Complete to run benchmark_app f32 for resample cpu impl

kelvinchoi-intel · Dec 24, 2024 · 219ebb1 · 219ebb1
1 parent 0f00606
commit 219ebb1
Show file tree

Hide file tree

Showing 4 changed files with 23 additions and 1 deletion.
diff --git a/src/plugins/intel_gpu/src/graph/impls/ocl/resample.hpp b/src/plugins/intel_gpu/src/graph/impls/ocl/resample.hpp
@@ -5,6 +5,7 @@
 #include "impls/registry/implementation_manager.hpp"
 #include "program_node.h"
 //#include "intel_gpu/primitives/resample.hpp"
+#include "resample_inst.h"
 
 #include <memory>
 namespace cldnn {

diff --git a/src/plugins/intel_gpu/src/graph/impls/registry/resample_impls.cpp b/src/plugins/intel_gpu/src/graph/impls/registry/resample_impls.cpp
@@ -37,7 +37,6 @@ const std::vector<std::shared_ptr<cldnn::ImplementationManager>>& Registry<resam
         //     [](const cldnn::program_node& node){
         //         return false;
         // })
-
         OV_GPU_GET_INSTANCE_CPU(resample, shape_types::static_shape,
             [](const cldnn::program_node& node){
                 return true;
@@ -46,6 +45,18 @@ const std::vector<std::shared_ptr<cldnn::ImplementationManager>>& Registry<resam
         //     [](const cldnn::program_node& node){
         //         return false;
         // })
+
+
+        // OV_GPU_GET_INSTANCE_OCL(resample, shape_types::static_shape, not_in_shape_flow())
+        // OV_GPU_GET_INSTANCE_OCL(resample, shape_types::dynamic_shape, not_in_shape_flow())
+        // OV_GPU_GET_INSTANCE_CPU(resample, shape_types::static_shape, in_shape_flow())
+        // OV_GPU_GET_INSTANCE_CPU(resample, shape_types::dynamic_shape, in_shape_flow())
+
+
+        // OV_GPU_CREATE_INSTANCE_OCL(ocl::ResampleImplementationManager, shape_types::static_shape, not_in_shape_flow())
+        // //OV_GPU_CREATE_INSTANCE_OCL(ocl::ResampleImplementationManager, shape_types::dynamic_shape, not_in_shape_flow())
+        // OV_GPU_GET_INSTANCE_CPU(resample, shape_types::static_shape, in_shape_flow())
+        // //OV_GPU_GET_INSTANCE_CPU(resample, shape_types::dynamic_shape, in_shape_flow())
     };
 
     return impls;

diff --git a/src/plugins/intel_gpu/src/graph/include/primitive_type_base.h b/src/plugins/intel_gpu/src/graph/include/primitive_type_base.h
@@ -46,6 +46,8 @@ struct primitive_type_base : primitive_type {
 
     std::shared_ptr<ImplementationManager> choose_impl(const program_node& node, shape_types requested_shape_type) const override {
         OPENVINO_ASSERT(node.type() == this, "[GPU] primitive_type_base::choose_impl: primitive type mismatch");
+        // if (node.id().find("interpolate:Interpolate") != std::string::npos)
+        //     std::cout << "=============================================" << node.id() << std::endl;
         for (auto& impl : get_supported_implementations(node)) {
             impl_types impl_type = impl->get_impl_type();
             if ((node.get_forced_impl_type() & impl_type) != impl_type)
@@ -58,6 +60,7 @@ struct primitive_type_base : primitive_type {
             if ((requested_shape_type & supported_shape_type) != requested_shape_type && requested_shape_type != shape_types::any)
                 continue;
 
+            // std::cout << "=== choose_impl " << node.id() << ", impl_type=" << impl_type << std::endl;
             return impl;
         }
         return nullptr;

diff --git a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
@@ -1792,6 +1792,10 @@ void primitive_inst::prepare_primitive() {
     }
     GPU_DEBUG_TRACE_DETAIL << "-----------------------------------------------------------------" << std::endl;
 
+    // if (this->get_node().id().find("__module.depth_head.scratch.refinenet1/aten::upsample_bilinear2d/Interpolate") != std::string::npos) {
+    //     std::cout << "=============" << std::endl;
+    //     std::cout << "=============primitive_inst::prepare_primitive() " << this->get_node().id() << ", impl_type=" << this->_impl->m_manager->get_impl_type() << std::endl;
+    // }
     // If it is optimized out or skipped for zero dimension at the previous iteration,
     // Set this flag true to reset output memory in realloc_if_needed.
     const bool prev_execution_skipped = can_be_optimized()
@@ -1972,6 +1976,9 @@ void primitive_inst::execute() {
 
     set_out_event(_impl->execute(_impl_params->dep_events, *this));
 
+    // if (this->get_node().id().find("__module.depth_head.scratch.refinenet1/aten::upsample_bilinear2d/Interpolate") != std::string::npos)
+    //     std::cout << "=============primitive_inst::execute() " << this->get_node().id() << ", impl_type=" << this->_impl->m_manager->get_impl_type() << std::endl;
+
     GPU_DEBUG_GET_INSTANCE(debug_config);
     GPU_DEBUG_IF(!debug_config->dump_profiling_data.empty()) {
         auto ev = _impl_params->out_event;