doc: capitalize headings

Author: huydo862003

README.md

@@ -14,13 +14,13 @@ This project ports lock-free Multiple-Producer Single-Consumer (MPSC) queue algo
   - [Why MPI-3 RMA?](#why-mpi-3-rma)
   - [Hybrid MPI+MPI](#hybrid-mpimpi)
   - [Hybrid MPI+MPI+C++11](#hybrid-mpimpic11)
-  - [Lock-free MPI porting](#lock-free-mpi-porting)
+  - [Lock-Free MPI Porting](#lock-free-mpi-porting)
 - [Literature Review](#literature-review)
   - [Known Problems](#known-problems)
   - [Trends](#trends)
 - [Evaluation Strategy](#evaluation-strategy)
   - [Correctness](#correctness)
-  - [Lock-freedom](#lock-freedom)
+  - [Lock-Freedom](#lock-freedom)
   - [Performance](#performance)
   - [Scalability](#scalability)
 
@@ -76,7 +76,7 @@ Pure MPI ignores intra-node locality. MPI-3 SHM provides `MPI_Win_allocate_share
 
 C++11 atomics outperform MPI synchronization for intra-node communication. Using C++11 within shared memory windows optimizes the intra-node path.
 
-### Lock-free MPI porting
+### Lock-Free MPI Porting
 
 MPI-3 RMA enables lock-free implementations:
 
@@ -149,7 +149,7 @@ We focus on the following criteria, in the order of decreasing importance:
 - ABA-freedom
 - Safe memory reclamation
 
-### Lock-freedom
+### Lock-Freedom
 
 No process may block system-wide progress. Note: lock-freedom depends on underlying primitives being lock-free on the target platform.
 

implementations/README.md

@@ -6,7 +6,7 @@ Porting challenges & strategies are discussed further in these algorithms' folde
 
 Some have accompanying sections in the typst report to prove their correctness when modification to the original algorithms is necessary.
 
-## Ported algorithms
+## Ported Algorithms
 
 - [`LTQueue` (Prasad Jayanti & Srdjan Petrovic)](/references/LTQueue/README.md): [implementation](/implementations/ltqueue)
 

implementations/ltqueue/README.md

@@ -1,8 +1,8 @@
-# LTQueue (Prasad Jayanti, Srdjan Petrovic, 2005) - MPI port
+# LTQueue (Prasad Jayanti, Srdjan Petrovic, 2005) - MPI Port
 
 Link: [paper](/references/LTQueue/README.md)
 
-## Naive port 
+## Naive Port 
 
 ### Replacing LL/SC
 
@@ -67,7 +67,7 @@ As LL/SC is not supported by MPI, we'll have to replace them using some other su
 
   We'll investigate first scheme for the time being, which means find a way to avoid ABA problem.
 
-#### Avoiding ABA problem
+#### Avoiding ABA Problem
 
 The simplest approach is to use a monotonic version tag: Reserve some bits in the shared variable to use as a monotonic counter, so the shared variable now consists of two parts:
 * Control bits: The bits that comprise the meaningful value of the shared variable.
@@ -109,7 +109,7 @@ There is a nuance though. In the original version, the `timestamp` at each inter
 Cons:
   - Each time we read the internal node, we have to dereference the rank at the node to access the timestamp. This doubles network activities when accessing the internal node. 
 
-### Pseudo code after removing LL/SC
+### Pseudo Code After Removing LL/SC
 
 SPSC is kept intact, and due to Prasad Jayanti and Srdjan Petrovic:
 
@@ -247,7 +247,7 @@ function refresh(mpsc_t* q, tree_node_t* node)
   return CAS(&current_node->min_timestamp_rank, current_rank, (min_timestamp_rank, current_rank.version + 1))
 ```
 
-### Linearizability, Wait-freedom, Memory-safety
+### Linearizability, Wait-Freedom, Memory-Safety
 
 ### Porting
 

implementations/slotqueue/README.md

@@ -1,4 +1,4 @@
-# Slot-queue - A tailor-made optimized and simplified LTQueue for distributed context
+# Slot-Queue - A Tailor-Made Optimized and Simplified LTQueue for Distributed Context
 
 This algorithm is inspired by both [Jiffy (Dolev Adas, Roy Friedman, 2022](/references/Jiffy/README.md) and [LTQueue (Prasad Jayanti, Srdjan Petrovic, 2005)](/references/LTQueue/README.md):
   - The shared timestamp and double refresh trick is inspired by LTQueue to help Slot-queue wait-free.

report/README.md

@@ -1,4 +1,4 @@
-# Thesis report
+# Thesis Report
 
 This folder stores my thesis report: The summarized, comprehensive and focused culmination of this whole project.