xref: /XiangShan/README.md (revision 2f256e1de47c0fa510ed811cb0cf2518eeca17c9)
1# XiangShan
2
3XiangShan (香山) is an open-source high-performance RISC-V processor project.
4
5Detailed documents will be released in the near future.
6
7Copyright 2020-2021 by Institute of Computing Technology, Chinese Academy of Sciences.
8
9
10
11## === Tutorial is Comming! ===
12
13A tutorial on XiangShan is held at RISC-V World Conference China 2021 in June.  Multiple technical reports will be included in the conference as well. Stay tuned!
14
15For more information, please refer to [this link](https://openxiangshan.github.io).
16
17
18
19## Architecture
20
21The first stable micro-architecture of XiangShan is called Yanqihu (雁栖湖) on this [branch](https://github.com/OpenXiangShan/XiangShan/tree/yanqihu), which has been developed since June, 2020. The current version of XiangShan, also known as Nanhu (南湖), is still under development on master branch.
22
23The micro-architecture overview is shown below.
24
25![xs-arch-single](xs-arch-simple.svg)
26
27
28
29## Sub-directories Overview
30
31Some of the key directories are shown below.
32
33```
34.
35├── fpga                   # supported FPGA boards and files to build a Vivado project
36├── read-to-run            # pre-built simulation images
37├── scripts                # scripts for agile development
38└── src
39    ├── test               # test files (including diff-test, module-test, etc.)
40    └── main/scala         # design files
41        ├── bus/tilelink   # tilelink utils
42        ├── device         # virtual device for simulation
43        ├── difftest       # diff-test chisel interface
44        ├── system         # SoC wrapper
45        ├── top            # top module
46        ├── utils          # utilization code
47        ├── xiangshan      # main design code
48        └── xstransforms   # some useful firrtl transforms
49```
50
51
52
53## Generate Verilog
54
55* Run `make verilog` to generate verilog code. The output file is `build/XSTop.v`.
56* Refer to `Makefile` for more information.
57
58
59
60## Run Programs by Simulation
61
62### Prepare environment
63
64* Set environment variable `NEMU_HOME` to the **absolute path** of the [NEMU project](https://github.com/OpenXiangShan/NEMU).
65* Set environment variable `NOOP_HOME` to the **absolute path** of the XiangShan project.
66* Set environment variable `AM_HOME` to the **absolute path** of the [AM project](https://github.com/OpenXiangShan/nexus-am).
67* Install `mill`. Refer to [the Manual section in this guide](https://com-lihaoyi.github.io/mill/mill/Intro_to_Mill.html#_installation).
68* Clone this project and run `make init` to initialize submodules.
69
70### Run with simulator
71
72* Install [Verilator](https://verilator.org/guide/latest/), the open-source Verilog simulator.
73* Run `make emu` to build the C++ simulator `./build/emu` with Verilator.
74* Refer to `./build/emu --help` for run-time arguments of the simulator.
75* Refer to `Makefile` and `verilator.mk` for more information.
76
77Example:
78```bash
79make emu CONFIG=MinimalConfig SIM_ARGS=--disable-log EMU_THREADS=2 -j10
80./build/emu -b 0 -e 0 -i ./ready-to-run/coremark-2-iteration.bin --diff ./ready-to-run/riscv64-nemu-interpreter-so
81```
82
83
84
85## Acknowledgement
86
87In the development of XiangShan, some sub-modules from open source community are employed. All relavant usage is listed below.
88
89| Sub-module         | Source                                                       | Detail                                                       |
90| ------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ |
91| L2 Cache/LLC       | [Sifive block-inclusivecache](https://github.com/ucb-bar/block-inclusivecache-sifive) | We enhance the function and the timing of the original module, finally turning it into a Cache generator that can be configured as L2/LLC. |
92| Diplomacy/TileLink | [Rocket-chip](https://github.com/chipsalliance/rocket-chip)  | We reused the diplomacy framework and TileLink utility that exist in rocket-chip to negotiate bus. |
93| FPU                | [Berkeley hardfloat](https://github.com/ucb-bar/berkeley-hardfloat) | We use Berkeley-hardfloat as our FPU and implement a SRT-4 div/sqrt unit for it. Additionally, we split the FMA pipeline to optimize the timing. |
94
95We are grateful for the support of the open source community and encourage other open source projects to reuse our code within the scope of the [license](LICENSE).
96
97