---
title: Computer Architecture Week1
tags: jserv
---

# On Fall 2021
> [Course online links](https://youtu.be/h_3ZakOBxTM)

## Introduction
- Compatibility is important on computer industry

## David Patterson: [A New Golden Age for Computer Architecture: History, Challenges and Opportunities](https://content.riscv.org/wp-content/uploads/2018/12/A-New-Golden-Age-for-Computer-Architecture-History-Challenges-and-Opportunities-David-Patterson-.pdf) / [video](https://youtu.be/uyc_pDBJotI)
> [Article](https://cacm.acm.org/magazines/2019/2/234352-a-new-golden-age-for-computer-architecture/fulltext), [note](https://github.com/cristiean/notes/blob/4c4ceb291ba68c55659ab0d4571385c8cccacd21/a-new-golden-age-for-computer-architecture.md)

- Software advances can inspire architecture
- Virtualization technology
- hardware support: [Intel vt-x vt-d vt-c](https://www.itread01.com/content/1550308865.html)
- MicroVAX
- simplify everything
- inspire **RISC**
- 
- 
- 

## David Patterson: [From Mainframe CPUs to Neural-Network TPUs and Open RISC-V](https://riscv.org/wp-content/uploads/2017/05/Mon0915-RISC-V-50-Years-Computer-Arch.pdf) / [video](https://youtu.be/eH9_ygsio3I)

## [Great Ideas in Computer Architecture, Intro](https://drive.google.com/open?id=1Y4yUaw9vryZcdc4Xu0RQxHfqjVbeHJLF) / [video](https://youtu.be/u_hXH5-tbM4?t=263)

## [Number Representation](https://drive.google.com/drive/folders/1_Zbt8KdwieLsqRdfc6aNBOa2_hEmFkxR)
- trick on switch of upper/lower cases in ASCII code
- one's complement: still be used in tcp/ip
- Sign extension

## [Floating Point](https://drive.google.com/drive/folders/147xm_DhklkevG6602mH4yoGuVMw-3GQZ?usp=sharing)
### Fixed point
- sometimes shortened to **fixpoint**
- need to think the scope of number you want to represent
#### fixed-point multiplication
> [Multiplication Examples Using the Fixed-Point Representation](https://www.allaboutcircuits.com/technical-articles/multiplication-examples-using-the-fixed-point-representation/)
- in a manner similar to the **decimal multiplication**
- Steps:
1. ignore the binary point of the multiplier and multiplicand
> by making the operands to be an integer
2. perform the multiplication treating the operands as [two’s complement](https://www.allaboutcircuits.com/technical-articles/twos-complement-representation-theory-and-examples/) numbers
3. determine the position of the binary point for the result

### Representation of Fractions
- We really want is to "float" the binary point. Why?
> Floating binary point most effective use of our limited bits (and thus more accuracy in our number representation)

### Floating Point Representation
- :question: Representation range
> 延伸閱讀：[Floating-Point Arithmetic](/amphuEo3SdGXbWCYMndTBg)

- *NaN* has many kinds of Not a Number
- float vs double
- precision
- efficient
- :warning: convert double to float data type
- limitation
- efficient
- int -> float -> int
- may have rounding error
### Floating Point Fallacy
- FP add associative: **FALSE!**
- Why? FP result __approximates__ real result!

## [C Memory Management](https://drive.google.com/drive/folders/1pHpHRvjEg5TdidIp_fkOrh6TvMlYl5Yj?usp=sharing) / [video](https://youtu.be/fhEE4J4QCps)

# On Fall 2022
> [Course online link](https://youtu.be/kchwRBGRIHk)

## Introduction
- The performance of Single-Thread Processor Performance is rapidly increased from year `1985` to `2000` and ended on year `2000`.

- Software developers *must* consider
- Parallel systems
- Heterogeneous systems

## David Patterson: [A New Golden Age for Computer Architecture: History, Challenges and Opportunities](https://content.riscv.org/wp-content/uploads/2018/12/A-New-Golden-Age-for-Computer-Architecture-History-Challenges-and-Opportunities-David-Patterson-.pdf) / [video](https://youtu.be/uyc_pDBJotI)
> [Article](https://cacm.acm.org/magazines/2019/2/234352-a-new-golden-age-for-computer-architecture/fulltext), [note](https://github.com/cristiean/notes/blob/4c4ceb291ba68c55659ab0d4571385c8cccacd21/a-new-golden-age-for-computer-architecture.md)

- By early 1960's

- `Intel iAPX 432` is a `RISC architecture`, not compitible to `x86`
- the architecture is finally failed:
- the ambition was punished by `performance problems` and ran late, so they had to come to their founder `Gordon Moore` and say: *look we're not going to be there in time with a 16-bit microprocessor*, so that set off alarm bells at Intel and they stared an emergency project to try to build a 16-bit computer in just 52 weeks.

- The complex ISA prevents a compiler being optimized and performing advance optimizing strategies
- :notes: So, the main concept behind `RISC` is that everything is from scratch and keep as simple as possible which makes optimizing compiler can do a good job

### CISC vs. RISC Today
**CISC**
- Pros
- For *comptibility* on `PC Era`, CISC can do better, because tight coupling between Intel x86 and Microsoft Windows OS
- You can also extend the existing microprocessor with help of the `microcode (μcode)`
- Cons
- It suffers on *performance problems* because it's hard to optimize the x86 program as mentioned before
- e.g., reorder the order of instructions

**PostPC Era: Client/Cloud**
- \> `20B` total / year in 2017
- e.g., IoT embedded devices
- these devices could be powered by RISC, because of the cost and optimization requirement
- 99% Processors today are RISC
- x86 could be regards as a RISC architecture
- CISC instruction will be decoded into `micro-ops` which are native instruction in RISC architecture, such as:
- [ ] (TO-READ) [What's MPU](https://semiengineering.com/mpu-vs-mcu/)