prg, "loadys", &err)); + s->loaduv_krnl = CL_CHECK_ERR(clCreateKernel(prg, "loaduv", &err)); + s->copy_krnl = CL_CHECK_ERR(clCreateKernel(prg, "copy", &err)); + + // done with this + CL_CHECK(clReleaseProgram(prg)); +} + +void loadyuv_destroy(LoadYUVState* s) { + CL_CHECK(clReleaseKernel(s->loadys_krnl)); + CL_CHECK(clReleaseKernel(s->loaduv_krnl)); + CL_CHECK(clReleaseKernel(s->copy_krnl)); +} + +void loadyuv_queue(LoadYUVState* s, cl_command_queue q, + cl_mem y_cl, cl_mem u_cl, cl_mem v_cl, + cl_mem out_cl) { + cl_int global_out_off = 0; + + CL_CHECK(clSetKernelArg(s->loadys_krnl, 0, sizeof(cl_mem), &y_cl)); + CL_CHECK(clSetKernelArg(s->loadys_krnl, 1, sizeof(cl_mem), &out_cl)); + CL_CHECK(clSetKernelArg(s->loadys_krnl, 2, sizeof(cl_int), &global_out_off)); + + const size_t loadys_work_size = (s->width*s->height)/8; + CL_CHECK(clEnqueueNDRangeKernel(q, s->loadys_krnl, 1, NULL, + &loadys_work_size, NULL, 0, 0, NULL)); + + const size_t loaduv_work_size = ((s->width/2)*(s->height/2))/8; + global_out_off += (s->width*s->height); + + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 0, sizeof(cl_mem), &u_cl)); + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 1, sizeof(cl_mem), &out_cl)); + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 2, sizeof(cl_int), &global_out_off)); + + CL_CHECK(clEnqueueNDRangeKernel(q, s->loaduv_krnl, 1, NULL, + &loaduv_work_size, NULL, 0, 0, NULL)); + + global_out_off += (s->width/2)*(s->height/2); + + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 0, sizeof(cl_mem), &v_cl)); + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 1, sizeof(cl_mem), &out_cl)); + CL_CHECK(clSetKernelArg(s->loaduv_krnl, 2, sizeof(cl_int), &global_out_off)); + + CL_CHECK(clEnqueueNDRangeKernel(q, s->loaduv_krnl, 1, NULL, + &loaduv_work_size, NULL, 0, 0, NULL)); +} + +void copy_queue(LoadYUVState* s, cl_command_queue q, cl_mem src, cl_mem dst, + size_t src_offset, size_t dst_offset, size_t size) { + CL_CHECK(clSetKernelArg(s->copy_krnl, 0, sizeof(cl_mem), &src)); + CL_CHECK(clSetKernelArg(s->copy_krnl, 1, sizeof(cl_mem), &dst)); + CL_CHECK(clSetKernelArg(s->copy_krnl, 2, sizeof(cl_int), &src_offset)); + CL_CHECK(clSetKernelArg(s->copy_krnl, 3, sizeof(cl_int), &dst_offset)); + const size_t copy_work_size = size/8; + CL_CHECK(clEnqueueNDRangeKernel(q, s->copy_krnl, 1, NULL, + ©_work_size, NULL, 0, 0, NULL)); +} \ No newline at end of file diff --git a/selfdrive/modeld/transforms/loadyuv.cl b/selfdrive/modeld/transforms/loadyuv.cl new file mode 100644 index 000000000..970187a6d --- /dev/null +++ b/selfdrive/modeld/transforms/loadyuv.cl @@ -0,0 +1,47 @@ +#define UV_SIZE ((TRANSFORMED_WIDTH/2)*(TRANSFORMED_HEIGHT/2)) + +__kernel void loadys(__global uchar8 const * const Y, + __global uchar * out, + int out_offset) +{ + const int gid = get_global_id(0); + const int ois = gid * 8; + const int oy = ois / TRANSFORMED_WIDTH; + const int ox = ois % TRANSFORMED_WIDTH; + + const uchar8 ys = Y[gid]; + + // 02 + // 13 + + __global uchar* outy0; + __global uchar* outy1; + if ((oy & 1) == 0) { + outy0 = out + out_offset; //y0 + outy1 = out + out_offset + UV_SIZE*2; //y2 + } else { + outy0 = out + out_offset + UV_SIZE; //y1 + outy1 = out + out_offset + UV_SIZE*3; //y3 + } + + vstore4(ys.s0246, 0, outy0 + (oy/2) * (TRANSFORMED_WIDTH/2) + ox/2); + vstore4(ys.s1357, 0, outy1 + (oy/2) * (TRANSFORMED_WIDTH/2) + ox/2); +} + +__kernel void loaduv(__global uchar8 const * const in, + __global uchar8 * out, + int out_offset) +{ + const int gid = get_global_id(0); + const uchar8 inv = in[gid]; + out[gid + out_offset / 8] = inv; +} + +__kernel void copy(__global uchar8 * in, + __global uchar8 * out, + int in_offset, + int out_offset) +{ + const int gid = get_global_id(0); + out[gid + out_offset / 8] = in[gid + in_offset / 8]; +} diff --git a/selfdrive/modeld/transforms/loadyuv.h b/selfdrive/modeld/transforms/loadyuv.h new file mode 100644 index 000000000..659059cd2 --- /dev/null +++ b/selfdrive/modeld/transforms/loadyuv.h @@ -0,0 +1,20 @@ +#pragma once + +#include "common/clutil.h" + +typedef struct { + int width, height; + cl_kernel loadys_krnl, loaduv_krnl, copy_krnl; +} LoadYUVState; + +void loadyuv_init(LoadYUVState* s, cl_context ctx, cl_device_id device_id, int width, int height); + +void loadyuv_destroy(LoadYUVState* s); + +void loadyuv_queue(LoadYUVState* s, cl_command_queue q, + cl_mem y_cl, cl_mem u_cl, cl_mem v_cl, + cl_mem out_cl); + + +void copy_queue(LoadYUVState* s, cl_command_queue q, cl_mem src, cl_mem dst, + size_t src_offset, size_t dst_offset, size_t size); \ No newline at end of file diff --git a/selfdrive/modeld/transforms/transform.cc b/selfdrive/modeld/transforms/transform.cc new file mode 100644 index 000000000..305643cf4 --- /dev/null +++ b/selfdrive/modeld/transforms/transform.cc @@ -0,0 +1,97 @@ +#include "selfdrive/modeld/transforms/transform.h" + +#include +#include + +#include "common/clutil.h" + +void transform_init(Transform* s, cl_context ctx, cl_device_id device_id) { + memset(s, 0, sizeof(*s)); + + cl_program prg = cl_program_from_file(ctx, device_id, TRANSFORM_PATH, ""); + s->krnl = CL_CHECK_ERR(clCreateKernel(prg, "warpPerspective", &err)); + // done with this + CL_CHECK(clReleaseProgram(prg)); + + s->m_y_cl = CL_CHECK_ERR(clCreateBuffer(ctx, CL_MEM_READ_WRITE, 3*3*sizeof(float), NULL, &err)); + s->m_uv_cl = CL_CHECK_ERR(clCreateBuffer(ctx, CL_MEM_READ_WRITE, 3*3*sizeof(float), NULL, &err)); +} + +void transform_destroy(Transform* s) { + CL_CHECK(clReleaseMemObject(s->m_y_cl)); + CL_CHECK(clReleaseMemObject(s->m_uv_cl)); + CL_CHECK(clReleaseKernel(s->krnl)); +} + +void transform_queue(Transform* s, + cl_command_queue q, + cl_mem in_yuv, int in_width, int in_height, int in_stride, int in_uv_offset, + cl_mem out_y, cl_mem out_u, cl_mem out_v, + int out_width, int out_height, + const mat3& projection) { + const int zero = 0; + + // sampled using pixel center origin + // (because that's how fastcv and opencv does it) + + mat3 projection_y = projection; + + // in and out uv is half the size of y. + mat3 projection_uv = transform_scale_buffer(projection, 0.5); + + CL_CHECK(clEnqueueWriteBuffer(q, s->m_y_cl, CL_TRUE, 0, 3*3*sizeof(float), (void*)projection_y.v, 0, NULL, NULL)); + CL_CHECK(clEnqueueWriteBuffer(q, s->m_uv_cl, CL_TRUE, 0, 3*3*sizeof(float), (void*)projection_uv.v, 0, NULL, NULL)); + + const int in_y_width = in_width; + const int in_y_height = in_height; + const int in_y_px_stride = 1; + const int in_uv_width = in_width/2; + const int in_uv_height = in_height/2; + const int in_uv_px_stride = 2; + const int in_u_offset = in_uv_offset; + const int in_v_offset = in_uv_offset + 1; + + const int out_y_width = out_width; + const int out_y_height = out_height; + const int out_uv_width = out_width/2; + const int out_uv_height = out_height/2; + + CL_CHECK(clSetKernelArg(s->krnl, 0, sizeof(cl_mem), &in_yuv)); // src + CL_CHECK(clSetKernelArg(s->krnl, 1, sizeof(cl_int), &in_stride)); // src_row_stride + CL_CHECK(clSetKernelArg(s->krnl, 2, sizeof(cl_int), &in_y_px_stride)); // src_px_stride + CL_CHECK(clSetKernelArg(s->krnl, 3, sizeof(cl_int), &zero)); // src_offset + CL_CHECK(clSetKernelArg(s->krnl, 4, sizeof(cl_int), &in_y_height)); // src_rows + CL_CHECK(clSetKernelArg(s->krnl, 5, sizeof(cl_int), &in_y_width)); // src_cols + CL_CHECK(clSetKernelArg(s->krnl, 6, sizeof(cl_mem), &out_y)); // dst + CL_CHECK(clSetKernelArg(s->krnl, 7, sizeof(cl_int), &out_y_width)); // dst_row_stride + CL_CHECK(clSetKernelArg(s->krnl, 8, sizeof(cl_int), &zero)); // dst_offset + CL_CHECK(clSetKernelArg(s->krnl, 9, sizeof(cl_int), &out_y_height)); // dst_rows + CL_CHECK(clSetKernelArg(s->krnl, 10, sizeof(cl_int), &out_y_width)); // dst_cols + CL_CHECK(clSetKernelArg(s->krnl, 11, sizeof(cl_mem), &s->m_y_cl)); // M + + const size_t work_size_y[2] = {(size_t)out_y_width, (size_t)out_y_height}; + + CL_CHECK(clEnqueueNDRangeKernel(q, s->krnl, 2, NULL, + (const size_t*)&work_size_y, NULL, 0, 0, NULL)); + + const size_t work_size_uv[2] = {(size_t)out_uv_width, (size_t)out_uv_height}; + + CL_CHECK(clSetKernelArg(s->krnl, 2, sizeof(cl_int), &in_uv_px_stride)); // src_px_stride + CL_CHECK(clSetKernelArg(s->krnl, 3, sizeof(cl_int), &in_u_offset)); // src_offset + CL_CHECK(clSetKernelArg(s->krnl, 4, sizeof(cl_int), &in_uv_height)); // src_rows + CL_CHECK(clSetKernelArg(s->krnl, 5, sizeof(cl_int), &in_uv_width)); // src_cols + CL_CHECK(clSetKernelArg(s->krnl, 6, sizeof(cl_mem), &out_u)); // dst + CL_CHECK(clSetKernelArg(s->krnl, 7, sizeof(cl_int), &out_uv_width)); // dst_row_stride + CL_CHECK(clSetKernelArg(s->krnl, 8, sizeof(cl_int), &zero)); // dst_offset + CL_CHECK(clSetKernelArg(s->krnl, 9, sizeof(cl_int), &out_uv_height)); // dst_rows + CL_CHECK(clSetKernelArg(s->krnl, 10, sizeof(cl_int), &out_uv_width)); // dst_cols + CL_CHECK(clSetKernelArg(s->krnl, 11, sizeof(cl_mem), &s->m_uv_cl)); // M + + CL_CHECK(clEnqueueNDRangeKernel(q, s->krnl, 2, NULL, + (const size_t*)&work_size_uv, NULL, 0, 0, NULL)); + CL_CHECK(clSetKernelArg(s->krnl, 3, sizeof(cl_int), &in_v_offset)); // src_ofset + CL_CHECK(clSetKernelArg(s->krnl, 6, sizeof(cl_mem), &out_v)); // dst + + CL_CHECK(clEnqueueNDRangeKernel(q, s->krnl, 2, NULL, + (const size_t*)&work_size_uv, NULL, 0, 0, NULL)); +} diff --git a/selfdrive/modeld/transforms/transform.cl b/selfdrive/modeld/transforms/transform.cl new file mode 100644 index 000000000..2ca25920c --- /dev/null +++ b/selfdrive/modeld/transforms/transform.cl @@ -0,0 +1,54 @@ +#define INTER_BITS 5 +#define INTER_TAB_SIZE (1 << INTER_BITS) +#define INTER_SCALE 1.f / INTER_TAB_SIZE + +#define INTER_REMAP_COEF_BITS 15 +#define INTER_REMAP_COEF_SCALE (1 << INTER_REMAP_COEF_BITS) + +__kernel void warpPerspective(__global const uchar * src, + int src_row_stride, int src_px_stride, int src_offset, int src_rows, int src_cols, + __global uchar * dst, + int dst_row_stride, int dst_offset, int dst_rows, int dst_cols, + __constant float * M) +{ + int dx = get_global_id(0); + int dy = get_global_id(1); + + if (dx < dst_cols && dy < dst_rows) + { + float X0 = M[0] * dx + M[1] * dy + M[2]; + float Y0 = M[3] * dx + M[4] * dy + M[5]; + float W = M[6] * dx + M[7] * dy + M[8]; + W = W != 0.0f ? INTER_TAB_SIZE / W : 0.0f; + int X = rint(X0 * W), Y = rint(Y0 * W); + + int sx = convert_short_sat(X >> INTER_BITS); + int sy = convert_short_sat(Y >> INTER_BITS); + + short sx_clamp = clamp(sx, 0, src_cols - 1); + short sx_p1_clamp = clamp(sx + 1, 0, src_cols - 1); + short sy_clamp = clamp(sy, 0, src_rows - 1); + short sy_p1_clamp = clamp(sy + 1, 0, src_rows - 1); + int v0 = convert_int(src[mad24(sy_clamp, src_row_stride, src_offset + sx_clamp*src_px_stride)]); + int v1 = convert_int(src[mad24(sy_clamp, src_row_stride, src_offset + sx_p1_clamp*src_px_stride)]); + int v2 = convert_int(src[mad24(sy_p1_clamp, src_row_stride, src_offset + sx_clamp*src_px_stride)]); + int v3 = convert_int(src[mad24(sy_p1_clamp, src_row_stride, src_offset + sx_p1_clamp*src_px_stride)]); + + short ay = (short)(Y & (INTER_TAB_SIZE - 1)); + short ax = (short)(X & (INTER_TAB_SIZE - 1)); + float taby = 1.f/INTER_TAB_SIZE*ay; + float tabx = 1.f/INTER_TAB_SIZE*ax; + + int dst_index = mad24(dy, dst_row_stride, dst_offset + dx); + + int itab0 = convert_short_sat_rte( (1.0f-taby)*(1.0f-tabx) * INTER_REMAP_COEF_SCALE ); + int itab1 = convert_short_sat_rte( (1.0f-taby)*tabx * INTER_REMAP_COEF_SCALE ); + int itab2 = convert_short_sat_rte( taby*(1.0f-tabx) * INTER_REMAP_COEF_SCALE ); + int itab3 = convert_short_sat_rte( taby*tabx * INTER_REMAP_COEF_SCALE ); + + int val = v0 * itab0 + v1 * itab1 + v2 * itab2 + v3 * itab3; + + uchar pix = convert_uchar_sat((val + (1 << (INTER_REMAP_COEF_BITS-1))) >> INTER_REMAP_COEF_BITS); + dst[dst_index] = pix; + } +} diff --git a/selfdrive/modeld/transforms/transform.h b/selfdrive/modeld/transforms/transform.h new file mode 100644 index 000000000..771a7054b --- /dev/null +++ b/selfdrive/modeld/transforms/transform.h @@ -0,0 +1,25 @@ +#pragma once + +#define CL_USE_DEPRECATED_OPENCL_1_2_APIS +#ifdef __APPLE__ +#include +#else +#include +#endif + +#include "common/mat.h" + +typedef struct { + cl_kernel krnl; + cl_mem m_y_cl, m_uv_cl; +} Transform; + +void transform_init(Transform* s, cl_context ctx, cl_device_id device_id); + +void transform_destroy(Transform* transform); + +void transform_queue(Transform* s, cl_command_queue q, + cl_mem yuv, int in_width, int in_height, int in_stride, int in_uv_offset, + cl_mem out_y, cl_mem out_u, cl_mem out_v, + int out_width, int out_height, + const mat3& projection); diff --git a/selfdrive/test/process_replay/model_replay.py b/selfdrive/test/process_replay/model_replay.py index c24232840..9ba599bac 100755 --- a/selfdrive/test/process_replay/model_replay.py +++ b/selfdrive/test/process_replay/model_replay.py @@ -34,7 +34,7 @@ GITHUB = GithubUtils(API_TOKEN, DATA_TOKEN) EXEC_TIMINGS = [ # model, instant max, average max - ("modelV2", 0.035, 0.028), + ("modelV2", 0.035, 0.025), ("driverStateV2", 0.02, 0.015), ] diff --git a/selfdrive/test/process_replay/process_replay.py b/selfdrive/test/process_replay/process_replay.py index 5143334bc..8af72e5f4 100755 --- a/selfdrive/test/process_replay/process_replay.py +++ b/selfdrive/test/process_replay/process_replay.py @@ -4,7 +4,6 @@ import time import copy import heapq import signal -import numpy as np from collections import Counter from dataclasses import dataclass, field from itertools import islice @@ -24,7 +23,6 @@ from openpilot.common.params import Params from openpilot.common.prefix import OpenpilotPrefix from openpilot.common.timeout import Timeout from openpilot.common.realtime import DT_CTRL -from openpilot.system.camerad.cameras.nv12_info import get_nv12_info from openpilot.system.manager.process_config import managed_processes from openpilot.selfdrive.test.process_replay.vision_meta import meta_from_camera_state, available_streams from openpilot.selfdrive.test.process_replay.migration import migrate_all @@ -205,8 +203,7 @@ class ProcessContainer: if meta.camera_state in self.cfg.vision_pubs: assert frs[meta.camera_state].pix_fmt == 'nv12' frame_size = (frs[meta.camera_state].w, frs[meta.camera_state].h) - stride, y_height, _, yuv_size = get_nv12_info(frame_size[0], frame_size[1]) - vipc_server.create_buffers_with_sizes(meta.stream, 2, frame_size[0], frame_size[1], yuv_size, stride, stride * y_height) + vipc_server.create_buffers(meta.stream, 2, *frame_size) vipc_server.start_listener() self.vipc_server = vipc_server @@ -303,15 +300,7 @@ class ProcessContainer: camera_meta = meta_from_camera_state(m.which()) assert frs is not None img = frs[m.which()].get(camera_state.frameId) - - h, w = frs[m.which()].h, frs[m.which()].w - stride, y_height, _, yuv_size = get_nv12_info(w, h) - uv_offset = stride * y_height - padded_img = np.zeros((yuv_size), dtype=np.uint8).reshape((-1, stride)) - padded_img[:h, :w] = img[:h * w].reshape((-1, w)) - padded_img[uv_offset // stride:uv_offset // stride + h // 2, :w] = img[h * w:].reshape((-1, w)) - - self.vipc_server.send(camera_meta.stream, padded_img.flatten().tobytes(), + self.vipc_server.send(camera_meta.stream, img.flatten().tobytes(), camera_state.frameId, camera_state.timestampSof, camera_state.timestampEof) self.msg_queue = [] diff --git a/system/camerad/SConscript b/system/camerad/SConscript index c28330b32..e288c6d8b 100644 --- a/system/camerad/SConscript +++ b/system/camerad/SConscript @@ -1,6 +1,6 @@ Import('env', 'arch', 'messaging', 'common', 'visionipc') -libs = [common, messaging, visionipc] +libs = [common, 'OpenCL', messaging, visionipc] if arch != "Darwin": camera_obj = env.Object(['cameras/camera_qcom2.cc', 'cameras/camera_common.cc', 'cameras/spectra.cc', diff --git a/system/camerad/cameras/camera_common.cc b/system/camerad/cameras/camera_common.cc index 329192b63..88bca7f77 100644 --- a/system/camerad/cameras/camera_common.cc +++ b/system/camerad/cameras/camera_common.cc @@ -7,7 +7,7 @@ #include "system/camerad/cameras/spectra.h" -void CameraBuf::init(SpectraCamera *cam, VisionIpcServer * v, int frame_cnt, VisionStreamType type) { +void CameraBuf::init(cl_device_id device_id, cl_context context, SpectraCamera *cam, VisionIpcServer * v, int frame_cnt, VisionStreamType type) { vipc_server = v; stream_type = type; frame_buf_count = frame_cnt; @@ -21,8 +21,9 @@ void CameraBuf::init(SpectraCamera *cam, VisionIpcServer * v, int frame_cnt, Vis const int raw_frame_size = (sensor->frame_height + sensor->extra_height) * sensor->frame_stride; for (int i = 0; i < frame_buf_count; i++) { camera_bufs_raw[i].allocate(raw_frame_size); + camera_bufs_raw[i].init_cl(device_id, context); } - LOGD("allocated %d buffers", frame_buf_count); + LOGD("allocated %d CL buffers", frame_buf_count); } vipc_server->create_buffers_with_sizes(stream_type, VIPC_BUFFER_COUNT, out_img_width, out_img_height, cam->yuv_size, cam->stride, cam->uv_offset); diff --git a/system/camerad/cameras/camera_common.h b/system/camerad/cameras/camera_common.h index 7f35e06a8..c26859cbc 100644 --- a/system/camerad/cameras/camera_common.h +++ b/system/camerad/cameras/camera_common.h @@ -36,7 +36,7 @@ public: CameraBuf() = default; ~CameraBuf(); - void init(SpectraCamera *cam, VisionIpcServer * v, int frame_cnt, VisionStreamType type); + void init(cl_device_id device_id, cl_context context, SpectraCamera *cam, VisionIpcServer * v, int frame_cnt, VisionStreamType type); void sendFrameToVipc(); }; diff --git a/system/camerad/cameras/camera_qcom2.cc b/system/camerad/cameras/camera_qcom2.cc index 6a7f599ab..d741e13cf 100644 --- a/system/camerad/cameras/camera_qcom2.cc +++ b/system/camerad/cameras/camera_qcom2.cc @@ -12,8 +12,16 @@ #include #include +#ifdef __TICI__ +#include "CL/cl_ext_qcom.h" +#else +#define CL_PRIORITY_HINT_HIGH_QCOM NULL +#define CL_CONTEXT_PRIORITY_HINT_QCOM NULL +#endif + #include "media/cam_sensor_cmn_header.h" +#include "common/clutil.h" #include "common/params.h" #include "common/swaglog.h" @@ -49,7 +57,7 @@ public: CameraState(SpectraMaster *master, const CameraConfig &config) : camera(master, config) {}; ~CameraState(); - void init(VisionIpcServer *v); + void init(VisionIpcServer *v, cl_device_id device_id, cl_context ctx); void update_exposure_score(float desired_ev, int exp_t, int exp_g_idx, float exp_gain); void set_camera_exposure(float grey_frac); void set_exposure_rect(); @@ -60,8 +68,8 @@ public: } }; -void CameraState::init(VisionIpcServer *v) { - camera.camera_open(v); +void CameraState::init(VisionIpcServer *v, cl_device_id device_id, cl_context ctx) { + camera.camera_open(v, device_id, ctx); if (!camera.enabled) return; @@ -249,7 +257,11 @@ void CameraState::sendState() { void camerad_thread() { // TODO: centralize enabled handling - VisionIpcServer v("camerad"); + cl_device_id device_id = cl_get_device_id(CL_DEVICE_TYPE_DEFAULT); + const cl_context_properties props[] = {CL_CONTEXT_PRIORITY_HINT_QCOM, CL_PRIORITY_HINT_HIGH_QCOM, 0}; + cl_context ctx = CL_CHECK_ERR(clCreateContext(props, 1, &device_id, NULL, NULL, &err)); + + VisionIpcServer v("camerad", device_id, ctx); // *** initial ISP init *** SpectraMaster m; @@ -259,7 +271,7 @@ void camerad_thread() { std::vector> cams; for (const auto &config : ALL_CAMERA_CONFIGS) { auto cam = std::make_unique(&m, config); - cam->init(&v); + cam->init(&v, device_id, ctx); cams.emplace_back(std::move(cam)); } diff --git a/system/camerad/cameras/spectra.cc b/system/camerad/cameras/spectra.cc index 73e0a78da..5c3e7a9d2 100644 --- a/system/camerad/cameras/spectra.cc +++ b/system/camerad/cameras/spectra.cc @@ -274,7 +274,7 @@ int SpectraCamera::clear_req_queue() { return ret; } -void SpectraCamera::camera_open(VisionIpcServer *v) { +void SpectraCamera::camera_open(VisionIpcServer *v, cl_device_id device_id, cl_context ctx) { if (!openSensor()) { return; } @@ -296,7 +296,7 @@ void SpectraCamera::camera_open(VisionIpcServer *v) { linkDevices(); LOGD("camera init %d", cc.camera_num); - buf.init(this, v, ife_buf_depth, cc.stream_type); + buf.init(device_id, ctx, this, v, ife_buf_depth, cc.stream_type); camera_map_bufs(); clearAndRequeue(1); } diff --git a/system/camerad/cameras/spectra.h b/system/camerad/cameras/spectra.h index a02b8a6ca..13cb13f98 100644 --- a/system/camerad/cameras/spectra.h +++ b/system/camerad/cameras/spectra.h @@ -113,7 +113,7 @@ public: SpectraCamera(SpectraMaster *master, const CameraConfig &config); ~SpectraCamera(); - void camera_open(VisionIpcServer *v); + void camera_open(VisionIpcServer *v, cl_device_id device_id, cl_context ctx); bool handle_camera_event(const cam_req_mgr_message *event_data); void camera_close(); void camera_map_bufs(); diff --git a/system/loggerd/SConscript b/system/loggerd/SConscript index cc8ef7c88..cf169f4dc 100644 --- a/system/loggerd/SConscript +++ b/system/loggerd/SConscript @@ -2,13 +2,16 @@ Import('env', 'arch', 'messaging', 'common', 'visionipc') libs = [common, messaging, visionipc, 'avformat', 'avcodec', 'avutil', - 'yuv', 'pthread', 'zstd'] + 'yuv', 'OpenCL', 'pthread', 'zstd'] src = ['logger.cc', 'zstd_writer.cc', 'video_writer.cc', 'encoder/encoder.cc', 'encoder/v4l_encoder.cc', 'encoder/jpeg_encoder.cc'] if arch != "larch64": src += ['encoder/ffmpeg_encoder.cc'] if arch == "Darwin": + # fix OpenCL + del libs[libs.index('OpenCL')] + env['FRAMEWORKS'] = ['OpenCL'] # exclude v4l del src[src.index('encoder/v4l_encoder.cc')] diff --git a/third_party/opencl/include/CL/cl.h b/third_party/opencl/include/CL/cl.h new file mode 100644 index 000000000..0086319f5 --- /dev/null +++ b/third_party/opencl/include/CL/cl.h @@ -0,0 +1,1452 @@ +/******************************************************************************* + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + ******************************************************************************/ + +#ifndef __OPENCL_CL_H +#define __OPENCL_CL_H + +#ifdef __APPLE__ +#include +#else +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************/ + +typedef struct _cl_platform_id * cl_platform_id; +typedef struct _cl_device_id * cl_device_id; +typedef struct _cl_context * cl_context; +typedef struct _cl_command_queue * cl_command_queue; +typedef struct _cl_mem * cl_mem; +typedef struct _cl_program * cl_program; +typedef struct _cl_kernel * cl_kernel; +typedef struct _cl_event * cl_event; +typedef struct _cl_sampler * cl_sampler; + +typedef cl_uint cl_bool; /* WARNING! Unlike cl_ types in cl_platform.h, cl_bool is not guaranteed to be the same size as the bool in kernels. */ +typedef cl_ulong cl_bitfield; +typedef cl_bitfield cl_device_type; +typedef cl_uint cl_platform_info; +typedef cl_uint cl_device_info; +typedef cl_bitfield cl_device_fp_config; +typedef cl_uint cl_device_mem_cache_type; +typedef cl_uint cl_device_local_mem_type; +typedef cl_bitfield cl_device_exec_capabilities; +typedef cl_bitfield cl_device_svm_capabilities; +typedef cl_bitfield cl_command_queue_properties; +typedef intptr_t cl_device_partition_property; +typedef cl_bitfield cl_device_affinity_domain; + +typedef intptr_t cl_context_properties; +typedef cl_uint cl_context_info; +typedef cl_bitfield cl_queue_properties; +typedef cl_uint cl_command_queue_info; +typedef cl_uint cl_channel_order; +typedef cl_uint cl_channel_type; +typedef cl_bitfield cl_mem_flags; +typedef cl_bitfield cl_svm_mem_flags; +typedef cl_uint cl_mem_object_type; +typedef cl_uint cl_mem_info; +typedef cl_bitfield cl_mem_migration_flags; +typedef cl_uint cl_image_info; +typedef cl_uint cl_buffer_create_type; +typedef cl_uint cl_addressing_mode; +typedef cl_uint cl_filter_mode; +typedef cl_uint cl_sampler_info; +typedef cl_bitfield cl_map_flags; +typedef intptr_t cl_pipe_properties; +typedef cl_uint cl_pipe_info; +typedef cl_uint cl_program_info; +typedef cl_uint cl_program_build_info; +typedef cl_uint cl_program_binary_type; +typedef cl_int cl_build_status; +typedef cl_uint cl_kernel_info; +typedef cl_uint cl_kernel_arg_info; +typedef cl_uint cl_kernel_arg_address_qualifier; +typedef cl_uint cl_kernel_arg_access_qualifier; +typedef cl_bitfield cl_kernel_arg_type_qualifier; +typedef cl_uint cl_kernel_work_group_info; +typedef cl_uint cl_kernel_sub_group_info; +typedef cl_uint cl_event_info; +typedef cl_uint cl_command_type; +typedef cl_uint cl_profiling_info; +typedef cl_bitfield cl_sampler_properties; +typedef cl_uint cl_kernel_exec_info; + +typedef struct _cl_image_format { + cl_channel_order image_channel_order; + cl_channel_type image_channel_data_type; +} cl_image_format; + +typedef struct _cl_image_desc { + cl_mem_object_type image_type; + size_t image_width; + size_t image_height; + size_t image_depth; + size_t image_array_size; + size_t image_row_pitch; + size_t image_slice_pitch; + cl_uint num_mip_levels; + cl_uint num_samples; +#ifdef __GNUC__ + __extension__ /* Prevents warnings about anonymous union in -pedantic builds */ +#endif + union { + cl_mem buffer; + cl_mem mem_object; + }; +} cl_image_desc; + +typedef struct _cl_buffer_region { + size_t origin; + size_t size; +} cl_buffer_region; + + +/******************************************************************************/ + +/* Error Codes */ +#define CL_SUCCESS 0 +#define CL_DEVICE_NOT_FOUND -1 +#define CL_DEVICE_NOT_AVAILABLE -2 +#define CL_COMPILER_NOT_AVAILABLE -3 +#define CL_MEM_OBJECT_ALLOCATION_FAILURE -4 +#define CL_OUT_OF_RESOURCES -5 +#define CL_OUT_OF_HOST_MEMORY -6 +#define CL_PROFILING_INFO_NOT_AVAILABLE -7 +#define CL_MEM_COPY_OVERLAP -8 +#define CL_IMAGE_FORMAT_MISMATCH -9 +#define CL_IMAGE_FORMAT_NOT_SUPPORTED -10 +#define CL_BUILD_PROGRAM_FAILURE -11 +#define CL_MAP_FAILURE -12 +#define CL_MISALIGNED_SUB_BUFFER_OFFSET -13 +#define CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST -14 +#define CL_COMPILE_PROGRAM_FAILURE -15 +#define CL_LINKER_NOT_AVAILABLE -16 +#define CL_LINK_PROGRAM_FAILURE -17 +#define CL_DEVICE_PARTITION_FAILED -18 +#define CL_KERNEL_ARG_INFO_NOT_AVAILABLE -19 + +#define CL_INVALID_VALUE -30 +#define CL_INVALID_DEVICE_TYPE -31 +#define CL_INVALID_PLATFORM -32 +#define CL_INVALID_DEVICE -33 +#define CL_INVALID_CONTEXT -34 +#define CL_INVALID_QUEUE_PROPERTIES -35 +#define CL_INVALID_COMMAND_QUEUE -36 +#define CL_INVALID_HOST_PTR -37 +#define CL_INVALID_MEM_OBJECT -38 +#define CL_INVALID_IMAGE_FORMAT_DESCRIPTOR -39 +#define CL_INVALID_IMAGE_SIZE -40 +#define CL_INVALID_SAMPLER -41 +#define CL_INVALID_BINARY -42 +#define CL_INVALID_BUILD_OPTIONS -43 +#define CL_INVALID_PROGRAM -44 +#define CL_INVALID_PROGRAM_EXECUTABLE -45 +#define CL_INVALID_KERNEL_NAME -46 +#define CL_INVALID_KERNEL_DEFINITION -47 +#define CL_INVALID_KERNEL -48 +#define CL_INVALID_ARG_INDEX -49 +#define CL_INVALID_ARG_VALUE -50 +#define CL_INVALID_ARG_SIZE -51 +#define CL_INVALID_KERNEL_ARGS -52 +#define CL_INVALID_WORK_DIMENSION -53 +#define CL_INVALID_WORK_GROUP_SIZE -54 +#define CL_INVALID_WORK_ITEM_SIZE -55 +#define CL_INVALID_GLOBAL_OFFSET -56 +#define CL_INVALID_EVENT_WAIT_LIST -57 +#define CL_INVALID_EVENT -58 +#define CL_INVALID_OPERATION -59 +#define CL_INVALID_GL_OBJECT -60 +#define CL_INVALID_BUFFER_SIZE -61 +#define CL_INVALID_MIP_LEVEL -62 +#define CL_INVALID_GLOBAL_WORK_SIZE -63 +#define CL_INVALID_PROPERTY -64 +#define CL_INVALID_IMAGE_DESCRIPTOR -65 +#define CL_INVALID_COMPILER_OPTIONS -66 +#define CL_INVALID_LINKER_OPTIONS -67 +#define CL_INVALID_DEVICE_PARTITION_COUNT -68 +#define CL_INVALID_PIPE_SIZE -69 +#define CL_INVALID_DEVICE_QUEUE -70 + +/* OpenCL Version */ +#define CL_VERSION_1_0 1 +#define CL_VERSION_1_1 1 +#define CL_VERSION_1_2 1 +#define CL_VERSION_2_0 1 +#define CL_VERSION_2_1 1 + +/* cl_bool */ +#define CL_FALSE 0 +#define CL_TRUE 1 +#define CL_BLOCKING CL_TRUE +#define CL_NON_BLOCKING CL_FALSE + +/* cl_platform_info */ +#define CL_PLATFORM_PROFILE 0x0900 +#define CL_PLATFORM_VERSION 0x0901 +#define CL_PLATFORM_NAME 0x0902 +#define CL_PLATFORM_VENDOR 0x0903 +#define CL_PLATFORM_EXTENSIONS 0x0904 +#define CL_PLATFORM_HOST_TIMER_RESOLUTION 0x0905 + +/* cl_device_type - bitfield */ +#define CL_DEVICE_TYPE_DEFAULT (1 << 0) +#define CL_DEVICE_TYPE_CPU (1 << 1) +#define CL_DEVICE_TYPE_GPU (1 << 2) +#define CL_DEVICE_TYPE_ACCELERATOR (1 << 3) +#define CL_DEVICE_TYPE_CUSTOM (1 << 4) +#define CL_DEVICE_TYPE_ALL 0xFFFFFFFF + +/* cl_device_info */ +#define CL_DEVICE_TYPE 0x1000 +#define CL_DEVICE_VENDOR_ID 0x1001 +#define CL_DEVICE_MAX_COMPUTE_UNITS 0x1002 +#define CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS 0x1003 +#define CL_DEVICE_MAX_WORK_GROUP_SIZE 0x1004 +#define CL_DEVICE_MAX_WORK_ITEM_SIZES 0x1005 +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR 0x1006 +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT 0x1007 +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT 0x1008 +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG 0x1009 +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT 0x100A +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE 0x100B +#define CL_DEVICE_MAX_CLOCK_FREQUENCY 0x100C +#define CL_DEVICE_ADDRESS_BITS 0x100D +#define CL_DEVICE_MAX_READ_IMAGE_ARGS 0x100E +#define CL_DEVICE_MAX_WRITE_IMAGE_ARGS 0x100F +#define CL_DEVICE_MAX_MEM_ALLOC_SIZE 0x1010 +#define CL_DEVICE_IMAGE2D_MAX_WIDTH 0x1011 +#define CL_DEVICE_IMAGE2D_MAX_HEIGHT 0x1012 +#define CL_DEVICE_IMAGE3D_MAX_WIDTH 0x1013 +#define CL_DEVICE_IMAGE3D_MAX_HEIGHT 0x1014 +#define CL_DEVICE_IMAGE3D_MAX_DEPTH 0x1015 +#define CL_DEVICE_IMAGE_SUPPORT 0x1016 +#define CL_DEVICE_MAX_PARAMETER_SIZE 0x1017 +#define CL_DEVICE_MAX_SAMPLERS 0x1018 +#define CL_DEVICE_MEM_BASE_ADDR_ALIGN 0x1019 +#define CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 0x101A +#define CL_DEVICE_SINGLE_FP_CONFIG 0x101B +#define CL_DEVICE_GLOBAL_MEM_CACHE_TYPE 0x101C +#define CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 0x101D +#define CL_DEVICE_GLOBAL_MEM_CACHE_SIZE 0x101E +#define CL_DEVICE_GLOBAL_MEM_SIZE 0x101F +#define CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE 0x1020 +#define CL_DEVICE_MAX_CONSTANT_ARGS 0x1021 +#define CL_DEVICE_LOCAL_MEM_TYPE 0x1022 +#define CL_DEVICE_LOCAL_MEM_SIZE 0x1023 +#define CL_DEVICE_ERROR_CORRECTION_SUPPORT 0x1024 +#define CL_DEVICE_PROFILING_TIMER_RESOLUTION 0x1025 +#define CL_DEVICE_ENDIAN_LITTLE 0x1026 +#define CL_DEVICE_AVAILABLE 0x1027 +#define CL_DEVICE_COMPILER_AVAILABLE 0x1028 +#define CL_DEVICE_EXECUTION_CAPABILITIES 0x1029 +#define CL_DEVICE_QUEUE_PROPERTIES 0x102A /* deprecated */ +#define CL_DEVICE_QUEUE_ON_HOST_PROPERTIES 0x102A +#define CL_DEVICE_NAME 0x102B +#define CL_DEVICE_VENDOR 0x102C +#define CL_DRIVER_VERSION 0x102D +#define CL_DEVICE_PROFILE 0x102E +#define CL_DEVICE_VERSION 0x102F +#define CL_DEVICE_EXTENSIONS 0x1030 +#define CL_DEVICE_PLATFORM 0x1031 +#define CL_DEVICE_DOUBLE_FP_CONFIG 0x1032 +/* 0x1033 reserved for CL_DEVICE_HALF_FP_CONFIG */ +#define CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 0x1034 +#define CL_DEVICE_HOST_UNIFIED_MEMORY 0x1035 /* deprecated */ +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 0x1036 +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 0x1037 +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 0x1038 +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 0x1039 +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 0x103A +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 0x103B +#define CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 0x103C +#define CL_DEVICE_OPENCL_C_VERSION 0x103D +#define CL_DEVICE_LINKER_AVAILABLE 0x103E +#define CL_DEVICE_BUILT_IN_KERNELS 0x103F +#define CL_DEVICE_IMAGE_MAX_BUFFER_SIZE 0x1040 +#define CL_DEVICE_IMAGE_MAX_ARRAY_SIZE 0x1041 +#define CL_DEVICE_PARENT_DEVICE 0x1042 +#define CL_DEVICE_PARTITION_MAX_SUB_DEVICES 0x1043 +#define CL_DEVICE_PARTITION_PROPERTIES 0x1044 +#define CL_DEVICE_PARTITION_AFFINITY_DOMAIN 0x1045 +#define CL_DEVICE_PARTITION_TYPE 0x1046 +#define CL_DEVICE_REFERENCE_COUNT 0x1047 +#define CL_DEVICE_PREFERRED_INTEROP_USER_SYNC 0x1048 +#define CL_DEVICE_PRINTF_BUFFER_SIZE 0x1049 +#define CL_DEVICE_IMAGE_PITCH_ALIGNMENT 0x104A +#define CL_DEVICE_IMAGE_BASE_ADDRESS_ALIGNMENT 0x104B +#define CL_DEVICE_MAX_READ_WRITE_IMAGE_ARGS 0x104C +#define CL_DEVICE_MAX_GLOBAL_VARIABLE_SIZE 0x104D +#define CL_DEVICE_QUEUE_ON_DEVICE_PROPERTIES 0x104E +#define CL_DEVICE_QUEUE_ON_DEVICE_PREFERRED_SIZE 0x104F +#define CL_DEVICE_QUEUE_ON_DEVICE_MAX_SIZE 0x1050 +#define CL_DEVICE_MAX_ON_DEVICE_QUEUES 0x1051 +#define CL_DEVICE_MAX_ON_DEVICE_EVENTS 0x1052 +#define CL_DEVICE_SVM_CAPABILITIES 0x1053 +#define CL_DEVICE_GLOBAL_VARIABLE_PREFERRED_TOTAL_SIZE 0x1054 +#define CL_DEVICE_MAX_PIPE_ARGS 0x1055 +#define CL_DEVICE_PIPE_MAX_ACTIVE_RESERVATIONS 0x1056 +#define CL_DEVICE_PIPE_MAX_PACKET_SIZE 0x1057 +#define CL_DEVICE_PREFERRED_PLATFORM_ATOMIC_ALIGNMENT 0x1058 +#define CL_DEVICE_PREFERRED_GLOBAL_ATOMIC_ALIGNMENT 0x1059 +#define CL_DEVICE_PREFERRED_LOCAL_ATOMIC_ALIGNMENT 0x105A +#define CL_DEVICE_IL_VERSION 0x105B +#define CL_DEVICE_MAX_NUM_SUB_GROUPS 0x105C +#define CL_DEVICE_SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS 0x105D + +/* cl_device_fp_config - bitfield */ +#define CL_FP_DENORM (1 << 0) +#define CL_FP_INF_NAN (1 << 1) +#define CL_FP_ROUND_TO_NEAREST (1 << 2) +#define CL_FP_ROUND_TO_ZERO (1 << 3) +#define CL_FP_ROUND_TO_INF (1 << 4) +#define CL_FP_FMA (1 << 5) +#define CL_FP_SOFT_FLOAT (1 << 6) +#define CL_FP_CORRECTLY_ROUNDED_DIVIDE_SQRT (1 << 7) + +/* cl_device_mem_cache_type */ +#define CL_NONE 0x0 +#define CL_READ_ONLY_CACHE 0x1 +#define CL_READ_WRITE_CACHE 0x2 + +/* cl_device_local_mem_type */ +#define CL_LOCAL 0x1 +#define CL_GLOBAL 0x2 + +/* cl_device_exec_capabilities - bitfield */ +#define CL_EXEC_KERNEL (1 << 0) +#define CL_EXEC_NATIVE_KERNEL (1 << 1) + +/* cl_command_queue_properties - bitfield */ +#define CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE (1 << 0) +#define CL_QUEUE_PROFILING_ENABLE (1 << 1) +#define CL_QUEUE_ON_DEVICE (1 << 2) +#define CL_QUEUE_ON_DEVICE_DEFAULT (1 << 3) + +/* cl_context_info */ +#define CL_CONTEXT_REFERENCE_COUNT 0x1080 +#define CL_CONTEXT_DEVICES 0x1081 +#define CL_CONTEXT_PROPERTIES 0x1082 +#define CL_CONTEXT_NUM_DEVICES 0x1083 + +/* cl_context_properties */ +#define CL_CONTEXT_PLATFORM 0x1084 +#define CL_CONTEXT_INTEROP_USER_SYNC 0x1085 + +/* cl_device_partition_property */ +#define CL_DEVICE_PARTITION_EQUALLY 0x1086 +#define CL_DEVICE_PARTITION_BY_COUNTS 0x1087 +#define CL_DEVICE_PARTITION_BY_COUNTS_LIST_END 0x0 +#define CL_DEVICE_PARTITION_BY_AFFINITY_DOMAIN 0x1088 + +/* cl_device_affinity_domain */ +#define CL_DEVICE_AFFINITY_DOMAIN_NUMA (1 << 0) +#define CL_DEVICE_AFFINITY_DOMAIN_L4_CACHE (1 << 1) +#define CL_DEVICE_AFFINITY_DOMAIN_L3_CACHE (1 << 2) +#define CL_DEVICE_AFFINITY_DOMAIN_L2_CACHE (1 << 3) +#define CL_DEVICE_AFFINITY_DOMAIN_L1_CACHE (1 << 4) +#define CL_DEVICE_AFFINITY_DOMAIN_NEXT_PARTITIONABLE (1 << 5) + +/* cl_device_svm_capabilities */ +#define CL_DEVICE_SVM_COARSE_GRAIN_BUFFER (1 << 0) +#define CL_DEVICE_SVM_FINE_GRAIN_BUFFER (1 << 1) +#define CL_DEVICE_SVM_FINE_GRAIN_SYSTEM (1 << 2) +#define CL_DEVICE_SVM_ATOMICS (1 << 3) + +/* cl_command_queue_info */ +#define CL_QUEUE_CONTEXT 0x1090 +#define CL_QUEUE_DEVICE 0x1091 +#define CL_QUEUE_REFERENCE_COUNT 0x1092 +#define CL_QUEUE_PROPERTIES 0x1093 +#define CL_QUEUE_SIZE 0x1094 +#define CL_QUEUE_DEVICE_DEFAULT 0x1095 + +/* cl_mem_flags and cl_svm_mem_flags - bitfield */ +#define CL_MEM_READ_WRITE (1 << 0) +#define CL_MEM_WRITE_ONLY (1 << 1) +#define CL_MEM_READ_ONLY (1 << 2) +#define CL_MEM_USE_HOST_PTR (1 << 3) +#define CL_MEM_ALLOC_HOST_PTR (1 << 4) +#define CL_MEM_COPY_HOST_PTR (1 << 5) +/* reserved (1 << 6) */ +#define CL_MEM_HOST_WRITE_ONLY (1 << 7) +#define CL_MEM_HOST_READ_ONLY (1 << 8) +#define CL_MEM_HOST_NO_ACCESS (1 << 9) +#define CL_MEM_SVM_FINE_GRAIN_BUFFER (1 << 10) /* used by cl_svm_mem_flags only */ +#define CL_MEM_SVM_ATOMICS (1 << 11) /* used by cl_svm_mem_flags only */ +#define CL_MEM_KERNEL_READ_AND_WRITE (1 << 12) + +/* cl_mem_migration_flags - bitfield */ +#define CL_MIGRATE_MEM_OBJECT_HOST (1 << 0) +#define CL_MIGRATE_MEM_OBJECT_CONTENT_UNDEFINED (1 << 1) + +/* cl_channel_order */ +#define CL_R 0x10B0 +#define CL_A 0x10B1 +#define CL_RG 0x10B2 +#define CL_RA 0x10B3 +#define CL_RGB 0x10B4 +#define CL_RGBA 0x10B5 +#define CL_BGRA 0x10B6 +#define CL_ARGB 0x10B7 +#define CL_INTENSITY 0x10B8 +#define CL_LUMINANCE 0x10B9 +#define CL_Rx 0x10BA +#define CL_RGx 0x10BB +#define CL_RGBx 0x10BC +#define CL_DEPTH 0x10BD +#define CL_DEPTH_STENCIL 0x10BE +#define CL_sRGB 0x10BF +#define CL_sRGBx 0x10C0 +#define CL_sRGBA 0x10C1 +#define CL_sBGRA 0x10C2 +#define CL_ABGR 0x10C3 + +/* cl_channel_type */ +#define CL_SNORM_INT8 0x10D0 +#define CL_SNORM_INT16 0x10D1 +#define CL_UNORM_INT8 0x10D2 +#define CL_UNORM_INT16 0x10D3 +#define CL_UNORM_SHORT_565 0x10D4 +#define CL_UNORM_SHORT_555 0x10D5 +#define CL_UNORM_INT_101010 0x10D6 +#define CL_SIGNED_INT8 0x10D7 +#define CL_SIGNED_INT16 0x10D8 +#define CL_SIGNED_INT32 0x10D9 +#define CL_UNSIGNED_INT8 0x10DA +#define CL_UNSIGNED_INT16 0x10DB +#define CL_UNSIGNED_INT32 0x10DC +#define CL_HALF_FLOAT 0x10DD +#define CL_FLOAT 0x10DE +#define CL_UNORM_INT24 0x10DF +#define CL_UNORM_INT_101010_2 0x10E0 + +/* cl_mem_object_type */ +#define CL_MEM_OBJECT_BUFFER 0x10F0 +#define CL_MEM_OBJECT_IMAGE2D 0x10F1 +#define CL_MEM_OBJECT_IMAGE3D 0x10F2 +#define CL_MEM_OBJECT_IMAGE2D_ARRAY 0x10F3 +#define CL_MEM_OBJECT_IMAGE1D 0x10F4 +#define CL_MEM_OBJECT_IMAGE1D_ARRAY 0x10F5 +#define CL_MEM_OBJECT_IMAGE1D_BUFFER 0x10F6 +#define CL_MEM_OBJECT_PIPE 0x10F7 + +/* cl_mem_info */ +#define CL_MEM_TYPE 0x1100 +#define CL_MEM_FLAGS 0x1101 +#define CL_MEM_SIZE 0x1102 +#define CL_MEM_HOST_PTR 0x1103 +#define CL_MEM_MAP_COUNT 0x1104 +#define CL_MEM_REFERENCE_COUNT 0x1105 +#define CL_MEM_CONTEXT 0x1106 +#define CL_MEM_ASSOCIATED_MEMOBJECT 0x1107 +#define CL_MEM_OFFSET 0x1108 +#define CL_MEM_USES_SVM_POINTER 0x1109 + +/* cl_image_info */ +#define CL_IMAGE_FORMAT 0x1110 +#define CL_IMAGE_ELEMENT_SIZE 0x1111 +#define CL_IMAGE_ROW_PITCH 0x1112 +#define CL_IMAGE_SLICE_PITCH 0x1113 +#define CL_IMAGE_WIDTH 0x1114 +#define CL_IMAGE_HEIGHT 0x1115 +#define CL_IMAGE_DEPTH 0x1116 +#define CL_IMAGE_ARRAY_SIZE 0x1117 +#define CL_IMAGE_BUFFER 0x1118 +#define CL_IMAGE_NUM_MIP_LEVELS 0x1119 +#define CL_IMAGE_NUM_SAMPLES 0x111A + +/* cl_pipe_info */ +#define CL_PIPE_PACKET_SIZE 0x1120 +#define CL_PIPE_MAX_PACKETS 0x1121 + +/* cl_addressing_mode */ +#define CL_ADDRESS_NONE 0x1130 +#define CL_ADDRESS_CLAMP_TO_EDGE 0x1131 +#define CL_ADDRESS_CLAMP 0x1132 +#define CL_ADDRESS_REPEAT 0x1133 +#define CL_ADDRESS_MIRRORED_REPEAT 0x1134 + +/* cl_filter_mode */ +#define CL_FILTER_NEAREST 0x1140 +#define CL_FILTER_LINEAR 0x1141 + +/* cl_sampler_info */ +#define CL_SAMPLER_REFERENCE_COUNT 0x1150 +#define CL_SAMPLER_CONTEXT 0x1151 +#define CL_SAMPLER_NORMALIZED_COORDS 0x1152 +#define CL_SAMPLER_ADDRESSING_MODE 0x1153 +#define CL_SAMPLER_FILTER_MODE 0x1154 +#define CL_SAMPLER_MIP_FILTER_MODE 0x1155 +#define CL_SAMPLER_LOD_MIN 0x1156 +#define CL_SAMPLER_LOD_MAX 0x1157 + +/* cl_map_flags - bitfield */ +#define CL_MAP_READ (1 << 0) +#define CL_MAP_WRITE (1 << 1) +#define CL_MAP_WRITE_INVALIDATE_REGION (1 << 2) + +/* cl_program_info */ +#define CL_PROGRAM_REFERENCE_COUNT 0x1160 +#define CL_PROGRAM_CONTEXT 0x1161 +#define CL_PROGRAM_NUM_DEVICES 0x1162 +#define CL_PROGRAM_DEVICES 0x1163 +#define CL_PROGRAM_SOURCE 0x1164 +#define CL_PROGRAM_BINARY_SIZES 0x1165 +#define CL_PROGRAM_BINARIES 0x1166 +#define CL_PROGRAM_NUM_KERNELS 0x1167 +#define CL_PROGRAM_KERNEL_NAMES 0x1168 +#define CL_PROGRAM_IL 0x1169 + +/* cl_program_build_info */ +#define CL_PROGRAM_BUILD_STATUS 0x1181 +#define CL_PROGRAM_BUILD_OPTIONS 0x1182 +#define CL_PROGRAM_BUILD_LOG 0x1183 +#define CL_PROGRAM_BINARY_TYPE 0x1184 +#define CL_PROGRAM_BUILD_GLOBAL_VARIABLE_TOTAL_SIZE 0x1185 + +/* cl_program_binary_type */ +#define CL_PROGRAM_BINARY_TYPE_NONE 0x0 +#define CL_PROGRAM_BINARY_TYPE_COMPILED_OBJECT 0x1 +#define CL_PROGRAM_BINARY_TYPE_LIBRARY 0x2 +#define CL_PROGRAM_BINARY_TYPE_EXECUTABLE 0x4 + +/* cl_build_status */ +#define CL_BUILD_SUCCESS 0 +#define CL_BUILD_NONE -1 +#define CL_BUILD_ERROR -2 +#define CL_BUILD_IN_PROGRESS -3 + +/* cl_kernel_info */ +#define CL_KERNEL_FUNCTION_NAME 0x1190 +#define CL_KERNEL_NUM_ARGS 0x1191 +#define CL_KERNEL_REFERENCE_COUNT 0x1192 +#define CL_KERNEL_CONTEXT 0x1193 +#define CL_KERNEL_PROGRAM 0x1194 +#define CL_KERNEL_ATTRIBUTES 0x1195 +#define CL_KERNEL_MAX_NUM_SUB_GROUPS 0x11B9 +#define CL_KERNEL_COMPILE_NUM_SUB_GROUPS 0x11BA + +/* cl_kernel_arg_info */ +#define CL_KERNEL_ARG_ADDRESS_QUALIFIER 0x1196 +#define CL_KERNEL_ARG_ACCESS_QUALIFIER 0x1197 +#define CL_KERNEL_ARG_TYPE_NAME 0x1198 +#define CL_KERNEL_ARG_TYPE_QUALIFIER 0x1199 +#define CL_KERNEL_ARG_NAME 0x119A + +/* cl_kernel_arg_address_qualifier */ +#define CL_KERNEL_ARG_ADDRESS_GLOBAL 0x119B +#define CL_KERNEL_ARG_ADDRESS_LOCAL 0x119C +#define CL_KERNEL_ARG_ADDRESS_CONSTANT 0x119D +#define CL_KERNEL_ARG_ADDRESS_PRIVATE 0x119E + +/* cl_kernel_arg_access_qualifier */ +#define CL_KERNEL_ARG_ACCESS_READ_ONLY 0x11A0 +#define CL_KERNEL_ARG_ACCESS_WRITE_ONLY 0x11A1 +#define CL_KERNEL_ARG_ACCESS_READ_WRITE 0x11A2 +#define CL_KERNEL_ARG_ACCESS_NONE 0x11A3 + +/* cl_kernel_arg_type_qualifer */ +#define CL_KERNEL_ARG_TYPE_NONE 0 +#define CL_KERNEL_ARG_TYPE_CONST (1 << 0) +#define CL_KERNEL_ARG_TYPE_RESTRICT (1 << 1) +#define CL_KERNEL_ARG_TYPE_VOLATILE (1 << 2) +#define CL_KERNEL_ARG_TYPE_PIPE (1 << 3) + +/* cl_kernel_work_group_info */ +#define CL_KERNEL_WORK_GROUP_SIZE 0x11B0 +#define CL_KERNEL_COMPILE_WORK_GROUP_SIZE 0x11B1 +#define CL_KERNEL_LOCAL_MEM_SIZE 0x11B2 +#define CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE 0x11B3 +#define CL_KERNEL_PRIVATE_MEM_SIZE 0x11B4 +#define CL_KERNEL_GLOBAL_WORK_SIZE 0x11B5 + +/* cl_kernel_sub_group_info */ +#define CL_KERNEL_MAX_SUB_GROUP_SIZE_FOR_NDRANGE 0x2033 +#define CL_KERNEL_SUB_GROUP_COUNT_FOR_NDRANGE 0x2034 +#define CL_KERNEL_LOCAL_SIZE_FOR_SUB_GROUP_COUNT 0x11B8 + +/* cl_kernel_exec_info */ +#define CL_KERNEL_EXEC_INFO_SVM_PTRS 0x11B6 +#define CL_KERNEL_EXEC_INFO_SVM_FINE_GRAIN_SYSTEM 0x11B7 + +/* cl_event_info */ +#define CL_EVENT_COMMAND_QUEUE 0x11D0 +#define CL_EVENT_COMMAND_TYPE 0x11D1 +#define CL_EVENT_REFERENCE_COUNT 0x11D2 +#define CL_EVENT_COMMAND_EXECUTION_STATUS 0x11D3 +#define CL_EVENT_CONTEXT 0x11D4 + +/* cl_command_type */ +#define CL_COMMAND_NDRANGE_KERNEL 0x11F0 +#define CL_COMMAND_TASK 0x11F1 +#define CL_COMMAND_NATIVE_KERNEL 0x11F2 +#define CL_COMMAND_READ_BUFFER 0x11F3 +#define CL_COMMAND_WRITE_BUFFER 0x11F4 +#define CL_COMMAND_COPY_BUFFER 0x11F5 +#define CL_COMMAND_READ_IMAGE 0x11F6 +#define CL_COMMAND_WRITE_IMAGE 0x11F7 +#define CL_COMMAND_COPY_IMAGE 0x11F8 +#define CL_COMMAND_COPY_IMAGE_TO_BUFFER 0x11F9 +#define CL_COMMAND_COPY_BUFFER_TO_IMAGE 0x11FA +#define CL_COMMAND_MAP_BUFFER 0x11FB +#define CL_COMMAND_MAP_IMAGE 0x11FC +#define CL_COMMAND_UNMAP_MEM_OBJECT 0x11FD +#define CL_COMMAND_MARKER 0x11FE +#define CL_COMMAND_ACQUIRE_GL_OBJECTS 0x11FF +#define CL_COMMAND_RELEASE_GL_OBJECTS 0x1200 +#define CL_COMMAND_READ_BUFFER_RECT 0x1201 +#define CL_COMMAND_WRITE_BUFFER_RECT 0x1202 +#define CL_COMMAND_COPY_BUFFER_RECT 0x1203 +#define CL_COMMAND_USER 0x1204 +#define CL_COMMAND_BARRIER 0x1205 +#define CL_COMMAND_MIGRATE_MEM_OBJECTS 0x1206 +#define CL_COMMAND_FILL_BUFFER 0x1207 +#define CL_COMMAND_FILL_IMAGE 0x1208 +#define CL_COMMAND_SVM_FREE 0x1209 +#define CL_COMMAND_SVM_MEMCPY 0x120A +#define CL_COMMAND_SVM_MEMFILL 0x120B +#define CL_COMMAND_SVM_MAP 0x120C +#define CL_COMMAND_SVM_UNMAP 0x120D + +/* command execution status */ +#define CL_COMPLETE 0x0 +#define CL_RUNNING 0x1 +#define CL_SUBMITTED 0x2 +#define CL_QUEUED 0x3 + +/* cl_buffer_create_type */ +#define CL_BUFFER_CREATE_TYPE_REGION 0x1220 + +/* cl_profiling_info */ +#define CL_PROFILING_COMMAND_QUEUED 0x1280 +#define CL_PROFILING_COMMAND_SUBMIT 0x1281 +#define CL_PROFILING_COMMAND_START 0x1282 +#define CL_PROFILING_COMMAND_END 0x1283 +#define CL_PROFILING_COMMAND_COMPLETE 0x1284 + +/********************************************************************************************************/ + +/* Platform API */ +extern CL_API_ENTRY cl_int CL_API_CALL +clGetPlatformIDs(cl_uint /* num_entries */, + cl_platform_id * /* platforms */, + cl_uint * /* num_platforms */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetPlatformInfo(cl_platform_id /* platform */, + cl_platform_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Device APIs */ +extern CL_API_ENTRY cl_int CL_API_CALL +clGetDeviceIDs(cl_platform_id /* platform */, + cl_device_type /* device_type */, + cl_uint /* num_entries */, + cl_device_id * /* devices */, + cl_uint * /* num_devices */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetDeviceInfo(cl_device_id /* device */, + cl_device_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clCreateSubDevices(cl_device_id /* in_device */, + const cl_device_partition_property * /* properties */, + cl_uint /* num_devices */, + cl_device_id * /* out_devices */, + cl_uint * /* num_devices_ret */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainDevice(cl_device_id /* device */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseDevice(cl_device_id /* device */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetDefaultDeviceCommandQueue(cl_context /* context */, + cl_device_id /* device */, + cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_2_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetDeviceAndHostTimer(cl_device_id /* device */, + cl_ulong* /* device_timestamp */, + cl_ulong* /* host_timestamp */) CL_API_SUFFIX__VERSION_2_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetHostTimer(cl_device_id /* device */, + cl_ulong * /* host_timestamp */) CL_API_SUFFIX__VERSION_2_1; + + +/* Context APIs */ +extern CL_API_ENTRY cl_context CL_API_CALL +clCreateContext(const cl_context_properties * /* properties */, + cl_uint /* num_devices */, + const cl_device_id * /* devices */, + void (CL_CALLBACK * /* pfn_notify */)(const char *, const void *, size_t, void *), + void * /* user_data */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_context CL_API_CALL +clCreateContextFromType(const cl_context_properties * /* properties */, + cl_device_type /* device_type */, + void (CL_CALLBACK * /* pfn_notify*/ )(const char *, const void *, size_t, void *), + void * /* user_data */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainContext(cl_context /* context */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseContext(cl_context /* context */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetContextInfo(cl_context /* context */, + cl_context_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Command Queue APIs */ +extern CL_API_ENTRY cl_command_queue CL_API_CALL +clCreateCommandQueueWithProperties(cl_context /* context */, + cl_device_id /* device */, + const cl_queue_properties * /* properties */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainCommandQueue(cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseCommandQueue(cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetCommandQueueInfo(cl_command_queue /* command_queue */, + cl_command_queue_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Memory Object APIs */ +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateBuffer(cl_context /* context */, + cl_mem_flags /* flags */, + size_t /* size */, + void * /* host_ptr */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateSubBuffer(cl_mem /* buffer */, + cl_mem_flags /* flags */, + cl_buffer_create_type /* buffer_create_type */, + const void * /* buffer_create_info */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateImage(cl_context /* context */, + cl_mem_flags /* flags */, + const cl_image_format * /* image_format */, + const cl_image_desc * /* image_desc */, + void * /* host_ptr */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreatePipe(cl_context /* context */, + cl_mem_flags /* flags */, + cl_uint /* pipe_packet_size */, + cl_uint /* pipe_max_packets */, + const cl_pipe_properties * /* properties */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainMemObject(cl_mem /* memobj */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseMemObject(cl_mem /* memobj */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetSupportedImageFormats(cl_context /* context */, + cl_mem_flags /* flags */, + cl_mem_object_type /* image_type */, + cl_uint /* num_entries */, + cl_image_format * /* image_formats */, + cl_uint * /* num_image_formats */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetMemObjectInfo(cl_mem /* memobj */, + cl_mem_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetImageInfo(cl_mem /* image */, + cl_image_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetPipeInfo(cl_mem /* pipe */, + cl_pipe_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_2_0; + + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetMemObjectDestructorCallback(cl_mem /* memobj */, + void (CL_CALLBACK * /*pfn_notify*/)( cl_mem /* memobj */, void* /*user_data*/), + void * /*user_data */ ) CL_API_SUFFIX__VERSION_1_1; + +/* SVM Allocation APIs */ +extern CL_API_ENTRY void * CL_API_CALL +clSVMAlloc(cl_context /* context */, + cl_svm_mem_flags /* flags */, + size_t /* size */, + cl_uint /* alignment */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY void CL_API_CALL +clSVMFree(cl_context /* context */, + void * /* svm_pointer */) CL_API_SUFFIX__VERSION_2_0; + +/* Sampler APIs */ +extern CL_API_ENTRY cl_sampler CL_API_CALL +clCreateSamplerWithProperties(cl_context /* context */, + const cl_sampler_properties * /* normalized_coords */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainSampler(cl_sampler /* sampler */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseSampler(cl_sampler /* sampler */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetSamplerInfo(cl_sampler /* sampler */, + cl_sampler_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Program Object APIs */ +extern CL_API_ENTRY cl_program CL_API_CALL +clCreateProgramWithSource(cl_context /* context */, + cl_uint /* count */, + const char ** /* strings */, + const size_t * /* lengths */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_program CL_API_CALL +clCreateProgramWithBinary(cl_context /* context */, + cl_uint /* num_devices */, + const cl_device_id * /* device_list */, + const size_t * /* lengths */, + const unsigned char ** /* binaries */, + cl_int * /* binary_status */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_program CL_API_CALL +clCreateProgramWithBuiltInKernels(cl_context /* context */, + cl_uint /* num_devices */, + const cl_device_id * /* device_list */, + const char * /* kernel_names */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_program CL_API_CALL +clCreateProgramWithIL(cl_context /* context */, + const void* /* il */, + size_t /* length */, + cl_int* /* errcode_ret */) CL_API_SUFFIX__VERSION_2_1; + + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainProgram(cl_program /* program */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseProgram(cl_program /* program */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clBuildProgram(cl_program /* program */, + cl_uint /* num_devices */, + const cl_device_id * /* device_list */, + const char * /* options */, + void (CL_CALLBACK * /* pfn_notify */)(cl_program /* program */, void * /* user_data */), + void * /* user_data */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clCompileProgram(cl_program /* program */, + cl_uint /* num_devices */, + const cl_device_id * /* device_list */, + const char * /* options */, + cl_uint /* num_input_headers */, + const cl_program * /* input_headers */, + const char ** /* header_include_names */, + void (CL_CALLBACK * /* pfn_notify */)(cl_program /* program */, void * /* user_data */), + void * /* user_data */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_program CL_API_CALL +clLinkProgram(cl_context /* context */, + cl_uint /* num_devices */, + const cl_device_id * /* device_list */, + const char * /* options */, + cl_uint /* num_input_programs */, + const cl_program * /* input_programs */, + void (CL_CALLBACK * /* pfn_notify */)(cl_program /* program */, void * /* user_data */), + void * /* user_data */, + cl_int * /* errcode_ret */ ) CL_API_SUFFIX__VERSION_1_2; + + +extern CL_API_ENTRY cl_int CL_API_CALL +clUnloadPlatformCompiler(cl_platform_id /* platform */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetProgramInfo(cl_program /* program */, + cl_program_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetProgramBuildInfo(cl_program /* program */, + cl_device_id /* device */, + cl_program_build_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Kernel Object APIs */ +extern CL_API_ENTRY cl_kernel CL_API_CALL +clCreateKernel(cl_program /* program */, + const char * /* kernel_name */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clCreateKernelsInProgram(cl_program /* program */, + cl_uint /* num_kernels */, + cl_kernel * /* kernels */, + cl_uint * /* num_kernels_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_kernel CL_API_CALL +clCloneKernel(cl_kernel /* source_kernel */, + cl_int* /* errcode_ret */) CL_API_SUFFIX__VERSION_2_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainKernel(cl_kernel /* kernel */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseKernel(cl_kernel /* kernel */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetKernelArg(cl_kernel /* kernel */, + cl_uint /* arg_index */, + size_t /* arg_size */, + const void * /* arg_value */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetKernelArgSVMPointer(cl_kernel /* kernel */, + cl_uint /* arg_index */, + const void * /* arg_value */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetKernelExecInfo(cl_kernel /* kernel */, + cl_kernel_exec_info /* param_name */, + size_t /* param_value_size */, + const void * /* param_value */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetKernelInfo(cl_kernel /* kernel */, + cl_kernel_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetKernelArgInfo(cl_kernel /* kernel */, + cl_uint /* arg_indx */, + cl_kernel_arg_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetKernelWorkGroupInfo(cl_kernel /* kernel */, + cl_device_id /* device */, + cl_kernel_work_group_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetKernelSubGroupInfo(cl_kernel /* kernel */, + cl_device_id /* device */, + cl_kernel_sub_group_info /* param_name */, + size_t /* input_value_size */, + const void* /*input_value */, + size_t /* param_value_size */, + void* /* param_value */, + size_t* /* param_value_size_ret */ ) CL_API_SUFFIX__VERSION_2_1; + + +/* Event Object APIs */ +extern CL_API_ENTRY cl_int CL_API_CALL +clWaitForEvents(cl_uint /* num_events */, + const cl_event * /* event_list */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetEventInfo(cl_event /* event */, + cl_event_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_event CL_API_CALL +clCreateUserEvent(cl_context /* context */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clRetainEvent(cl_event /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clReleaseEvent(cl_event /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetUserEventStatus(cl_event /* event */, + cl_int /* execution_status */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetEventCallback( cl_event /* event */, + cl_int /* command_exec_callback_type */, + void (CL_CALLBACK * /* pfn_notify */)(cl_event, cl_int, void *), + void * /* user_data */) CL_API_SUFFIX__VERSION_1_1; + +/* Profiling APIs */ +extern CL_API_ENTRY cl_int CL_API_CALL +clGetEventProfilingInfo(cl_event /* event */, + cl_profiling_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +/* Flush and Finish APIs */ +extern CL_API_ENTRY cl_int CL_API_CALL +clFlush(cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clFinish(cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_1_0; + +/* Enqueued Commands APIs */ +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueReadBuffer(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + cl_bool /* blocking_read */, + size_t /* offset */, + size_t /* size */, + void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueReadBufferRect(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + cl_bool /* blocking_read */, + const size_t * /* buffer_offset */, + const size_t * /* host_offset */, + const size_t * /* region */, + size_t /* buffer_row_pitch */, + size_t /* buffer_slice_pitch */, + size_t /* host_row_pitch */, + size_t /* host_slice_pitch */, + void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueWriteBuffer(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + cl_bool /* blocking_write */, + size_t /* offset */, + size_t /* size */, + const void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueWriteBufferRect(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + cl_bool /* blocking_write */, + const size_t * /* buffer_offset */, + const size_t * /* host_offset */, + const size_t * /* region */, + size_t /* buffer_row_pitch */, + size_t /* buffer_slice_pitch */, + size_t /* host_row_pitch */, + size_t /* host_slice_pitch */, + const void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueFillBuffer(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + const void * /* pattern */, + size_t /* pattern_size */, + size_t /* offset */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueCopyBuffer(cl_command_queue /* command_queue */, + cl_mem /* src_buffer */, + cl_mem /* dst_buffer */, + size_t /* src_offset */, + size_t /* dst_offset */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueCopyBufferRect(cl_command_queue /* command_queue */, + cl_mem /* src_buffer */, + cl_mem /* dst_buffer */, + const size_t * /* src_origin */, + const size_t * /* dst_origin */, + const size_t * /* region */, + size_t /* src_row_pitch */, + size_t /* src_slice_pitch */, + size_t /* dst_row_pitch */, + size_t /* dst_slice_pitch */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_1; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueReadImage(cl_command_queue /* command_queue */, + cl_mem /* image */, + cl_bool /* blocking_read */, + const size_t * /* origin[3] */, + const size_t * /* region[3] */, + size_t /* row_pitch */, + size_t /* slice_pitch */, + void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueWriteImage(cl_command_queue /* command_queue */, + cl_mem /* image */, + cl_bool /* blocking_write */, + const size_t * /* origin[3] */, + const size_t * /* region[3] */, + size_t /* input_row_pitch */, + size_t /* input_slice_pitch */, + const void * /* ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueFillImage(cl_command_queue /* command_queue */, + cl_mem /* image */, + const void * /* fill_color */, + const size_t * /* origin[3] */, + const size_t * /* region[3] */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueCopyImage(cl_command_queue /* command_queue */, + cl_mem /* src_image */, + cl_mem /* dst_image */, + const size_t * /* src_origin[3] */, + const size_t * /* dst_origin[3] */, + const size_t * /* region[3] */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueCopyImageToBuffer(cl_command_queue /* command_queue */, + cl_mem /* src_image */, + cl_mem /* dst_buffer */, + const size_t * /* src_origin[3] */, + const size_t * /* region[3] */, + size_t /* dst_offset */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueCopyBufferToImage(cl_command_queue /* command_queue */, + cl_mem /* src_buffer */, + cl_mem /* dst_image */, + size_t /* src_offset */, + const size_t * /* dst_origin[3] */, + const size_t * /* region[3] */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY void * CL_API_CALL +clEnqueueMapBuffer(cl_command_queue /* command_queue */, + cl_mem /* buffer */, + cl_bool /* blocking_map */, + cl_map_flags /* map_flags */, + size_t /* offset */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY void * CL_API_CALL +clEnqueueMapImage(cl_command_queue /* command_queue */, + cl_mem /* image */, + cl_bool /* blocking_map */, + cl_map_flags /* map_flags */, + const size_t * /* origin[3] */, + const size_t * /* region[3] */, + size_t * /* image_row_pitch */, + size_t * /* image_slice_pitch */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueUnmapMemObject(cl_command_queue /* command_queue */, + cl_mem /* memobj */, + void * /* mapped_ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueMigrateMemObjects(cl_command_queue /* command_queue */, + cl_uint /* num_mem_objects */, + const cl_mem * /* mem_objects */, + cl_mem_migration_flags /* flags */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueNDRangeKernel(cl_command_queue /* command_queue */, + cl_kernel /* kernel */, + cl_uint /* work_dim */, + const size_t * /* global_work_offset */, + const size_t * /* global_work_size */, + const size_t * /* local_work_size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueNativeKernel(cl_command_queue /* command_queue */, + void (CL_CALLBACK * /*user_func*/)(void *), + void * /* args */, + size_t /* cb_args */, + cl_uint /* num_mem_objects */, + const cl_mem * /* mem_list */, + const void ** /* args_mem_loc */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueMarkerWithWaitList(cl_command_queue /* command_queue */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueBarrierWithWaitList(cl_command_queue /* command_queue */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMFree(cl_command_queue /* command_queue */, + cl_uint /* num_svm_pointers */, + void *[] /* svm_pointers[] */, + void (CL_CALLBACK * /*pfn_free_func*/)(cl_command_queue /* queue */, + cl_uint /* num_svm_pointers */, + void *[] /* svm_pointers[] */, + void * /* user_data */), + void * /* user_data */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMMemcpy(cl_command_queue /* command_queue */, + cl_bool /* blocking_copy */, + void * /* dst_ptr */, + const void * /* src_ptr */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMMemFill(cl_command_queue /* command_queue */, + void * /* svm_ptr */, + const void * /* pattern */, + size_t /* pattern_size */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMMap(cl_command_queue /* command_queue */, + cl_bool /* blocking_map */, + cl_map_flags /* flags */, + void * /* svm_ptr */, + size_t /* size */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMUnmap(cl_command_queue /* command_queue */, + void * /* svm_ptr */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueSVMMigrateMem(cl_command_queue /* command_queue */, + cl_uint /* num_svm_pointers */, + const void ** /* svm_pointers */, + const size_t * /* sizes */, + cl_mem_migration_flags /* flags */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_2_1; + + +/* Extension function access + * + * Returns the extension function address for the given function name, + * or NULL if a valid function can not be found. The client must + * check to make sure the address is not NULL, before using or + * calling the returned function address. + */ +extern CL_API_ENTRY void * CL_API_CALL +clGetExtensionFunctionAddressForPlatform(cl_platform_id /* platform */, + const char * /* func_name */) CL_API_SUFFIX__VERSION_1_2; + + +/* Deprecated OpenCL 1.1 APIs */ +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL +clCreateImage2D(cl_context /* context */, + cl_mem_flags /* flags */, + const cl_image_format * /* image_format */, + size_t /* image_width */, + size_t /* image_height */, + size_t /* image_row_pitch */, + void * /* host_ptr */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL +clCreateImage3D(cl_context /* context */, + cl_mem_flags /* flags */, + const cl_image_format * /* image_format */, + size_t /* image_width */, + size_t /* image_height */, + size_t /* image_depth */, + size_t /* image_row_pitch */, + size_t /* image_slice_pitch */, + void * /* host_ptr */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_int CL_API_CALL +clEnqueueMarker(cl_command_queue /* command_queue */, + cl_event * /* event */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_int CL_API_CALL +clEnqueueWaitForEvents(cl_command_queue /* command_queue */, + cl_uint /* num_events */, + const cl_event * /* event_list */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_int CL_API_CALL +clEnqueueBarrier(cl_command_queue /* command_queue */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_int CL_API_CALL +clUnloadCompiler(void) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED void * CL_API_CALL +clGetExtensionFunctionAddress(const char * /* func_name */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +/* Deprecated OpenCL 2.0 APIs */ +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_2_DEPRECATED cl_command_queue CL_API_CALL +clCreateCommandQueue(cl_context /* context */, + cl_device_id /* device */, + cl_command_queue_properties /* properties */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED; + + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_2_DEPRECATED cl_sampler CL_API_CALL +clCreateSampler(cl_context /* context */, + cl_bool /* normalized_coords */, + cl_addressing_mode /* addressing_mode */, + cl_filter_mode /* filter_mode */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_2_DEPRECATED cl_int CL_API_CALL +clEnqueueTask(cl_command_queue /* command_queue */, + cl_kernel /* kernel */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED; + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_H */ + diff --git a/third_party/opencl/include/CL/cl_d3d10.h b/third_party/opencl/include/CL/cl_d3d10.h new file mode 100644 index 000000000..d5960a43f --- /dev/null +++ b/third_party/opencl/include/CL/cl_d3d10.h @@ -0,0 +1,131 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */ + +#ifndef __OPENCL_CL_D3D10_H +#define __OPENCL_CL_D3D10_H + +#include +#include +#include + +#ifdef __cplusplus +extern "C" { +#endif + +/****************************************************************************** + * cl_khr_d3d10_sharing */ +#define cl_khr_d3d10_sharing 1 + +typedef cl_uint cl_d3d10_device_source_khr; +typedef cl_uint cl_d3d10_device_set_khr; + +/******************************************************************************/ + +/* Error Codes */ +#define CL_INVALID_D3D10_DEVICE_KHR -1002 +#define CL_INVALID_D3D10_RESOURCE_KHR -1003 +#define CL_D3D10_RESOURCE_ALREADY_ACQUIRED_KHR -1004 +#define CL_D3D10_RESOURCE_NOT_ACQUIRED_KHR -1005 + +/* cl_d3d10_device_source_nv */ +#define CL_D3D10_DEVICE_KHR 0x4010 +#define CL_D3D10_DXGI_ADAPTER_KHR 0x4011 + +/* cl_d3d10_device_set_nv */ +#define CL_PREFERRED_DEVICES_FOR_D3D10_KHR 0x4012 +#define CL_ALL_DEVICES_FOR_D3D10_KHR 0x4013 + +/* cl_context_info */ +#define CL_CONTEXT_D3D10_DEVICE_KHR 0x4014 +#define CL_CONTEXT_D3D10_PREFER_SHARED_RESOURCES_KHR 0x402C + +/* cl_mem_info */ +#define CL_MEM_D3D10_RESOURCE_KHR 0x4015 + +/* cl_image_info */ +#define CL_IMAGE_D3D10_SUBRESOURCE_KHR 0x4016 + +/* cl_command_type */ +#define CL_COMMAND_ACQUIRE_D3D10_OBJECTS_KHR 0x4017 +#define CL_COMMAND_RELEASE_D3D10_OBJECTS_KHR 0x4018 + +/******************************************************************************/ + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromD3D10KHR_fn)( + cl_platform_id platform, + cl_d3d10_device_source_khr d3d_device_source, + void * d3d_object, + cl_d3d10_device_set_khr d3d_device_set, + cl_uint num_entries, + cl_device_id * devices, + cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10BufferKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D10Buffer * resource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10Texture2DKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D10Texture2D * resource, + UINT subresource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D10Texture3DKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D10Texture3D * resource, + UINT subresource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireD3D10ObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseD3D10ObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_0; + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_D3D10_H */ + diff --git a/third_party/opencl/include/CL/cl_d3d11.h b/third_party/opencl/include/CL/cl_d3d11.h new file mode 100644 index 000000000..39f907239 --- /dev/null +++ b/third_party/opencl/include/CL/cl_d3d11.h @@ -0,0 +1,131 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */ + +#ifndef __OPENCL_CL_D3D11_H +#define __OPENCL_CL_D3D11_H + +#include +#include +#include + +#ifdef __cplusplus +extern "C" { +#endif + +/****************************************************************************** + * cl_khr_d3d11_sharing */ +#define cl_khr_d3d11_sharing 1 + +typedef cl_uint cl_d3d11_device_source_khr; +typedef cl_uint cl_d3d11_device_set_khr; + +/******************************************************************************/ + +/* Error Codes */ +#define CL_INVALID_D3D11_DEVICE_KHR -1006 +#define CL_INVALID_D3D11_RESOURCE_KHR -1007 +#define CL_D3D11_RESOURCE_ALREADY_ACQUIRED_KHR -1008 +#define CL_D3D11_RESOURCE_NOT_ACQUIRED_KHR -1009 + +/* cl_d3d11_device_source */ +#define CL_D3D11_DEVICE_KHR 0x4019 +#define CL_D3D11_DXGI_ADAPTER_KHR 0x401A + +/* cl_d3d11_device_set */ +#define CL_PREFERRED_DEVICES_FOR_D3D11_KHR 0x401B +#define CL_ALL_DEVICES_FOR_D3D11_KHR 0x401C + +/* cl_context_info */ +#define CL_CONTEXT_D3D11_DEVICE_KHR 0x401D +#define CL_CONTEXT_D3D11_PREFER_SHARED_RESOURCES_KHR 0x402D + +/* cl_mem_info */ +#define CL_MEM_D3D11_RESOURCE_KHR 0x401E + +/* cl_image_info */ +#define CL_IMAGE_D3D11_SUBRESOURCE_KHR 0x401F + +/* cl_command_type */ +#define CL_COMMAND_ACQUIRE_D3D11_OBJECTS_KHR 0x4020 +#define CL_COMMAND_RELEASE_D3D11_OBJECTS_KHR 0x4021 + +/******************************************************************************/ + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromD3D11KHR_fn)( + cl_platform_id platform, + cl_d3d11_device_source_khr d3d_device_source, + void * d3d_object, + cl_d3d11_device_set_khr d3d_device_set, + cl_uint num_entries, + cl_device_id * devices, + cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11BufferKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D11Buffer * resource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11Texture2DKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D11Texture2D * resource, + UINT subresource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromD3D11Texture3DKHR_fn)( + cl_context context, + cl_mem_flags flags, + ID3D11Texture3D * resource, + UINT subresource, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireD3D11ObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseD3D11ObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_2; + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_D3D11_H */ + diff --git a/third_party/opencl/include/CL/cl_dx9_media_sharing.h b/third_party/opencl/include/CL/cl_dx9_media_sharing.h new file mode 100644 index 000000000..2729e8b9e --- /dev/null +++ b/third_party/opencl/include/CL/cl_dx9_media_sharing.h @@ -0,0 +1,132 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */ + +#ifndef __OPENCL_CL_DX9_MEDIA_SHARING_H +#define __OPENCL_CL_DX9_MEDIA_SHARING_H + +#include +#include + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************/ +/* cl_khr_dx9_media_sharing */ +#define cl_khr_dx9_media_sharing 1 + +typedef cl_uint cl_dx9_media_adapter_type_khr; +typedef cl_uint cl_dx9_media_adapter_set_khr; + +#if defined(_WIN32) +#include +typedef struct _cl_dx9_surface_info_khr +{ + IDirect3DSurface9 *resource; + HANDLE shared_handle; +} cl_dx9_surface_info_khr; +#endif + + +/******************************************************************************/ + +/* Error Codes */ +#define CL_INVALID_DX9_MEDIA_ADAPTER_KHR -1010 +#define CL_INVALID_DX9_MEDIA_SURFACE_KHR -1011 +#define CL_DX9_MEDIA_SURFACE_ALREADY_ACQUIRED_KHR -1012 +#define CL_DX9_MEDIA_SURFACE_NOT_ACQUIRED_KHR -1013 + +/* cl_media_adapter_type_khr */ +#define CL_ADAPTER_D3D9_KHR 0x2020 +#define CL_ADAPTER_D3D9EX_KHR 0x2021 +#define CL_ADAPTER_DXVA_KHR 0x2022 + +/* cl_media_adapter_set_khr */ +#define CL_PREFERRED_DEVICES_FOR_DX9_MEDIA_ADAPTER_KHR 0x2023 +#define CL_ALL_DEVICES_FOR_DX9_MEDIA_ADAPTER_KHR 0x2024 + +/* cl_context_info */ +#define CL_CONTEXT_ADAPTER_D3D9_KHR 0x2025 +#define CL_CONTEXT_ADAPTER_D3D9EX_KHR 0x2026 +#define CL_CONTEXT_ADAPTER_DXVA_KHR 0x2027 + +/* cl_mem_info */ +#define CL_MEM_DX9_MEDIA_ADAPTER_TYPE_KHR 0x2028 +#define CL_MEM_DX9_MEDIA_SURFACE_INFO_KHR 0x2029 + +/* cl_image_info */ +#define CL_IMAGE_DX9_MEDIA_PLANE_KHR 0x202A + +/* cl_command_type */ +#define CL_COMMAND_ACQUIRE_DX9_MEDIA_SURFACES_KHR 0x202B +#define CL_COMMAND_RELEASE_DX9_MEDIA_SURFACES_KHR 0x202C + +/******************************************************************************/ + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetDeviceIDsFromDX9MediaAdapterKHR_fn)( + cl_platform_id platform, + cl_uint num_media_adapters, + cl_dx9_media_adapter_type_khr * media_adapter_type, + void * media_adapters, + cl_dx9_media_adapter_set_khr media_adapter_set, + cl_uint num_entries, + cl_device_id * devices, + cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromDX9MediaSurfaceKHR_fn)( + cl_context context, + cl_mem_flags flags, + cl_dx9_media_adapter_type_khr adapter_type, + void * surface_info, + cl_uint plane, + cl_int * errcode_ret) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireDX9MediaSurfacesKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseDX9MediaSurfacesKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) CL_API_SUFFIX__VERSION_1_2; + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_DX9_MEDIA_SHARING_H */ + diff --git a/third_party/opencl/include/CL/cl_egl.h b/third_party/opencl/include/CL/cl_egl.h new file mode 100644 index 000000000..a765bd526 --- /dev/null +++ b/third_party/opencl/include/CL/cl_egl.h @@ -0,0 +1,136 @@ +/******************************************************************************* + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + ******************************************************************************/ + +#ifndef __OPENCL_CL_EGL_H +#define __OPENCL_CL_EGL_H + +#ifdef __APPLE__ + +#else +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + + +/* Command type for events created with clEnqueueAcquireEGLObjectsKHR */ +#define CL_COMMAND_EGL_FENCE_SYNC_OBJECT_KHR 0x202F +#define CL_COMMAND_ACQUIRE_EGL_OBJECTS_KHR 0x202D +#define CL_COMMAND_RELEASE_EGL_OBJECTS_KHR 0x202E + +/* Error type for clCreateFromEGLImageKHR */ +#define CL_INVALID_EGL_OBJECT_KHR -1093 +#define CL_EGL_RESOURCE_NOT_ACQUIRED_KHR -1092 + +/* CLeglImageKHR is an opaque handle to an EGLImage */ +typedef void* CLeglImageKHR; + +/* CLeglDisplayKHR is an opaque handle to an EGLDisplay */ +typedef void* CLeglDisplayKHR; + +/* CLeglSyncKHR is an opaque handle to an EGLSync object */ +typedef void* CLeglSyncKHR; + +/* properties passed to clCreateFromEGLImageKHR */ +typedef intptr_t cl_egl_image_properties_khr; + + +#define cl_khr_egl_image 1 + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateFromEGLImageKHR(cl_context /* context */, + CLeglDisplayKHR /* egldisplay */, + CLeglImageKHR /* eglimage */, + cl_mem_flags /* flags */, + const cl_egl_image_properties_khr * /* properties */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromEGLImageKHR_fn)( + cl_context context, + CLeglDisplayKHR egldisplay, + CLeglImageKHR eglimage, + cl_mem_flags flags, + const cl_egl_image_properties_khr * properties, + cl_int * errcode_ret); + + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueAcquireEGLObjectsKHR(cl_command_queue /* command_queue */, + cl_uint /* num_objects */, + const cl_mem * /* mem_objects */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireEGLObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event); + + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueReleaseEGLObjectsKHR(cl_command_queue /* command_queue */, + cl_uint /* num_objects */, + const cl_mem * /* mem_objects */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseEGLObjectsKHR_fn)( + cl_command_queue command_queue, + cl_uint num_objects, + const cl_mem * mem_objects, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event); + + +#define cl_khr_egl_event 1 + +extern CL_API_ENTRY cl_event CL_API_CALL +clCreateEventFromEGLSyncKHR(cl_context /* context */, + CLeglSyncKHR /* sync */, + CLeglDisplayKHR /* display */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_event (CL_API_CALL *clCreateEventFromEGLSyncKHR_fn)( + cl_context context, + CLeglSyncKHR sync, + CLeglDisplayKHR display, + cl_int * errcode_ret); + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_EGL_H */ diff --git a/third_party/opencl/include/CL/cl_ext.h b/third_party/opencl/include/CL/cl_ext.h new file mode 100644 index 000000000..794158389 --- /dev/null +++ b/third_party/opencl/include/CL/cl_ext.h @@ -0,0 +1,391 @@ +/******************************************************************************* + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + ******************************************************************************/ + +/* $Revision: 11928 $ on $Date: 2010-07-13 09:04:56 -0700 (Tue, 13 Jul 2010) $ */ + +/* cl_ext.h contains OpenCL extensions which don't have external */ +/* (OpenGL, D3D) dependencies. */ + +#ifndef __CL_EXT_H +#define __CL_EXT_H + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef __APPLE__ + #include + #include +#else + #include +#endif + +/* cl_khr_fp16 extension - no extension #define since it has no functions */ +#define CL_DEVICE_HALF_FP_CONFIG 0x1033 + +/* Memory object destruction + * + * Apple extension for use to manage externally allocated buffers used with cl_mem objects with CL_MEM_USE_HOST_PTR + * + * Registers a user callback function that will be called when the memory object is deleted and its resources + * freed. Each call to clSetMemObjectCallbackFn registers the specified user callback function on a callback + * stack associated with memobj. The registered user callback functions are called in the reverse order in + * which they were registered. The user callback functions are called and then the memory object is deleted + * and its resources freed. This provides a mechanism for the application (and libraries) using memobj to be + * notified when the memory referenced by host_ptr, specified when the memory object is created and used as + * the storage bits for the memory object, can be reused or freed. + * + * The application may not call CL api's with the cl_mem object passed to the pfn_notify. + * + * Please check for the "cl_APPLE_SetMemObjectDestructor" extension using clGetDeviceInfo(CL_DEVICE_EXTENSIONS) + * before using. + */ +#define cl_APPLE_SetMemObjectDestructor 1 +cl_int CL_API_ENTRY clSetMemObjectDestructorAPPLE( cl_mem /* memobj */, + void (* /*pfn_notify*/)( cl_mem /* memobj */, void* /*user_data*/), + void * /*user_data */ ) CL_EXT_SUFFIX__VERSION_1_0; + + +/* Context Logging Functions + * + * The next three convenience functions are intended to be used as the pfn_notify parameter to clCreateContext(). + * Please check for the "cl_APPLE_ContextLoggingFunctions" extension using clGetDeviceInfo(CL_DEVICE_EXTENSIONS) + * before using. + * + * clLogMessagesToSystemLog fowards on all log messages to the Apple System Logger + */ +#define cl_APPLE_ContextLoggingFunctions 1 +extern void CL_API_ENTRY clLogMessagesToSystemLogAPPLE( const char * /* errstr */, + const void * /* private_info */, + size_t /* cb */, + void * /* user_data */ ) CL_EXT_SUFFIX__VERSION_1_0; + +/* clLogMessagesToStdout sends all log messages to the file descriptor stdout */ +extern void CL_API_ENTRY clLogMessagesToStdoutAPPLE( const char * /* errstr */, + const void * /* private_info */, + size_t /* cb */, + void * /* user_data */ ) CL_EXT_SUFFIX__VERSION_1_0; + +/* clLogMessagesToStderr sends all log messages to the file descriptor stderr */ +extern void CL_API_ENTRY clLogMessagesToStderrAPPLE( const char * /* errstr */, + const void * /* private_info */, + size_t /* cb */, + void * /* user_data */ ) CL_EXT_SUFFIX__VERSION_1_0; + + +/************************ +* cl_khr_icd extension * +************************/ +#define cl_khr_icd 1 + +/* cl_platform_info */ +#define CL_PLATFORM_ICD_SUFFIX_KHR 0x0920 + +/* Additional Error Codes */ +#define CL_PLATFORM_NOT_FOUND_KHR -1001 + +extern CL_API_ENTRY cl_int CL_API_CALL +clIcdGetPlatformIDsKHR(cl_uint /* num_entries */, + cl_platform_id * /* platforms */, + cl_uint * /* num_platforms */); + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clIcdGetPlatformIDsKHR_fn)( + cl_uint /* num_entries */, + cl_platform_id * /* platforms */, + cl_uint * /* num_platforms */); + + +/* Extension: cl_khr_image2D_buffer + * + * This extension allows a 2D image to be created from a cl_mem buffer without a copy. + * The type associated with a 2D image created from a buffer in an OpenCL program is image2d_t. + * Both the sampler and sampler-less read_image built-in functions are supported for 2D images + * and 2D images created from a buffer. Similarly, the write_image built-ins are also supported + * for 2D images created from a buffer. + * + * When the 2D image from buffer is created, the client must specify the width, + * height, image format (i.e. channel order and channel data type) and optionally the row pitch + * + * The pitch specified must be a multiple of CL_DEVICE_IMAGE_PITCH_ALIGNMENT pixels. + * The base address of the buffer must be aligned to CL_DEVICE_IMAGE_BASE_ADDRESS_ALIGNMENT pixels. + */ + +/************************************* + * cl_khr_initalize_memory extension * + *************************************/ + +#define CL_CONTEXT_MEMORY_INITIALIZE_KHR 0x2030 + + +/************************************** + * cl_khr_terminate_context extension * + **************************************/ + +#define CL_DEVICE_TERMINATE_CAPABILITY_KHR 0x2031 +#define CL_CONTEXT_TERMINATE_KHR 0x2032 + +#define cl_khr_terminate_context 1 +extern CL_API_ENTRY cl_int CL_API_CALL clTerminateContextKHR(cl_context /* context */) CL_EXT_SUFFIX__VERSION_1_2; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clTerminateContextKHR_fn)(cl_context /* context */) CL_EXT_SUFFIX__VERSION_1_2; + + +/* + * Extension: cl_khr_spir + * + * This extension adds support to create an OpenCL program object from a + * Standard Portable Intermediate Representation (SPIR) instance + */ + +#define CL_DEVICE_SPIR_VERSIONS 0x40E0 +#define CL_PROGRAM_BINARY_TYPE_INTERMEDIATE 0x40E1 + + +/****************************************** +* cl_nv_device_attribute_query extension * +******************************************/ +/* cl_nv_device_attribute_query extension - no extension #define since it has no functions */ +#define CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV 0x4000 +#define CL_DEVICE_COMPUTE_CAPABILITY_MINOR_NV 0x4001 +#define CL_DEVICE_REGISTERS_PER_BLOCK_NV 0x4002 +#define CL_DEVICE_WARP_SIZE_NV 0x4003 +#define CL_DEVICE_GPU_OVERLAP_NV 0x4004 +#define CL_DEVICE_KERNEL_EXEC_TIMEOUT_NV 0x4005 +#define CL_DEVICE_INTEGRATED_MEMORY_NV 0x4006 + +/********************************* +* cl_amd_device_attribute_query * +*********************************/ +#define CL_DEVICE_PROFILING_TIMER_OFFSET_AMD 0x4036 + +/********************************* +* cl_arm_printf extension +*********************************/ +#define CL_PRINTF_CALLBACK_ARM 0x40B0 +#define CL_PRINTF_BUFFERSIZE_ARM 0x40B1 + +#ifdef CL_VERSION_1_1 + /*********************************** + * cl_ext_device_fission extension * + ***********************************/ + #define cl_ext_device_fission 1 + + extern CL_API_ENTRY cl_int CL_API_CALL + clReleaseDeviceEXT( cl_device_id /*device*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + typedef CL_API_ENTRY cl_int + (CL_API_CALL *clReleaseDeviceEXT_fn)( cl_device_id /*device*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + extern CL_API_ENTRY cl_int CL_API_CALL + clRetainDeviceEXT( cl_device_id /*device*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + typedef CL_API_ENTRY cl_int + (CL_API_CALL *clRetainDeviceEXT_fn)( cl_device_id /*device*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + typedef cl_ulong cl_device_partition_property_ext; + extern CL_API_ENTRY cl_int CL_API_CALL + clCreateSubDevicesEXT( cl_device_id /*in_device*/, + const cl_device_partition_property_ext * /* properties */, + cl_uint /*num_entries*/, + cl_device_id * /*out_devices*/, + cl_uint * /*num_devices*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + typedef CL_API_ENTRY cl_int + ( CL_API_CALL * clCreateSubDevicesEXT_fn)( cl_device_id /*in_device*/, + const cl_device_partition_property_ext * /* properties */, + cl_uint /*num_entries*/, + cl_device_id * /*out_devices*/, + cl_uint * /*num_devices*/ ) CL_EXT_SUFFIX__VERSION_1_1; + + /* cl_device_partition_property_ext */ + #define CL_DEVICE_PARTITION_EQUALLY_EXT 0x4050 + #define CL_DEVICE_PARTITION_BY_COUNTS_EXT 0x4051 + #define CL_DEVICE_PARTITION_BY_NAMES_EXT 0x4052 + #define CL_DEVICE_PARTITION_BY_AFFINITY_DOMAIN_EXT 0x4053 + + /* clDeviceGetInfo selectors */ + #define CL_DEVICE_PARENT_DEVICE_EXT 0x4054 + #define CL_DEVICE_PARTITION_TYPES_EXT 0x4055 + #define CL_DEVICE_AFFINITY_DOMAINS_EXT 0x4056 + #define CL_DEVICE_REFERENCE_COUNT_EXT 0x4057 + #define CL_DEVICE_PARTITION_STYLE_EXT 0x4058 + + /* error codes */ + #define CL_DEVICE_PARTITION_FAILED_EXT -1057 + #define CL_INVALID_PARTITION_COUNT_EXT -1058 + #define CL_INVALID_PARTITION_NAME_EXT -1059 + + /* CL_AFFINITY_DOMAINs */ + #define CL_AFFINITY_DOMAIN_L1_CACHE_EXT 0x1 + #define CL_AFFINITY_DOMAIN_L2_CACHE_EXT 0x2 + #define CL_AFFINITY_DOMAIN_L3_CACHE_EXT 0x3 + #define CL_AFFINITY_DOMAIN_L4_CACHE_EXT 0x4 + #define CL_AFFINITY_DOMAIN_NUMA_EXT 0x10 + #define CL_AFFINITY_DOMAIN_NEXT_FISSIONABLE_EXT 0x100 + + /* cl_device_partition_property_ext list terminators */ + #define CL_PROPERTIES_LIST_END_EXT ((cl_device_partition_property_ext) 0) + #define CL_PARTITION_BY_COUNTS_LIST_END_EXT ((cl_device_partition_property_ext) 0) + #define CL_PARTITION_BY_NAMES_LIST_END_EXT ((cl_device_partition_property_ext) 0 - 1) + +/********************************* +* cl_qcom_ext_host_ptr extension +*********************************/ + +#define CL_MEM_EXT_HOST_PTR_QCOM (1 << 29) + +#define CL_DEVICE_EXT_MEM_PADDING_IN_BYTES_QCOM 0x40A0 +#define CL_DEVICE_PAGE_SIZE_QCOM 0x40A1 +#define CL_IMAGE_ROW_ALIGNMENT_QCOM 0x40A2 +#define CL_IMAGE_SLICE_ALIGNMENT_QCOM 0x40A3 +#define CL_MEM_HOST_UNCACHED_QCOM 0x40A4 +#define CL_MEM_HOST_WRITEBACK_QCOM 0x40A5 +#define CL_MEM_HOST_WRITETHROUGH_QCOM 0x40A6 +#define CL_MEM_HOST_WRITE_COMBINING_QCOM 0x40A7 + +typedef cl_uint cl_image_pitch_info_qcom; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetDeviceImageInfoQCOM(cl_device_id device, + size_t image_width, + size_t image_height, + const cl_image_format *image_format, + cl_image_pitch_info_qcom param_name, + size_t param_value_size, + void *param_value, + size_t *param_value_size_ret); + +typedef struct _cl_mem_ext_host_ptr +{ + /* Type of external memory allocation. */ + /* Legal values will be defined in layered extensions. */ + cl_uint allocation_type; + + /* Host cache policy for this external memory allocation. */ + cl_uint host_cache_policy; + +} cl_mem_ext_host_ptr; + +/********************************* +* cl_qcom_ion_host_ptr extension +*********************************/ + +#define CL_MEM_ION_HOST_PTR_QCOM 0x40A8 + +typedef struct _cl_mem_ion_host_ptr +{ + /* Type of external memory allocation. */ + /* Must be CL_MEM_ION_HOST_PTR_QCOM for ION allocations. */ + cl_mem_ext_host_ptr ext_host_ptr; + + /* ION file descriptor */ + int ion_filedesc; + + /* Host pointer to the ION allocated memory */ + void* ion_hostptr; + +} cl_mem_ion_host_ptr; + +#endif /* CL_VERSION_1_1 */ + + +#ifdef CL_VERSION_2_0 +/********************************* +* cl_khr_sub_groups extension +*********************************/ +#define cl_khr_sub_groups 1 + +typedef cl_uint cl_kernel_sub_group_info_khr; + +/* cl_khr_sub_group_info */ +#define CL_KERNEL_MAX_SUB_GROUP_SIZE_FOR_NDRANGE_KHR 0x2033 +#define CL_KERNEL_SUB_GROUP_COUNT_FOR_NDRANGE_KHR 0x2034 + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetKernelSubGroupInfoKHR(cl_kernel /* in_kernel */, + cl_device_id /*in_device*/, + cl_kernel_sub_group_info_khr /* param_name */, + size_t /*input_value_size*/, + const void * /*input_value*/, + size_t /*param_value_size*/, + void* /*param_value*/, + size_t* /*param_value_size_ret*/ ) CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED; + +typedef CL_API_ENTRY cl_int + ( CL_API_CALL * clGetKernelSubGroupInfoKHR_fn)(cl_kernel /* in_kernel */, + cl_device_id /*in_device*/, + cl_kernel_sub_group_info_khr /* param_name */, + size_t /*input_value_size*/, + const void * /*input_value*/, + size_t /*param_value_size*/, + void* /*param_value*/, + size_t* /*param_value_size_ret*/ ) CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED; +#endif /* CL_VERSION_2_0 */ + +#ifdef CL_VERSION_2_1 +/********************************* +* cl_khr_priority_hints extension +*********************************/ +#define cl_khr_priority_hints 1 + +typedef cl_uint cl_queue_priority_khr; + +/* cl_command_queue_properties */ +#define CL_QUEUE_PRIORITY_KHR 0x1096 + +/* cl_queue_priority_khr */ +#define CL_QUEUE_PRIORITY_HIGH_KHR (1<<0) +#define CL_QUEUE_PRIORITY_MED_KHR (1<<1) +#define CL_QUEUE_PRIORITY_LOW_KHR (1<<2) + +#endif /* CL_VERSION_2_1 */ + +#ifdef CL_VERSION_2_1 +/********************************* +* cl_khr_throttle_hints extension +*********************************/ +#define cl_khr_throttle_hints 1 + +typedef cl_uint cl_queue_throttle_khr; + +/* cl_command_queue_properties */ +#define CL_QUEUE_THROTTLE_KHR 0x1097 + +/* cl_queue_throttle_khr */ +#define CL_QUEUE_THROTTLE_HIGH_KHR (1<<0) +#define CL_QUEUE_THROTTLE_MED_KHR (1<<1) +#define CL_QUEUE_THROTTLE_LOW_KHR (1<<2) + +#endif /* CL_VERSION_2_1 */ + +#ifdef __cplusplus +} +#endif + + +#endif /* __CL_EXT_H */ diff --git a/third_party/opencl/include/CL/cl_ext_qcom.h b/third_party/opencl/include/CL/cl_ext_qcom.h new file mode 100644 index 000000000..6328a1cd9 --- /dev/null +++ b/third_party/opencl/include/CL/cl_ext_qcom.h @@ -0,0 +1,255 @@ +/* Copyright (c) 2009-2017 Qualcomm Technologies, Inc. All Rights Reserved. + * Qualcomm Technologies Proprietary and Confidential. + */ + +#ifndef __OPENCL_CL_EXT_QCOM_H +#define __OPENCL_CL_EXT_QCOM_H + +// Needed by cl_khr_egl_event extension +#include +#include +#include + +#ifdef __cplusplus +extern "C" { +#endif + + +/************************************ + * cl_qcom_create_buffer_from_image * + ************************************/ + +#define CL_BUFFER_FROM_IMAGE_ROW_PITCH_QCOM 0x40C0 +#define CL_BUFFER_FROM_IMAGE_SLICE_PITCH_QCOM 0x40C1 + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateBufferFromImageQCOM(cl_mem image, + cl_mem_flags flags, + cl_int *errcode_ret); + + +/************************************ + * cl_qcom_limited_printf extension * + ************************************/ + +/* Builtin printf function buffer size in bytes. */ +#define CL_DEVICE_PRINTF_BUFFER_SIZE_QCOM 0x1049 + + +/************************************* + * cl_qcom_extended_images extension * + *************************************/ + +#define CL_CONTEXT_ENABLE_EXTENDED_IMAGES_QCOM 0x40AA +#define CL_DEVICE_EXTENDED_IMAGE2D_MAX_WIDTH_QCOM 0x40AB +#define CL_DEVICE_EXTENDED_IMAGE2D_MAX_HEIGHT_QCOM 0x40AC +#define CL_DEVICE_EXTENDED_IMAGE3D_MAX_WIDTH_QCOM 0x40AD +#define CL_DEVICE_EXTENDED_IMAGE3D_MAX_HEIGHT_QCOM 0x40AE +#define CL_DEVICE_EXTENDED_IMAGE3D_MAX_DEPTH_QCOM 0x40AF + +/************************************* + * cl_qcom_perf_hint extension * + *************************************/ + +typedef cl_uint cl_perf_hint; + +#define CL_CONTEXT_PERF_HINT_QCOM 0x40C2 + +/*cl_perf_hint*/ +#define CL_PERF_HINT_HIGH_QCOM 0x40C3 +#define CL_PERF_HINT_NORMAL_QCOM 0x40C4 +#define CL_PERF_HINT_LOW_QCOM 0x40C5 + +extern CL_API_ENTRY cl_int CL_API_CALL +clSetPerfHintQCOM(cl_context context, + cl_perf_hint perf_hint); + +// This extension is published at Khronos, so its definitions are made in cl_ext.h. +// This duplication is for backward compatibility. + +#ifndef CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM + +/********************************* +* cl_qcom_android_native_buffer_host_ptr extension +*********************************/ + +#define CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM 0x40C6 + + +typedef struct _cl_mem_android_native_buffer_host_ptr +{ + // Type of external memory allocation. + // Must be CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM for Android native buffers. + cl_mem_ext_host_ptr ext_host_ptr; + + // Virtual pointer to the android native buffer + void* anb_ptr; + +} cl_mem_android_native_buffer_host_ptr; + +#endif //#ifndef CL_MEM_ANDROID_NATIVE_BUFFER_HOST_PTR_QCOM + +/*********************************** +* cl_img_egl_image extension * +************************************/ +typedef void* CLeglImageIMG; +typedef void* CLeglDisplayIMG; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateFromEGLImageIMG(cl_context context, + cl_mem_flags flags, + CLeglImageIMG image, + CLeglDisplayIMG display, + cl_int *errcode_ret); + + +/********************************* +* cl_qcom_other_image extension +*********************************/ + +// Extended flag for creating/querying QCOM non-standard images +#define CL_MEM_OTHER_IMAGE_QCOM (1<<25) + +// cl_channel_type +#define CL_QCOM_UNORM_MIPI10 0x4159 +#define CL_QCOM_UNORM_MIPI12 0x415A +#define CL_QCOM_UNSIGNED_MIPI10 0x415B +#define CL_QCOM_UNSIGNED_MIPI12 0x415C +#define CL_QCOM_UNORM_INT10 0x415D +#define CL_QCOM_UNORM_INT12 0x415E +#define CL_QCOM_UNSIGNED_INT16 0x415F + +// cl_channel_order +// Dedicate 0x4130-0x415F range for QCOM extended image formats +// 0x4130 - 0x4132 range is assigned to pixel-oriented compressed format +#define CL_QCOM_BAYER 0x414E + +#define CL_QCOM_NV12 0x4133 +#define CL_QCOM_NV12_Y 0x4134 +#define CL_QCOM_NV12_UV 0x4135 + +#define CL_QCOM_TILED_NV12 0x4136 +#define CL_QCOM_TILED_NV12_Y 0x4137 +#define CL_QCOM_TILED_NV12_UV 0x4138 + +#define CL_QCOM_P010 0x413C +#define CL_QCOM_P010_Y 0x413D +#define CL_QCOM_P010_UV 0x413E + +#define CL_QCOM_TILED_P010 0x413F +#define CL_QCOM_TILED_P010_Y 0x4140 +#define CL_QCOM_TILED_P010_UV 0x4141 + + +#define CL_QCOM_TP10 0x4145 +#define CL_QCOM_TP10_Y 0x4146 +#define CL_QCOM_TP10_UV 0x4147 + +#define CL_QCOM_TILED_TP10 0x4148 +#define CL_QCOM_TILED_TP10_Y 0x4149 +#define CL_QCOM_TILED_TP10_UV 0x414A + +/********************************* +* cl_qcom_compressed_image extension +*********************************/ + +// Extended flag for creating/querying QCOM non-planar compressed images +#define CL_MEM_COMPRESSED_IMAGE_QCOM (1<<27) + +// Extended image format +// cl_channel_order +#define CL_QCOM_COMPRESSED_RGBA 0x4130 +#define CL_QCOM_COMPRESSED_RGBx 0x4131 + +#define CL_QCOM_COMPRESSED_NV12_Y 0x413A +#define CL_QCOM_COMPRESSED_NV12_UV 0x413B + +#define CL_QCOM_COMPRESSED_P010 0x4142 +#define CL_QCOM_COMPRESSED_P010_Y 0x4143 +#define CL_QCOM_COMPRESSED_P010_UV 0x4144 + +#define CL_QCOM_COMPRESSED_TP10 0x414B +#define CL_QCOM_COMPRESSED_TP10_Y 0x414C +#define CL_QCOM_COMPRESSED_TP10_UV 0x414D + +#define CL_QCOM_COMPRESSED_NV12_4R 0x414F +#define CL_QCOM_COMPRESSED_NV12_4R_Y 0x4150 +#define CL_QCOM_COMPRESSED_NV12_4R_UV 0x4151 +/********************************* +* cl_qcom_compressed_yuv_image_read extension +*********************************/ + +// Extended flag for creating/querying QCOM compressed images +#define CL_MEM_COMPRESSED_YUV_IMAGE_QCOM (1<<28) + +// Extended image format +#define CL_QCOM_COMPRESSED_NV12 0x10C4 + +// Extended flag for setting ION buffer allocation type +#define CL_MEM_ION_HOST_PTR_COMPRESSED_YUV_QCOM 0x40CD +#define CL_MEM_ION_HOST_PTR_PROTECTED_COMPRESSED_YUV_QCOM 0x40CE + +/********************************* +* cl_qcom_accelerated_image_ops +*********************************/ +#define CL_MEM_OBJECT_WEIGHT_IMAGE_QCOM 0x4110 +#define CL_DEVICE_HOF_MAX_NUM_PHASES_QCOM 0x4111 +#define CL_DEVICE_HOF_MAX_FILTER_SIZE_X_QCOM 0x4112 +#define CL_DEVICE_HOF_MAX_FILTER_SIZE_Y_QCOM 0x4113 +#define CL_DEVICE_BLOCK_MATCHING_MAX_REGION_SIZE_X_QCOM 0x4114 +#define CL_DEVICE_BLOCK_MATCHING_MAX_REGION_SIZE_Y_QCOM 0x4115 + +//Extended flag for specifying weight image type +#define CL_WEIGHT_IMAGE_SEPARABLE_QCOM (1<<0) + +// Box Filter +typedef struct _cl_box_filter_size_qcom +{ + // Width of box filter on X direction. + float box_filter_width; + + // Height of box filter on Y direction. + float box_filter_height; +} cl_box_filter_size_qcom; + +// HOF Weight Image Desc +typedef struct _cl_weight_desc_qcom +{ + /** Coordinate of the "center" point of the weight image, + based on the weight image's top-left corner as the origin. */ + size_t center_coord_x; + size_t center_coord_y; + cl_bitfield flags; +} cl_weight_desc_qcom; + +typedef struct _cl_weight_image_desc_qcom +{ + cl_image_desc image_desc; + cl_weight_desc_qcom weight_desc; +} cl_weight_image_desc_qcom; + +/************************************* + * cl_qcom_protected_context extension * + *************************************/ + +#define CL_CONTEXT_PROTECTED_QCOM 0x40C7 +#define CL_MEM_ION_HOST_PTR_PROTECTED_QCOM 0x40C8 + +/************************************* + * cl_qcom_priority_hint extension * + *************************************/ +#define CL_PRIORITY_HINT_NONE_QCOM 0 +typedef cl_uint cl_priority_hint; + +#define CL_CONTEXT_PRIORITY_HINT_QCOM 0x40C9 + +/*cl_priority_hint*/ +#define CL_PRIORITY_HINT_HIGH_QCOM 0x40CA +#define CL_PRIORITY_HINT_NORMAL_QCOM 0x40CB +#define CL_PRIORITY_HINT_LOW_QCOM 0x40CC + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_EXT_QCOM_H */ diff --git a/third_party/opencl/include/CL/cl_gl.h b/third_party/opencl/include/CL/cl_gl.h new file mode 100644 index 000000000..945daa83d --- /dev/null +++ b/third_party/opencl/include/CL/cl_gl.h @@ -0,0 +1,167 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +#ifndef __OPENCL_CL_GL_H +#define __OPENCL_CL_GL_H + +#ifdef __APPLE__ +#include +#else +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +typedef cl_uint cl_gl_object_type; +typedef cl_uint cl_gl_texture_info; +typedef cl_uint cl_gl_platform_info; +typedef struct __GLsync *cl_GLsync; + +/* cl_gl_object_type = 0x2000 - 0x200F enum values are currently taken */ +#define CL_GL_OBJECT_BUFFER 0x2000 +#define CL_GL_OBJECT_TEXTURE2D 0x2001 +#define CL_GL_OBJECT_TEXTURE3D 0x2002 +#define CL_GL_OBJECT_RENDERBUFFER 0x2003 +#define CL_GL_OBJECT_TEXTURE2D_ARRAY 0x200E +#define CL_GL_OBJECT_TEXTURE1D 0x200F +#define CL_GL_OBJECT_TEXTURE1D_ARRAY 0x2010 +#define CL_GL_OBJECT_TEXTURE_BUFFER 0x2011 + +/* cl_gl_texture_info */ +#define CL_GL_TEXTURE_TARGET 0x2004 +#define CL_GL_MIPMAP_LEVEL 0x2005 +#define CL_GL_NUM_SAMPLES 0x2012 + + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateFromGLBuffer(cl_context /* context */, + cl_mem_flags /* flags */, + cl_GLuint /* bufobj */, + int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateFromGLTexture(cl_context /* context */, + cl_mem_flags /* flags */, + cl_GLenum /* target */, + cl_GLint /* miplevel */, + cl_GLuint /* texture */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_2; + +extern CL_API_ENTRY cl_mem CL_API_CALL +clCreateFromGLRenderbuffer(cl_context /* context */, + cl_mem_flags /* flags */, + cl_GLuint /* renderbuffer */, + cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetGLObjectInfo(cl_mem /* memobj */, + cl_gl_object_type * /* gl_object_type */, + cl_GLuint * /* gl_object_name */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetGLTextureInfo(cl_mem /* memobj */, + cl_gl_texture_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueAcquireGLObjects(cl_command_queue /* command_queue */, + cl_uint /* num_objects */, + const cl_mem * /* mem_objects */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + +extern CL_API_ENTRY cl_int CL_API_CALL +clEnqueueReleaseGLObjects(cl_command_queue /* command_queue */, + cl_uint /* num_objects */, + const cl_mem * /* mem_objects */, + cl_uint /* num_events_in_wait_list */, + const cl_event * /* event_wait_list */, + cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0; + + +/* Deprecated OpenCL 1.1 APIs */ +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL +clCreateFromGLTexture2D(cl_context /* context */, + cl_mem_flags /* flags */, + cl_GLenum /* target */, + cl_GLint /* miplevel */, + cl_GLuint /* texture */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +extern CL_API_ENTRY CL_EXT_PREFIX__VERSION_1_1_DEPRECATED cl_mem CL_API_CALL +clCreateFromGLTexture3D(cl_context /* context */, + cl_mem_flags /* flags */, + cl_GLenum /* target */, + cl_GLint /* miplevel */, + cl_GLuint /* texture */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED; + +/* cl_khr_gl_sharing extension */ + +#define cl_khr_gl_sharing 1 + +typedef cl_uint cl_gl_context_info; + +/* Additional Error Codes */ +#define CL_INVALID_GL_SHAREGROUP_REFERENCE_KHR -1000 + +/* cl_gl_context_info */ +#define CL_CURRENT_DEVICE_FOR_GL_CONTEXT_KHR 0x2006 +#define CL_DEVICES_FOR_GL_CONTEXT_KHR 0x2007 + +/* Additional cl_context_properties */ +#define CL_GL_CONTEXT_KHR 0x2008 +#define CL_EGL_DISPLAY_KHR 0x2009 +#define CL_GLX_DISPLAY_KHR 0x200A +#define CL_WGL_HDC_KHR 0x200B +#define CL_CGL_SHAREGROUP_KHR 0x200C + +extern CL_API_ENTRY cl_int CL_API_CALL +clGetGLContextInfoKHR(const cl_context_properties * /* properties */, + cl_gl_context_info /* param_name */, + size_t /* param_value_size */, + void * /* param_value */, + size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0; + +typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetGLContextInfoKHR_fn)( + const cl_context_properties * properties, + cl_gl_context_info param_name, + size_t param_value_size, + void * param_value, + size_t * param_value_size_ret); + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_GL_H */ diff --git a/third_party/opencl/include/CL/cl_gl_ext.h b/third_party/opencl/include/CL/cl_gl_ext.h new file mode 100644 index 000000000..e3c14c640 --- /dev/null +++ b/third_party/opencl/include/CL/cl_gl_ext.h @@ -0,0 +1,74 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */ + +/* cl_gl_ext.h contains vendor (non-KHR) OpenCL extensions which have */ +/* OpenGL dependencies. */ + +#ifndef __OPENCL_CL_GL_EXT_H +#define __OPENCL_CL_GL_EXT_H + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef __APPLE__ + #include +#else + #include +#endif + +/* + * For each extension, follow this template + * cl_VEN_extname extension */ +/* #define cl_VEN_extname 1 + * ... define new types, if any + * ... define new tokens, if any + * ... define new APIs, if any + * + * If you need GLtypes here, mirror them with a cl_GLtype, rather than including a GL header + * This allows us to avoid having to decide whether to include GL headers or GLES here. + */ + +/* + * cl_khr_gl_event extension + * See section 9.9 in the OpenCL 1.1 spec for more information + */ +#define CL_COMMAND_GL_FENCE_SYNC_OBJECT_KHR 0x200D + +extern CL_API_ENTRY cl_event CL_API_CALL +clCreateEventFromGLsyncKHR(cl_context /* context */, + cl_GLsync /* cl_GLsync */, + cl_int * /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1; + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_CL_GL_EXT_H */ diff --git a/third_party/opencl/include/CL/cl_platform.h b/third_party/opencl/include/CL/cl_platform.h new file mode 100644 index 000000000..4e334a291 --- /dev/null +++ b/third_party/opencl/include/CL/cl_platform.h @@ -0,0 +1,1333 @@ +/********************************************************************************** + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + **********************************************************************************/ + +/* $Revision: 11803 $ on $Date: 2010-06-25 10:02:12 -0700 (Fri, 25 Jun 2010) $ */ + +#ifndef __CL_PLATFORM_H +#define __CL_PLATFORM_H + +#ifdef __APPLE__ + /* Contains #defines for AVAILABLE_MAC_OS_X_VERSION_10_6_AND_LATER below */ + #include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +#if defined(_WIN32) + #define CL_API_ENTRY + #define CL_API_CALL __stdcall + #define CL_CALLBACK __stdcall +#else + #define CL_API_ENTRY + #define CL_API_CALL + #define CL_CALLBACK +#endif + +/* + * Deprecation flags refer to the last version of the header in which the + * feature was not deprecated. + * + * E.g. VERSION_1_1_DEPRECATED means the feature is present in 1.1 without + * deprecation but is deprecated in versions later than 1.1. + */ + +#ifdef __APPLE__ + #define CL_EXTENSION_WEAK_LINK __attribute__((weak_import)) + #define CL_API_SUFFIX__VERSION_1_0 AVAILABLE_MAC_OS_X_VERSION_10_6_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_0 CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_6_AND_LATER + #define CL_API_SUFFIX__VERSION_1_1 AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define GCL_API_SUFFIX__VERSION_1_1 AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_1 CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_6_AND_LATER_BUT_DEPRECATED_IN_MAC_OS_X_VERSION_10_7 + + #ifdef AVAILABLE_MAC_OS_X_VERSION_10_8_AND_LATER + #define CL_API_SUFFIX__VERSION_1_2 AVAILABLE_MAC_OS_X_VERSION_10_8_AND_LATER + #define GCL_API_SUFFIX__VERSION_1_2 AVAILABLE_MAC_OS_X_VERSION_10_8_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_2 CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_8_AND_LATER + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER_BUT_DEPRECATED_IN_MAC_OS_X_VERSION_10_8 + #else + #warning This path should never happen outside of internal operating system development. AvailabilityMacros do not function correctly here! + #define CL_API_SUFFIX__VERSION_1_2 AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define GCL_API_SUFFIX__VERSION_1_2 AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_2 CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED CL_EXTENSION_WEAK_LINK AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER + #endif +#else + #define CL_EXTENSION_WEAK_LINK + #define CL_API_SUFFIX__VERSION_1_0 + #define CL_EXT_SUFFIX__VERSION_1_0 + #define CL_API_SUFFIX__VERSION_1_1 + #define CL_EXT_SUFFIX__VERSION_1_1 + #define CL_API_SUFFIX__VERSION_1_2 + #define CL_EXT_SUFFIX__VERSION_1_2 + #define CL_API_SUFFIX__VERSION_2_0 + #define CL_EXT_SUFFIX__VERSION_2_0 + #define CL_API_SUFFIX__VERSION_2_1 + #define CL_EXT_SUFFIX__VERSION_2_1 + + #ifdef __GNUC__ + #ifdef CL_USE_DEPRECATED_OPENCL_1_0_APIS + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_0_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED __attribute__((deprecated)) + #define CL_EXT_PREFIX__VERSION_1_0_DEPRECATED + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_1_1_APIS + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED __attribute__((deprecated)) + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_1_2_APIS + #define CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_2_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED __attribute__((deprecated)) + #define CL_EXT_PREFIX__VERSION_1_2_DEPRECATED + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_2_0_APIS + #define CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_2_0_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED __attribute__((deprecated)) + #define CL_EXT_PREFIX__VERSION_2_0_DEPRECATED + #endif + #elif _WIN32 + #ifdef CL_USE_DEPRECATED_OPENCL_1_0_APIS + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_0_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_0_DEPRECATED __declspec(deprecated) + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_1_1_APIS + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED __declspec(deprecated) + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_1_2_APIS + #define CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_2_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_2_DEPRECATED __declspec(deprecated) + #endif + + #ifdef CL_USE_DEPRECATED_OPENCL_2_0_APIS + #define CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_2_0_DEPRECATED + #else + #define CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_2_0_DEPRECATED __declspec(deprecated) + #endif + #else + #define CL_EXT_SUFFIX__VERSION_1_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_0_DEPRECATED + + #define CL_EXT_SUFFIX__VERSION_1_1_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_1_DEPRECATED + + #define CL_EXT_SUFFIX__VERSION_1_2_DEPRECATED + #define CL_EXT_PREFIX__VERSION_1_2_DEPRECATED + + #define CL_EXT_SUFFIX__VERSION_2_0_DEPRECATED + #define CL_EXT_PREFIX__VERSION_2_0_DEPRECATED + #endif +#endif + +#if (defined (_WIN32) && defined(_MSC_VER)) + +/* scalar types */ +typedef signed __int8 cl_char; +typedef unsigned __int8 cl_uchar; +typedef signed __int16 cl_short; +typedef unsigned __int16 cl_ushort; +typedef signed __int32 cl_int; +typedef unsigned __int32 cl_uint; +typedef signed __int64 cl_long; +typedef unsigned __int64 cl_ulong; + +typedef unsigned __int16 cl_half; +typedef float cl_float; +typedef double cl_double; + +/* Macro names and corresponding values defined by OpenCL */ +#define CL_CHAR_BIT 8 +#define CL_SCHAR_MAX 127 +#define CL_SCHAR_MIN (-127-1) +#define CL_CHAR_MAX CL_SCHAR_MAX +#define CL_CHAR_MIN CL_SCHAR_MIN +#define CL_UCHAR_MAX 255 +#define CL_SHRT_MAX 32767 +#define CL_SHRT_MIN (-32767-1) +#define CL_USHRT_MAX 65535 +#define CL_INT_MAX 2147483647 +#define CL_INT_MIN (-2147483647-1) +#define CL_UINT_MAX 0xffffffffU +#define CL_LONG_MAX ((cl_long) 0x7FFFFFFFFFFFFFFFLL) +#define CL_LONG_MIN ((cl_long) -0x7FFFFFFFFFFFFFFFLL - 1LL) +#define CL_ULONG_MAX ((cl_ulong) 0xFFFFFFFFFFFFFFFFULL) + +#define CL_FLT_DIG 6 +#define CL_FLT_MANT_DIG 24 +#define CL_FLT_MAX_10_EXP +38 +#define CL_FLT_MAX_EXP +128 +#define CL_FLT_MIN_10_EXP -37 +#define CL_FLT_MIN_EXP -125 +#define CL_FLT_RADIX 2 +#define CL_FLT_MAX 340282346638528859811704183484516925440.0f +#define CL_FLT_MIN 1.175494350822287507969e-38f +#define CL_FLT_EPSILON 0x1.0p-23f + +#define CL_DBL_DIG 15 +#define CL_DBL_MANT_DIG 53 +#define CL_DBL_MAX_10_EXP +308 +#define CL_DBL_MAX_EXP +1024 +#define CL_DBL_MIN_10_EXP -307 +#define CL_DBL_MIN_EXP -1021 +#define CL_DBL_RADIX 2 +#define CL_DBL_MAX 179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.0 +#define CL_DBL_MIN 2.225073858507201383090e-308 +#define CL_DBL_EPSILON 2.220446049250313080847e-16 + +#define CL_M_E 2.718281828459045090796 +#define CL_M_LOG2E 1.442695040888963387005 +#define CL_M_LOG10E 0.434294481903251816668 +#define CL_M_LN2 0.693147180559945286227 +#define CL_M_LN10 2.302585092994045901094 +#define CL_M_PI 3.141592653589793115998 +#define CL_M_PI_2 1.570796326794896557999 +#define CL_M_PI_4 0.785398163397448278999 +#define CL_M_1_PI 0.318309886183790691216 +#define CL_M_2_PI 0.636619772367581382433 +#define CL_M_2_SQRTPI 1.128379167095512558561 +#define CL_M_SQRT2 1.414213562373095145475 +#define CL_M_SQRT1_2 0.707106781186547572737 + +#define CL_M_E_F 2.71828174591064f +#define CL_M_LOG2E_F 1.44269502162933f +#define CL_M_LOG10E_F 0.43429449200630f +#define CL_M_LN2_F 0.69314718246460f +#define CL_M_LN10_F 2.30258512496948f +#define CL_M_PI_F 3.14159274101257f +#define CL_M_PI_2_F 1.57079637050629f +#define CL_M_PI_4_F 0.78539818525314f +#define CL_M_1_PI_F 0.31830987334251f +#define CL_M_2_PI_F 0.63661974668503f +#define CL_M_2_SQRTPI_F 1.12837922573090f +#define CL_M_SQRT2_F 1.41421353816986f +#define CL_M_SQRT1_2_F 0.70710676908493f + +#define CL_NAN (CL_INFINITY - CL_INFINITY) +#define CL_HUGE_VALF ((cl_float) 1e50) +#define CL_HUGE_VAL ((cl_double) 1e500) +#define CL_MAXFLOAT CL_FLT_MAX +#define CL_INFINITY CL_HUGE_VALF + +#else + +#include + +/* scalar types */ +typedef int8_t cl_char; +typedef uint8_t cl_uchar; +typedef int16_t cl_short __attribute__((aligned(2))); +typedef uint16_t cl_ushort __attribute__((aligned(2))); +typedef int32_t cl_int __attribute__((aligned(4))); +typedef uint32_t cl_uint __attribute__((aligned(4))); +typedef int64_t cl_long __attribute__((aligned(8))); +typedef uint64_t cl_ulong __attribute__((aligned(8))); + +typedef uint16_t cl_half __attribute__((aligned(2))); +typedef float cl_float __attribute__((aligned(4))); +typedef double cl_double __attribute__((aligned(8))); + +/* Macro names and corresponding values defined by OpenCL */ +#define CL_CHAR_BIT 8 +#define CL_SCHAR_MAX 127 +#define CL_SCHAR_MIN (-127-1) +#define CL_CHAR_MAX CL_SCHAR_MAX +#define CL_CHAR_MIN CL_SCHAR_MIN +#define CL_UCHAR_MAX 255 +#define CL_SHRT_MAX 32767 +#define CL_SHRT_MIN (-32767-1) +#define CL_USHRT_MAX 65535 +#define CL_INT_MAX 2147483647 +#define CL_INT_MIN (-2147483647-1) +#define CL_UINT_MAX 0xffffffffU +#define CL_LONG_MAX ((cl_long) 0x7FFFFFFFFFFFFFFFLL) +#define CL_LONG_MIN ((cl_long) -0x7FFFFFFFFFFFFFFFLL - 1LL) +#define CL_ULONG_MAX ((cl_ulong) 0xFFFFFFFFFFFFFFFFULL) + +#define CL_FLT_DIG 6 +#define CL_FLT_MANT_DIG 24 +#define CL_FLT_MAX_10_EXP +38 +#define CL_FLT_MAX_EXP +128 +#define CL_FLT_MIN_10_EXP -37 +#define CL_FLT_MIN_EXP -125 +#define CL_FLT_RADIX 2 +#define CL_FLT_MAX 0x1.fffffep127f +#define CL_FLT_MIN 0x1.0p-126f +#define CL_FLT_EPSILON 0x1.0p-23f + +#define CL_DBL_DIG 15 +#define CL_DBL_MANT_DIG 53 +#define CL_DBL_MAX_10_EXP +308 +#define CL_DBL_MAX_EXP +1024 +#define CL_DBL_MIN_10_EXP -307 +#define CL_DBL_MIN_EXP -1021 +#define CL_DBL_RADIX 2 +#define CL_DBL_MAX 0x1.fffffffffffffp1023 +#define CL_DBL_MIN 0x1.0p-1022 +#define CL_DBL_EPSILON 0x1.0p-52 + +#define CL_M_E 2.718281828459045090796 +#define CL_M_LOG2E 1.442695040888963387005 +#define CL_M_LOG10E 0.434294481903251816668 +#define CL_M_LN2 0.693147180559945286227 +#define CL_M_LN10 2.302585092994045901094 +#define CL_M_PI 3.141592653589793115998 +#define CL_M_PI_2 1.570796326794896557999 +#define CL_M_PI_4 0.785398163397448278999 +#define CL_M_1_PI 0.318309886183790691216 +#define CL_M_2_PI 0.636619772367581382433 +#define CL_M_2_SQRTPI 1.128379167095512558561 +#define CL_M_SQRT2 1.414213562373095145475 +#define CL_M_SQRT1_2 0.707106781186547572737 + +#define CL_M_E_F 2.71828174591064f +#define CL_M_LOG2E_F 1.44269502162933f +#define CL_M_LOG10E_F 0.43429449200630f +#define CL_M_LN2_F 0.69314718246460f +#define CL_M_LN10_F 2.30258512496948f +#define CL_M_PI_F 3.14159274101257f +#define CL_M_PI_2_F 1.57079637050629f +#define CL_M_PI_4_F 0.78539818525314f +#define CL_M_1_PI_F 0.31830987334251f +#define CL_M_2_PI_F 0.63661974668503f +#define CL_M_2_SQRTPI_F 1.12837922573090f +#define CL_M_SQRT2_F 1.41421353816986f +#define CL_M_SQRT1_2_F 0.70710676908493f + +#if defined( __GNUC__ ) + #define CL_HUGE_VALF __builtin_huge_valf() + #define CL_HUGE_VAL __builtin_huge_val() + #define CL_NAN __builtin_nanf( "" ) +#else + #define CL_HUGE_VALF ((cl_float) 1e50) + #define CL_HUGE_VAL ((cl_double) 1e500) + float nanf( const char * ); + #define CL_NAN nanf( "" ) +#endif +#define CL_MAXFLOAT CL_FLT_MAX +#define CL_INFINITY CL_HUGE_VALF + +#endif + +#include + +/* Mirror types to GL types. Mirror types allow us to avoid deciding which 87s to load based on whether we are using GL or GLES here. */ +typedef unsigned int cl_GLuint; +typedef int cl_GLint; +typedef unsigned int cl_GLenum; + +/* + * Vector types + * + * Note: OpenCL requires that all types be naturally aligned. + * This means that vector types must be naturally aligned. + * For example, a vector of four floats must be aligned to + * a 16 byte boundary (calculated as 4 * the natural 4-byte + * alignment of the float). The alignment qualifiers here + * will only function properly if your compiler supports them + * and if you don't actively work to defeat them. For example, + * in order for a cl_float4 to be 16 byte aligned in a struct, + * the start of the struct must itself be 16-byte aligned. + * + * Maintaining proper alignment is the user's responsibility. + */ + +/* Define basic vector types */ +#if defined( __VEC__ ) + #include /* may be omitted depending on compiler. AltiVec spec provides no way to detect whether the header is required. */ + typedef vector unsigned char __cl_uchar16; + typedef vector signed char __cl_char16; + typedef vector unsigned short __cl_ushort8; + typedef vector signed short __cl_short8; + typedef vector unsigned int __cl_uint4; + typedef vector signed int __cl_int4; + typedef vector float __cl_float4; + #define __CL_UCHAR16__ 1 + #define __CL_CHAR16__ 1 + #define __CL_USHORT8__ 1 + #define __CL_SHORT8__ 1 + #define __CL_UINT4__ 1 + #define __CL_INT4__ 1 + #define __CL_FLOAT4__ 1 +#endif + +#if defined( __SSE__ ) + #if defined( __MINGW64__ ) + #include + #else + #include + #endif + #if defined( __GNUC__ ) + typedef float __cl_float4 __attribute__((vector_size(16))); + #else + typedef __m128 __cl_float4; + #endif + #define __CL_FLOAT4__ 1 +#endif + +#if defined( __SSE2__ ) + #if defined( __MINGW64__ ) + #include + #else + #include + #endif + #if defined( __GNUC__ ) + typedef cl_uchar __cl_uchar16 __attribute__((vector_size(16))); + typedef cl_char __cl_char16 __attribute__((vector_size(16))); + typedef cl_ushort __cl_ushort8 __attribute__((vector_size(16))); + typedef cl_short __cl_short8 __attribute__((vector_size(16))); + typedef cl_uint __cl_uint4 __attribute__((vector_size(16))); + typedef cl_int __cl_int4 __attribute__((vector_size(16))); + typedef cl_ulong __cl_ulong2 __attribute__((vector_size(16))); + typedef cl_long __cl_long2 __attribute__((vector_size(16))); + typedef cl_double __cl_double2 __attribute__((vector_size(16))); + #else + typedef __m128i __cl_uchar16; + typedef __m128i __cl_char16; + typedef __m128i __cl_ushort8; + typedef __m128i __cl_short8; + typedef __m128i __cl_uint4; + typedef __m128i __cl_int4; + typedef __m128i __cl_ulong2; + typedef __m128i __cl_long2; + typedef __m128d __cl_double2; + #endif + #define __CL_UCHAR16__ 1 + #define __CL_CHAR16__ 1 + #define __CL_USHORT8__ 1 + #define __CL_SHORT8__ 1 + #define __CL_INT4__ 1 + #define __CL_UINT4__ 1 + #define __CL_ULONG2__ 1 + #define __CL_LONG2__ 1 + #define __CL_DOUBLE2__ 1 +#endif + +#if defined( __MMX__ ) + #include + #if defined( __GNUC__ ) + typedef cl_uchar __cl_uchar8 __attribute__((vector_size(8))); + typedef cl_char __cl_char8 __attribute__((vector_size(8))); + typedef cl_ushort __cl_ushort4 __attribute__((vector_size(8))); + typedef cl_short __cl_short4 __attribute__((vector_size(8))); + typedef cl_uint __cl_uint2 __attribute__((vector_size(8))); + typedef cl_int __cl_int2 __attribute__((vector_size(8))); + typedef cl_ulong __cl_ulong1 __attribute__((vector_size(8))); + typedef cl_long __cl_long1 __attribute__((vector_size(8))); + typedef cl_float __cl_float2 __attribute__((vector_size(8))); + #else + typedef __m64 __cl_uchar8; + typedef __m64 __cl_char8; + typedef __m64 __cl_ushort4; + typedef __m64 __cl_short4; + typedef __m64 __cl_uint2; + typedef __m64 __cl_int2; + typedef __m64 __cl_ulong1; + typedef __m64 __cl_long1; + typedef __m64 __cl_float2; + #endif + #define __CL_UCHAR8__ 1 + #define __CL_CHAR8__ 1 + #define __CL_USHORT4__ 1 + #define __CL_SHORT4__ 1 + #define __CL_INT2__ 1 + #define __CL_UINT2__ 1 + #define __CL_ULONG1__ 1 + #define __CL_LONG1__ 1 + #define __CL_FLOAT2__ 1 +#endif + +#if defined( __AVX__ ) + #if defined( __MINGW64__ ) + #include + #else + #include + #endif + #if defined( __GNUC__ ) + typedef cl_float __cl_float8 __attribute__((vector_size(32))); + typedef cl_double __cl_double4 __attribute__((vector_size(32))); + #else + typedef __m256 __cl_float8; + typedef __m256d __cl_double4; + #endif + #define __CL_FLOAT8__ 1 + #define __CL_DOUBLE4__ 1 +#endif + +/* Define capabilities for anonymous struct members. */ +#if defined( __GNUC__) && ! defined( __STRICT_ANSI__ ) +#define __CL_HAS_ANON_STRUCT__ 1 +#define __CL_ANON_STRUCT__ __extension__ +#elif defined( _WIN32) && (_MSC_VER >= 1500) + /* Microsoft Developer Studio 2008 supports anonymous structs, but + * complains by default. */ +#define __CL_HAS_ANON_STRUCT__ 1 +#define __CL_ANON_STRUCT__ + /* Disable warning C4201: nonstandard extension used : nameless + * struct/union */ +#pragma warning( push ) +#pragma warning( disable : 4201 ) +#else +#define __CL_HAS_ANON_STRUCT__ 0 +#define __CL_ANON_STRUCT__ +#endif + +/* Define alignment keys */ +#if defined( __GNUC__ ) + #define CL_ALIGNED(_x) __attribute__ ((aligned(_x))) +#elif defined( _WIN32) && (_MSC_VER) + /* Alignment keys neutered on windows because MSVC can't swallow function arguments with alignment requirements */ + /* http://msdn.microsoft.com/en-us/library/373ak2y1%28VS.71%29.aspx */ + /* #include */ + /* #define CL_ALIGNED(_x) _CRT_ALIGN(_x) */ + #define CL_ALIGNED(_x) +#else + #warning Need to implement some method to align data here + #define CL_ALIGNED(_x) +#endif + +/* Indicate whether .xyzw, .s0123 and .hi.lo are supported */ +#if __CL_HAS_ANON_STRUCT__ + /* .xyzw and .s0123...{f|F} are supported */ + #define CL_HAS_NAMED_VECTOR_FIELDS 1 + /* .hi and .lo are supported */ + #define CL_HAS_HI_LO_VECTOR_FIELDS 1 +#endif + +/* Define cl_vector types */ + +/* ---- cl_charn ---- */ +typedef union +{ + cl_char CL_ALIGNED(2) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_char x, y; }; + __CL_ANON_STRUCT__ struct{ cl_char s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_char lo, hi; }; +#endif +#if defined( __CL_CHAR2__) + __cl_char2 v2; +#endif +}cl_char2; + +typedef union +{ + cl_char CL_ALIGNED(4) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_char x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_char s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_char2 lo, hi; }; +#endif +#if defined( __CL_CHAR2__) + __cl_char2 v2[2]; +#endif +#if defined( __CL_CHAR4__) + __cl_char4 v4; +#endif +}cl_char4; + +/* cl_char3 is identical in size, alignment and behavior to cl_char4. See section 6.1.5. */ +typedef cl_char4 cl_char3; + +typedef union +{ + cl_char CL_ALIGNED(8) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_char x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_char s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_char4 lo, hi; }; +#endif +#if defined( __CL_CHAR2__) + __cl_char2 v2[4]; +#endif +#if defined( __CL_CHAR4__) + __cl_char4 v4[2]; +#endif +#if defined( __CL_CHAR8__ ) + __cl_char8 v8; +#endif +}cl_char8; + +typedef union +{ + cl_char CL_ALIGNED(16) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_char x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_char s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_char8 lo, hi; }; +#endif +#if defined( __CL_CHAR2__) + __cl_char2 v2[8]; +#endif +#if defined( __CL_CHAR4__) + __cl_char4 v4[4]; +#endif +#if defined( __CL_CHAR8__ ) + __cl_char8 v8[2]; +#endif +#if defined( __CL_CHAR16__ ) + __cl_char16 v16; +#endif +}cl_char16; + + +/* ---- cl_ucharn ---- */ +typedef union +{ + cl_uchar CL_ALIGNED(2) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uchar x, y; }; + __CL_ANON_STRUCT__ struct{ cl_uchar s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_uchar lo, hi; }; +#endif +#if defined( __cl_uchar2__) + __cl_uchar2 v2; +#endif +}cl_uchar2; + +typedef union +{ + cl_uchar CL_ALIGNED(4) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uchar x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_uchar s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_uchar2 lo, hi; }; +#endif +#if defined( __CL_UCHAR2__) + __cl_uchar2 v2[2]; +#endif +#if defined( __CL_UCHAR4__) + __cl_uchar4 v4; +#endif +}cl_uchar4; + +/* cl_uchar3 is identical in size, alignment and behavior to cl_uchar4. See section 6.1.5. */ +typedef cl_uchar4 cl_uchar3; + +typedef union +{ + cl_uchar CL_ALIGNED(8) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uchar x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_uchar s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_uchar4 lo, hi; }; +#endif +#if defined( __CL_UCHAR2__) + __cl_uchar2 v2[4]; +#endif +#if defined( __CL_UCHAR4__) + __cl_uchar4 v4[2]; +#endif +#if defined( __CL_UCHAR8__ ) + __cl_uchar8 v8; +#endif +}cl_uchar8; + +typedef union +{ + cl_uchar CL_ALIGNED(16) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uchar x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_uchar s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_uchar8 lo, hi; }; +#endif +#if defined( __CL_UCHAR2__) + __cl_uchar2 v2[8]; +#endif +#if defined( __CL_UCHAR4__) + __cl_uchar4 v4[4]; +#endif +#if defined( __CL_UCHAR8__ ) + __cl_uchar8 v8[2]; +#endif +#if defined( __CL_UCHAR16__ ) + __cl_uchar16 v16; +#endif +}cl_uchar16; + + +/* ---- cl_shortn ---- */ +typedef union +{ + cl_short CL_ALIGNED(4) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_short x, y; }; + __CL_ANON_STRUCT__ struct{ cl_short s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_short lo, hi; }; +#endif +#if defined( __CL_SHORT2__) + __cl_short2 v2; +#endif +}cl_short2; + +typedef union +{ + cl_short CL_ALIGNED(8) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_short x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_short s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_short2 lo, hi; }; +#endif +#if defined( __CL_SHORT2__) + __cl_short2 v2[2]; +#endif +#if defined( __CL_SHORT4__) + __cl_short4 v4; +#endif +}cl_short4; + +/* cl_short3 is identical in size, alignment and behavior to cl_short4. See section 6.1.5. */ +typedef cl_short4 cl_short3; + +typedef union +{ + cl_short CL_ALIGNED(16) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_short x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_short s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_short4 lo, hi; }; +#endif +#if defined( __CL_SHORT2__) + __cl_short2 v2[4]; +#endif +#if defined( __CL_SHORT4__) + __cl_short4 v4[2]; +#endif +#if defined( __CL_SHORT8__ ) + __cl_short8 v8; +#endif +}cl_short8; + +typedef union +{ + cl_short CL_ALIGNED(32) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_short x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_short s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_short8 lo, hi; }; +#endif +#if defined( __CL_SHORT2__) + __cl_short2 v2[8]; +#endif +#if defined( __CL_SHORT4__) + __cl_short4 v4[4]; +#endif +#if defined( __CL_SHORT8__ ) + __cl_short8 v8[2]; +#endif +#if defined( __CL_SHORT16__ ) + __cl_short16 v16; +#endif +}cl_short16; + + +/* ---- cl_ushortn ---- */ +typedef union +{ + cl_ushort CL_ALIGNED(4) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ushort x, y; }; + __CL_ANON_STRUCT__ struct{ cl_ushort s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_ushort lo, hi; }; +#endif +#if defined( __CL_USHORT2__) + __cl_ushort2 v2; +#endif +}cl_ushort2; + +typedef union +{ + cl_ushort CL_ALIGNED(8) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ushort x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_ushort s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_ushort2 lo, hi; }; +#endif +#if defined( __CL_USHORT2__) + __cl_ushort2 v2[2]; +#endif +#if defined( __CL_USHORT4__) + __cl_ushort4 v4; +#endif +}cl_ushort4; + +/* cl_ushort3 is identical in size, alignment and behavior to cl_ushort4. See section 6.1.5. */ +typedef cl_ushort4 cl_ushort3; + +typedef union +{ + cl_ushort CL_ALIGNED(16) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ushort x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_ushort s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_ushort4 lo, hi; }; +#endif +#if defined( __CL_USHORT2__) + __cl_ushort2 v2[4]; +#endif +#if defined( __CL_USHORT4__) + __cl_ushort4 v4[2]; +#endif +#if defined( __CL_USHORT8__ ) + __cl_ushort8 v8; +#endif +}cl_ushort8; + +typedef union +{ + cl_ushort CL_ALIGNED(32) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ushort x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_ushort s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_ushort8 lo, hi; }; +#endif +#if defined( __CL_USHORT2__) + __cl_ushort2 v2[8]; +#endif +#if defined( __CL_USHORT4__) + __cl_ushort4 v4[4]; +#endif +#if defined( __CL_USHORT8__ ) + __cl_ushort8 v8[2]; +#endif +#if defined( __CL_USHORT16__ ) + __cl_ushort16 v16; +#endif +}cl_ushort16; + +/* ---- cl_intn ---- */ +typedef union +{ + cl_int CL_ALIGNED(8) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_int x, y; }; + __CL_ANON_STRUCT__ struct{ cl_int s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_int lo, hi; }; +#endif +#if defined( __CL_INT2__) + __cl_int2 v2; +#endif +}cl_int2; + +typedef union +{ + cl_int CL_ALIGNED(16) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_int x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_int s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_int2 lo, hi; }; +#endif +#if defined( __CL_INT2__) + __cl_int2 v2[2]; +#endif +#if defined( __CL_INT4__) + __cl_int4 v4; +#endif +}cl_int4; + +/* cl_int3 is identical in size, alignment and behavior to cl_int4. See section 6.1.5. */ +typedef cl_int4 cl_int3; + +typedef union +{ + cl_int CL_ALIGNED(32) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_int x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_int s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_int4 lo, hi; }; +#endif +#if defined( __CL_INT2__) + __cl_int2 v2[4]; +#endif +#if defined( __CL_INT4__) + __cl_int4 v4[2]; +#endif +#if defined( __CL_INT8__ ) + __cl_int8 v8; +#endif +}cl_int8; + +typedef union +{ + cl_int CL_ALIGNED(64) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_int x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_int8 lo, hi; }; +#endif +#if defined( __CL_INT2__) + __cl_int2 v2[8]; +#endif +#if defined( __CL_INT4__) + __cl_int4 v4[4]; +#endif +#if defined( __CL_INT8__ ) + __cl_int8 v8[2]; +#endif +#if defined( __CL_INT16__ ) + __cl_int16 v16; +#endif +}cl_int16; + + +/* ---- cl_uintn ---- */ +typedef union +{ + cl_uint CL_ALIGNED(8) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uint x, y; }; + __CL_ANON_STRUCT__ struct{ cl_uint s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_uint lo, hi; }; +#endif +#if defined( __CL_UINT2__) + __cl_uint2 v2; +#endif +}cl_uint2; + +typedef union +{ + cl_uint CL_ALIGNED(16) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uint x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_uint s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_uint2 lo, hi; }; +#endif +#if defined( __CL_UINT2__) + __cl_uint2 v2[2]; +#endif +#if defined( __CL_UINT4__) + __cl_uint4 v4; +#endif +}cl_uint4; + +/* cl_uint3 is identical in size, alignment and behavior to cl_uint4. See section 6.1.5. */ +typedef cl_uint4 cl_uint3; + +typedef union +{ + cl_uint CL_ALIGNED(32) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uint x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_uint s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_uint4 lo, hi; }; +#endif +#if defined( __CL_UINT2__) + __cl_uint2 v2[4]; +#endif +#if defined( __CL_UINT4__) + __cl_uint4 v4[2]; +#endif +#if defined( __CL_UINT8__ ) + __cl_uint8 v8; +#endif +}cl_uint8; + +typedef union +{ + cl_uint CL_ALIGNED(64) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_uint x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_uint s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_uint8 lo, hi; }; +#endif +#if defined( __CL_UINT2__) + __cl_uint2 v2[8]; +#endif +#if defined( __CL_UINT4__) + __cl_uint4 v4[4]; +#endif +#if defined( __CL_UINT8__ ) + __cl_uint8 v8[2]; +#endif +#if defined( __CL_UINT16__ ) + __cl_uint16 v16; +#endif +}cl_uint16; + +/* ---- cl_longn ---- */ +typedef union +{ + cl_long CL_ALIGNED(16) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_long x, y; }; + __CL_ANON_STRUCT__ struct{ cl_long s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_long lo, hi; }; +#endif +#if defined( __CL_LONG2__) + __cl_long2 v2; +#endif +}cl_long2; + +typedef union +{ + cl_long CL_ALIGNED(32) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_long x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_long s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_long2 lo, hi; }; +#endif +#if defined( __CL_LONG2__) + __cl_long2 v2[2]; +#endif +#if defined( __CL_LONG4__) + __cl_long4 v4; +#endif +}cl_long4; + +/* cl_long3 is identical in size, alignment and behavior to cl_long4. See section 6.1.5. */ +typedef cl_long4 cl_long3; + +typedef union +{ + cl_long CL_ALIGNED(64) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_long x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_long s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_long4 lo, hi; }; +#endif +#if defined( __CL_LONG2__) + __cl_long2 v2[4]; +#endif +#if defined( __CL_LONG4__) + __cl_long4 v4[2]; +#endif +#if defined( __CL_LONG8__ ) + __cl_long8 v8; +#endif +}cl_long8; + +typedef union +{ + cl_long CL_ALIGNED(128) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_long x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_long s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_long8 lo, hi; }; +#endif +#if defined( __CL_LONG2__) + __cl_long2 v2[8]; +#endif +#if defined( __CL_LONG4__) + __cl_long4 v4[4]; +#endif +#if defined( __CL_LONG8__ ) + __cl_long8 v8[2]; +#endif +#if defined( __CL_LONG16__ ) + __cl_long16 v16; +#endif +}cl_long16; + + +/* ---- cl_ulongn ---- */ +typedef union +{ + cl_ulong CL_ALIGNED(16) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ulong x, y; }; + __CL_ANON_STRUCT__ struct{ cl_ulong s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_ulong lo, hi; }; +#endif +#if defined( __CL_ULONG2__) + __cl_ulong2 v2; +#endif +}cl_ulong2; + +typedef union +{ + cl_ulong CL_ALIGNED(32) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ulong x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_ulong s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_ulong2 lo, hi; }; +#endif +#if defined( __CL_ULONG2__) + __cl_ulong2 v2[2]; +#endif +#if defined( __CL_ULONG4__) + __cl_ulong4 v4; +#endif +}cl_ulong4; + +/* cl_ulong3 is identical in size, alignment and behavior to cl_ulong4. See section 6.1.5. */ +typedef cl_ulong4 cl_ulong3; + +typedef union +{ + cl_ulong CL_ALIGNED(64) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ulong x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_ulong s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_ulong4 lo, hi; }; +#endif +#if defined( __CL_ULONG2__) + __cl_ulong2 v2[4]; +#endif +#if defined( __CL_ULONG4__) + __cl_ulong4 v4[2]; +#endif +#if defined( __CL_ULONG8__ ) + __cl_ulong8 v8; +#endif +}cl_ulong8; + +typedef union +{ + cl_ulong CL_ALIGNED(128) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_ulong x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_ulong s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_ulong8 lo, hi; }; +#endif +#if defined( __CL_ULONG2__) + __cl_ulong2 v2[8]; +#endif +#if defined( __CL_ULONG4__) + __cl_ulong4 v4[4]; +#endif +#if defined( __CL_ULONG8__ ) + __cl_ulong8 v8[2]; +#endif +#if defined( __CL_ULONG16__ ) + __cl_ulong16 v16; +#endif +}cl_ulong16; + + +/* --- cl_floatn ---- */ + +typedef union +{ + cl_float CL_ALIGNED(8) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_float x, y; }; + __CL_ANON_STRUCT__ struct{ cl_float s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_float lo, hi; }; +#endif +#if defined( __CL_FLOAT2__) + __cl_float2 v2; +#endif +}cl_float2; + +typedef union +{ + cl_float CL_ALIGNED(16) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_float x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_float s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_float2 lo, hi; }; +#endif +#if defined( __CL_FLOAT2__) + __cl_float2 v2[2]; +#endif +#if defined( __CL_FLOAT4__) + __cl_float4 v4; +#endif +}cl_float4; + +/* cl_float3 is identical in size, alignment and behavior to cl_float4. See section 6.1.5. */ +typedef cl_float4 cl_float3; + +typedef union +{ + cl_float CL_ALIGNED(32) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_float x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_float s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_float4 lo, hi; }; +#endif +#if defined( __CL_FLOAT2__) + __cl_float2 v2[4]; +#endif +#if defined( __CL_FLOAT4__) + __cl_float4 v4[2]; +#endif +#if defined( __CL_FLOAT8__ ) + __cl_float8 v8; +#endif +}cl_float8; + +typedef union +{ + cl_float CL_ALIGNED(64) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_float x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_float s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_float8 lo, hi; }; +#endif +#if defined( __CL_FLOAT2__) + __cl_float2 v2[8]; +#endif +#if defined( __CL_FLOAT4__) + __cl_float4 v4[4]; +#endif +#if defined( __CL_FLOAT8__ ) + __cl_float8 v8[2]; +#endif +#if defined( __CL_FLOAT16__ ) + __cl_float16 v16; +#endif +}cl_float16; + +/* --- cl_doublen ---- */ + +typedef union +{ + cl_double CL_ALIGNED(16) s[2]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_double x, y; }; + __CL_ANON_STRUCT__ struct{ cl_double s0, s1; }; + __CL_ANON_STRUCT__ struct{ cl_double lo, hi; }; +#endif +#if defined( __CL_DOUBLE2__) + __cl_double2 v2; +#endif +}cl_double2; + +typedef union +{ + cl_double CL_ALIGNED(32) s[4]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_double x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_double s0, s1, s2, s3; }; + __CL_ANON_STRUCT__ struct{ cl_double2 lo, hi; }; +#endif +#if defined( __CL_DOUBLE2__) + __cl_double2 v2[2]; +#endif +#if defined( __CL_DOUBLE4__) + __cl_double4 v4; +#endif +}cl_double4; + +/* cl_double3 is identical in size, alignment and behavior to cl_double4. See section 6.1.5. */ +typedef cl_double4 cl_double3; + +typedef union +{ + cl_double CL_ALIGNED(64) s[8]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_double x, y, z, w; }; + __CL_ANON_STRUCT__ struct{ cl_double s0, s1, s2, s3, s4, s5, s6, s7; }; + __CL_ANON_STRUCT__ struct{ cl_double4 lo, hi; }; +#endif +#if defined( __CL_DOUBLE2__) + __cl_double2 v2[4]; +#endif +#if defined( __CL_DOUBLE4__) + __cl_double4 v4[2]; +#endif +#if defined( __CL_DOUBLE8__ ) + __cl_double8 v8; +#endif +}cl_double8; + +typedef union +{ + cl_double CL_ALIGNED(128) s[16]; +#if __CL_HAS_ANON_STRUCT__ + __CL_ANON_STRUCT__ struct{ cl_double x, y, z, w, __spacer4, __spacer5, __spacer6, __spacer7, __spacer8, __spacer9, sa, sb, sc, sd, se, sf; }; + __CL_ANON_STRUCT__ struct{ cl_double s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, sA, sB, sC, sD, sE, sF; }; + __CL_ANON_STRUCT__ struct{ cl_double8 lo, hi; }; +#endif +#if defined( __CL_DOUBLE2__) + __cl_double2 v2[8]; +#endif +#if defined( __CL_DOUBLE4__) + __cl_double4 v4[4]; +#endif +#if defined( __CL_DOUBLE8__ ) + __cl_double8 v8[2]; +#endif +#if defined( __CL_DOUBLE16__ ) + __cl_double16 v16; +#endif +}cl_double16; + +/* Macro to facilitate debugging + * Usage: + * Place CL_PROGRAM_STRING_DEBUG_INFO on the line before the first line of your source. + * The first line ends with: CL_PROGRAM_STRING_DEBUG_INFO \" + * Each line thereafter of OpenCL C source must end with: \n\ + * The last line ends in "; + * + * Example: + * + * const char *my_program = CL_PROGRAM_STRING_DEBUG_INFO "\ + * kernel void foo( int a, float * b ) \n\ + * { \n\ + * // my comment \n\ + * *b[ get_global_id(0)] = a; \n\ + * } \n\ + * "; + * + * This should correctly set up the line, (column) and file information for your source + * string so you can do source level debugging. + */ +#define __CL_STRINGIFY( _x ) # _x +#define _CL_STRINGIFY( _x ) __CL_STRINGIFY( _x ) +#define CL_PROGRAM_STRING_DEBUG_INFO "#line " _CL_STRINGIFY(__LINE__) " \"" __FILE__ "\" \n\n" + +#ifdef __cplusplus +} +#endif + +#undef __CL_HAS_ANON_STRUCT__ +#undef __CL_ANON_STRUCT__ +#if defined( _WIN32) && (_MSC_VER >= 1500) +#pragma warning( pop ) +#endif + +#endif /* __CL_PLATFORM_H */ diff --git a/third_party/opencl/include/CL/opencl.h b/third_party/opencl/include/CL/opencl.h new file mode 100644 index 000000000..9855cd75e --- /dev/null +++ b/third_party/opencl/include/CL/opencl.h @@ -0,0 +1,59 @@ +/******************************************************************************* + * Copyright (c) 2008-2015 The Khronos Group Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and/or associated documentation files (the + * "Materials"), to deal in the Materials without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Materials, and to + * permit persons to whom the Materials are furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Materials. + * + * MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS + * KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS + * SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT + * https://www.khronos.org/registry/ + * + * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY + * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS. + ******************************************************************************/ + +/* $Revision: 11708 $ on $Date: 2010-06-13 23:36:24 -0700 (Sun, 13 Jun 2010) $ */ + +#ifndef __OPENCL_H +#define __OPENCL_H + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef __APPLE__ + +#include +#include +#include +#include + +#else + +#include +#include +#include +#include + +#endif + +#ifdef __cplusplus +} +#endif + +#endif /* __OPENCL_H */ + diff --git a/tools/cabana/SConscript b/tools/cabana/SConscript index b79d046fc..d4cfc6728 100644 --- a/tools/cabana/SConscript +++ b/tools/cabana/SConscript @@ -57,9 +57,11 @@ base_frameworks = qt_env['FRAMEWORKS'] base_libs = [common, messaging, cereal, visionipc, 'm', 'ssl', 'crypto', 'pthread'] + qt_env["LIBS"] if arch == "Darwin": + base_frameworks.append('OpenCL') base_frameworks.append('QtCharts') base_frameworks.append('QtSerialBus') else: + base_libs.append('OpenCL') base_libs.append('Qt5Charts') base_libs.append('Qt5SerialBus') diff --git a/tools/install_ubuntu_dependencies.sh b/tools/install_ubuntu_dependencies.sh index f428e9972..5c2131d4b 100755 --- a/tools/install_ubuntu_dependencies.sh +++ b/tools/install_ubuntu_dependencies.sh @@ -58,6 +58,9 @@ function install_ubuntu_common_requirements() { libzmq3-dev \ libzstd-dev \ libsqlite3-dev \ + opencl-headers \ + ocl-icd-libopencl1 \ + ocl-icd-opencl-dev \ portaudio19-dev \ qttools5-dev-tools \ libqt5svg5-dev \ diff --git a/tools/replay/SConscript b/tools/replay/SConscript index 698ab9885..136c4119f 100644 --- a/tools/replay/SConscript +++ b/tools/replay/SConscript @@ -6,6 +6,11 @@ replay_env['CCFLAGS'] += ['-Wno-deprecated-declarations'] base_frameworks = [] base_libs = [common, messaging, cereal, visionipc, 'm', 'ssl', 'crypto', 'pthread'] +if arch == "Darwin": + base_frameworks.append('OpenCL') +else: + base_libs.append('OpenCL') + replay_lib_src = ["replay.cc", "consoleui.cc", "camera.cc", "filereader.cc", "logreader.cc", "framereader.cc", "route.cc", "util.cc", "seg_mgr.cc", "timeline.cc", "api.cc"] if arch != "Darwin": diff --git a/tools/webcam/README.md b/tools/webcam/README.md index 6abbc4793..67ad2cc8c 100644 --- a/tools/webcam/README.md +++ b/tools/webcam/README.md @@ -10,6 +10,10 @@ What's needed: ## Setup openpilot - Follow [this readme](../README.md) to install and build the requirements +- Install OpenCL Driver (Ubuntu) +``` +sudo apt install pocl-opencl-icd +``` ## Connect the hardware - Connect the camera first