CS 233 MicroLectures

A computer can do two things: store state and manipulate state

Applying logic operations to multi-bit state information

Exploring the difference between data and control with the XOR function

One way to translate truth tables to Boolean expressions

Designing combinational logic circuits

Using hexadecimal to make it easier to read/write binary bit strings

Performing addition with fixed-width binary representations

Interpreting and using bit shift operations

Building circuits that can perform addition and substraction

We use multiplexers to choose between multiple streams of data

Creating a component that controls what arithmetic and logic operations we perform on system state

Basic concepts and notation for finite state machines

A datapath for performing binary multiplication

Creating a programmable datapath

How to convert MIPS assembly instructions into machine code and machine code into control signals

Instruction Set Architecture make high-level languages (like C and Python) possible, but they are possible only because we have programmable datapaths like our arithmetic machine.

Overview of how to read the MIPS Green Sheet

Creating a programmable finite state machine 

Adding the lui instruction to the MIPS datapath

Adding the slt and slti instructions to the MIPS datapath

MIPS provides pseudoinstructions to make your life easier

Live coding example using caller save register conventions

Manipulating pointers in MIPS

Using the spiral rule to parse C variable declarations

Manipulating more complex data types and more complex data structures

Live coding example: Fib recursion (Callee save)

Different methods for receiving and sending inputs and outputs

Using the status and cause registers to handle interrupts and exceptions

Exceptions are errors in our programs

Introduction to SPIMbot

Defining and giving an example of latency and throughput

Estimating the time it takes for a program to run

Comparing the throughput and latency of single-cycle datapath and pipelined datapath

Example of data propagating through a pipelined datapath

Data dependencies are instructions the read values written to the register file by previous instructions

Data forwarding to fix data hazards

Stalling the Pipeline

Counting how many cycle a program takes to execute when stalls are involved

Control Hazards and Flushes

Using temporal and spatial locality in code to improve memory performance

One method for mapping a large main memory to small caches

Taking advantage of spatial locality using multi-byte cache blocks

Taking advantage of temporal locality using set associativity

Overview of why we have multi-level cache designs and how they work.

Writing to caches and back to memory

Using prefetching to minimize the effect of cache misses

Analyzing a baseline piece of code without tiling

Virtual memory provides access to increased data storage and data security

Vectorization and Single Instruction Multiple Data (SIMD)

Vectorization and Dependencies