Update 01-parallel-programming.md

Author: csccva

intro-day/docs/01-parallel-programming.md

@@ -270,16 +270,13 @@ EuroHPC JU LUMI (HPE Cray EX) <br>at CSC (2021–)
 # Summary
 
 - Parallel Computing: data is splitted in small independent parts
-  - Tightly coupled vs. Embarrassingly parallel
-  - Data Parallelism vs. Task Parallelism
+  - Tightly coupled vs. Embarrassingly parallel, Data  vs. Task Parallelism
 - Parallelization causes an extra cost not present in a serial program
   - Competition over the same data can cause race conditions or deadlocks
-  - Data distribution: local vs. global dependencies
-  - Data distribution: load balance
+  - Data distribution: local vs. global dependencies, load balance
 - Parallel programming models:
   - The hardware target  affect the algorithm choice
-  - Processes vs. Threads
-      - GPU programming models are threads based
+  - Processes vs. Threads (include GPU programming model)
 - Parallel Programs run on Supecomputers
   - Nodes, Fast Interconect, Parallel Storage Servers
   - LUMI: 2064 CPU nodes (~260,000 CPU cores), 2978 GPU nodes (~12,000 MI250X GPUs)