<div id="readability-page-1" class="page"><article><header><p>🖥️ CPU, Memory, OS, Kernel</p><small>Created: 2020, 03 16 &gt;<a href="https://github.com/shdkej/shdkej.github.io/blame/master/content/Computer_Architecture.md">Updated: 2021, 05 23</a></small></header><nav><strong>Table of Contents</strong><div><ul>
<li><a href="/computer_architecture/#summary">Summary</a></li>
<li>
<p><a href="/computer_architecture/#cpu">CPU</a></p>
<ul>
<li>
<ul>
<li><a href="/computer_architecture/#process-thread">Process, Thread</a></li>
<li><a href="/computer_architecture/#process">process</a></li>
</ul>
</li>
</ul>
</li>
<li>
<p><a href="/computer_architecture/#memory">MEMORY</a></p>
<ul>
<li>
<ul>
<li><a href="/computer_architecture/#memory-management">memory management</a></li>
<li><a href="/computer_architecture/#%EB%A9%94%EB%AA%A8%EB%A6%AC%EC%99%80-cpu-%EC%BA%90%EC%8B%9C%EB%8A%94-%EA%B5%AC%EC%A1%B0%EA%B0%80-%EB%8B%A4%EB%A5%B4%EB%8B%A4">메모리와 cpu 캐시는 구조가 다르다</a></li>
</ul>
</li>
</ul>
</li>
<li>
<p><a href="/computer_architecture/#cache">Cache</a></p>
<ul>
<li>
<ul>
<li><a href="/computer_architecture/#cache-1">cache</a></li>
</ul>
</li>
</ul>
</li>
<li>
<p><a href="/computer_architecture/#ipc">IPC</a></p>
<ul>
<li>
<ul>
<li><a href="/computer_architecture/#disk-io">Disk I/O</a></li>
<li><a href="/computer_architecture/#%EC%84%B8%EB%A7%88%ED%8F%AC%EC%96%B4%EC%8B%A0%ED%98%B8%EA%B8%B0">세마포어(신호기)</a></li>
<li><a href="/computer_architecture/#iocpinputoutput-completion-port">IOCP(input/output Completion Port)</a></li>
</ul>
</li>
</ul>
</li>
<li>
<p><a href="/computer_architecture/#os">OS</a></p>
<ul>
<li>
<ul>
<li><a href="/computer_architecture/#in-os-micro-kernel-vs-monolithic-kernel">In OS, micro kernel vs monolithic kernel</a></li>
</ul>
</li>
</ul>
</li>
</ul></div></nav><p>---</p><section><h2 id="summary"><a href="#summary" aria-label="summary permalink"></a>Summary</h2>
<p>디스크에 있는 프로그램을 메모리로 부르고,
메모리에 있는 데이터를 CPU가 읽어가면서 프로그램이 동작한다.</p>
<h2 id="cpu"><a href="#cpu" aria-label="cpu permalink"></a>CPU</h2>
<p>불 대수 -&gt; 폰 노이만
Transistor -&gt; Flipflop -&gt; IC -&gt; CPU or RAM</p>
<ul>
<li>AND OR XOR</li>
<li>ARU MBR MAR -- John von neumann</li>
<li>low level - Assembly - complier - high level language</li>
<li>low level (CISC - intel, RISC - arm)</li>
<li>USER want to running program: HDD -&gt; RAM -&gt; CPU (process) -&gt; OUTPUT</li>
<li>really running CPU with RAM and process scheduler</li>
</ul>
<h4 id="process-thread"><a href="#process-thread" aria-label="process thread permalink"></a>Process, Thread</h4>
<p>in 1970, computer has only 1 process, can run one program. process can make
run multiple program. but people need multi-work. so multi-processing
invented.</p>
<h4 id="process"><a href="#process" aria-label="process permalink"></a>process</h4>
<p>디스크에 저장된 프로그램이 코드와 데이터를 갖고 있으면 이를 실행시키면
프로세스를 RAM에 적재하면서 코드와 데이터를 가져오고, 스택과 힙을 할당 받는다
코드를 읽는 것은 CPU에서 메모리를 불러와 명령을 수행한다.</p>
</section><section>
<hr>
<h2 id="memory"><a href="#memory" aria-label="memory permalink"></a>MEMORY</h2>
<ul>
<li>why cpu is fast, memory is slow</li>
<li>Data - Code text</li>
<li>Heap - Global variable</li>
<li>Stack - region variable</li>
</ul>
</section><section>
<h4 id="memory-management"><a href="#memory-management" aria-label="memory management permalink"></a>memory management</h4>
<ul>
<li>memory leak</li>
<li>internal fragmentation: 100 size but only use 99. 1 is fragmentation.</li>
<li>External fragmentation: 100 -- 100 -- 0 -- 100 -- 100, 0 is fragmentation.</li>
<li>Paging: memory separate to static size(it called page) by logical memory in
physical memory.</li>
<li>Segmentation: memory separate dynamic size. even though physical location.</li>
</ul>
<h4 id="메모리와-cpu-캐시는-구조가-다르다"><a href="#%EB%A9%94%EB%AA%A8%EB%A6%AC%EC%99%80-cpu-%EC%BA%90%EC%8B%9C%EB%8A%94-%EA%B5%AC%EC%A1%B0%EA%B0%80-%EB%8B%A4%EB%A5%B4%EB%8B%A4" aria-label="메모리와 cpu 캐시는 구조가 다르다 permalink"></a>메모리와 cpu 캐시는 구조가 다르다</h4>
<p>메모리는 dram, 캐시는 sram
cpu와 메모리의 발전 속도가 차이나서 지금도 차이가 있다</p>
<p>메모리는 cpu에 비해 데이터 보관이 필요하다</p>
<h2 id="cache"><a href="#cache" aria-label="cache permalink"></a>Cache</h2>
<ul>
<li>SRAM(static) 으로 만들어진다</li>
<li>일반적으로 사용되는 RAM은 DRAM이다 (dynamic random access memory)</li>
<li>캐시는 키:벨류 타입의 해시테이블로 구성되어 있어서 시간복잡도가 O(1)이다</li>
</ul>
<h4 id="cache-1"><a href="#cache-1" aria-label="cache 1 permalink"></a>cache</h4>
<ul>
<li>batch</li>
<li>if need store log to db, firstly store log to redis, and batch to db</li>
</ul>
<h2 id="ipc"><a href="#ipc" aria-label="ipc permalink"></a>IPC</h2>
<p>socket, file handling, pipe, message queue, semaphore, shared memory</p>
<h4 id="disk-io"><a href="#disk-io" aria-label="disk io permalink"></a>Disk I/O</h4>
<h4 id="세마포어신호기"><a href="#%EC%84%B8%EB%A7%88%ED%8F%AC%EC%96%B4%EC%8B%A0%ED%98%B8%EA%B8%B0" aria-label="세마포어신호기 permalink"></a>세마포어(신호기)</h4>
<p>it's like a counter for check the processor is working.
by dijkstra for multiprocessor's deadlock
it looks like golang's waitgroup</p>
<p>mutex is binary semaphore for thread.
GIL is mutex for python to prevent use object by multi thread.</p>
<ul>
<li>멀티 스레드 사용 시 스레드들의 간섭을 막기 위해 Mutex라는 키를 이용하여 접근을 제어한다.</li>
<li>한 스레드가 mutex를 lock 시키면 lock을 해제할 때까지 다른 스레드는 작업을 하지 못하게 해준다.</li>
</ul>
<h4 id="iocpinputoutput-completion-port"><a href="#iocpinputoutput-completion-port" aria-label="iocpinputoutput completion port permalink"></a>IOCP(input/output Completion Port)</h4>
<p>epoll
kqueue
it is for asynchronization I/O.
every computer use I/O for computing.
nowadays computer has multiple processor. So windows has IOCP, linux has epoll.
networking also used with socket generally.</p>
<h2 id="os"><a href="#os" aria-label="os permalink"></a>OS</h2>
<ul>
<li>multics - unix - minix - linux</li>
<li>zenix - Macintosh - windows</li>
<li>unix, mac, windows are not first os. but it is now most used OS.</li>
</ul>
<h4 id="in-os-micro-kernel-vs-monolithic-kernel"><a href="#in-os-micro-kernel-vs-monolithic-kernel" aria-label="in os micro kernel vs monolithic kernel permalink"></a>In OS, micro kernel vs monolithic kernel</h4>
<ul>
<li>unix, linux = monolithic. base on kernel</li>
<li>micro kernel is base on each process. and communicate with other messenger</li>
<li>micro kernel can use general hardware. linux made for 1 target hardware.
it is good to use more hardware performance. but it depends on 1 hardware.
micro kernel's problem is don't use hardware performance efficiently.
micro kernel has portability.</li>
</ul>
</section><hr></article></div>

shdkej

Summary
디스크에 있는 프로그램을 메모리로 부르고,
메모리에 있는 데이터를 CPU가 읽어가면서 프로그램이 동작한다.
CPU
불 대수 -> 폰 노이만
Transistor -> Flipflop -> IC -> CPU or RAM
AND OR XOR
ARU MBR MAR -- John von neumann
low level - Assembly - complier - high level language
low level (CISC - intel, RISC - arm)
USER want to running program: HDD -> RAM -> CPU (process) -> OUTPUT
really running CPU with RAM and process scheduler
Process, Thread
in 1970, computer has only 1 process, can run one program. process can make
run multiple program. but people need multi-work. so multi-processing
invented.
process
디스크에 저장된 프로그램이 코드와 데이터를 갖고 있으면 이를 실행시키면
프로세스를 RAM에 적재하면서 코드와 데이터를 가져오고, 스택과 힙을 할당 받는다
코드를 읽는 것은 CPU에서 메모리를 불러와 명령을 수행한다.
[[Software#History]]
MEMORY
why cpu is fast, memory is slow
Data - Code text
Heap - Global variable
Stack - region variable
[[Data_Structure]]
memory management
memory leak
internal fragmentation: 100 size but only use 99. 1 is fragmentation.
External fragmentation: 100 -- 100 -- 0 -- 100 -- 100, 0 is fragmentation.
Paging: memory separate to static size(it called page) by logical memory in
physical memory.
Segmentation: memory separate dynamic size. even though physical location.
메모리와 cpu 캐시는 구조가 다르다
메모리는 dram, 캐시는 sram
cpu와 메모리의 발전 속도가 차이나서 지금도 차이가 있다
메모리는 cpu에 비해 데이터 보관이 필요하다
Cache
SRAM(static) 으로 만들어진다
일반적으로 사용되는 RAM은 DRAM이다 (dynamic random access memory)
캐시는 키:벨류 타입의 해시테이블로 구성되어 있어서 시간복잡도가 O(1)이다
cache
batch
if need store log to db, firstly store log to redis, and batch to db
IPC
socket, file handling, pipe, message queue, semaphore, shared memory
Disk I/O
세마포어(신호기)
it's like a counter for check the processor is working.
by dijkstra for multiprocessor's deadlock
it looks like golang's waitgroup
mutex is binary semaphore for thread.
GIL is mutex for python to prevent use object by multi thread.
멀티 스레드 사용 시 스레드들의 간섭을 막기 위해 Mutex라는 키를 이용하여 접근을 제어한다.
한 스레드가 mutex를 lock 시키면 lock을 해제할 때까지 다른 스레드는 작업을 하지 못하게 해준다.
IOCP(input/output Completion Port)
epoll
kqueue
it is for asynchronization I/O.
every computer use I/O for computing.
nowadays computer has multiple processor. So windows has IOCP, linux has epoll.
networking also used with socket generally.
OS
multics - unix - minix - linux
zenix - Macintosh - windows
unix, mac, windows are not first os. but it is now most used OS.
In OS, micro kernel vs monolithic kernel
unix, linux = monolithic. base on kernel
micro kernel is base on each process. and communicate with other messenger
micro kernel can use general hardware. linux made for 1 target hardware.
it is good to use more hardware performance. but it depends on 1 hardware.
micro kernel's problem is don't use hardware performance efficiently.
micro kernel has portability.
[[Software#desktop os history]]

Computer Architecture

댓글