Skip to content

P
Paddle-Lite

PaddlePaddle / Paddle-Lite

通知 332
Star 4
Fork 1
P
Paddle-Lite
提交 13f66d39 编写于 作者: qnqinan's avatar qnqinan
浏览文件
操作

update, test=develop

上级 a8577694
展开全部 隐藏空白更改
内联 并排
Showing with 428 addition and 5868 deletion
cmake/cuda.cmake
浏览文件 @ 13f66d39
......@@ -4,9 +4,9 @@ endif()
set(paddle_known_gpu_archs "30 35 50 52 60 61 70")
set(paddle_known_gpu_archs7 "30 35 50 52")
set(paddle_known_gpu_archs8 "30 35 50 52 60 61")
set(paddle_known_gpu_archs9 "30 35 50 52 60 61 62 70")
set(paddle_known_gpu_archs10 "30 35 50 52 60 61 62 70 75")
set(paddle_known_gpu_archs8 "30 35 50 52 53 60 61 62")
set(paddle_known_gpu_archs9 "30 35 50 52 53 60 61 62 70")
set(paddle_known_gpu_archs10 "30 35 50 52 53 60 61 62 70 72 75")
######################################################################################
# A function for automatic detection of GPUs installed (if autodetection is enabled)
......
cmake/xpu.cmake
浏览文件 @ 13f66d39
......@@ -99,5 +99,7 @@ else()
set_property(TARGET xpu_sdk_llvm PROPERTY IMPORTED_LOCATION ${XPU_SDK_LLVM_FILE})
endif()
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DDMLC_USE_GLOG=1")
set(xpu_runtime_libs xpu_sdk_xtcl xpu_sdk_tvm xpu_sdk_xpu_api xpu_sdk_xpu_rt xpu_sdk_xpu_jitc xpu_sdk_llvm CACHE INTERNAL "xpu runtime libs")
set(xpu_builder_libs xpu_sdk_xtcl xpu_sdk_tvm xpu_sdk_xpu_api xpu_sdk_xpu_rt xpu_sdk_xpu_jitc xpu_sdk_llvm CACHE INTERNAL "xpu builder libs")
lite/CMakeLists.txt
浏览文件 @ 13f66d39
......@@ -112,10 +112,7 @@ if (LITE_WITH_X86)
COMMAND mkdir -p "${INFER_LITE_PUBLISH_ROOT}/demo/cxx"
)
add_dependencies(publish_inference_x86_cxx_lib publish_inference_x86_cxx_demos)
add_dependencies(publish_inference_x86_cxx_demos gflags eigen3 xbyak xxhash glog)
if(WITH_MKL)
add_dependencies(publish_inference_x86_cxx_demos mklml)
endif()
add_dependencies(publish_inference_x86_cxx_demos paddle_full_api_shared eigen3)
endif()
if (LITE_WITH_LIGHT_WEIGHT_FRAMEWORK AND LITE_WITH_ARM)
......
lite/api/benchmark.cc
浏览文件 @ 13f66d39
......@@ -32,7 +32,9 @@ DEFINE_string(input_shape,
DEFINE_string(result_filename, "", "save test result");
DEFINE_bool(run_model_optimize,
false,
"apply model_optimize_tool to model, use optimized model to test");
"if set true, apply model_optimize_tool to model, use optimized "
"model to test");
DEFINE_bool(is_quantized_model, false, "if set true, test the quantized model");
namespace paddle {
namespace lite_api {
......@@ -42,12 +44,14 @@ void OutputOptModel(const std::string& load_model_dir,
const std::vector<std::vector<int64_t>>& input_shapes) {
lite_api::CxxConfig config;
config.set_model_dir(load_model_dir);
config.set_valid_places({
Place{TARGET(kARM), PRECISION(kFloat)},
Place{TARGET(kARM), PRECISION(kInt8)},
Place{TARGET(kX86), PRECISION(kFloat)},
Place{TARGET(kOpenCL), PRECISION(kFloat)},
});
std::vector<Place> vaild_places = {Place{TARGET(kARM), PRECISION(kFloat)},
Place{TARGET(kX86), PRECISION(kFloat)},
Place{TARGET(kOpenCL), PRECISION(kFloat)}};
if (FLAGS_is_quantized_model) {
vaild_places.insert(vaild_places.begin(),
Place{TARGET(kARM), PRECISION(kInt8)});
}
config.set_valid_places(vaild_places);
auto predictor = lite_api::CreatePaddlePredictor(config);
int ret = system(
......
lite/api/test_googlenet_lite.cc
浏览文件 @ 13f66d39
......@@ -59,8 +59,7 @@ TEST(CXXApi, test_lite_googlenet) {
<< ", warmup: " << FLAGS_warmup << ", repeats: " << FLAGS_repeats
<< ", spend " << (GetCurrentUS() - start) / FLAGS_repeats / 1000.0
<< " ms in average.";
// TODO(FrostML): open this check.
/*auto out = predictor->GetOutput(0);
auto out = predictor->GetOutput(0);
std::vector<float> results(
{0.00034298553, 0.0008200012, 0.0005046297, 0.000839279,
0.00052616704, 0.0003447803, 0.0010877076, 0.00081762316,
......@@ -72,7 +71,7 @@ TEST(CXXApi, test_lite_googlenet) {
}
ASSERT_EQ(out->shape().size(), 2);
ASSERT_EQ(out->shape()[0], 1);
ASSERT_EQ(out->shape()[1], 1000);*/
ASSERT_EQ(out->shape()[1], 1000);
}
#endif
} // namespace lite
......
lite/api/test_inceptionv4_lite_x86.cc
浏览文件 @ 13f66d39
......@@ -59,8 +59,7 @@ TEST(InceptionV4, test_inceptionv4_lite_x86) {
<< ", repeats: " << FLAGS_repeats << ", spend "
<< (GetCurrentUS() - start) / FLAGS_repeats / 1000.0
<< " ms in average.";
// TODO(FrostML): open this check.
/*std::vector<std::vector<float>> results;
std::vector<std::vector<float>> results;
// i = 1
results.emplace_back(std::vector<float>(
{0.0011684548, 0.0010390386, 0.0011301535, 0.0010133048,
......@@ -81,7 +80,7 @@ TEST(InceptionV4, test_inceptionv4_lite_x86) {
results[i][j],
1e-6);
}
}*/
}
}
} // namespace lite
......
lite/api/test_mobilenetv1_lite_x86.cc
浏览文件 @ 13f66d39
......@@ -59,8 +59,7 @@ TEST(Mobilenet_v1, test_mobilenetv1_lite_x86) {
<< ", repeats: " << FLAGS_repeats << ", spend "
<< (GetCurrentUS() - start) / FLAGS_repeats / 1000.0
<< " ms in average.";
// TODO(FrostML): open this check.
/*std::vector<std::vector<float>> results;
std::vector<std::vector<float>> results;
// i = 1
results.emplace_back(std::vector<float>(
{0.00019130898, 9.467885e-05, 0.00015971427, 0.0003650665,
......@@ -80,7 +79,7 @@ TEST(Mobilenet_v1, test_mobilenetv1_lite_x86) {
results[i][j],
1e-6);
}
}*/
}
}
} // namespace lite
......
lite/api/test_mobilenetv2_lite_x86.cc
浏览文件 @ 13f66d39
......@@ -60,8 +60,7 @@ TEST(Mobilenet_v2, test_mobilenetv2_lite_x86) {
<< ", repeats: " << FLAGS_repeats << ", spend "
<< (GetCurrentUS() - start) / FLAGS_repeats / 1000.0
<< " ms in average.";
// TODO(FrostML): open this check.
/*std::vector<std::vector<float>> results;
std::vector<std::vector<float>> results;
// i = 1
results.emplace_back(std::vector<float>(
{0.00017082224, 5.699624e-05, 0.000260885, 0.00016412718,
......@@ -81,7 +80,7 @@ TEST(Mobilenet_v2, test_mobilenetv2_lite_x86) {
results[i][j],
1e-6);
}
}*/
}
}
} // namespace lite
......
lite/api/test_resnet50_lite_x86.cc
浏览文件 @ 13f66d39
......@@ -60,8 +60,7 @@ TEST(Resnet50, test_resnet50_lite_x86) {
<< ", spend " << (GetCurrentUS() - start) / FLAGS_repeats / 1000.0
<< " ms in average.";
// TODO(FrostML): open this check.
/*std::vector<std::vector<float>> results;
std::vector<std::vector<float>> results;
// i = 1
results.emplace_back(std::vector<float>(
{0.00024139918, 0.00020566184, 0.00022418296, 0.00041731037,
......@@ -81,7 +80,7 @@ TEST(Resnet50, test_resnet50_lite_x86) {
results[i][j],
1e-6);
}
}*/
}
}
} // namespace lite
......
lite/kernels/arm/CMakeLists.txt
浏览文件 @ 13f66d39
# for conv op
add_kernel(conv_depthwise ARM basic SRCS conv_depthwise.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_direct ARM basic SRCS conv_direct.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_gemmlike ARM basic SRCS conv_gemmlike.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_winograd ARM basic SRCS conv_winograd.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_compute_arm ARM basic SRCS conv_compute.cc DEPS ${lite_kernel_deps}
add_kernel(conv_depthwise ARM extra SRCS conv_depthwise.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_direct ARM extra SRCS conv_direct.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_gemmlike ARM extra SRCS conv_gemmlike.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_winograd ARM extra SRCS conv_winograd.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(conv_compute_arm ARM extra SRCS conv_compute.cc DEPS ${lite_kernel_deps}
conv_depthwise conv_direct conv_gemmlike conv_winograd)
add_kernel(fc_compute_arm ARM basic SRCS fc_compute.cc DEPS ${lite_kernel_deps} math_arm)
......@@ -40,24 +40,24 @@ add_kernel(shape_compute_arm ARM basic SRCS shape_compute.cc DEPS ${lite_kernel_
add_kernel(slice_compute_arm ARM basic SRCS slice_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(cast_compute_arm ARM basic SRCS cast_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(squeeze_compute_arm ARM basic SRCS squeeze_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(unsqueeze_compute_arm ARM basic SRCS unsqueeze_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(unsqueeze_compute_arm ARM extra SRCS unsqueeze_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(expand_compute_arm ARM basic SRCS expand_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(reduce_max_compute_arm ARM basic SRCS reduce_max_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(sequence_expand_compute_arm ARM basic SRCS sequence_expand_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(im2sequence_compute_arm ARM basic SRCS im2sequence_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(sequence_pool_compute_arm ARM basic SRCS sequence_pool_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(layer_norm_compute_arm ARM basic SRCS layer_norm_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(gather_compute_arm ARM basic SRCS gather_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(reduce_mean_compute_arm ARM basic SRCS reduce_mean_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(stack_compute_arm ARM basic SRCS stack_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(assign_compute_arm ARM basic SRCS assign_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(affine_channel_compute_arm ARM basic SRCS affine_channel_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(anchor_generator_compute_arm ARM basic SRCS anchor_generator_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(generate_proposals_compute_arm ARM basic SRCS generate_proposals_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(roi_align_compute_arm ARM basic SRCS roi_align_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(box_clip_compute_arm ARM basic SRCS box_clip_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(range_compute_arm ARM basic SRCS range_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(assign_value_compute_arm ARM basic SRCS assign_value_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(layer_norm_compute_arm ARM extra SRCS layer_norm_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(gather_compute_arm ARM extra SRCS gather_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(reduce_mean_compute_arm ARM extra SRCS reduce_mean_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(stack_compute_arm ARM extra SRCS stack_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(assign_compute_arm ARM extra SRCS assign_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(affine_channel_compute_arm ARM extra SRCS affine_channel_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(anchor_generator_compute_arm ARM extra SRCS anchor_generator_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(generate_proposals_compute_arm ARM extra SRCS generate_proposals_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(roi_align_compute_arm ARM extra SRCS roi_align_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(box_clip_compute_arm ARM extra SRCS box_clip_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(range_compute_arm ARM extra SRCS range_compute.cc DEPS ${lite_kernel_deps} math_arm)
add_kernel(assign_value_compute_arm ARM extra SRCS assign_value_compute.cc DEPS ${lite_kernel_deps} math_arm)
# for OCR specific
add_kernel(gru_unit_compute_arm ARM extra SRCS gru_unit_compute.cc DEPS ${lite_kernel_deps} math_arm)
......
lite/kernels/host/CMakeLists.txt
浏览文件 @ 13f66d39
......@@ -3,7 +3,7 @@ message(STATUS "compile with lite host kernels")
add_kernel(feed_compute_host Host basic SRCS feed_compute.cc DEPS ${lite_kernel_deps})
add_kernel(fetch_compute_host Host basic SRCS fetch_compute.cc DEPS ${lite_kernel_deps})
add_kernel(reshape_compute_host Host basic SRCS reshape_compute.cc DEPS ${lite_kernel_deps} reshape_op)
add_kernel(multiclass_nms_compute_host Host basic SRCS multiclass_nms_compute.cc DEPS ${lite_kernel_deps})
add_kernel(multiclass_nms_compute_host Host extra SRCS multiclass_nms_compute.cc DEPS ${lite_kernel_deps})
#lite_cc_test(test_reshape_compute_host SRCS reshape_compute_test.cc DEPS reshape_compute_host any)
#lite_cc_test(test_multiclass_nms_compute_host SRCS multiclass_nms_compute_test.cc DEPS multiclass_nms_compute_host any)
lite/kernels/xpu/bridges/CMakeLists.txt
浏览文件 @ 13f66d39
......@@ -8,6 +8,7 @@ lite_cc_library(xpu_bridge_elementwise_ops SRCS elementwise_ops.cc DEPS ${xpu_br
lite_cc_library(xpu_bridge_pool_op SRCS pool_op.cc DEPS ${xpu_bridge_deps})
lite_cc_library(xpu_bridge_softmax_op SRCS softmax_op.cc DEPS ${xpu_bridge_deps})
lite_cc_library(xpu_bridge_mul_op SRCS mul_op.cc DEPS ${xpu_bridge_deps})
lite_cc_library(xpu_bridge_batch_norm_op SRCS batch_norm_op.cc DEPS ${xpu_bridge_deps})
set(xpu_bridges
xpu_bridge_registry
......@@ -17,6 +18,7 @@ set(xpu_bridges
xpu_bridge_pool_op
xpu_bridge_softmax_op
xpu_bridge_mul_op
xpu_bridge_batch_norm_op
CACHE INTERNAL "xpu_bridges")
set(xpu_bridge_test_deps ${xpu_bridges} ${xpu_kernels} ${ops})
......@@ -27,3 +29,4 @@ lite_cc_test(test_xpu_bridge_elementwise_ops SRCS elementwise_ops_test.cc test_h
lite_cc_test(test_xpu_bridge_pool_op SRCS pool_op_test.cc test_helper.cc DEPS ${xpu_bridge_test_deps})
lite_cc_test(test_xpu_bridge_softmax_op SRCS softmax_op_test.cc test_helper.cc DEPS ${xpu_bridge_test_deps})
lite_cc_test(test_xpu_bridge_mul_op SRCS mul_op_test.cc test_helper.cc DEPS ${xpu_bridge_test_deps})
lite_cc_test(test_xpu_bridge_batch_norm_op SRCS batch_norm_op_test.cc test_helper.cc DEPS ${xpu_bridge_test_deps})
lite/kernels/xpu/bridges/batch_norm_op.cc 0 → 100644
浏览文件 @ 13f66d39
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "lite/backends/xpu/builder.h"
#include "lite/kernels/xpu/bridges/registry.h"
namespace paddle {
namespace lite {
namespace kernels {
namespace xpu {
namespace bridges {
node_map_type BatchNormConverter(const std::shared_ptr<lite::OpLite> op,
graph_ctx_type* graph_ctx,
const node_map_type& input_nodes) {
auto scope = op->scope();
auto op_info = op->op_info();
auto op_type = op_info->Type();
auto unique_op_type = lite::xpu::UniqueName(op_type);
LOG(INFO) << "[XPU] Converting " + op_type + "...";
// check context
CHECK(graph_ctx != nullptr);
CHECK(graph_ctx->builder != nullptr);
CHECK(graph_ctx->params != nullptr);
// get input, and attributes
auto x_var_name = op_info->Input("X").front();
auto scale_var_name = op_info->Input("Scale").front();
auto* scale = scope->FindMutableTensor(scale_var_name);
auto bias_var_name = op_info->Input("Bias").front();
auto* bias = scope->FindMutableTensor(bias_var_name);
auto mean_var_name = op_info->Input("Mean").front();
auto* mean = scope->FindMutableTensor(mean_var_name);
auto variance_var_name = op_info->Input("Variance").front();
auto* variance = scope->FindMutableTensor(variance_var_name);
auto epsilon = op_info->GetAttr<float>("epsilon");
// create scale node
CHECK(!input_nodes.count(scale_var_name));
auto scale_const_node = std::make_shared<xtcl::xExpr>(
graph_ctx->builder->CreateTensor(scale_var_name,
lite::xpu::CvtShape(scale->dims()),
::xtcl::Float(32)));
auto scale_const_tensor = lite::xpu::CvtTensor(scale);
graph_ctx->params->emplace(
std::make_pair(scale_var_name, *scale_const_tensor));
// create bias node
CHECK(!input_nodes.count(bias_var_name));
auto bias_const_node =
std::make_shared<xtcl::xExpr>(graph_ctx->builder->CreateTensor(
bias_var_name, lite::xpu::CvtShape(bias->dims()), ::xtcl::Float(32)));
auto bias_const_tensor = lite::xpu::CvtTensor(bias);
graph_ctx->params->emplace(std::make_pair(bias_var_name, *bias_const_tensor));
// create mean node
CHECK(!input_nodes.count(mean_var_name));
auto mean_const_node =
std::make_shared<xtcl::xExpr>(graph_ctx->builder->CreateTensor(
mean_var_name, lite::xpu::CvtShape(mean->dims()), ::xtcl::Float(32)));
auto mean_const_tensor = lite::xpu::CvtTensor(mean);
graph_ctx->params->emplace(std::make_pair(mean_var_name, *mean_const_tensor));
// create variance node
CHECK(!input_nodes.count(variance_var_name));
auto variance_const_node = std::make_shared<xtcl::xExpr>(
graph_ctx->builder->CreateTensor(variance_var_name,
lite::xpu::CvtShape(variance->dims()),
::xtcl::Float(32)));
auto variance_const_tensor = lite::xpu::CvtTensor(variance);
graph_ctx->params->emplace(
std::make_pair(variance_var_name, *variance_const_tensor));
// create batch_norm node and set params from op
CHECK(input_nodes.count(x_var_name));
auto batch_norm_node = std::make_shared<xtcl::xExpr>(
graph_ctx->builder->CreateBatchNorm(*input_nodes.at(x_var_name),
*scale_const_node,
*bias_const_node,
*mean_const_node,
*variance_const_node,
1,
epsilon));
batch_norm_node = std::make_shared<xtcl::xExpr>(
graph_ctx->builder->GetField(*batch_norm_node, 0));
graph_ctx->builder->SetLayer(unique_op_type);
// output converted nodes
node_map_type output_nodes;
output_nodes[op_info->Output("Y").front()] = batch_norm_node;
return output_nodes;
}
} // namespace bridges
} // namespace xpu
} // namespace kernels
} // namespace lite
} // namespace paddle
REGISTER_XPU_BRIDGE(batch_norm,
paddle::lite::kernels::xpu::bridges::BatchNormConverter);
lite/kernels/xpu/bridges/batch_norm_op_test.cc 0 → 100644
浏览文件 @ 13f66d39
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "lite/operators/batch_norm_op.h"
#include <gtest/gtest.h>
#include "lite/core/op_registry.h"
#include "lite/kernels/xpu/bridges/registry.h"
#include "lite/kernels/xpu/bridges/test_helper.h"
namespace paddle {
namespace lite {
namespace kernels {
namespace xpu {
namespace bridges {
template <typename dtype>
void batch_norm_ref(const std::shared_ptr<operators::BatchNormOp> op) {
Scope* scope = op->scope();
const OpInfo* op_info = op->op_info();
auto x = scope->FindVar(op_info->Input("X").front())->GetMutable<Tensor>();
auto y = scope->FindVar(op_info->Output("Y").front())->GetMutable<Tensor>();
auto bias =
scope->FindVar(op_info->Input("Bias").front())->GetMutable<Tensor>();
auto scale =
scope->FindVar(op_info->Input("Scale").front())->GetMutable<Tensor>();
auto mean =
scope->FindVar(op_info->Input("Mean").front())->GetMutable<Tensor>();
auto variance =
scope->FindVar(op_info->Input("Variance").front())->GetMutable<Tensor>();
auto x_data = x->data<dtype>();
auto y_data = y->mutable_data<dtype>();
auto scale_data = scale->mutable_data<dtype>();
auto bias_data = bias->mutable_data<dtype>();
auto mean_data = mean->mutable_data<dtype>();
auto variance_data = variance->mutable_data<dtype>();
DDim x_dims = x->dims();
float epsilon = op_info->GetAttr<float>("epsilon");
auto data_layout = op_info->GetAttr<std::string>("data_layout");
bool global_stats = op_info->GetAttr<bool>("use_global_stats");
if (global_stats) {
int64_t outer_size = 0;
int64_t channel_size = 0;
int64_t inner_size = 0;
if (data_layout == "NCHW") {
outer_size = x_dims[0];
channel_size = x_dims[1];
inner_size = x_dims.Slice(2, x_dims.size()).production();
} else {
LOG(FATAL) << "Unknown storage order: " << data_layout;
}
auto x_ptr = x_data;
auto y_ptr = y_data;
for (int o = 0; o < outer_size; o++) {
for (int c = 0; c < channel_size; c++) {
for (int i = 0; i < inner_size; i++) {
dtype norm_x =
(*x_ptr - mean_data[c]) / std::sqrt(variance_data[c] + epsilon);
*y_ptr = norm_x * scale_data[c] + bias_data[c];
x_ptr++;
y_ptr++;
}
}
}
}
}
void test_batch_norm(int bs, int ic, int ih, int iw, float epsilon) {
// prepare input&output variables
Scope scope;
std::string x_var_name = "x";
std::string out_var_name = "out";
std::string out_ref_var_name = "out_ref";
std::string scale_var_name = "scale";
std::string bias_var_name = "bias";
std::string mean_var_name = "mean";
std::string variance_var_name = "variance";
auto* x = scope.Var(x_var_name)->GetMutable<Tensor>();
auto* scale = scope.Var(scale_var_name)->GetMutable<Tensor>();
auto* bias = scope.Var(bias_var_name)->GetMutable<Tensor>();
auto* mean = scope.Var(mean_var_name)->GetMutable<Tensor>();
auto* variance = scope.Var(variance_var_name)->GetMutable<Tensor>();
auto* out = scope.Var(out_var_name)->GetMutable<Tensor>();
auto* out_ref = scope.Var(out_ref_var_name)->GetMutable<Tensor>();
x->Resize({bs, ic, ih, iw});
scale->Resize({ic});
bias->Resize({ic});
mean->Resize({ic});
variance->Resize({ic});
// initialize input&output data
FillTensor<float>(x);
FillTensor<float>(scale);
FillTensor<float>(bias);
FillTensor<float>(mean);
// variance > 0
FillTensor<float>(variance, 1.f, 5.f);
// initialize op desc
cpp::OpDesc opdesc;
opdesc.SetType("batch_norm");
opdesc.SetInput("X", {x_var_name});
opdesc.SetInput("Scale", {scale_var_name});
opdesc.SetInput("Bias", {bias_var_name});
opdesc.SetInput("Mean", {mean_var_name});
opdesc.SetInput("Variance", {variance_var_name});
opdesc.SetOutput("Y", {out_var_name});
opdesc.SetAttr("is_test", 1);
opdesc.SetAttr("use_global_stats", true);
opdesc.SetAttr("epsilon", epsilon);
opdesc.SetAttr("momentum", 0.9f);
opdesc.SetAttr("data_layout", std::string("NCHW"));
// create and convert op to XPU model, then run it on XPU
auto op = CreateOp<operators::BatchNormOp>(opdesc, &scope);
LauchOp(op, {x_var_name}, {out_var_name});
out_ref->CopyDataFrom(*out);
// execute reference implementation and save to output tensor
batch_norm_ref<float>(op);
// compare results
auto* out_data = out->mutable_data<float>();
auto* out_ref_data = out_ref->mutable_data<float>();
for (int i = 0; i < out->dims().production(); i++) {
EXPECT_NEAR(out_data[i], out_ref_data[i], 1e-5);
}
}
TEST(NPUBridges, batch_norm) {
for (auto bs : {1, 3}) {
for (auto ic : {2, 3}) {
for (auto ih : {4}) {
for (auto iw : {5}) {
for (auto epsilon : {1e-5f}) {
test_batch_norm(bs, ic, ih, iw, epsilon);
}
}
}
}
}
}
} // namespace bridges
} // namespace xpu
} // namespace kernels
} // namespace lite
} // namespace paddle
USE_LITE_OP(batch_norm);
USE_XPU_BRIDGE(batch_norm);
lite/kernels/xpu/bridges/paddle_use_xpu_bridges.h
浏览文件 @ 13f66d39
......@@ -23,3 +23,4 @@ USE_XPU_BRIDGE(elementwise_add);
USE_XPU_BRIDGE(pool2d);
USE_XPU_BRIDGE(softmax);
USE_XPU_BRIDGE(mul);
USE_XPU_BRIDGE(batch_norm);
lite/operators/CMakeLists.txt
浏览文件 @ 13f66d39
......@@ -5,7 +5,7 @@ lite_cc_library(op_params SRCS op_params.cc DEPS tensor any)
add_operator(conv_op basic SRCS conv_op.cc DEPS ${op_DEPS})
add_operator(pool_op basic SRCS pool_op.cc DEPS ${op_DEPS})
add_operator(fc_op basic SRCS fc_op.cc DEPS ${op_DEPS})
add_operator(assign_op basic SRCS assign_op.cc DEPS ${op_DEPS})
add_operator(assign_op extra SRCS assign_op.cc DEPS ${op_DEPS})
add_operator(relu_op basic SRCS relu_op.cc DEPS ${op_DEPS})
add_operator(mul_op basic SRCS mul_op.cc DEPS ${op_DEPS})
add_operator(matmul_op basic SRCS matmul_op.cc DEPS ${op_DEPS})
......@@ -59,23 +59,23 @@ add_operator(norm_op basic SRCS norm_op.cc DEPS ${op_DEPS})
add_operator(shape_op_lite basic SRCS shape_op.cc DEPS ${op_DEPS})
add_operator(sequence_expand_op_lite basic SRCS sequence_expand_op.cc DEPS ${op_DEPS})
add_operator(squeeze_op_lite basic SRCS squeeze_op.cc DEPS ${op_DEPS})
add_operator(unsqueeze_op_lite basic SRCS unsqueeze_op.cc DEPS ${op_DEPS})
add_operator(unsqueeze_op_lite extra SRCS unsqueeze_op.cc DEPS ${op_DEPS})
add_operator(im2sequence_op basic SRCS im2sequence_op.cc DEPS ${op_DEPS})
add_operator(gather_op basic SRCS gather_op.cc DEPS ${op_DEPS})
add_operator(reduce_mean_op basic SRCS reduce_mean_op.cc DEPS ${op_DEPS})
add_operator(stack_op basic SRCS stack_op.cc DEPS ${op_DEPS})
add_operator(cast_op_lite basic SRCS cast_op.cc DEPS ${op_DEPS})
add_operator(affine_channel_op basic SRCS affine_channel_op.cc DEPS ${op_DEPS})
add_operator(anchor_generator_op basic SRCS anchor_generator_op.cc DEPS ${op_DEPS})
add_operator(generate_proposals_op basic SRCS generate_proposals_op.cc DEPS ${op_DEPS})
add_operator(roi_align_op basic SRCS roi_align_op.cc DEPS ${op_DEPS})
add_operator(box_clip_op basic SRCS box_clip_op.cc DEPS ${op_DEPS})
add_operator(flatten_op basic SRCS flatten_op.cc DEPS ${op_DEPS})
add_operator(fake_quantize_range_abs_max_op basic SRCS fake_quantize_range_abs_max.cc DEPS ${op_DEPS})
add_operator(sequence_expand_as_op_lite basic SRCS sequence_expand_as_op.cc DEPS ${op_DEPS})
add_operator(range_op basic SRCS range_op.cc DEPS ${op_DEPS})
add_operator(assign_value_op basic SRCS assign_value_op.cc DEPS ${op_DEPS})
add_operator(fake_quantize_dequantize_moving_avg_abs_max_op basic SRCS fake_quantize_dequantize_moving_avg_max_abs.cc DEPS ${op_DEPS})
add_operator(gather_op extra SRCS gather_op.cc DEPS ${op_DEPS})
add_operator(reduce_mean_op extra SRCS reduce_mean_op.cc DEPS ${op_DEPS})
add_operator(stack_op extra SRCS stack_op.cc DEPS ${op_DEPS})
add_operator(cast_op_lite extra SRCS cast_op.cc DEPS ${op_DEPS})
add_operator(affine_channel_op extra SRCS affine_channel_op.cc DEPS ${op_DEPS})
add_operator(anchor_generator_op extra SRCS anchor_generator_op.cc DEPS ${op_DEPS})
add_operator(generate_proposals_op extra SRCS generate_proposals_op.cc DEPS ${op_DEPS})
add_operator(roi_align_op extra SRCS roi_align_op.cc DEPS ${op_DEPS})
add_operator(box_clip_op extra SRCS box_clip_op.cc DEPS ${op_DEPS})
add_operator(flatten_op extra SRCS flatten_op.cc DEPS ${op_DEPS})
add_operator(fake_quantize_range_abs_max_op extra SRCS fake_quantize_range_abs_max.cc DEPS ${op_DEPS})
add_operator(sequence_expand_as_op_lite extra SRCS sequence_expand_as_op.cc DEPS ${op_DEPS})
add_operator(range_op extra SRCS range_op.cc DEPS ${op_DEPS})
add_operator(assign_value_op extra SRCS assign_value_op.cc DEPS ${op_DEPS})
add_operator(fake_quantize_dequantize_moving_avg_abs_max_op extra SRCS fake_quantize_dequantize_moving_avg_max_abs.cc DEPS ${op_DEPS})
# for OCR specific
add_operator(while_op extra SRCS while_op.cc DEPS ${op_DEPS})
......
lite/tools/benchmark.sh
浏览文件 @ 13f66d39
......@@ -8,6 +8,7 @@ then
echo "Usage:"
echo " sh benchmark.sh <benchmark_bin_path> <benchmark_models_path> <result_filename>"
echo " sh benchmark.sh <benchmark_bin_path> <benchmark_models_path> <result_filename> <is_run_model_optimize: [true|false]>"
echo " sh benchmark.sh <benchmark_bin_path> <benchmark_models_path> <result_filename> <is_run_model_optimize: [true|false]> <is_run_quantized_model: [trur|false]>"
exit
fi
......@@ -20,6 +21,7 @@ RESULT_FILENAME=$3
WARMUP=10
REPEATS=30
IS_RUN_MODEL_OPTIMIZE=false
IS_RUN_QUANTIZED_MODEL=false
NUM_THREADS_LIST=(1 2 4)
MODELS_LIST=$(ls $MODELS_DIR)
......@@ -28,6 +30,10 @@ if [ $# -gt 3 ];
then
IS_RUN_MODEL_OPTIMIZE=$4
fi
if [ $# -gt 4 ];
then
IS_RUN_QUANTIZED_MODEL=$5
fi
# Adb push benchmark_bin, models
adb push $BENCHMARK_BIN $ANDROID_DIR/benchmark_bin
......@@ -46,7 +52,8 @@ for threads in ${NUM_THREADS_LIST[@]}; do
--repeats=$REPEATS \
--threads=$threads \
--result_filename=$ANDROID_DIR/$RESULT_FILENAME \
--run_model_optimize=$IS_RUN_MODEL_OPTIMIZE"
--run_model_optimize=$IS_RUN_MODEL_OPTIMIZE \
--is_quantized_model=$IS_RUN_QUANTIZED_MODEL"
done
adb shell "echo >> $ANDROID_DIR/$RESULT_FILENAME"
done
......
lite/tools/build.sh
浏览文件 @ 13f66d39
......@@ -231,7 +231,8 @@ function make_cuda {
-DLITE_WITH_LIGHT_WEIGHT_FRAMEWORK=OFF \
-DWITH_TESTING=OFF \
-DLITE_WITH_ARM=OFF \
-DLITE_WITH_PYTHON=ON
-DLITE_WITH_PYTHON=ON \
-DLITE_BUILD_EXTRA=ON
make publish_inference_python_lib -j8
cd -
......
mobile/.gitignore
浏览文件 @ 13f66d39
......@@ -101,3 +101,4 @@ metal/paddle-mobile-demo/paddle-mobile-demo/Resources
metal/paddle-mobile-demo/paddle-mobile-demo/Resources/images
metal/paddle-mobile-demo/paddle-mobile-demo/Resources/models
metal/MobileNetDemo/MobileNetDemo/Resources
third_party/opencl/OpenCL-Headers
mobile/src/fpga/V2/bias_scale.cpp
浏览文件 @ 13f66d39
......@@ -14,6 +14,7 @@ limitations under the License. */
#include "fpga/V2/bias_scale.h"
#include <memory.h>
#include <math.h>
#include "fpga/common/fpga_common.h"
namespace paddle_mobile {
......@@ -55,10 +56,22 @@ void align_element(float **data_in, int num_per_div_before_alignment, int num) {
*data_in = ptr_aligned;
}
void fixed_scale_bias_new(void*data_in, int data_len) {
int* data_tmp = static_cast<int*>(data_in);
for (int idx = 0; idx < data_len/2; ++idx) {
float tmp = (static_cast<float*>(data_in))[idx];
data_tmp[idx] = static_cast<int>(round(tmp*pow(2.0, 23.0)));
tmp = (static_cast<float*>(data_in))[idx+data_len/2];
data_tmp[idx+data_len/2] = static_cast<int>(round(tmp*pow(2.0, 30.0)));
}
return;
}
void interleave(float **data_in, int num_after_alignment) {
// num_after_alignment: number of bias after alignment
float *ptr_uninterleaved = *data_in;
// fixed_scale_bias_new(ptr_uninterleaved, 2 * num_after_alignment);
float *ptr_interleaved =
(float *)fpga_malloc(2 * num_after_alignment * sizeof(float)); // NOLINT
int num = num_after_alignment / 4;
......
mobile/src/fpga/V2/pe.cpp
浏览文件 @ 13f66d39
......@@ -79,7 +79,8 @@ using namespace std; // NOLINT
#define REG_CONVERT_CMD 0x400
#define REG_CONVERT_SRC_ADDR 0x408
#define REG_CONVERT_DST_ADDR 0x410
#define REG_CONVERT_LENGTH 0x418
#define REG_CONVERT_RD_LENGTH 0x418
#define REG_CONVERT_WR_LENGTH 0x420
/*resize*/
#define REG_RESIZE_CMD 0x600
......@@ -693,7 +694,8 @@ int PerformBypass(const struct BypassArgs &args) {
reg_writeq(output_scale, REG_SCALE_PARAMETER);
reg_writeq(input_address_phy, REG_CONVERT_SRC_ADDR);
reg_writeq(output_address_phy, REG_CONVERT_DST_ADDR);
reg_writeq(datalen, REG_CONVERT_LENGTH);
reg_writeq(datalen, REG_CONVERT_RD_LENGTH);
reg_writeq(datalen, REG_CONVERT_WR_LENGTH);
reg_writeq(cmd, REG_CONVERT_CMD);
DLOG << "before reg poll";
if (0 != fpga_regpoll(REG_INTERRUPT, INTERRUPT_BYPASS, PE_IRQ_TIMEOUT)) {
......
mobile/src/fpga/common/driver.cpp
浏览文件 @ 13f66d39
......@@ -134,6 +134,7 @@ int fpga_regpoll(uint64_t reg, uint64_t val, int time) {
uint64_t i = 0;
/*timeout精确性待确认*/
int64_t timeout = time * 6;
usleep(1);
for (i = 0; i < timeout; i++) {
if (val == reg_readq(reg)) {
......
mobile/src/framework/cl/cl_image.h
浏览文件 @ 13f66d39
......@@ -135,7 +135,7 @@ class CLImage {
// CLImageConverterFolder *folder_converter = new
// CLImageConverterFolder();
CLImageConverterNormal *normal_converter = new CLImageConverterNormal();
PADDLE_MOBILE_ENFORCE(!shared_mem_, "do not init mem after shared .")
DLOG << " to get image dims ";
image_dims_ = normal_converter->InitImageDimInfoWith(dim);
DLOG << " end get image dims " << image_dims_;
......@@ -176,7 +176,9 @@ class CLImage {
image_converter_ = normal_converter;
cl_event_ = CLEngine::Instance()->CreateEvent(context);
initialized_ = true;
DLOG << " end init cl image";
shared_mem_ = true;
DLOG << " end init FakeSizeImage";
}
/**
* init cl mem with a exist cl mem
......@@ -194,17 +196,18 @@ class CLImage {
DLOG << "InitWithExistMem ... ";
DLOG << "real_image_dims: " << real_image_dims_;
DLOG << "image_dims_: " << image_dims_;
// PADDLE_MOBILE_ENFORCE(real_image_dims[0] >= image_dims_[0] &&
// real_image_dims[1] >= image_dims_[1],
// "real image is not enough!");
if (real_image_dims_[0] < image_dims_[0] ||
real_image_dims_[1] < image_dims_[1]) {
DLOG << "real image is not enough!";
DLOG << "real_image_dims: " << real_image_dims_;
DLOG << "image_dims_: " << image_dims_;
}
PADDLE_MOBILE_ENFORCE(real_image_dims_[0] >= image_dims_[0] &&
real_image_dims_[1] >= image_dims_[1],
"real image is not enough!");
if (cl_image_ != src.cl_image_) {
cl_image_.reset(src.cl_image_.get());
cl_image_.reset(src.cl_image_.get(), CLMemDeleter());
}
tensor_dims_ = need_dims;
......@@ -212,7 +215,9 @@ class CLImage {
image_converter_ = normal_converter;
cl_event_ = CLEngine::Instance()->CreateEvent(context);
initialized_ = true;
DLOG << " end init cl image";
shared_mem_ = true;
DLOG << " end init WithExistMem";
}
void InitConv2dTransposeFilterCLImage(cl_context context,
......@@ -231,7 +236,7 @@ class CLImage {
"Tensor holds no memory. Call Tensor::mutable_data first.")
if (cl_image_ != src.cl_image_) {
cl_image_.reset(src.cl_image_.get());
cl_image_.reset(src.cl_image_.get(), CLMemDeleter());
}
return *this;
}
......@@ -281,6 +286,8 @@ class CLImage {
private:
void InitCLImage(cl_context context, size_t width, size_t height,
void *data) {
PADDLE_MOBILE_ENFORCE(!shared_mem_, "do not init mem after shared .")
cl_image_format cf = {.image_channel_order = CL_RGBA,
.image_channel_data_type = CL_HALF_FLOAT};
cl_image_desc cid = {
......@@ -303,7 +310,7 @@ class CLImage {
&cid, // const cl_image_desc *image_desc
data, // void *host_ptr
&err);
cl_image_.reset(cl_image);
cl_image_.reset(cl_image, CLMemDeleter());
if (err != CL_SUCCESS) {
CL_CHECK_ERRORS(err);
PADDLE_MOBILE_THROW_EXCEPTION(" create image 2d error ");
......@@ -311,7 +318,7 @@ class CLImage {
}
bool initialized_ = false;
std::unique_ptr<_cl_mem, CLMemDeleter> cl_image_;
std::shared_ptr<_cl_mem> cl_image_;
std::unique_ptr<_cl_event, CLEventDeleter> cl_event_;
DDim tensor_dims_;
DDim image_dims_;
......@@ -321,6 +328,7 @@ class CLImage {
cl_context context_;
cl_command_queue command_queue_;
CLImageConverterBase *image_converter_ = nullptr;
bool shared_mem_ = false;
};
void TensorToCLImage(Tensor *tensor, CLImage *image, cl_context context,
......
mobile/src/operators/kernel/cl/fusion_fc_kernel.cpp
浏览文件 @ 13f66d39
......@@ -98,7 +98,7 @@ void FusionFcCompute(const FusionFcParam<GPU_CL> &param, cl_context context,
static_cast<float>(1), out, static_cast<float>(1),
false);
out_image->InitEmptyImage(context, commandQueue, out->dims());
// out_image->InitEmptyImage(context, commandQueue, out->dims());
framework::TensorToCLImage(out, out_image, context, commandQueue, kernel1);
delete (input_x);
......
mobile/src/operators/kernel/cl/mul_kernel.cpp
浏览文件 @ 13f66d39
......@@ -63,7 +63,7 @@ void MulCompute(const MulParam<GPU_CL> &param, cl_context context,
static_cast<float>(1), output_tensor,
static_cast<float>(0));
output->InitEmptyImage(context, commandQueue, output_tensor->dims());
// output->InitEmptyImage(context, commandQueue, output_tensor->dims());
framework::TensorToCLImage(output_tensor, output, context, commandQueue,
kernel1);
......
mobile/src/operators/kernel/fpga/V2/anchor_generator_kernel.cpp
浏览文件 @ 13f66d39
......@@ -45,9 +45,9 @@ bool AnchorGeneratorKernel<FPGA, float>::Init(
if (offset > 0.6) {
memcpy(anchors_offset, anchors_offset2, sizeof(anchors_offset));
std::cout << "anchor generator marker" << std::endl;
DLOG << "anchor generator marker";
} else {
std::cout << "anchor generator rfcn" << std::endl;
DLOG << "anchor generator rfcn";
}
int num_anchors = sizeof(anchors_offset) / (sizeof(int) * 4);
......
mobile/src/operators/kernel/fpga/V2/proposal_kernel.cpp
浏览文件 @ 13f66d39
......@@ -83,7 +83,7 @@ void AppendProposals(Tensor *dst, int64_t offset, const Tensor &src) {
}
template <class T>
static inline void BoxCoder(Tensor *all_anchors, Tensor *bbox_deltas,
static inline void BoxCoder(Tensor *all_anchors, Tensor *bbox_deltas,
Tensor *proposals) {
T *proposals_data = proposals->mutable_data<T>();
......@@ -368,8 +368,8 @@ void ProposalKernel<FPGA, float>::Compute(const ProposalParam<FPGA> &param) {
score_height * alignedCW * sizeof(int8_t));
Tensor score_tensor = *input_score;
for(int h = 0; h < score_height; h++){
for(int w = 0; w < score_width; w++){
for (int h = 0; h < score_height; h++) {
for (int w = 0; w < score_width; w++) {
for (int c = 0; c < score_channels; ++c) {
int dstidx = h*unalignedCW + w*score_channels + c;
int srcidx = h*alignedCW + w*score_channels + c;
......@@ -385,12 +385,14 @@ void ProposalKernel<FPGA, float>::Compute(const ProposalParam<FPGA> &param) {
bbox_height * alignedCW * sizeof(int8_t));
auto bbox_tensor = param.float_bbox.get();
for(int h = 0; h < bbox_height; h++){
for(int w = 0; w < bbox_width; w++){
for (int h = 0; h < bbox_height; h++) {
for (int w = 0; w < bbox_width; w++) {
for (int c = 0; c < bbox_channels; ++c) {
int dstidx = h*unalignedCW + w*bbox_channels + c;
int srcidx = h*alignedCW + w*bbox_channels + c;
bbox_tensor->data<float>()[dstidx] = ((int)(input_bbox_data[srcidx])) / 127.0 * input_bbox->scale[0];
bbox_tensor->data<float>()[dstidx] =
(static_cast<int>(input_bbox_data[srcidx]))/127.0*
input_bbox->scale[0];
}
}
}
......
mobile/src/operators/kernel/fpga/V2/psroi_pool_kernel.cpp
浏览文件 @ 13f66d39
......@@ -176,8 +176,8 @@ void PSRoiPoolKernel<FPGA, float>::Compute(const PSRoiPoolParam<FPGA>& param) {
PSROIPoolingForward<float>(
input_data, height, width, input_channels, offset_output_data,
pooled_height, pooled_width, output_channels, input_rois,
bin_size_h, bin_size_w, roi_start_h, roi_start_w, pw, ph,
scale, roi_batch_ind);
bin_size_h, bin_size_w, roi_start_h, roi_start_w, pw, ph,
scale, roi_batch_ind);
}
}
}
......
mobile/src/operators/kernel/fpga/V2/reshape2_kernel.cpp
浏览文件 @ 13f66d39
......@@ -93,29 +93,6 @@ void reshape(LoDTensor *input, LoDTensor *output) {
fpga::fpga_flush(output_ptr, Hr * WCr_align * sizeof(int8_t));
}
static inline bool reshape2_judge(const framework::DDim input_dims,const framework::DDim output_dims){
int input_dims_size = input_dims.size();
int output_dims_size = output_dims.size();
bool dims_flag2 = true;
auto temp_dims = input_dims_size > output_dims_size ? input_dims : output_dims;
int short_dims = input_dims_size > output_dims_size ? output_dims_size : input_dims_size;
for(int i = 0; i < temp_dims.size(); ++i){
if(i < short_dims){
if(input_dims[i] != output_dims[i]){
dims_flag2 = false;
break;
}
}
else{
if(temp_dims[i] != 1){
dims_flag2 = false;
break;
}
}
}
return dims_flag2;
}
template <>
void Reshape2Kernel<FPGA, float>::Compute(const Reshape2Param<FPGA> &param) {
auto input = const_cast<LoDTensor *>(param.InputX());
......@@ -133,17 +110,7 @@ void Reshape2Kernel<FPGA, float>::Compute(const Reshape2Param<FPGA> &param) {
}
}
output->Resize(framework::make_ddim(shape));
auto input_dims = input->dims();
auto output_dims = output->dims();
bool dims_flags = input_dims == output_dims;
bool dims_flag2 = true;
if(!dims_flags){
dims_flag2 = reshape2_judge(input_dims, output_dims);
}
if (dims_flags || dims_flag2) {
if (output->dims() == input->dims()) {
DLOG << "No need to reshape";
output->ShareDataWith(*input);
framework::LoD lod = input->lod();
......
mobile/src/operators/kernel/fpga/V2/sigmoid_kernel.cpp
浏览文件 @ 13f66d39
......@@ -21,11 +21,12 @@ namespace operators {
template <>
bool SigmoidKernel<FPGA, float>::Init(SigmoidParam<FPGA> *param) {
paddle_mobile::fpga::ActivationType activation_enable =
paddle_mobile::fpga::SIGMOID;
int16_t leaky_relu_negative_slope = 0;
auto input = const_cast<LoDTensor *>(param->InputX());
auto input_ptr = input->data<int8_t>();
paddle_mobile::fpga::ActivationType activation_enable =
paddle_mobile::fpga::SIGMOID;
int16_t leaky_relu_negative_slope =
fpga::fp32_2_fp16(input->scale[0] / 127.0);
auto out = param->Out();
fpga::format_ofm(out);
......
mobile/third_party/opencl/OpenCL-Headers/CL/cl_d3d10.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* 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 <d3d10.h>
#include <CL/cl.h>
#include <CL/cl_platform.h>
#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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_d3d11.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* 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 <d3d11.h>
#include <CL/cl.h>
#include <CL/cl_platform.h>
#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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_dx9_media_sharing.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* 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 <CL/cl.h>
#include <CL/cl_platform.h>
#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 <d3d9.h>
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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_dx9_media_sharing_intel.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* Copyright (c) 2008-2016 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.
**********************************************************************************/
/*****************************************************************************\
Copyright (c) 2013-2016 Intel Corporation All Rights Reserved.
THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
File Name: cl_dx9_media_sharing_intel.h
Abstract:
Notes:
\*****************************************************************************/
#ifndef __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H
#define __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H
#include <CL/cl.h>
#include <CL/cl_platform.h>
#include <d3d9.h>
#include <dxvahd.h>
#include <wtypes.h>
#include <d3d9types.h>
#ifdef __cplusplus
extern "C" {
#endif
/***************************************
* cl_intel_dx9_media_sharing extension *
****************************************/
#define cl_intel_dx9_media_sharing 1
typedef cl_uint cl_dx9_device_source_intel;
typedef cl_uint cl_dx9_device_set_intel;
/* error codes */
#define CL_INVALID_DX9_DEVICE_INTEL -1010
#define CL_INVALID_DX9_RESOURCE_INTEL -1011
#define CL_DX9_RESOURCE_ALREADY_ACQUIRED_INTEL -1012
#define CL_DX9_RESOURCE_NOT_ACQUIRED_INTEL -1013
/* cl_dx9_device_source_intel */
#define CL_D3D9_DEVICE_INTEL 0x4022
#define CL_D3D9EX_DEVICE_INTEL 0x4070
#define CL_DXVA_DEVICE_INTEL 0x4071
/* cl_dx9_device_set_intel */
#define CL_PREFERRED_DEVICES_FOR_DX9_INTEL 0x4024
#define CL_ALL_DEVICES_FOR_DX9_INTEL 0x4025
/* cl_context_info */
#define CL_CONTEXT_D3D9_DEVICE_INTEL 0x4026
#define CL_CONTEXT_D3D9EX_DEVICE_INTEL 0x4072
#define CL_CONTEXT_DXVA_DEVICE_INTEL 0x4073
/* cl_mem_info */
#define CL_MEM_DX9_RESOURCE_INTEL 0x4027
#define CL_MEM_DX9_SHARED_HANDLE_INTEL 0x4074
/* cl_image_info */
#define CL_IMAGE_DX9_PLANE_INTEL 0x4075
/* cl_command_type */
#define CL_COMMAND_ACQUIRE_DX9_OBJECTS_INTEL 0x402A
#define CL_COMMAND_RELEASE_DX9_OBJECTS_INTEL 0x402B
/******************************************************************************/
extern CL_API_ENTRY cl_int CL_API_CALL
clGetDeviceIDsFromDX9INTEL(
cl_platform_id /* platform */,
cl_dx9_device_source_intel /* dx9_device_source */,
void* /* dx9_object */,
cl_dx9_device_set_intel /* dx9_device_set */,
cl_uint /* num_entries */,
cl_device_id* /* devices */,
cl_uint* /* num_devices */) CL_EXT_SUFFIX__VERSION_1_1;
typedef CL_API_ENTRY cl_int (CL_API_CALL* clGetDeviceIDsFromDX9INTEL_fn)(
cl_platform_id /* platform */,
cl_dx9_device_source_intel /* dx9_device_source */,
void* /* dx9_object */,
cl_dx9_device_set_intel /* dx9_device_set */,
cl_uint /* num_entries */,
cl_device_id* /* devices */,
cl_uint* /* num_devices */) CL_EXT_SUFFIX__VERSION_1_1;
extern CL_API_ENTRY cl_mem CL_API_CALL
clCreateFromDX9MediaSurfaceINTEL(
cl_context /* context */,
cl_mem_flags /* flags */,
IDirect3DSurface9* /* resource */,
HANDLE /* sharedHandle */,
UINT /* plane */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1;
typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateFromDX9MediaSurfaceINTEL_fn)(
cl_context /* context */,
cl_mem_flags /* flags */,
IDirect3DSurface9* /* resource */,
HANDLE /* sharedHandle */,
UINT /* plane */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_1;
extern CL_API_ENTRY cl_int CL_API_CALL
clEnqueueAcquireDX9ObjectsINTEL(
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_EXT_SUFFIX__VERSION_1_1;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireDX9ObjectsINTEL_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_EXT_SUFFIX__VERSION_1_1;
extern CL_API_ENTRY cl_int CL_API_CALL
clEnqueueReleaseDX9ObjectsINTEL(
cl_command_queue /* command_queue */,
cl_uint /* num_objects */,
cl_mem* /* mem_objects */,
cl_uint /* num_events_in_wait_list */,
const cl_event* /* event_wait_list */,
cl_event* /* event */) CL_EXT_SUFFIX__VERSION_1_1;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseDX9ObjectsINTEL_fn)(
cl_command_queue /* command_queue */,
cl_uint /* num_objects */,
cl_mem* /* mem_objects */,
cl_uint /* num_events_in_wait_list */,
const cl_event* /* event_wait_list */,
cl_event* /* event */) CL_EXT_SUFFIX__VERSION_1_1;
#ifdef __cplusplus
}
#endif
#endif /* __OPENCL_CL_DX9_MEDIA_SHARING_INTEL_H */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_egl.h 已删除 100644 → 0
浏览文件 @ a8577694
/*******************************************************************************
* 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 <CL/cl.h>
#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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_ext_intel.h 已删除 100644 → 0
浏览文件 @ a8577694
/*******************************************************************************
* Copyright (c) 2008-2017 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.
******************************************************************************/
/*****************************************************************************\
Copyright (c) 2013-2017 Intel Corporation All Rights Reserved.
THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
File Name: cl_ext_intel.h
Abstract:
Notes:
\*****************************************************************************/
#ifndef __CL_EXT_INTEL_H
#define __CL_EXT_INTEL_H
#ifdef __APPLE__
#include <OpenCL/cl.h>
#include <OpenCL/cl_platform.h>
#else
#include <CL/cl.h>
#include <CL/cl_platform.h>
#endif
#ifdef __cplusplus
extern "C" {
#endif
/***************************************
* cl_intel_thread_local_exec extension *
****************************************/
#define cl_intel_thread_local_exec 1
#define CL_QUEUE_THREAD_LOCAL_EXEC_ENABLE_INTEL (((cl_bitfield)1) << 31)
/***********************************************
* cl_intel_device_partition_by_names extension *
************************************************/
#define cl_intel_device_partition_by_names 1
#define CL_DEVICE_PARTITION_BY_NAMES_INTEL 0x4052
#define CL_PARTITION_BY_NAMES_LIST_END_INTEL -1
/************************************************
* cl_intel_accelerator extension *
* cl_intel_motion_estimation extension *
* cl_intel_advanced_motion_estimation extension *
*************************************************/
#define cl_intel_accelerator 1
#define cl_intel_motion_estimation 1
#define cl_intel_advanced_motion_estimation 1
typedef struct _cl_accelerator_intel* cl_accelerator_intel;
typedef cl_uint cl_accelerator_type_intel;
typedef cl_uint cl_accelerator_info_intel;
typedef struct _cl_motion_estimation_desc_intel {
cl_uint mb_block_type;
cl_uint subpixel_mode;
cl_uint sad_adjust_mode;
cl_uint search_path_type;
} cl_motion_estimation_desc_intel;
/* error codes */
#define CL_INVALID_ACCELERATOR_INTEL -1094
#define CL_INVALID_ACCELERATOR_TYPE_INTEL -1095
#define CL_INVALID_ACCELERATOR_DESCRIPTOR_INTEL -1096
#define CL_ACCELERATOR_TYPE_NOT_SUPPORTED_INTEL -1097
/* cl_accelerator_type_intel */
#define CL_ACCELERATOR_TYPE_MOTION_ESTIMATION_INTEL 0x0
/* cl_accelerator_info_intel */
#define CL_ACCELERATOR_DESCRIPTOR_INTEL 0x4090
#define CL_ACCELERATOR_REFERENCE_COUNT_INTEL 0x4091
#define CL_ACCELERATOR_CONTEXT_INTEL 0x4092
#define CL_ACCELERATOR_TYPE_INTEL 0x4093
/* cl_motion_detect_desc_intel flags */
#define CL_ME_MB_TYPE_16x16_INTEL 0x0
#define CL_ME_MB_TYPE_8x8_INTEL 0x1
#define CL_ME_MB_TYPE_4x4_INTEL 0x2
#define CL_ME_SUBPIXEL_MODE_INTEGER_INTEL 0x0
#define CL_ME_SUBPIXEL_MODE_HPEL_INTEL 0x1
#define CL_ME_SUBPIXEL_MODE_QPEL_INTEL 0x2
#define CL_ME_SAD_ADJUST_MODE_NONE_INTEL 0x0
#define CL_ME_SAD_ADJUST_MODE_HAAR_INTEL 0x1
#define CL_ME_SEARCH_PATH_RADIUS_2_2_INTEL 0x0
#define CL_ME_SEARCH_PATH_RADIUS_4_4_INTEL 0x1
#define CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL 0x5
#define CL_ME_SKIP_BLOCK_TYPE_16x16_INTEL 0x0
#define CL_ME_CHROMA_INTRA_PREDICT_ENABLED_INTEL 0x1
#define CL_ME_LUMA_INTRA_PREDICT_ENABLED_INTEL 0x2
#define CL_ME_SKIP_BLOCK_TYPE_8x8_INTEL 0x4
#define CL_ME_FORWARD_INPUT_MODE_INTEL 0x1
#define CL_ME_BACKWARD_INPUT_MODE_INTEL 0x2
#define CL_ME_BIDIRECTION_INPUT_MODE_INTEL 0x3
#define CL_ME_BIDIR_WEIGHT_QUARTER_INTEL 16
#define CL_ME_BIDIR_WEIGHT_THIRD_INTEL 21
#define CL_ME_BIDIR_WEIGHT_HALF_INTEL 32
#define CL_ME_BIDIR_WEIGHT_TWO_THIRD_INTEL 43
#define CL_ME_BIDIR_WEIGHT_THREE_QUARTER_INTEL 48
#define CL_ME_COST_PENALTY_NONE_INTEL 0x0
#define CL_ME_COST_PENALTY_LOW_INTEL 0x1
#define CL_ME_COST_PENALTY_NORMAL_INTEL 0x2
#define CL_ME_COST_PENALTY_HIGH_INTEL 0x3
#define CL_ME_COST_PRECISION_QPEL_INTEL 0x0
#define CL_ME_COST_PRECISION_HPEL_INTEL 0x1
#define CL_ME_COST_PRECISION_PEL_INTEL 0x2
#define CL_ME_COST_PRECISION_DPEL_INTEL 0x3
#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_INTEL 0x0
#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_INTEL 0x1
#define CL_ME_LUMA_PREDICTOR_MODE_DC_INTEL 0x2
#define CL_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_LEFT_INTEL 0x3
#define CL_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_RIGHT_INTEL 0x4
#define CL_ME_LUMA_PREDICTOR_MODE_PLANE_INTEL 0x4
#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_RIGHT_INTEL 0x5
#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_DOWN_INTEL 0x6
#define CL_ME_LUMA_PREDICTOR_MODE_VERTICAL_LEFT_INTEL 0x7
#define CL_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_UP_INTEL 0x8
#define CL_ME_CHROMA_PREDICTOR_MODE_DC_INTEL 0x0
#define CL_ME_CHROMA_PREDICTOR_MODE_HORIZONTAL_INTEL 0x1
#define CL_ME_CHROMA_PREDICTOR_MODE_VERTICAL_INTEL 0x2
#define CL_ME_CHROMA_PREDICTOR_MODE_PLANE_INTEL 0x3
/* cl_device_info */
#define CL_DEVICE_ME_VERSION_INTEL 0x407E
#define CL_ME_VERSION_LEGACY_INTEL 0x0
#define CL_ME_VERSION_ADVANCED_VER_1_INTEL 0x1
#define CL_ME_VERSION_ADVANCED_VER_2_INTEL 0x2
extern CL_API_ENTRY cl_accelerator_intel CL_API_CALL
clCreateAcceleratorINTEL(
cl_context /* context */,
cl_accelerator_type_intel /* accelerator_type */,
size_t /* descriptor_size */,
const void* /* descriptor */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_accelerator_intel (CL_API_CALL *clCreateAcceleratorINTEL_fn)(
cl_context /* context */,
cl_accelerator_type_intel /* accelerator_type */,
size_t /* descriptor_size */,
const void* /* descriptor */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_int CL_API_CALL
clGetAcceleratorInfoINTEL(
cl_accelerator_intel /* accelerator */,
cl_accelerator_info_intel /* param_name */,
size_t /* param_value_size */,
void* /* param_value */,
size_t* /* param_value_size_ret */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetAcceleratorInfoINTEL_fn)(
cl_accelerator_intel /* accelerator */,
cl_accelerator_info_intel /* param_name */,
size_t /* param_value_size */,
void* /* param_value */,
size_t* /* param_value_size_ret */) CL_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_int CL_API_CALL
clRetainAcceleratorINTEL(
cl_accelerator_intel /* accelerator */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clRetainAcceleratorINTEL_fn)(
cl_accelerator_intel /* accelerator */) CL_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_int CL_API_CALL
clReleaseAcceleratorINTEL(
cl_accelerator_intel /* accelerator */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clReleaseAcceleratorINTEL_fn)(
cl_accelerator_intel /* accelerator */) CL_EXT_SUFFIX__VERSION_1_2;
/******************************************
* cl_intel_simultaneous_sharing extension *
*******************************************/
#define cl_intel_simultaneous_sharing 1
#define CL_DEVICE_SIMULTANEOUS_INTEROPS_INTEL 0x4104
#define CL_DEVICE_NUM_SIMULTANEOUS_INTEROPS_INTEL 0x4105
/***********************************
* cl_intel_egl_image_yuv extension *
************************************/
#define cl_intel_egl_image_yuv 1
#define CL_EGL_YUV_PLANE_INTEL 0x4107
/********************************
* cl_intel_packed_yuv extension *
*********************************/
#define cl_intel_packed_yuv 1
#define CL_YUYV_INTEL 0x4076
#define CL_UYVY_INTEL 0x4077
#define CL_YVYU_INTEL 0x4078
#define CL_VYUY_INTEL 0x4079
/********************************************
* cl_intel_required_subgroup_size extension *
*********************************************/
#define cl_intel_required_subgroup_size 1
#define CL_DEVICE_SUB_GROUP_SIZES_INTEL 0x4108
#define CL_KERNEL_SPILL_MEM_SIZE_INTEL 0x4109
#define CL_KERNEL_COMPILE_SUB_GROUP_SIZE_INTEL 0x410A
/****************************************
* cl_intel_driver_diagnostics extension *
*****************************************/
#define cl_intel_driver_diagnostics 1
typedef cl_uint cl_diagnostics_verbose_level;
#define CL_CONTEXT_SHOW_DIAGNOSTICS_INTEL 0x4106
#define CL_CONTEXT_DIAGNOSTICS_LEVEL_ALL_INTEL ( 0xff )
#define CL_CONTEXT_DIAGNOSTICS_LEVEL_GOOD_INTEL ( 1 )
#define CL_CONTEXT_DIAGNOSTICS_LEVEL_BAD_INTEL ( 1 << 1 )
#define CL_CONTEXT_DIAGNOSTICS_LEVEL_NEUTRAL_INTEL ( 1 << 2 )
/********************************
* cl_intel_planar_yuv extension *
*********************************/
#define CL_NV12_INTEL 0x410E
#define CL_MEM_NO_ACCESS_INTEL ( 1 << 24 )
#define CL_MEM_ACCESS_FLAGS_UNRESTRICTED_INTEL ( 1 << 25 )
#define CL_DEVICE_PLANAR_YUV_MAX_WIDTH_INTEL 0x417E
#define CL_DEVICE_PLANAR_YUV_MAX_HEIGHT_INTEL 0x417F
/*******************************************************
* cl_intel_device_side_avc_motion_estimation extension *
********************************************************/
#define CL_DEVICE_AVC_ME_VERSION_INTEL 0x410B
#define CL_DEVICE_AVC_ME_SUPPORTS_TEXTURE_SAMPLER_USE_INTEL 0x410C
#define CL_DEVICE_AVC_ME_SUPPORTS_PREEMPTION_INTEL 0x410D
#define CL_AVC_ME_VERSION_0_INTEL 0x0; // No support.
#define CL_AVC_ME_VERSION_1_INTEL 0x1; // First supported version.
#define CL_AVC_ME_MAJOR_16x16_INTEL 0x0
#define CL_AVC_ME_MAJOR_16x8_INTEL 0x1
#define CL_AVC_ME_MAJOR_8x16_INTEL 0x2
#define CL_AVC_ME_MAJOR_8x8_INTEL 0x3
#define CL_AVC_ME_MINOR_8x8_INTEL 0x0
#define CL_AVC_ME_MINOR_8x4_INTEL 0x1
#define CL_AVC_ME_MINOR_4x8_INTEL 0x2
#define CL_AVC_ME_MINOR_4x4_INTEL 0x3
#define CL_AVC_ME_MAJOR_FORWARD_INTEL 0x0
#define CL_AVC_ME_MAJOR_BACKWARD_INTEL 0x1
#define CL_AVC_ME_MAJOR_BIDIRECTIONAL_INTEL 0x2
#define CL_AVC_ME_PARTITION_MASK_ALL_INTEL 0x0
#define CL_AVC_ME_PARTITION_MASK_16x16_INTEL 0x7E
#define CL_AVC_ME_PARTITION_MASK_16x8_INTEL 0x7D
#define CL_AVC_ME_PARTITION_MASK_8x16_INTEL 0x7B
#define CL_AVC_ME_PARTITION_MASK_8x8_INTEL 0x77
#define CL_AVC_ME_PARTITION_MASK_8x4_INTEL 0x6F
#define CL_AVC_ME_PARTITION_MASK_4x8_INTEL 0x5F
#define CL_AVC_ME_PARTITION_MASK_4x4_INTEL 0x3F
#define CL_AVC_ME_SEARCH_WINDOW_EXHAUSTIVE_INTEL 0x0
#define CL_AVC_ME_SEARCH_WINDOW_SMALL_INTEL 0x1
#define CL_AVC_ME_SEARCH_WINDOW_TINY_INTEL 0x2
#define CL_AVC_ME_SEARCH_WINDOW_EXTRA_TINY_INTEL 0x3
#define CL_AVC_ME_SEARCH_WINDOW_DIAMOND_INTEL 0x4
#define CL_AVC_ME_SEARCH_WINDOW_LARGE_DIAMOND_INTEL 0x5
#define CL_AVC_ME_SEARCH_WINDOW_RESERVED0_INTEL 0x6
#define CL_AVC_ME_SEARCH_WINDOW_RESERVED1_INTEL 0x7
#define CL_AVC_ME_SEARCH_WINDOW_CUSTOM_INTEL 0x8
#define CL_AVC_ME_SEARCH_WINDOW_16x12_RADIUS_INTEL 0x9
#define CL_AVC_ME_SEARCH_WINDOW_4x4_RADIUS_INTEL 0x2
#define CL_AVC_ME_SEARCH_WINDOW_2x2_RADIUS_INTEL 0xa
#define CL_AVC_ME_SAD_ADJUST_MODE_NONE_INTEL 0x0
#define CL_AVC_ME_SAD_ADJUST_MODE_HAAR_INTEL 0x2
#define CL_AVC_ME_SUBPIXEL_MODE_INTEGER_INTEL 0x0
#define CL_AVC_ME_SUBPIXEL_MODE_HPEL_INTEL 0x1
#define CL_AVC_ME_SUBPIXEL_MODE_QPEL_INTEL 0x3
#define CL_AVC_ME_COST_PRECISION_QPEL_INTEL 0x0
#define CL_AVC_ME_COST_PRECISION_HPEL_INTEL 0x1
#define CL_AVC_ME_COST_PRECISION_PEL_INTEL 0x2
#define CL_AVC_ME_COST_PRECISION_DPEL_INTEL 0x3
#define CL_AVC_ME_BIDIR_WEIGHT_QUARTER_INTEL 0x10
#define CL_AVC_ME_BIDIR_WEIGHT_THIRD_INTEL 0x15
#define CL_AVC_ME_BIDIR_WEIGHT_HALF_INTEL 0x20
#define CL_AVC_ME_BIDIR_WEIGHT_TWO_THIRD_INTEL 0x2B
#define CL_AVC_ME_BIDIR_WEIGHT_THREE_QUARTER_INTEL 0x30
#define CL_AVC_ME_BORDER_REACHED_LEFT_INTEL 0x0
#define CL_AVC_ME_BORDER_REACHED_RIGHT_INTEL 0x2
#define CL_AVC_ME_BORDER_REACHED_TOP_INTEL 0x4
#define CL_AVC_ME_BORDER_REACHED_BOTTOM_INTEL 0x8
#define CL_AVC_ME_SKIP_BLOCK_PARTITION_16x16_INTEL 0x0
#define CL_AVC_ME_SKIP_BLOCK_PARTITION_8x8_INTEL 0x4000
#define CL_AVC_ME_SKIP_BLOCK_16x16_FORWARD_ENABLE_INTEL ( 0x1 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_16x16_BACKWARD_ENABLE_INTEL ( 0x2 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_16x16_DUAL_ENABLE_INTEL ( 0x3 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_FORWARD_ENABLE_INTEL ( 0x55 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_BACKWARD_ENABLE_INTEL ( 0xAA << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_DUAL_ENABLE_INTEL ( 0xFF << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_0_FORWARD_ENABLE_INTEL ( 0x1 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_0_BACKWARD_ENABLE_INTEL ( 0x2 << 24 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_1_FORWARD_ENABLE_INTEL ( 0x1 << 26 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_1_BACKWARD_ENABLE_INTEL ( 0x2 << 26 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_2_FORWARD_ENABLE_INTEL ( 0x1 << 28 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_2_BACKWARD_ENABLE_INTEL ( 0x2 << 28 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_3_FORWARD_ENABLE_INTEL ( 0x1 << 30 )
#define CL_AVC_ME_SKIP_BLOCK_8x8_3_BACKWARD_ENABLE_INTEL ( 0x2 << 30 )
#define CL_AVC_ME_BLOCK_BASED_SKIP_4x4_INTEL 0x00
#define CL_AVC_ME_BLOCK_BASED_SKIP_8x8_INTEL 0x80
#define CL_AVC_ME_INTRA_16x16_INTEL 0x0
#define CL_AVC_ME_INTRA_8x8_INTEL 0x1
#define CL_AVC_ME_INTRA_4x4_INTEL 0x2
#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_16x16_INTEL 0x6
#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_8x8_INTEL 0x5
#define CL_AVC_ME_INTRA_LUMA_PARTITION_MASK_4x4_INTEL 0x3
#define CL_AVC_ME_INTRA_NEIGHBOR_LEFT_MASK_ENABLE_INTEL 0x60
#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_MASK_ENABLE_INTEL 0x10
#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_RIGHT_MASK_ENABLE_INTEL 0x8
#define CL_AVC_ME_INTRA_NEIGHBOR_UPPER_LEFT_MASK_ENABLE_INTEL 0x4
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_INTEL 0x0
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_INTEL 0x1
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DC_INTEL 0x2
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_LEFT_INTEL 0x3
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_DIAGONAL_DOWN_RIGHT_INTEL 0x4
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_PLANE_INTEL 0x4
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_RIGHT_INTEL 0x5
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_DOWN_INTEL 0x6
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_VERTICAL_LEFT_INTEL 0x7
#define CL_AVC_ME_LUMA_PREDICTOR_MODE_HORIZONTAL_UP_INTEL 0x8
#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_DC_INTEL 0x0
#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_HORIZONTAL_INTEL 0x1
#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_VERTICAL_INTEL 0x2
#define CL_AVC_ME_CHROMA_PREDICTOR_MODE_PLANE_INTEL 0x3
#define CL_AVC_ME_FRAME_FORWARD_INTEL 0x1
#define CL_AVC_ME_FRAME_BACKWARD_INTEL 0x2
#define CL_AVC_ME_FRAME_DUAL_INTEL 0x3
#define CL_AVC_ME_SLICE_TYPE_PRED_INTEL 0x0
#define CL_AVC_ME_SLICE_TYPE_BPRED_INTEL 0x1
#define CL_AVC_ME_SLICE_TYPE_INTRA_INTEL 0x2
#define CL_AVC_ME_INTERLACED_SCAN_TOP_FIELD_INTEL 0x0
#define CL_AVC_ME_INTERLACED_SCAN_BOTTOM_FIELD_INTEL 0x1
#ifdef __cplusplus
}
#endif
#endif /* __CL_EXT_INTEL_H */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_gl.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* Copyright (c) 2008-2018 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 <OpenCL/cl.h>
#else
#include <CL/cl.h>
#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
#ifdef CL_VERSION_1_2
#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
#endif
/* cl_gl_texture_info */
#define CL_GL_TEXTURE_TARGET 0x2004
#define CL_GL_MIPMAP_LEVEL 0x2005
#ifdef CL_VERSION_1_2
#define CL_GL_NUM_SAMPLES 0x2012
#endif
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;
#ifdef CL_VERSION_1_2
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;
#endif
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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_gl_ext.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* 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 <OpenCL/cl_gl.h>
#else
#include <CL/cl_gl.h>
#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 */
mobile/third_party/opencl/OpenCL-Headers/CL/cl_va_api_media_sharing_intel.h 已删除 100644 → 0
浏览文件 @ a8577694
/**********************************************************************************
* Copyright (c) 2008-2016 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.
**********************************************************************************/
/*****************************************************************************\
Copyright (c) 2013-2016 Intel Corporation All Rights Reserved.
THESE MATERIALS ARE PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THESE
MATERIALS, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
File Name: cl_va_api_media_sharing_intel.h
Abstract:
Notes:
\*****************************************************************************/
#ifndef __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H
#define __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H
#include <CL/cl.h>
#include <CL/cl_platform.h>
#include <va/va.h>
#ifdef __cplusplus
extern "C" {
#endif
/******************************************
* cl_intel_va_api_media_sharing extension *
*******************************************/
#define cl_intel_va_api_media_sharing 1
/* error codes */
#define CL_INVALID_VA_API_MEDIA_ADAPTER_INTEL -1098
#define CL_INVALID_VA_API_MEDIA_SURFACE_INTEL -1099
#define CL_VA_API_MEDIA_SURFACE_ALREADY_ACQUIRED_INTEL -1100
#define CL_VA_API_MEDIA_SURFACE_NOT_ACQUIRED_INTEL -1101
/* cl_va_api_device_source_intel */
#define CL_VA_API_DISPLAY_INTEL 0x4094
/* cl_va_api_device_set_intel */
#define CL_PREFERRED_DEVICES_FOR_VA_API_INTEL 0x4095
#define CL_ALL_DEVICES_FOR_VA_API_INTEL 0x4096
/* cl_context_info */
#define CL_CONTEXT_VA_API_DISPLAY_INTEL 0x4097
/* cl_mem_info */
#define CL_MEM_VA_API_MEDIA_SURFACE_INTEL 0x4098
/* cl_image_info */
#define CL_IMAGE_VA_API_PLANE_INTEL 0x4099
/* cl_command_type */
#define CL_COMMAND_ACQUIRE_VA_API_MEDIA_SURFACES_INTEL 0x409A
#define CL_COMMAND_RELEASE_VA_API_MEDIA_SURFACES_INTEL 0x409B
typedef cl_uint cl_va_api_device_source_intel;
typedef cl_uint cl_va_api_device_set_intel;
extern CL_API_ENTRY cl_int CL_API_CALL
clGetDeviceIDsFromVA_APIMediaAdapterINTEL(
cl_platform_id /* platform */,
cl_va_api_device_source_intel /* media_adapter_type */,
void* /* media_adapter */,
cl_va_api_device_set_intel /* media_adapter_set */,
cl_uint /* num_entries */,
cl_device_id* /* devices */,
cl_uint* /* num_devices */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL * clGetDeviceIDsFromVA_APIMediaAdapterINTEL_fn)(
cl_platform_id /* platform */,
cl_va_api_device_source_intel /* media_adapter_type */,
void* /* media_adapter */,
cl_va_api_device_set_intel /* media_adapter_set */,
cl_uint /* num_entries */,
cl_device_id* /* devices */,
cl_uint* /* num_devices */) CL_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_mem CL_API_CALL
clCreateFromVA_APIMediaSurfaceINTEL(
cl_context /* context */,
cl_mem_flags /* flags */,
VASurfaceID* /* surface */,
cl_uint /* plane */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_mem (CL_API_CALL * clCreateFromVA_APIMediaSurfaceINTEL_fn)(
cl_context /* context */,
cl_mem_flags /* flags */,
VASurfaceID* /* surface */,
cl_uint /* plane */,
cl_int* /* errcode_ret */) CL_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_int CL_API_CALL
clEnqueueAcquireVA_APIMediaSurfacesINTEL(
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_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueAcquireVA_APIMediaSurfacesINTEL_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_EXT_SUFFIX__VERSION_1_2;
extern CL_API_ENTRY cl_int CL_API_CALL
clEnqueueReleaseVA_APIMediaSurfacesINTEL(
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_EXT_SUFFIX__VERSION_1_2;
typedef CL_API_ENTRY cl_int (CL_API_CALL *clEnqueueReleaseVA_APIMediaSurfacesINTEL_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_EXT_SUFFIX__VERSION_1_2;
#ifdef __cplusplus
}
#endif
#endif /* __OPENCL_CL_VA_API_MEDIA_SHARING_INTEL_H */
mobile/third_party/opencl/OpenCL-Headers/CL/opencl.h 已删除 100644 → 0
浏览文件 @ a8577694
/*******************************************************************************
* 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 <OpenCL/cl.h>
#include <OpenCL/cl_gl.h>
#include <OpenCL/cl_gl_ext.h>
#include <OpenCL/cl_ext.h>
#else
#include <CL/cl.h>
#include <CL/cl_gl.h>
#include <CL/cl_gl_ext.h>
#include <CL/cl_ext.h>
#endif
#ifdef __cplusplus
}
#endif
#endif /* __OPENCL_H */
mobile/tools/build.sh
浏览文件 @ 13f66d39
此差异已折叠。
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或