NPU-SIM is a lightweight, large-scale, and multi-level simulation framework designed for multi-core Neural Processing Units (NPUs). It supports both transaction-level and performance-model-based simulation, providing powerful system-level analysis capabilities for large-scale models, such as Large Language Models (LLMs).

The framework is highly flexible and extensible, allowing for detailed simulation of various hardware and model configurations. Key features include:

• Flexible Parallelism: Exploration of various tensor parallelism strategies.
• Customizable Core Placement: Support for user-defined core placement policies.
• Advanced Memory Management: Simulation of diverse memory management methods.
• Configurable Dataflow: Selection between PD-disaggregation and PD-fusion on multi-core NPUs.

Furthermore, NPU-SIM extends its capabilities to the modeling of next-generation hardware by supporting wafer-scale simulation, enabling the analysis of systems that utilize hybrid bonding and distributed memory architectures

NPU-SIM provides an interactive GUI for real-time visualization of the simulation process:

🎬 Demo Video:

📘 Documentation: Click here to access the NPU-SIM documentation.

• OS: Linux

• SystemC: 2.3.3

• Cmake: 3.31.3

• G++: 9.4.0

wget https://github.com/accellera-official/systemc/archive/refs/tags/2.3.3.tar.gz
tar -zxvf 2.3.3.tar.gz
cd systemc-2.3.3/
mkdir tmp && cd tmp
../configure --prefix=/path/to/install/systemc-2.3.3 CXXFLAGS="-std=c++17"
sudo make -j8
make install

Add the following to your ~/.bashrc file:

export SYSTEMC_HOME=/path/to/install/systemc-2.3.3/
export LD_LIBRARY_PATH=/path/to/install/systemc-2.3.3/lib-linux64/:$LD_LIBRARY_PATH

#https://cmake.org/download/ Download the source code or corresponding binary file from the Cmake official website
wget https://cmake.org/files/v3.31/cmake-3.31.3-linux-x86_64.tar.gz
tar -zxvf cmake-3.31.3-linux-x86_64.tar.gz

git clone --branch=v3.11.3 --single-branch --depth=1 https://github.com/nlohmann/json.git
cd json
mkdir build && cd build
cmake ..
make
sudo make install

# 安装 SFML
sudo apt-get install libsfml-dev

# 安装 CAIRO
sudo apt install libcairo2-dev

# 安装 X11（服务器环境可能需要）
sudo apt install xorg

# 安装字体（源文件已包含必要的 ttf 文件）
sudo apt install ttf-mscorefonts-installer  # 需要在弹出界面选择 OK

cd /path/to/NPU-SIM/src
mkdir build && cd build
cmake ..
make -j8

./train_gpt2 --workload-config /path/to/NPU-SIM/src/llm/test/workload_config/config_gpt2_small_tp_24_new.json --use-dramsys true