From 0a5a4814fa0a19173f7485721dcae5dc3562e258 Mon Sep 17 00:00:00 2001
From: czj0xyz <93525193+czj0xyz@users.noreply.github.com>
Date: Sat, 17 Jun 2023 02:08:36 +0800
Subject: [PATCH] W3D2 & W7D2 & W11D2: Chen Zejia

---
 W11D2/W11D2Notes_ChenZejia.md | 42 +++++++++++++++++++++++++++++++++++
 W3D2/W3D2Notes_ChenZejia.md   | 36 ++++++++++++++++++++++++++++++
 W7D2/W7D2Notes_ChenZejia.md   | 34 ++++++++++++++++++++++++++++
 3 files changed, 112 insertions(+)
 create mode 100644 W11D2/W11D2Notes_ChenZejia.md
 create mode 100644 W3D2/W3D2Notes_ChenZejia.md
 create mode 100644 W7D2/W7D2Notes_ChenZejia.md

diff --git a/W11D2/W11D2Notes_ChenZejia.md b/W11D2/W11D2Notes_ChenZejia.md
new file mode 100644
index 0000000..7c511d0
--- /dev/null
+++ b/W11D2/W11D2Notes_ChenZejia.md
@@ -0,0 +1,42 @@
+## W11D2 notes
+
+**written by Chen Zejia.**
+
+## Multi-core/processer
+
+* e.g. SMC/SMP/cluster/Grid/Cloud...
+* UMA: provides equal access to memory for all processors and has lower latency.
+* NUMA: provides higher scalability and better memory utilization.
+
+### Multiprocessor Synchronization
+
+* Multiple spinlock: locks form a list.
+* Use of multiple locks to avoid cache thrashing
+
+### Time/Space Sharing 
+
+### Gang Scheduling
+
+* Groups sof related threads are scheduled as a unit, a gang.
+* All member of a gang run at once on different timeshared CPUs.
+* All gang members start and end their time slices together.
+
+### Blocking vs Nonblocking Calls
+
+### Distributed Shared Memory
+
+* Like page table
+* "False" Sharing
+
+### Graph computing
+
+* Use graph to express show the relation(the edges represent the correlation of the data).
+* Graph-Theoretic Deterministic Algorithm
+* GraphChi
+* less space, higher locality.
+
+### Object-Based Middleware
+
+* CORBA
+* DCOM: Distributed Component Object Model
+
diff --git a/W3D2/W3D2Notes_ChenZejia.md b/W3D2/W3D2Notes_ChenZejia.md
new file mode 100644
index 0000000..4266370
--- /dev/null
+++ b/W3D2/W3D2Notes_ChenZejia.md
@@ -0,0 +1,36 @@
+## W3D2 notes
+
+**written by Chen Zejia.**
+
+## Process/Thread
+
+### Sync
+
+* lock, mutex, semaphore
+  * mutex: shared, exclusive. It might lead to Race Condition.
+* CLI: Clear Interrupt.
+* Busywating: keep testing and setting(might be no atomic).
+* Sleep and wake up
+* Semaphone by Dijkstra
+* RCU: read-copy-update
+
+### producer-consumer problem
+
+* using semaphores: use a counter/waiting queue. The order of up/down cannot be revered.
+* mutex in pthread: condition variable and mutex
+* Monitors: try to solve this in language level. Every time there are at most one procedure running at a time. Use condition variable to switch.
+
+### Barriers
+
+* Memory barriers: it is used to avoid the reverse of the operation of memory(eg: Tomasulo in ARH, 
+* Memory consistency.
+
+### Schedule
+
+* Priority, time slice(time-use unit)
+* RR 
+* O(1): use to two queues to maintain the processes
+* Staircase Down -> RSDL
+* CFS
+* Brain Fuck
+
diff --git a/W7D2/W7D2Notes_ChenZejia.md b/W7D2/W7D2Notes_ChenZejia.md
new file mode 100644
index 0000000..3e23ea5
--- /dev/null
+++ b/W7D2/W7D2Notes_ChenZejia.md
@@ -0,0 +1,34 @@
+## W7D2 notes
+
+**written by Chen Zejia.**
+
+## Deadlock
+
+### Definition
+
+A set of process is deadlocked if:
+
+* Each process is waiting for an event
+* That event can be caused only by another process
+
+### Conditions:
+
+* Mutual exclusion
+* Hold and wait
+* No preemption
+* Circular wait condition
+
+### Solutions
+
+* Don't care, ignore the problem.
+* Detection and recovery
+  * use something(e.g. matrix/graph) to maintain the resources.
+* Dynamic avoidance by careful resource allocation.
+* Prevention, by structurally negating one of the four required conditions:
+  * Mutual exclusion 
+  * Hold and wait: Request all resources initially
+  * No preemption:  Take resources away
+  * Circular wait: Order resources numerically
+
+### Banker's algorithm
+