lattice_elim1.cpp

// Generated by the nGraph GPU backend
#include <cublas_v2.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <cudnn.h>

#include "ngraph/descriptor/input.hpp"
#include "ngraph/descriptor/layout/dense_tensor_view_layout.hpp"
#include "ngraph/descriptor/output.hpp"
#include "ngraph/descriptor/primary_tensor_view.hpp"
#include "ngraph/file_util.hpp"
#include "ngraph/function.hpp"
#include "ngraph/graph_util.hpp"
#include "ngraph/node.hpp"
#include "ngraph/pass/assign_layout.hpp"
#include "ngraph/pass/dump_sorted.hpp"
#include "ngraph/pass/liveness.hpp"
#include "ngraph/pass/manager.hpp"
#include "ngraph/pass/memory_layout.hpp"
#include "ngraph/runtime/aligned_buffer.hpp"
#include "ngraph/runtime/gpu/cudnn_descriptors.hpp"
#include "ngraph/runtime/gpu/gpu_cuda_kernel_ops.hpp"
#include "ngraph/runtime/gpu/gpu_invoke.hpp"
#include "ngraph/runtime/gpu/gpu_runtime_context.hpp"
#include "ngraph/runtime/gpu/gpu_util.hpp"
#include "ngraph/util.hpp"

using namespace ngraph;
using namespace ngraph::runtime;
using namespace std;
void *__dso_handle = 0;

static gpu::GPURuntimeContext* m_runtime_context = nullptr;
// Declare all constants
static size_t Constant_18_0_idx = 47;
static float* Constant_18_0 = nullptr;
static size_t Constant_14_0_idx = 48;
static float* Constant_14_0 = nullptr;
static size_t Constant_11_0_idx = 49;
static float* Constant_11_0 = nullptr;
static size_t Constant_12_0_idx = 50;
static float* Constant_12_0 = nullptr;

static bool is_constant_mem_ptr_null = true;

static void invoke_constant_mem_ptr()
{
    if(is_constant_mem_ptr_null)
    {
        Constant_18_0 = reinterpret_cast<float*>(runtime::gpu::invoke_memory_primitive(m_runtime_context, Constant_18_0_idx));
        Constant_14_0 = reinterpret_cast<float*>(runtime::gpu::invoke_memory_primitive(m_runtime_context, Constant_14_0_idx));
        Constant_11_0 = reinterpret_cast<float*>(runtime::gpu::invoke_memory_primitive(m_runtime_context, Constant_11_0_idx));
        Constant_12_0 = reinterpret_cast<float*>(runtime::gpu::invoke_memory_primitive(m_runtime_context, Constant_12_0_idx));
        is_constant_mem_ptr_null = false;
    }
}
// Declare all functions
extern "C" void Function_0(void** inputs, void** outputs, gpu::GPURuntimeContext* ctx);

static void func_Concat_1024(
    float* _arg0,
    float* _arg1,
    float* _out2,
    gpu::GPURuntimeContext* ctx
)
{
    gpu::invoke_primitive(ctx, 1, std::vector<void*>{_arg0, _arg1}.data(), std::vector<void*>{_out2}.data());
}

static void func_Concat_1022(
    float* _arg0,
    float* _arg1,
    float* _out2,
    gpu::GPURuntimeContext* ctx
)
{
    gpu::invoke_primitive(ctx, 2, std::vector<void*>{_arg0, _arg1}.data(), std::vector<void*>{_out2}.data());
}

static void func_Broadcast_174(
    float* _arg0,
    float* _out1,
    gpu::GPURuntimeContext* ctx
)
{
    gpu::invoke_primitive(ctx, 13, std::vector<void*>{_arg0}.data(), std::vector<void*>{_out1}.data());
}

static void func_Reshape_231(
    float* _arg0,
    float* _out1,
    gpu::GPURuntimeContext* ctx
)
{
    runtime::gpu::cuda_memcpyDtD(_out1, _arg0, 6400 * 4);
}

static void func_Concat_1020(
    float* _arg0,
    float* _arg1,
    float* _out2,
    gpu::GPURuntimeContext* ctx
)
{
    gpu::invoke_primitive(ctx, 14, std::vector<void*>{_arg0, _arg1}.data(), std::vector<void*>{_out2}.data());
}

static void func_Reshape_1012(
    float* _arg0,
    float* _out1,
    gpu::GPURuntimeContext* ctx
)
{
    runtime::gpu::cuda_memcpyDtD(_out1, _arg0, 64000 * 4);
}


extern "C" void Function_0(void** inputs, void** outputs, gpu::GPURuntimeContext* ctx)
{
    m_runtime_context = ctx;
    invoke_constant_mem_ptr();
    // Allocate the memory pool
    void* pool_base_ptr = ngraph::runtime::gpu::invoke_memory_primitive(ctx, 0);

    // Broadcast_75(Parameter_0_0, Broadcast_75_0)
    gpu::invoke_primitive(ctx, 0, std::vector<void*>{((float*)(inputs[10]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 0))}.data());

    // Reshape_76(Parameter_1_0, Reshape_76_0)
    {
        const float alpha = 1.0;
        const float beta = 0;
        CUBLAS_SAFE_CALL(cublasSetPointerMode(*ctx->cublas_handle, CUBLAS_POINTER_MODE_HOST));
        CUBLAS_SAFE_CALL(cublasSgeam(*ctx->cublas_handle,CUBLAS_OP_T,CUBLAS_OP_T,10000,200,&alpha,((float*)(inputs[9])),200,&beta,((float*)(inputs[9])),200,((float*)((char *)pool_base_ptr + 12800000)),10000));
        CUBLAS_SAFE_CALL(cublasSetPointerMode(*ctx->cublas_handle, CUBLAS_POINTER_MODE_DEVICE));
    }

    // Concat_1024(Parameter_6_0, Parameter_2_0, Concat_1024_0)
    func_Concat_1024(((float*)(inputs[4])), ((float*)(inputs[8])), ((float*)((char *)pool_base_ptr + 20800000)), ctx);

    // Concat_1022(Parameter_7_0, Parameter_3_0, Concat_1022_0)
    func_Concat_1022(((float*)(inputs[3])), ((float*)(inputs[7])), ((float*)((char *)pool_base_ptr + 20806400)), ctx);

    // Concat_1023(Parameter_8_0, Parameter_4_0, Concat_1023_0)
    func_Concat_1024(((float*)(inputs[2])), ((float*)(inputs[6])), ((float*)((char *)pool_base_ptr + 22086400)), ctx);

    // Concat_1021(Parameter_9_0, Parameter_5_0, Concat_1021_0)
    func_Concat_1022(((float*)(inputs[1])), ((float*)(inputs[5])), ((float*)((char *)pool_base_ptr + 22092800)), ctx);

    // Slice_157(Parameter_10_0, Slice_157_0)
    {
        gpu::invoke_primitive(ctx, 3, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23372800))}.data());
    }

    // Slice_158(Parameter_10_0, Slice_158_0)
    {
        gpu::invoke_primitive(ctx, 4, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23398400))}.data());
    }

    // Slice_159(Parameter_10_0, Slice_159_0)
    {
        gpu::invoke_primitive(ctx, 5, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23424000))}.data());
    }

    // Slice_160(Parameter_10_0, Slice_160_0)
    {
        gpu::invoke_primitive(ctx, 6, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23449600))}.data());
    }

    // Slice_161(Parameter_10_0, Slice_161_0)
    {
        gpu::invoke_primitive(ctx, 7, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23475200))}.data());
    }

    // Slice_162(Parameter_10_0, Slice_162_0)
    {
        gpu::invoke_primitive(ctx, 8, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23500800))}.data());
    }

    // Slice_163(Parameter_10_0, Slice_163_0)
    {
        gpu::invoke_primitive(ctx, 9, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23526400))}.data());
    }

    // Slice_164(Parameter_10_0, Slice_164_0)
    {
        gpu::invoke_primitive(ctx, 10, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23552000))}.data());
    }

    // Slice_165(Parameter_10_0, Slice_165_0)
    {
        gpu::invoke_primitive(ctx, 11, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23577600))}.data());
    }

    // Slice_166(Parameter_10_0, Slice_166_0)
    {
        gpu::invoke_primitive(ctx, 12, std::vector<void*>{((float*)(inputs[0]))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23603200))}.data());
    }

    // Broadcast_174(Constant_18_0, Broadcast_174_0)
    func_Broadcast_174(Constant_18_0, ((float*)((char *)pool_base_ptr + 23628800)), ctx);

    // Broadcast_170(Constant_14_0, Broadcast_170_0)
    func_Broadcast_174(Constant_14_0, ((float*)((char *)pool_base_ptr + 23654400)), ctx);

    // Broadcast_167(Constant_11_0, Broadcast_167_0)
    func_Broadcast_174(Constant_11_0, ((float*)((char *)pool_base_ptr + 23680000)), ctx);

    // Broadcast_168(Constant_12_0, Broadcast_168_0)
    func_Broadcast_174(Constant_12_0, ((float*)((char *)pool_base_ptr + 23705600)), ctx);

    // Reshape_231(Slice_157_0, Reshape_231_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23372800)), ((float*)((char *)pool_base_ptr + 23731200)), ctx);

    // Reshape_232(Slice_158_0, Reshape_232_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23398400)), ((float*)((char *)pool_base_ptr + 23372800)), ctx);

    // Reshape_233(Slice_159_0, Reshape_233_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23424000)), ((float*)((char *)pool_base_ptr + 23398400)), ctx);

    // Reshape_234(Slice_160_0, Reshape_234_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23449600)), ((float*)((char *)pool_base_ptr + 23424000)), ctx);

    // Reshape_235(Slice_161_0, Reshape_235_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23475200)), ((float*)((char *)pool_base_ptr + 23449600)), ctx);

    // Reshape_236(Slice_162_0, Reshape_236_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23500800)), ((float*)((char *)pool_base_ptr + 23475200)), ctx);

    // Reshape_237(Slice_163_0, Reshape_237_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23526400)), ((float*)((char *)pool_base_ptr + 23500800)), ctx);

    // Reshape_238(Slice_164_0, Reshape_238_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23552000)), ((float*)((char *)pool_base_ptr + 23526400)), ctx);

    // Reshape_239(Slice_165_0, Reshape_239_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23577600)), ((float*)((char *)pool_base_ptr + 23552000)), ctx);

    // Reshape_240(Slice_166_0, Reshape_240_0)
    func_Reshape_231(((float*)((char *)pool_base_ptr + 23603200)), ((float*)((char *)pool_base_ptr + 23577600)), ctx);

    // Concat_1020(Broadcast_170_0, Broadcast_174_0, Concat_1020_0)
    func_Concat_1020(((float*)((char *)pool_base_ptr + 23654400)), ((float*)((char *)pool_base_ptr + 23628800)), ((float*)((char *)pool_base_ptr + 23756800)), ctx);

    // Concat_1019(Broadcast_168_0, Broadcast_167_0, Concat_1019_0)
    func_Concat_1020(((float*)((char *)pool_base_ptr + 23705600)), ((float*)((char *)pool_base_ptr + 23680000)), ((float*)((char *)pool_base_ptr + 23603200)), ctx);

    // Concat_989(Reshape_240_0, Reshape_231_0, Reshape_232_0, Reshape_233_0, Reshape_234_0, Reshape_235_0, Reshape_236_0, Reshape_237_0, Reshape_238_0, Reshape_239_0, Concat_989_0)
    {
        gpu::invoke_primitive(ctx, 15, std::vector<void*>{((float*)((char *)pool_base_ptr + 23577600)), ((float*)((char *)pool_base_ptr + 23731200)), ((float*)((char *)pool_base_ptr + 23372800)), ((float*)((char *)pool_base_ptr + 23398400)), ((float*)((char *)pool_base_ptr + 23424000)), ((float*)((char *)pool_base_ptr + 23449600)), ((float*)((char *)pool_base_ptr + 23475200)), ((float*)((char *)pool_base_ptr + 23500800)), ((float*)((char *)pool_base_ptr + 23526400)), ((float*)((char *)pool_base_ptr + 23552000))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 23808000))}.data());
    }

    // Rnn_1025(Concat_989_0, Concat_1019_0, Concat_1021_0, Concat_1022_0, Concat_1023_0, Concat_1024_0, Concat_1020_0, Rnn_1025_0, Rnn_1025_1, Rnn_1025_2)
    gpu::invoke_primitive(ctx, 16, std::vector<void*>{((float*)((char *)pool_base_ptr + 23808000)), ((float*)((char *)pool_base_ptr + 23603200)), ((float*)((char *)pool_base_ptr + 22092800)), ((float*)((char *)pool_base_ptr + 20806400)), ((float*)((char *)pool_base_ptr + 22086400)), ((float*)((char *)pool_base_ptr + 20800000)), ((float*)((char *)pool_base_ptr + 23756800))}.data(), std::vector<void*>{((float*)((char *)pool_base_ptr + 24064000)), ((float*)((char *)pool_base_ptr + 23372800)), ((float*)((char *)pool_base_ptr + 23424000))}.data());

    // GetOutputElement_1026(Rnn_1025_0, Rnn_1025_1, Rnn_1025_2, GetOutputElement_1026_0)
    {
        runtime::gpu::cuda_memcpyDtH(((float*)((char *)pool_base_ptr + 20800000)), ((float*)((char *)pool_base_ptr + 24064000)), 256000);
    }

    // Reshape_1012(GetOutputElement_1026_0, Reshape_1012_0)
    func_Reshape_1012(((float*)((char *)pool_base_ptr + 20800000)), ((float*)((char *)pool_base_ptr + 21056000)), ctx);

    // Reshape_792(Reshape_1012_0, Reshape_792_0)
    func_Reshape_1012(((float*)((char *)pool_base_ptr + 21056000)), ((float*)((char *)pool_base_ptr + 20800000)), ctx);

    // Dot_793(Reshape_792_0, Reshape_76_0, Dot_793_0)
    {
        const float alpha = 1.0;
        const float beta  = 0.0;
        int m = 320;
        int n = 10000;
        int k = 200;
        CUBLAS_SAFE_CALL(cublasSetPointerMode(*ctx->cublas_handle, CUBLAS_POINTER_MODE_HOST));
        CUBLAS_SAFE_CALL(cublasSgemm(*ctx->cublas_handle,CUBLAS_OP_N,CUBLAS_OP_N,n,m,k,&alpha,((float*)((char *)pool_base_ptr + 12800000)),n,((float*)((char *)pool_base_ptr + 20800000)),k,&beta,((float*)((char *)pool_base_ptr + 21056000)),n));
        CUBLAS_SAFE_CALL(cublasSetPointerMode(*ctx->cublas_handle, CUBLAS_POINTER_MODE_DEVICE));
    }

    // Add_794(Dot_793_0, Broadcast_75_0, Add_794_0)
    {
        gpu::invoke_primitive(ctx, 17, std::vector<void*>{((float*)((char *)pool_base_ptr + 21056000)), ((float*)((char *)pool_base_ptr + 0))}.data(), std::vector<void*>{((float*)(outputs[0]))}.data());
    }

    // Result_795(Add_794_0, Result_795_0)
    {
        runtime::gpu::cuda_memcpyDtD(((float*)(outputs[0])), ((float*)(outputs[0])), 3200000 * 4);
    }
}