Convlusion Cl Codes vstore2 and vload2 function __kernel void default_function_kernel0(__global float* restrict A, __global float* restrict W, __global float* restrict B) { float B_local[64]; __local float Apad_shared[8]; __local float W_shared[128]; float Apad_shared_local[8]; float W_shared_local[8]; for (int ff_c_init = 0; ff_c_init < 4; ++ff_c_init) {

解析紅黑樹（ RED-BLACK TREE ） 概述 RBtree 是一種特別的平衡樹，既然身為平衡樹它在新增、刪除、搜尋的複雜度都能維持在 O(log n)，而它的特別之處在於樹高 h 的定義與一般平衡樹不同(用的是 black height )，此特別之處讓紅黑樹在重新平衡所需的"代價"較其他平衡樹小 :::info 黑高度( black height) bh(x):從 x 到任何一個它的子孫葉子 leaf 的路徑上，遇到的 black node 個數(不包含自己） ::: 定義

Simulation Platform: https://drive.google.com/file/d/16xIMCTcmbXTVWEEe6Ru0KnbB37kOYD_j/view?usp=sharing Architecture Overview Manual Password Ubuntu password: caslabgpu

REQUIREMENTS The processor core operating speed is targeting at least 75 MHz for post-synthesized netlist. Its instruction set shall have at least 45 instructions, including branch, I/O instructions. IM + DM cannot be larger than 320KB The silicon area of the CPU+ICache+DCache+IM+DM shall be confined within 110 mm2 in total EPU shall be synthesized and constrained within 3 mm2 The kernel of the chip shall be less than 120 mm2 The read/write access time of an off-chip not-synthesized memory, usually DRAM, is 60ns The main memory only has one read/write port with bit width of 32

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

Note 在 systemc2.3.2 版本 (2.3.3感覺沒問題) ， dma.h 宣告的 socket 的順序，在建構子初始化的順序需要相同，否則會有 error b_transport 需要加 wait() ，否則沒辦法寫進 control reg

Week 1 (Introductiion & the Cool Programming Language) Structure of compiler Lexical Analysis (詞彙分析) 空白鍵為 token 之一 Parsing (語法分析) 將一段程式碼分成 parsing tree

TCP TCP (Transmission Control Protocol) 使用wireshark分析tcpdump出来的pcap文件 TCP 三次交握 實驗八 TCP協定分析 DDos

論文架構 Abstract INTRODUCTION BACKGROUND AND MOTIVATION 2.1 Sparsity-Centric Optimization for DNNs 2.2 Existing Sparsifying Methods on GPUs 2.3 The Characterization of Sparsity

1. 製作 Ubuntu 開機隨身碟 下載 ubuntu 18.04.03 ISO image ubuntu 18.04.03 下載 balenaEtcher ，以用來製作開機隨身碟 balenaEtcher

EASY 1. Hello_CTF 解題思路 這題應該是要告訴我們密碼的格式都是EENS{XXX} ，答案就寫在題目上 2. You can't type 問題分析 這個題目提供一個 korea.PNG 檔案給我們下載，下載下來後將圖片打開後會發現密碼就寫在圖片中，但密碼中包含韓文，大部分的同學不認得韓文且又不能直接複製中的文字。 解題思路

使用 valgrind 效能分析工具分析 lenet , yolov3-tiny 在 MDFI 以及 Darknet 所消耗 heap 的比較 指令參考 5.3.3. 使用 MASSIF 側寫堆積與堆疊空間 Darknet for yolov3-tiny 在 makefile 中要把 DEBUG 參數改 1 並重新 make 執行 darkent 並產生 massif.out

Outline AI 課程 計算機組織實驗 ARC Contest Code generation AI 課程 第一周

# 論文 ## ISCA 2018 ### GPUs - [ ] [RegMutex: Inter-Warp GPU Register Time-Sharing](https://drive.google.com/open?id=1hXG25sjvY-7Ldf6Q5GlkNwoksV450PtI) - share a subset of physical registers between warps during the GPU kernel execution. >warp: 是 CUDA 中，每一個 SM 執行的最小單位 >SM: Streaming Multiprocessor >參考： [CUDA 的 Threading：Block 和 Grid 的設定與 Warp](https://kheresy.wordpress.com/2008/07/09/cuda-%E7%9A%84-threading%EF%BC%9Ablock-%E5%92%8C-grid-%E7%9A%84%E8%A8%AD%E5%AE%9A%E8%

# Assignment3: SoftCPU ## Modify the assembly programs for Reindeer Simulation with Verilator. ### A new Makefile In order to run the assembly program with Verilator, we have to generate the `.elf` file, and we can find some clues in the `Makefile` of `riscv-compliance` for generating `.elf` file. There are the messages when i command `make` in `riscv-compliance` folder.  In this picture we can find some path of libraries and `link` file we need when compi

# Assignment2: RISC-V Toolchain ## 1.pick up two assembly programs from [Assignment1: RISC-V Assembly and Instruction Pipeline](https://hackmd.io/ZHsZ-7HBTvqtH8SoDHxQoQ) and rewrite into C implementations which can execute with [rv32emu](https://github.com/sysprog21/rv32emu). ### Bit Reverse Giving an input number and reverse every bit of the number. #### C code for emu-rv32 ```cpp= void _start() { volatile char* tx = (volatile char*) 0x40002000; unsigned int example = 0x12345678

# Assignment1: RISC-V Assembly and Instruction Pipeline -- Bubble Sort contributed by < `Shengyuu` > ## C code Sort integer array `arr[]` which includes 5 integer with bublle sort algorithm ```cpp #include <stdio.h> int main() { int arr[5]; arr[0] = 3; arr[1] = 5; arr[2] = 1; arr[3] = 2; arr[4] = 4; int tmp; for(int i = 0; i < 4; i++){ for(int j = 0; j < 4 - i; j++){ if(arr[j] > arr[j + 1]){ tmp = arr[j];

# 2019q3 Homework2 (lab0) contributed by < `Shengyuu` > ## 說明 - [作業要求](https://hackmd.io/s/BJA8EgFB4) - [C Programming Lab](http://www.cs.cmu.edu/~213/labs/cprogramminglab.pdf) ## 實驗環境 ```shell $ uname -a Linux yao 4.15.0-45-generic #48-Ubuntu SMP Tue Jan 29 16:28:13 UTC 2019 x86_64 x86_64 x86_64 GNU/Linu $ gcc --version gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0 ``` ## 作業需求 ### 實作 FIFO & LIFO Queue - 要求 - `q_new`: 新增一空的 queue - `q_free`: 清除 queue - `q_insert_head`: 在 queue 的頭插入一元

# 2019q3 Homework1 (review) contributed by < `Shengyuu` > ## 題組 1 考慮以下 C 程式的 align4 巨集的作用是，針對 4-byte alignment，找到給定地址的 round up alignment address。 ```cpp #include <stdio.h> #define align4(x) (((x) + K) & (-4)) int main(void) { int p = 0x1997; printf("align4(p) is %08x\n", align4(p)); return 0; } ``` 預期程式輸出 `align4(p) is 00001998` ### 解題思路 題目中 4-byte-alignment 的意思是要將給定的地址做 align ，讓地址永遠維持 4 的倍數，要維持 4 的倍數有兩種可能，分別是 `round up` 以及 `round down` ，例如：將題目中的 `p = 0x1997` align 成 `0x1998` 及為

# Code Generation ## 定義 Definition - What does Code Generation mean? Code generation is a mechanism where a compiler takes the source code as an input and converts it into machine code. This machine code is actually executed by the system. **Code generation is generally considered the last phase of compilation, although there are multiple intermediate steps performed before the final executable is produced.** These intermediate steps are used to perform optimization and other relevant processe