EngineX-Ascend/enginex-ascend-910-vllm
11
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init v0.23.0

Browse Source 
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
This commit is contained in:
Sun Ruoxi
2026-08-27 15:11:51 +08:00
parent b582a8e7d1
commit 7f8a1b1f7a
2849 changed files with 712887 additions and 22001 deletions
Download Patch File Download Diff File Expand all files Collapse all files

163
csrc/scripts/opgen/opgen_standalone.py Normal file
View File

@@ -0,0 +1,163 @@
# -----------------------------------------------------------------------------------------------------------
# Copyright (c) 2025 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.
# -----------------------------------------------------------------------------------------------------------
import argparse
import logging
import os
import shutil
import sys
import regex as re
class OpGenerator:
"""算子工程生成器"""
def __init__(self, op_type, op_name, output_path):
self.op_type = op_type
self.op_name = op_name
self.output_path = output_path
self.template_name = "add_example"
self.script_dir = os.path.dirname(os.path.abspath(__file__))
self.template_dir = os.path.abspath(os.path.join(self.script_dir, "template", "add"))
self.dest_dir = os.path.abspath(os.path.join(self.output_path, self.op_type, self.op_name))
def run(self):
"""执行生成流程"""
self._validate_inputs()
self._copy_template()
self._rename_files()
self._replace_content()
logging.info("成功为 %s/%s 创建算子工程！", self.op_type, self.op_name)
logging.info("工程路径: %s", self.dest_dir)
logging.info("Create the initial directory for %s under %s success", self.op_name, self.op_type)
def _validate_inputs(self):
"""校验输入参数的有效性和安全性"""
if not self.op_type or not self.op_name:
raise ValueError("算子类型和算子名称均不能为空。")
if not re.match(r"^[a-zA-Z0-9_]+$", self.op_type):
raise ValueError(f"算子类型 '{self.op_type}' 包含无效字符。只允许字母、数字和下划线。")
if not re.match(r"^[a-zA-Z0-9_]+$", self.op_name):
raise ValueError(f"算子名称 '{self.op_name}' 包含无效字符。只允许字母、数字和下划线。")
if os.path.exists(self.dest_dir):
raise FileExistsError(f"目标目录 '{self.dest_dir}' 已存在。")
def _copy_template(self):
"""复制模板文件到目标目录"""
logging.info("使用模板在 '%s' 创建算子工程...", self.dest_dir)
if not os.path.exists(self.template_dir):
raise FileNotFoundError(f"找不到模板目录 '{self.template_dir}'。请确保 'template/add' 目录存在。")
try:
shutil.copytree(self.template_dir, self.dest_dir)
if not os.path.isfile(os.path.join(os.path.dirname(self.dest_dir), "CMakeLists.txt")):
cmake_src = os.path.join(os.path.dirname(self.template_dir), "CMakeLists.txt")
cmake_dest = os.path.join(os.path.dirname(self.dest_dir), "CMakeLists.txt")
shutil.copy2(cmake_src, cmake_dest)
except OSError as e:
raise OSError(f"复制模板文件失败: {e}") from e
def _rename_files(self):
"""重命名文件和目录中的占位符"""
for root, dirs, files in os.walk(self.dest_dir, topdown=False):
for name in files + dirs:
if self.template_name not in name:
continue
old_path = os.path.join(root, name)
new_name = name.replace(self.template_name, self.op_name)
new_path = os.path.join(root, new_name)
try:
os.rename(old_path, new_path)
except OSError as e:
raise OSError(f"重命名 '{old_path}' 到 '{new_path}' 失败: {e}") from e
def _replace_content_in_file(self, file_path, replacements):
"""Helper to replace content in a single file."""
try:
with open(file_path, encoding="utf-8", errors="ignore") as f:
content = f.read()
except OSError as e:
logging.warning("读取文件 '%s' 失败: %s", file_path, e)
return
original_content = content
for old, new in replacements.items():
content = content.replace(old, new)
if content == original_content:
return
try:
with open(file_path, "w", encoding="utf-8") as f:
f.write(content)
except OSError as e:
logging.warning("写入文件 '%s' 失败: %s", file_path, e)
def _replace_content(self):
"""替换文件内容中的占位符"""
op_name_capitalized = "".join(word.capitalize() for word in self.op_name.split("_"))
template_name_capitalized = "".join(word.capitalize() for word in self.template_name.split("_"))
replacements = {
self.template_name: self.op_name,
self.template_name.upper(): self.op_name.upper(),
template_name_capitalized: op_name_capitalized,
"add_example": self.op_name,
}
for root, _, files in os.walk(self.dest_dir):
for file in files:
if file.endswith((".pyc", ".pyo")):
continue
file_path = os.path.join(root, file)
self._replace_content_in_file(file_path, replacements)
def execute(args):
"""根据命令行参数执行算子生成"""
generator = OpGenerator(op_type=args.op_type, op_name=args.op_name, output_path=args.output_path)
generator.run()
def register_parser(subparsers):
"""为 opgen 命令注册解析器。"""
parser_opgen = subparsers.add_parser("opgen", help="生成项目骨架")
parser_opgen.add_argument("--op_type", "-t", required=True, help="算子分类，例如 math")
parser_opgen.add_argument("--op_name", "-n", required=True, help="新算子的名称，例如 asinh")
parser_opgen.add_argument("--output_path", "-p", default=".", help="生成工程的根路径")
parser_opgen.set_defaults(func=execute)
def main():
"""主函数，用于独立执行"""
logging.basicConfig(level=logging.INFO, format="%(levelname)s: %(message)s", stream=sys.stdout)
parser = argparse.ArgumentParser(description="生成项目骨架")
parser.add_argument("--op_type", "-t", required=True, help="算子分类，例如 math")
parser.add_argument("--op_name", "-n", required=True, help="新算子的名称，例如 asinh")
parser.add_argument("--output_path", "-p", default=".", help="生成工程的根路径")
args = parser.parse_args()
try:
execute(args)
except Exception as e:
logging.error("发生非预期的错误，退出。错误信息: %s", e)
sys.exit(1)
if __name__ == "__main__":
main()

29
csrc/scripts/opgen/template/CMakeLists.txt Normal file
View File

@@ -0,0 +1,29 @@
# -----------------------------------------------------------------------------------------------------------
# Copyright (c) 2025 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.
# -----------------------------------------------------------------------------------------------------------
set(OPTEST_NAME optest_${PKG_NAME})
file(GLOB CURRENT_DIRS RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/*)
foreach(SUB_DIR ${CURRENT_DIRS})
if (DEFINED ASCEND_OP_NAME AND NOT "${ASCEND_OP_NAME}" STREQUAL "")
if (NOT "${ASCEND_OP_NAME}" STREQUAL "all" AND NOT "${ASCEND_OP_NAME}" STREQUAL "ALL")
if (NOT ${SUB_DIR} IN_LIST ASCEND_OP_NAME)
continue()
endif ()
endif ()
endif ()
if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/${SUB_DIR}/CMakeLists.txt")
add_subdirectory(${SUB_DIR})
else()
if (EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/${SUB_DIR}/op_host/CMakeLists.txt")
add_subdirectory(${SUB_DIR}/op_host)
endif()
endif()
endforeach()

19
csrc/scripts/opgen/template/add/CMakeLists.txt Normal file
View File

@@ -0,0 +1,19 @@
# -----------------------------------------------------------------------------------------------------------
# Copyright (c) 2025 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(GLOB CURRENT_DIRS RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/*)
if(NOT ENABLE_TEST AND NOT BENCHMARK)
list(REMOVE_ITEM CURRENT_DIRS tests)
endif()
foreach(SUB_DIR ${CURRENT_DIRS})
if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/${SUB_DIR}/CMakeLists.txt")
add_subdirectory(${SUB_DIR})
endif()
endforeach()

166
csrc/scripts/opgen/template/add/examples/test_aclnn_add_example.cpp Normal file
View File

@@ -0,0 +1,166 @@
/**
 * Copyright (c) 2025 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.
 */
// 本文件参考example下示例自动生成
// 您可自由修改此文件满足需求
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_add_example.h"
#define CHECK_RET(cond, return_expr) \
do { \
if (!(cond)) { \
return_expr; \
} \
} while (0)
#define LOG_PRINT(message, ...) \
do { \
printf(message, ##__VA_ARGS__); \
} while (0)
int64_t GetShapeSize(const std::vector<int64_t>& shape)
{
int64_t shapeSize = 1;
for (auto i : shape) {
shapeSize *= i;
}
return shapeSize;
}
void PrintOutResult(std::vector<int64_t>& shape, void** deviceAddr)
{
auto size = GetShapeSize(shape);
std::vector<float> resultData(size, 0);
auto ret = aclrtMemcpy(
resultData.data(), resultData.size() * sizeof(resultData[0]), *deviceAddr, size * sizeof(resultData[0]),
ACL_MEMCPY_DEVICE_TO_HOST);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return);
for (int64_t i = 0; i < size; i++) {
LOG_PRINT("mean result[%ld] is: %f\n", i, resultData[i]);
}
}
int Init(int32_t deviceId, aclrtStream* stream)
{
// 固定写法，初始化
auto ret = aclInit(nullptr);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
ret = aclrtSetDevice(deviceId);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
ret = aclrtCreateStream(stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
return 0;
}
template <typename T>
int CreateAclTensor(
const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType,
aclTensor** tensor)
{
auto size = GetShapeSize(shape) * sizeof(T);
// 2. 申请device侧内存
auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
// 3. 调用aclrtMemcpy将host侧数据拷贝到device侧内存上
ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);
// 计算连续tensor的strides
std::vector<int64_t> strides(shape.size(), 1);
for (int64_t i = shape.size() - 2; i >= 0; i--) {
strides[i] = shape[i + 1] * strides[i + 1];
}
// 调用aclCreateTensor接口创建aclTensor
*tensor = aclCreateTensor(
shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
*deviceAddr);
return 0;
}
int main()
{
// 1. 调用acl进行device/stream初始化
int32_t deviceId = 0;
aclrtStream stream;
auto ret = Init(deviceId, &stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);
// 2. 构造输入与输出，需要根据API的接口自定义构造
aclTensor* selfX = nullptr;
void* selfXDeviceAddr = nullptr;
std::vector<int64_t> selfXShape = {32, 4, 4, 4};
std::vector<float> selfXHostData(2048, 1); // 2048：创建包含32*4*4*4=2048个元素的向量
ret = CreateAclTensor(selfXHostData, selfXShape, &selfXDeviceAddr, aclDataType::ACL_FLOAT, &selfX);
CHECK_RET(ret == ACL_SUCCESS, return ret);
aclTensor* selfY = nullptr;
void* selfYDeviceAddr = nullptr;
std::vector<int64_t> selfYShape = {32, 4, 4, 4};
std::vector<float> selfYHostData(2048, 1);
ret = CreateAclTensor(selfYHostData, selfYShape, &selfYDeviceAddr, aclDataType::ACL_FLOAT, &selfY);
CHECK_RET(ret == ACL_SUCCESS, return ret);
aclTensor* out = nullptr;
void* outDeviceAddr = nullptr;
std::vector<int64_t> outShape = {32, 4, 4, 4};
std::vector<float> outHostData(2048, 1);
ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 3. 调用CANN算子库API，需要修改为具体的Api名称
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
// 4. 调用aclnnAddExample第一段接口
ret = aclnnAddExampleGetWorkspaceSize(selfX, selfY, out, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddExampleGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
// 根据第一段接口计算出的workspaceSize申请device内存
void* workspaceAddr = nullptr;
if (workspaceSize > static_cast<uint64_t>(0)) {
ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
}
// 5. 调用aclnnAddExample第二段接口
ret = aclnnAddExample(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddExample failed. ERROR: %d\n", ret); return ret);
// 6. （固定写法）同步等待任务执行结束
ret = aclrtSynchronizeStream(stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);
// 5. 获取输出的值，将device侧内存上的结果拷贝至host侧，需要根据具体API的接口定义修改
PrintOutResult(outShape, &outDeviceAddr);
// 7. 释放aclTensor，需要根据具体API的接口定义修改
aclDestroyTensor(selfX);
aclDestroyTensor(selfY);
aclDestroyTensor(out);
// 8. 释放device资源
aclrtFree(selfXDeviceAddr);
aclrtFree(selfYDeviceAddr);
aclrtFree(outDeviceAddr);
if (workspaceSize > static_cast<uint64_t>(0)) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
// 9. acl去初始化
aclFinalize();
return 0;
}

18
csrc/scripts/opgen/template/add/op_host/CMakeLists.txt Normal file
View File

@@ -0,0 +1,18 @@
# -----------------------------------------------------------------------------------------------------------
# Copyright (c) 2025 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.
# -----------------------------------------------------------------------------------------------------------
add_op_to_compiled_list()
if (BUILD_OPEN_PROJECT)
target_sources(op_host_aclnn PRIVATE
add_example_def.cpp
)
endif()
add_modules_sources(OPTYPE add_example ACLNNTYPE aclnn)

55
csrc/scripts/opgen/template/add/op_host/add_example_def.cpp Normal file
View File

@@ -0,0 +1,55 @@
/**
 * Copyright (c) 2025 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 add_example_def.cpp
* \brief
*/
#include "register/op_def_registry.h"
namespace ops {
class AddExample : public OpDef {
public:
explicit AddExample(const char* name) : OpDef(name)
{
// 输入参数说明
this->Input("x1") // 输入x1定义
.ParamType(REQUIRED) // 必选输入
.DataType({ge::DT_FLOAT, ge::DT_INT32}) // 支持数据类型
.Format({ge::FORMAT_ND, ge::FORMAT_ND}) // 支持format格式
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND}) // 未确定大小shape对应format格式
.AutoContiguous(); // 内存自动连续化
this->Input("x2") // 输入x2定义
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT, ge::DT_INT32})
.Format({ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND})
.AutoContiguous();
this->Output("y") // 输出y定义
.ParamType(REQUIRED)
.DataType({ge::DT_FLOAT, ge::DT_INT32})
.Format({ge::FORMAT_ND, ge::FORMAT_ND})
.UnknownShapeFormat({ge::FORMAT_ND, ge::FORMAT_ND})
.AutoContiguous();
OpAICoreConfig aicoreConfig;
aicoreConfig.DynamicCompileStaticFlag(true)
.DynamicFormatFlag(false)
.DynamicRankSupportFlag(true)
.DynamicShapeSupportFlag(true)
.NeedCheckSupportFlag(false)
.PrecisionReduceFlag(true)
.ExtendCfgInfo("opFile.value", "add_example"); // 这里制定的值会对应到kernel入口文件名.cpp
this->AICore().AddConfig("ascend910b", aicoreConfig); // 其他的soc版本补充部分配置项
this->AICore().AddConfig("ascend910_93", aicoreConfig);
}
};
OP_ADD(AddExample); // 添加算子信息库
} // namespace ops

48
csrc/scripts/opgen/template/add/op_host/add_example_infershape.cpp Normal file
View File

@@ -0,0 +1,48 @@
/**
 * Copyright (c) 2025 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 add_example_infer.cpp
* \brief
*/
#include "register/op_impl_registry.h"
#include "log/log.h"
using namespace ge;
namespace ops {
static constexpr int64_t IDX_0 = 0;
static ge::graphStatus InferShapeAddExample(gert::InferShapeContext* context)
{
OP_LOGD(context->GetNodeName(), "Begin to do InferShapeAddExample");
// get input shapes
const gert::Shape* xShape = context->GetInputShape(IDX_0);
OP_CHECK_NULL_WITH_CONTEXT(context, xShape);
// get output shapes
gert::Shape* yShape = context->GetOutputShape(IDX_0);
OP_CHECK_NULL_WITH_CONTEXT(context, yShape);
// 填充输出shape大小
auto xShapeSize = xShape->GetDimNum();
yShape->SetDimNum(xShapeSize);
for (size_t i = 0; i < xShapeSize; i++) {
int64_t dim = xShape->GetDim(i);
yShape->SetDim(i, dim);
}
OP_LOGD(context->GetNodeName(), "End to do InferShapeAddExample");
return GRAPH_SUCCESS;
}
IMPL_OP_INFERSHAPE(AddExample).InferShape(InferShapeAddExample);
} // namespace ops

156
csrc/scripts/opgen/template/add/op_host/add_example_tiling.cpp Normal file
View File

@@ -0,0 +1,156 @@
/**
 * Copyright (c) 2025 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 add_example_tiling.cpp
* \brief
*/
#include "log/log.h"
#include "util/math_util.h"
#include "tiling_base/tiling_util.h"
#include "tiling_base/tiling_templates_registry.h"
#include "../op_kernel/add_example_tiling_data.h"
#include "../op_kernel/add_example_tiling_key.h"
namespace optiling {
using namespace Ops::Transformer::OpTiling;
const uint32_t BLOCK_DIM = 8;
const int64_t TILE_NUM = 8;
const uint32_t WS_SYS_SIZE = 16U * 1024U * 1024U;
const int32_t DIMS_LIMIT = 4;
constexpr int32_t ATTRPOS0 = 0;
constexpr uint32_t INDEXZERO = 0;
constexpr uint32_t INDEXONE = 1;
constexpr uint32_t INDEXTWO = 2;
constexpr uint32_t INDEXTHREE = 3;
struct AddExampleCompileInfo {};
// 获取平台信息如ubSize, coreNum
static ge::graphStatus GetPlatformInfo(gert::TilingContext* context, uint64_t& ubSize, int64_t& coreNum)
{
// 获取ubsize coreNum
fe::PlatFormInfos* platformInfoPtr = context->GetPlatformInfo();
OP_CHECK_NULL_WITH_CONTEXT(context, platformInfoPtr);
auto ascendcPlatform = platform_ascendc::PlatformAscendC(platformInfoPtr);
coreNum = ascendcPlatform.GetCoreNumAiv();
OP_CHECK_IF(coreNum == 0, OP_LOGE(context, "coreNum is 0"), return ge::GRAPH_FAILED);
ascendcPlatform.GetCoreMemSize(platform_ascendc::CoreMemType::UB, ubSize);
OP_CHECK_IF(ubSize == 0, OP_LOGE(context, "ubSize is 0"), return ge::GRAPH_FAILED);
return ge::GRAPH_SUCCESS;
}
// 获取属性，shape信息
ge::graphStatus GetShapeAttrsInfo(gert::TilingContext* context, int64_t& totalIdx, ge::DataType& dataType)
{
// 获取输入shape信息
auto inputX = context->GetInputShape(0);
OP_CHECK_NULL_WITH_CONTEXT(context, inputX);
// 如果输入shape 是标量 转换为{1}，否则保持原 shape 不变
auto inputShapeX = EnsureNotScalar(inputX->GetStorageShape());
auto inputY = context->GetInputShape(1);
OP_CHECK_NULL_WITH_CONTEXT(context, inputY);
auto inputShapeY = EnsureNotScalar(inputY->GetStorageShape());
auto outZ = context->GetOutputShape(0);
OP_CHECK_NULL_WITH_CONTEXT(context, outZ);
auto outShapeZ = EnsureNotScalar(outZ->GetStorageShape());
// shape校验
OP_CHECK_IF(
inputShapeX.GetDimNum() != DIMS_LIMIT || inputShapeY.GetDimNum() != DIMS_LIMIT ||
outShapeZ.GetDimNum() != DIMS_LIMIT,
OP_LOGE(
context, "AddExample: inputx,inputy,outputz shape dim = %zu, %zu, %zu, should be equal 4",
inputShapeX.GetDimNum(), inputShapeY.GetDimNum(), outShapeZ.GetDimNum()),
return ge::GRAPH_FAILED);
// 获取shape dim值
auto nDim = inputShapeX.GetDim(INDEXZERO);
auto cDim = inputShapeX.GetDim(INDEXONE);
auto hDim = inputShapeX.GetDim(INDEXTWO);
auto wDim = inputShapeX.GetDim(INDEXTHREE);
totalIdx = nDim * cDim * hDim * wDim;
// dtype校验
const std::set<ge::DataType> supportedDtype = {ge::DT_FLOAT, ge::DT_INT32};
auto inputDesc = context->GetInputDesc(0);
OP_CHECK_NULL_WITH_CONTEXT(context, inputDesc);
dataType = inputDesc->GetDataType();
if (supportedDtype.count(dataType) == 0) {
OP_LOGE(context, "invalid dtype");
return ge::GRAPH_FAILED;
}
return ge::GRAPH_SUCCESS;
}
ge::graphStatus GetWorkspaceSize(gert::TilingContext* context)
{
size_t* currentWorkspace = context->GetWorkspaceSizes(1);
OP_CHECK_NULL_WITH_CONTEXT(context, currentWorkspace);
currentWorkspace[0] = WS_SYS_SIZE;
return ge::GRAPH_SUCCESS;
}
// tiling 分发入口
static ge::graphStatus AddExampleTilingFunc(gert::TilingContext* context)
{
// 1、获取平台运行信息
uint64_t ubSize;
int64_t coreNum;
OP_CHECK_IF(
GetPlatformInfo(context, ubSize, coreNum) != ge::GRAPH_SUCCESS, OP_LOGE(context, "GetPlatformInfo error"),
return ge::GRAPH_FAILED);
// 2、获取shape、属性信息
int64_t totalIdx;
ge::DataType dataType;
OP_CHECK_IF(
GetShapeAttrsInfo(context, totalIdx, dataType) != ge::GRAPH_SUCCESS,
OP_LOGE(context, "GetShapeAttrsInfo error"), return ge::GRAPH_FAILED);
// 3、获取WorkspaceSize信息
OP_CHECK_IF(
GetWorkspaceSize(context) != ge::GRAPH_SUCCESS, OP_LOGE(context, "GetWorkspaceSize error"),
return ge::GRAPH_FAILED);
// 4、设置tiling信息
AddExampleTilingData* tiling = context->GetTilingData<AddExampleTilingData>();
OP_CHECK_NULL_WITH_CONTEXT(context, tiling);
OP_CHECK_IF(
memset_s(tiling, sizeof(AddExampleTilingData), 0, sizeof(AddExampleTilingData)) != EOK,
OP_LOGE(context, "set tiling data error"), return ge::GRAPH_FAILED);
tiling->totalLength = totalIdx;
tiling->tileNum = TILE_NUM;
context->SetBlockDim(BLOCK_DIM);
uint64_t tilingKey = 0;
// 区分dtype走不同得tiling key分支.
if (dataType == ge::DT_FLOAT) {
tilingKey = GET_TPL_TILING_KEY(ELEMENTWISE_TPL_SCH_MODE_0);
context->SetTilingKey(tilingKey);
} else if (dataType == ge::DT_INT32) {
tilingKey = GET_TPL_TILING_KEY(ELEMENTWISE_TPL_SCH_MODE_1);
context->SetTilingKey(tilingKey);
} else {
OP_LOGE(context, "get dtype error");
return ge::GRAPH_FAILED;
}
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus TilingParseForAddExample([[maybe_unused]] gert::TilingParseContext* context)
{
return ge::GRAPH_SUCCESS;
}
// tiling注册入口.
IMPL_OP_OPTILING(AddExample).Tiling(AddExampleTilingFunc).TilingParse<AddExampleCompileInfo>(TilingParseForAddExample);
} // namespace optiling

43
csrc/scripts/opgen/template/add/op_kernel/add_example.cpp Normal file
View File

@@ -0,0 +1,43 @@
/**
 * Copyright (c) 2025 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 add_example.cpp
* \brief
*/
#include "add_example.h"
enum class AddExampleTilingKey : uint32_t
{
TILING_KEY_EXAMPLE_FLOAT = 0,
TILING_KEY_EXAMPLE_INT32 = 1,
};
template <uint32_t schMode>
__global__ __aicore__ void add_example(GM_ADDR x, GM_ADDR y, GM_ADDR z, GM_ADDR workspace, GM_ADDR tiling)
{
REGISTER_TILING_DEFAULT(AddExampleTilingData);
GET_TILING_DATA_WITH_STRUCT(AddExampleTilingData, tilingData, tiling);
// 场景1
if constexpr (schMode == static_cast<uint32_t>(AddExampleTilingKey::TILING_KEY_EXAMPLE_FLOAT)) {
NsAddExample::AddExample<float> op; // 算子kernel实例获取
op.Init(x, y, z, &tilingData); // 算子kernel实例初始化
op.Process(); // 算子kernel实例执行
}
// 场景2
if constexpr (schMode == static_cast<uint32_t>(AddExampleTilingKey::TILING_KEY_EXAMPLE_INT32)) {
NsAddExample::AddExample<int32_t> op; // 算子kernel实例获取
op.Init(x, y, z, &tilingData); // 算子kernel实例初始化
op.Process(); // 算子kernel实例执行
}
}

117
csrc/scripts/opgen/template/add/op_kernel/add_example.h Normal file
View File

@@ -0,0 +1,117 @@
/**
 * Copyright (c) 2025 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 add_example.h
* \brief
*/
#ifndef ADD_EXAMPLE_H
#define ADD_EXAMPLE_H
#include "kernel_operator.h"
#include "kernel_tiling/kernel_tiling.h"
#include "add_example_tiling_data.h"
#include "add_example_tiling_key.h"
namespace NsAddExample {
using namespace AscendC;
constexpr int32_t BUFFER_NUM = 2;
template <typename T>
class AddExample
{
public:
__aicore__ inline AddExample(){};
__aicore__ inline void Init(GM_ADDR x, GM_ADDR y, GM_ADDR z, const AddExampleTilingData* tilingData);
__aicore__ inline void Process();
private:
__aicore__ inline void CopyIn(int32_t progress);
__aicore__ inline void CopyOut(int32_t progress);
__aicore__ inline void Compute(const int32_t dataLength);
private:
TPipe pipe;
TQue<QuePosition::VECIN, BUFFER_NUM> inputQueueX;
TQue<QuePosition::VECIN, BUFFER_NUM> inputQueueY;
TQue<QuePosition::VECOUT, BUFFER_NUM> outputQueueZ;
GlobalTensor<T> inputGMX;
GlobalTensor<T> inputGMY;
GlobalTensor<T> outputGMZ;
int64_t blockLength_ = 0;
int64_t tileNum_ = 0;
uint32_t tileLength_ = 0;
};
template <typename T>
__aicore__ inline void AddExample<T>::Init(GM_ADDR x, GM_ADDR y, GM_ADDR z, const AddExampleTilingData* tilingData)
{
blockLength_ = tilingData->totalLength / AscendC::GetBlockNum();
tileNum_ = tilingData->tileNum;
tileLength_ = blockLength_ / tileNum_ / BUFFER_NUM;
inputGMX.SetGlobalBuffer((__gm__ T*)x + blockLength_ * AscendC::GetBlockIdx(), blockLength_);
inputGMY.SetGlobalBuffer((__gm__ T*)y + blockLength_ * AscendC::GetBlockIdx(), blockLength_);
outputGMZ.SetGlobalBuffer((__gm__ T*)z + blockLength_ * AscendC::GetBlockIdx(), blockLength_);
pipe.InitBuffer(inputQueueX, BUFFER_NUM, tileLength_ * sizeof(T));
pipe.InitBuffer(inputQueueY, BUFFER_NUM, tileLength_ * sizeof(T));
pipe.InitBuffer(outputQueueZ, BUFFER_NUM, tileLength_ * sizeof(T));
}
template <typename T>
__aicore__ inline void AddExample<T>::CopyIn(int32_t progress)
{
AscendC::LocalTensor<T> xLocal = inputQueueX.AllocTensor<T>();
AscendC::LocalTensor<T> yLocal = inputQueueY.AllocTensor<T>();
AscendC::DataCopy(xLocal, inputGMX[progress * tileLength_], tileLength_);
AscendC::DataCopy(yLocal, inputGMY[progress * tileLength_], tileLength_);
inputQueueX.EnQue(xLocal);
inputQueueY.EnQue(yLocal);
}
template <typename T>
__aicore__ inline void AddExample<T>::CopyOut(int32_t progress)
{
AscendC::LocalTensor<T> zLocal = outputQueueZ.DeQue<T>();
AscendC::DataCopy(outputGMZ[progress * tileLength_], zLocal, tileLength_);
outputQueueZ.FreeTensor(zLocal);
}
template <typename T>
__aicore__ inline void AddExample<T>::Compute(int32_t progress)
{
AscendC::LocalTensor<T> xLocal = inputQueueX.DeQue<T>();
AscendC::LocalTensor<T> yLocal = inputQueueY.DeQue<T>();
AscendC::LocalTensor<T> zLocal = outputQueueZ.AllocTensor<T>();
AscendC::Add(zLocal, xLocal, yLocal, tileLength_);
outputQueueZ.EnQue<T>(zLocal);
inputQueueX.FreeTensor(xLocal);
inputQueueY.FreeTensor(yLocal);
}
template <typename T>
__aicore__ inline void AddExample<T>::Process()
{
int32_t loopCount = tileNum_ * BUFFER_NUM;
for (int32_t i = 0; i < loopCount; i++) {
CopyIn(i);
Compute(i);
CopyOut(i);
}
}
} // namespace NsAddExample
#endif // ADD_EXAMPLE_H

23
csrc/scripts/opgen/template/add/op_kernel/add_example_tiling_data.h Normal file
View File

@@ -0,0 +1,23 @@
/**
 * Copyright (c) 2025 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 add_example_tiling_data.h
* \brief tiling data struct
*/
#ifndef __ADD_EXAMPLE_TILLING_DATA_H__
#define __ADD_EXAMPLE_TILLING_DATA_H__
struct AddExampleTilingData {
int64_t totalLength;
int64_t tileNum;
};
#endif

35
csrc/scripts/opgen/template/add/op_kernel/add_example_tiling_key.h Normal file
View File

@@ -0,0 +1,35 @@
/**
 * Copyright (c) 2025 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 add_example_tiling_key.h
* \brief add_example tiling key declare
*/
#ifndef __ADD_EXAMPLE_TILING_KEY_H__
#define __ADD_EXAMPLE_TILING_KEY_H__
#include "ascendc/host_api/tiling/template_argument.h"
/* Mode场景定义 */
#define ELEMENTWISE_TPL_SCH_MODE_0 0
#define ELEMENTWISE_TPL_SCH_MODE_1 1
/* 继续定义其他Mode场景... */
/* 模板参数 */
ASCENDC_TPL_ARGS_DECL(AddExample,
ASCENDC_TPL_UINT_DECL(schMode, 1, ASCENDC_TPL_UI_LIST, ELEMENTWISE_TPL_SCH_MODE_0, ELEMENTWISE_TPL_SCH_MODE_1)
);
ASCENDC_TPL_SEL(
ASCENDC_TPL_ARGS_SEL(
ASCENDC_TPL_UINT_SEL(schMode, ASCENDC_TPL_UI_LIST, ELEMENTWISE_TPL_SCH_MODE_0, ELEMENTWISE_TPL_SCH_MODE_1)));
#endif

0
csrc/scripts/opgen/template/add/tests/ut/.gitkeep Normal file
View File