###### tags: `Lab`
# GPU 國科會計畫
## 標題: 前瞻 GPGPU XXXXXXX (待定)
## 目標: 完成 GPGPU IP Design
### 內容
- 目標介紹時可以先簡介目前市面上現有技術 (Ex: NVIDIA, AMD)，再來介紹我們要達成的目標以及實際上要如何達成，不用介紹到很 detail ，但大方向和如何實作要解釋的合理且看得懂，多附圖說明 orz
- 現有技術介紹很多文件可以參考，整理排版一下複製貼上就好
- 能加的幹話可以加多一點
- 目標介紹篇幅要長一點，一個項目約 3~5 頁 word
- 請大家在 ==**12/7**== 前給我初版
### 目標介紹
- 硬體
1. Debugger/ISR support
2. ~~Internal data bus (between SM)~~
3. IP Verification (Verification method, ex. formal verification)
4. TPU (Accelarate Matrix multiplication, ref. 峰銘)
- 軟體
1. TVM support (replace tf-coriander)
- 其他
1. Risc-V + CasLAB-GPU SoC Verification (By Qemu/FPGA, ref. 聖洋)
2. Integrate RTL version into CIC platform
3. Demo with an application (with `1.`)
4. GCN ISA
### 現有項目介紹
- 硬體
1. CasLAB-GPU Overall Architecture introduction (with Dispatcher unit)
2. SM Architecture introduction
3. RTL post-sim results
4. Verification on FPGA
- 軟體
1. Tensorflow/OpenCL/HSA Runtime support
2. Tf-coriander
- 其他
1. Benchmark we have run
2. 旺宏得獎
3. 科技部得獎
## 分工
- **YAO**
- 目標
- 軟體
- 現有
- 軟體
- **阿呆樺、Yanting ER**
- 目標
- 硬體(1)
- 現有
- 硬體(1, 2)
- **沛辰**
- 目標
- 其他(2)
- 現有
- 硬體(3, 4)
- 其他(2)
- **Tommy**
- 目標
- 硬體(4)
- 其他(1)
- **Kai Kai**
- 目標
- 其他(4)
- 現有
- 其他(3)
- **玉婷**
- 目標
- 硬體(3)
- https://docs.google.com/document/d/1XpcpBZxxqoi_rgsDITX9UdBjC9V2EwZpB_Y8DtSS1CQ/edit#
- 現有
- 其他(1)
- https://docs.google.com/document/d/1DjLySFn4DlKI44dxLEsqjfSYy9m5j24MwlfmQuTmqac/edit
- **Parker**
- 目標
- 其他(3)
## Question
- [ ] Describe Warp Schedualing policy or not

## 研究方法
- [ ] Debugger hardware
- ISR
- [ ] TPU hardware
- [ ] TPU Software
- Compiler
- OpenCL runtime
- HSA Runtime
- [ ] Soft IP
- [ ] Porting TVM
- Porting OpenCL Runtime
- Optimization
- OpenCL Codgen
- [ ] Debugger Software
- GDB
- [ ] Qemu + ESL simulation platform
- Device driver
- [ ] RISC-V/GPGPU SoC platform
- MMU
- Shared memory or not
- [ ] FPGA Demo
- [ ] NGS

## Ref
### 驗證
- [Verification of automation software by model checking](https://www.simulationstore.com/node/52)
- [軟／硬體的正規（formal）驗證](https://www.ctimes.com.tw/DispArt/tw/0303051424RD.shtml)
- [Introduction to Formal Verification](https://www.eeweb.com/introduction-to-formal-verification/)
### TVM
- [Apache TVM](https://tvm.apache.org/)
- [一步一步解讀神經網絡編譯器TVM(二)——利用TVM完成C++端的部署](https://read01.com/3z4KaA3.html#.X9dGQ9gzaUk)
### Tensor Core
- [NVIDIA深度學習Tensor Core全面解析（上篇）](https://kknews.cc/zh-tw/tech/blnnnn9.html)
### HSA
- [Heterogeneous System Architecture](https://en.wikipedia.org/wiki/Heterogeneous_System_Architecture)
- [Heterogeneous Systems Architecture (HSA) – the TL;DR](https://streamhpc.com/blog/2014-02-05/heterogeneous-systems-architecture-hsa-the-tldr/)
- [Basic concepts to keep in mind for GPU debugging](https://software.intel.com/content/www/us/en/develop/articles/gpu-debugging-challenges-and-opportunities.html)
# ==為了飯碗，大家加油 orz !==