pass args instead of pointer to array (#1591)

src/ngraph/runtime/gpu/cuda_emitter.cpp

@@ -588,19 +588,6 @@ size_t runtime::gpu::CUDAEmitter::build_reshape(const std::array<std::string, 2>
         return primitive_index;
     }
 
-    // check if the kernel has already been compiled. if so, create
-    // a launch primitive for it based on the input tensor shape
-    // but do not recompile the kernel. otherwise, do it all:
-    // recompile the kernel and then create the primitive
-    auto compiled_kernel = m_ctx->compiled_kernel_pool->get(kernel_name.str());
-    if (compiled_kernel == nullptr)
-    {
-        codegen::CodeWriter writer;
-        CudaKernelBuilder::add_pod_typedefs(writer);
-        CudaKernelBuilder::get_reshape_op(writer, kernel_name.str(), dtypes, rank);
-        compiled_kernel = m_ctx->compiled_kernel_pool->set(kernel_name.str(), writer.get_code());
-    }
-
     uint32_t nthreads = static_cast<uint32_t>(shape_size(input_shape));
     // TODO: currently we set it to 64, will add tuning method later
     uint32_t block_size_x = 64;
@@ -620,19 +607,32 @@ size_t runtime::gpu::CUDAEmitter::build_reshape(const std::array<std::string, 2>
     }
 
     // get an allocator for transient per kernel gpu memory
-    GPUAllocator allocator = this->m_primitive_emitter->get_memory_allocator();
-    size_t idx_input_strides =
-        allocator.reserve_argspace(input_strides.data(), input_strides.size() * sizeof(uint32_t));
-    size_t idx_trans_strides =
-        allocator.reserve_argspace(trans_strides.data(), trans_strides.size() * sizeof(uint32_t));
+    auto args = m_primitive_emitter->add_kernel_args();
+    args.add_placeholder(dtypes[0], "in")
+        .add_placeholder(dtypes[1], "out")
+        .add("input_strides", input_strides)
+        .add("trans_strides", trans_strides)
+        .add("n", nthreads);
+
+    // check if the kernel has already been compiled. if so, create
+    // a launch primitive for it based on the input tensor shape
+    // but do not recompile the kernel. otherwise, do it all:
+    // recompile the kernel and then create the primitive
+    auto compiled_kernel = m_ctx->compiled_kernel_pool->get(kernel_name.str());
+    if (compiled_kernel == nullptr)
+    {
+        codegen::CodeWriter writer;
+        CudaKernelBuilder::add_pod_typedefs(writer);
+        CudaKernelBuilder::get_reshape_op(writer, kernel_name.str(), args, dtypes, rank);
+        compiled_kernel = m_ctx->compiled_kernel_pool->set(kernel_name.str(), writer.get_code());
+    }
 
     // create the launch primitive
-    std::unique_ptr<gpu::primitive> kernel_launch(new gpu::primitive{[=](void** inputs,
-                                                                         void** outputs) mutable {
-        void* param_input_strides = runtime::gpu::invoke_memory_primitive(m_ctx, idx_input_strides);
-        void* param_trans_strides = runtime::gpu::invoke_memory_primitive(m_ctx, idx_trans_strides);
-        std::vector<void*> args_list{
-            &inputs[0], &outputs[0], &param_input_strides, &param_trans_strides, &nthreads};
+    std::unique_ptr<gpu::primitive> kernel_launch(
+        new gpu::primitive{[=](void** inputs, void** outputs) mutable {
+            void** args_list = args.resolve_placeholder(0, &inputs[0])
+                                   .resolve_placeholder(1, &outputs[0])
+                                   .get_argument_list();
 
             CUDA_SAFE_CALL(cuLaunchKernel(*compiled_kernel.get(),
                                           aligned_grid_size_x,
@@ -643,7 +643,7 @@ size_t runtime::gpu::CUDAEmitter::build_reshape(const std::array<std::string, 2>
                                           1, // block dim
                                           0,
                                           NULL, // shared mem and stream
-                                      args_list.data(),
+                                          args_list,
                                           0)); // arguments
             debug_sync();
         }});

src/ngraph/runtime/gpu/gpu_cuda_kernel_builder.cpp

@@ -452,12 +452,11 @@ void runtime::gpu::CudaKernelBuilder::get_onehot_op(codegen::CodeWriter& writer,
 
 void runtime::gpu::CudaKernelBuilder::get_reshape_op(codegen::CodeWriter& writer,
                                                      const std::string& name,
+                                                     runtime::gpu::GPUKernelArgs& args,
                                                      const std::array<std::string, 2>& data_types,
                                                      size_t rank)
 {
-    writer << "extern \"C\" __global__ void cuda_" << name << "(" << data_types[0] << "* in, "
-           << data_types[1]
-           << "* out, uint32_t* input_strides, uint32_t* trans_strides, uint32_t n)\n";
+    writer << "extern \"C\" __global__ void cuda_" << name << args.get_input_signature();
     writer.block_begin();
     {
         writer << "uint32_t tid = blockIdx.x * blockDim.x + threadIdx.x;\n";
@@ -469,12 +468,12 @@ void runtime::gpu::CudaKernelBuilder::get_reshape_op(codegen::CodeWriter& writer
             size_t i = 0;
             for (; i < rank - 1; i++)
             {
-                writer << "output_idx += (input_idx / input_strides[" << i << "]) * trans_strides["
-                       << i << "];\n";
-                writer << "input_idx %= input_strides[" << i << "];\n";
+                writer << "output_idx += (input_idx / input_strides" << i << ") * trans_strides"
+                       << i << ";\n";
+                writer << "input_idx %= input_strides" << i << ";\n";
             }
-            writer << "output_idx += (input_idx / input_strides[" << i << "]) * trans_strides[" << i
-                   << "];\n";
+            writer << "output_idx += (input_idx / input_strides" << i << ") * trans_strides" << i
+                   << ";\n";
             writer << "out[output_idx] = in[tid];\n";
         }
         writer.block_end();

src/ngraph/runtime/gpu/gpu_cuda_kernel_builder.hpp

@@ -56,6 +56,7 @@ namespace ngraph
 
                 static void get_reshape_op(codegen::CodeWriter& writer,
                                            const std::string& name,
+                                           runtime::gpu::GPUKernelArgs& args,
                                            const std::array<std::string, 2>& data_types,
                                            size_t rank);