* Copyright (c) 2026 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
* \file test_geir_get_shape.cpp
* \brief GEIR test for TBE-implemented GetShape operator.
*
* GetShape (TBE) reads each input as a 128-element int64 shape descriptor:
* descriptor[0..2] = reserved (0)
* descriptor[3] = ndim (number of dimensions)
* descriptor[4..4+ndim-1] = dim values
* descriptor[rest] = 0
*
* InferShape (array_ops.cc) computes output size = sum of each input's dim count.
* So the TensorDesc shape of each input must be the shape being queried
* (e.g. {2,3,4} → 3 dims → contributes 3 to output), while the actual data
* buffer must contain the descriptor format above.
*/
#include <iostream>
#include <fstream>
#include <string.h>
#include <stdint.h>
#include <vector>
#include <string>
#include <map>
#include "assert.h"
#include "graph.h"
#include "types.h"
#include "tensor.h"
#include "ge_error_codes.h"
#include "ge_api_types.h"
#include "ge_api.h"
#include "array_ops.h"
#include "ge_ir_build.h"
#define FAILED -1
#define SUCCESS 0
using namespace ge;
using std::map;
using std::string;
using std::vector;
#define DESCRIPTOR_SIZE 128
#define ADD_DYNAMIC_INPUT(inputIndex, inputDtype, inputShape, dataName) \
do { \
vector<int64_t> shape##dataName = inputShape; \
auto placeholder##dataName = op::Data("placeholder" #dataName).set_attr_index(0); \
TensorDesc desc##dataName(ge::Shape(shape##dataName), FORMAT_ND, inputDtype); \
desc##dataName.SetFormat(FORMAT_ND); \
placeholder##dataName.update_input_desc_x(desc##dataName); \
placeholder##dataName.update_output_desc_y(desc##dataName); \
Tensor tensor##dataName; \
ret = GenShapeDescriptor(shape##dataName, tensor##dataName, desc##dataName); \
if (ret != SUCCESS) { \
printf("%s - ERROR - [XIR]: Generate input data failed\n", GetTime().c_str()); \
return FAILED; \
} \
getShapeOp.UpdateDynamicInputDesc("x", inputIndex, desc##dataName); \
getShapeOp.set_dynamic_input_x(inputIndex, placeholder##dataName); \
input.push_back(tensor##dataName); \
graph.AddOp(placeholder##dataName); \
inputs.push_back(placeholder##dataName); \
} while (0)
#define ADD_OUTPUT(outputIndex, outputName, outputDtype, outputShape) \
do { \
TensorDesc outputName##outputIndex##_desc = TensorDesc(ge::Shape(outputShape), FORMAT_ND, outputDtype); \
getShapeOp.update_output_desc_##outputName(outputName##outputIndex##_desc); \
} while (0)
#define LOG_PRINT(message, ...) \
do { \
printf(message, ##__VA_ARGS__); \
} while (0)
string GetTime()
{
time_t timep;
time(&timep);
char tmp[64];
strftime(tmp, sizeof(tmp), "%Y-%m-%d %H:%M:%S,000", localtime(&timep));
return tmp;
}
uint32_t GetDataTypeSize(DataType dt)
{
if (dt == ge::DT_FLOAT || dt == ge::DT_INT32 || dt == ge::DT_UINT32) {
return 4;
}
if (dt == ge::DT_FLOAT16 || dt == ge::DT_BF16 || dt == ge::DT_INT16 || dt == ge::DT_UINT16) {
return 2;
}
if (dt == ge::DT_INT64 || dt == ge::DT_UINT64 || dt == ge::DT_DOUBLE) {
return 8;
}
return 1;
}
* \brief Generate a 128-element int64 shape descriptor as input tensor data.
*
* The TensorDesc shape (e.g. {2,3,4}) determines:
* - GM allocation size (product(shape) * sizeof(int64) bytes)
* - InferShape output contribution (shape.size() dims)
*
* The actual data buffer is filled with the descriptor format:
* [0, 0, 0, ndim, dim0, dim1, ..., 0, 0, ...]
*
* The TBE kernel reads 128 int64 values (1024 bytes) via data_move from GM,
* but only accesses index 3 (ndim) and index 4..4+ndim-1 (dim values).
* As long as these indices fall within the allocated GM region, the kernel
* produces correct results. For shape {2,3,4}: 24 int64 = 192 bytes,
* index 3 and 4-6 are within range.
*/
int32_t GenShapeDescriptor(vector<int64_t> shapes, Tensor& input_tensor, TensorDesc& input_tensor_desc)
{
input_tensor_desc.SetRealDimCnt(shapes.size());
input_tensor_desc.SetShape(ge::Shape(shapes));
size_t elemCount = 1;
for (uint32_t i = 0; i < shapes.size(); i++) {
elemCount *= static_cast<size_t>(shapes[i]);
}
size_t dataLen = elemCount * sizeof(int64_t);
int64_t* pData = new (std::nothrow) int64_t[elemCount]();
int64_t ndim = static_cast<int64_t>(shapes.size());
if (elemCount >= 4) {
pData[3] = ndim;
}
for (size_t i = 0; i < static_cast<size_t>(ndim) && (4 + i) < elemCount; ++i) {
pData[4 + i] = shapes[i];
}
printf("GenShapeDescriptor: TensorDesc shape = {");
for (size_t i = 0; i < shapes.size(); ++i) {
printf("%ld%s", shapes[i], (i + 1 < shapes.size()) ? ", " : "");
}
printf("}, ndim = %ld, elemCount = %zu, dataLen = %zu\n", ndim, elemCount, dataLen);
printf(" descriptor[3] = %ld (ndim)\n", pData[3]);
for (size_t i = 0; i < static_cast<size_t>(ndim); ++i) {
printf(" descriptor[%zu] = %ld (dim%zu)\n", 4 + i, pData[4 + i], i);
}
input_tensor = Tensor(input_tensor_desc, reinterpret_cast<uint8_t*>(pData), dataLen);
return SUCCESS;
}
int32_t WriteDataToFile(string bin_file, uint64_t data_size, uint8_t* inputData)
{
FILE* fp = fopen(bin_file.c_str(), "wb");
if (fp == nullptr) {
return FAILED;
}
size_t written = fwrite(inputData, 1, data_size, fp);
fclose(fp);
if (written != data_size) {
return FAILED;
}
return SUCCESS;
}
int CreateOppInGraph(DataType inDtype, std::vector<ge::Tensor>& input, std::vector<Operator>& inputs,
std::vector<Operator>& outputs, Graph& graph)
{
Status ret = SUCCESS;
auto getShapeOp = op::GetShape("get_shape_1").create_dynamic_input_x(2, false);
std::vector<int64_t> x1Shape = {2, 3, 4};
ADD_DYNAMIC_INPUT(0, ge::DT_INT64, x1Shape, X1);
std::vector<int64_t> x2Shape = {5, 6};
ADD_DYNAMIC_INPUT(1, ge::DT_INT64, x2Shape, X2);
vector<int64_t> yShape = {5};
ADD_OUTPUT(1, y, ge::DT_INT32, yShape);
outputs.push_back(getShapeOp);
return SUCCESS;
}
void SaveInputOutput(std::vector<ge::Tensor>& input, std::vector<ge::Tensor>& output)
{
int input_num = input.size();
for (int i = 0; i < input_num; i++) {
std::cout << "input " << i << " dtype : " << input[i].GetTensorDesc().GetDataType() << std::endl;
string input_file = "./tc_ge_irrun_test_0008_npu_input_" + std::to_string(i) + ".bin";
uint8_t* input_data_i = input[i].GetData();
int64_t input_shape = input[i].GetTensorDesc().GetShape().GetShapeSize();
std::cout << "this is " << i << "th input, input shape size =" << input_shape << std::endl;
uint32_t data_size = input_shape * GetDataTypeSize(input[i].GetTensorDesc().GetDataType());
WriteDataToFile((const char*)input_file.c_str(), data_size, input_data_i);
int64_t* descData = (int64_t*)input_data_i;
printf(" input[%d] descriptor dump (first 8 int64): ", i);
for (int j = 0; j < 8 && j < input_shape; ++j) {
printf("[%d]=%ld ", j, descData[j]);
}
printf("\n");
}
int output_num = output.size();
for (int i = 0; i < output_num; i++) {
std::cout << "output " << i << " dtype : " << output[i].GetTensorDesc().GetDataType() << std::endl;
string output_file = "./tc_ge_irrun_test_0008_npu_output_" + std::to_string(i) + ".bin";
uint8_t* output_data_i = output[i].GetData();
int64_t output_shape = output[i].GetTensorDesc().GetShape().GetShapeSize();
std::cout << "this is " << i << "th output, output shape size =" << output_shape << std::endl;
uint32_t data_size = output_shape * GetDataTypeSize(output[i].GetTensorDesc().GetDataType());
WriteDataToFile((const char*)output_file.c_str(), data_size, output_data_i);
int32_t* resultData = (int32_t*)output_data_i;
printf(" Expected output: [2, 3, 4, 5, 6]\n");
printf(" Actual output: [");
for (int64_t j = 0; j < output_shape; ++j) {
printf("%d%s", resultData[j], (j + 1 < output_shape) ? ", " : "");
}
printf("]\n");
bool pass = true;
int32_t expected[] = {2, 3, 4, 5, 6};
for (int64_t j = 0; j < output_shape && j < 5; ++j) {
if (resultData[j] != expected[j]) {
pass = false;
break;
}
}
printf(" Result: %s\n", pass ? "PASS" : "FAIL");
}
}
int main(int argc, char* argv[])
{
const char* graph_name = "tc_ge_irrun_test";
Graph graph(graph_name);
std::vector<ge::Tensor> input;
printf("%s - INFO - [XIR]: Start to initialize ge using ge global options\n", GetTime().c_str());
std::map<AscendString, AscendString> global_options = {{"ge.exec.deviceId", "0"}, {"ge.graphRunMode", "1"}};
Status ret = ge::GEInitialize(global_options);
if (ret != SUCCESS) {
printf("%s - ERROR - [XIR]: Initialize ge using ge global options failed\n", GetTime().c_str());
return FAILED;
}
printf("%s - INFO - [XIR]: Initialize ge using ge global options success\n", GetTime().c_str());
std::vector<Operator> inputs{};
std::vector<Operator> outputs{};
if (argc > 1) {
std::cout << argv[1] << std::endl;
}
DataType inDtype = DT_INT64;
std::cout << "Input dtype: " << inDtype << std::endl;
ret = CreateOppInGraph(inDtype, input, inputs, outputs, graph);
if (ret != SUCCESS) {
printf("%s - ERROR - [XIR]: Create graph failed\n", GetTime().c_str());
return FAILED;
}
if (!inputs.empty() && !outputs.empty()) {
graph.SetInputs(inputs).SetOutputs(outputs);
}
std::map<AscendString, AscendString> build_options = {};
printf("%s - INFO - [XIR]: Start to create ir session using build options\n", GetTime().c_str());
ge::Session* session = new Session(build_options);
if (session == nullptr) {
printf("%s - ERROR - [XIR]: Create ir session using build options failed\n", GetTime().c_str());
return FAILED;
}
printf("%s - INFO - [XIR]: Create ir session using build options success\n", GetTime().c_str());
printf("%s - INFO - [XIR]: Start to add compute graph to ir session\n", GetTime().c_str());
std::map<AscendString, AscendString> graph_options = {};
uint32_t graph_id = 0;
ret = session->AddGraph(graph_id, graph, graph_options);
printf("%s - INFO - [XIR]: Session add ir compute graph to ir session success\n", GetTime().c_str());
printf("%s - INFO - [XIR]: dump graph to txt\n", GetTime().c_str());
std::string file_path = "./dump";
aclgrphDumpGraph(graph, file_path.c_str(), file_path.length());
printf("%s - INFO - [XIR]: Start to run ir compute graph\n", GetTime().c_str());
std::vector<ge::Tensor> output;
ret = session->RunGraph(graph_id, input, output);
if (ret != SUCCESS) {
printf("%s - ERROR - [XIR]: Run graph failed\n", GetTime().c_str());
delete session;
GEFinalize();
return FAILED;
}
printf("%s - INFO - [XIR]: Session run ir compute graph success\n", GetTime().c_str());
SaveInputOutput(input, output);
ge::AscendString error_msg = ge::GEGetErrorMsgV2();
std::string error_str(error_msg.GetString());
std::cout << "Error message: " << error_str << std::endl;
ge::AscendString warning_msg = ge::GEGetWarningMsgV2();
std::string warning_str(warning_msg.GetString());
std::cout << "Warning message: " << warning_str << std::endl;
printf("%s - INFO - [XIR]: Start to finalize ir graph session\n", GetTime().c_str());
ret = ge::GEFinalize();
if (ret != SUCCESS) {
printf("%s - ERROR - [XIR]: Finalize ir graph session failed\n", GetTime().c_str());
return FAILED;
}
printf("%s - INFO - [XIR]: Finalize ir graph session success\n", GetTime().c_str());
return SUCCESS;
}
