User:Blevando

I am an undergraduate at Valparaiso University pursuing a Bachelor of Science in Computer Science.

E-mail: benjamin.levandowski@marquette.edu

Work log

Week 1

• General Orientation of REU and projects, along with Marquette University and its facilities
• Met with mentor and student researchers to discuss potential projects
• Read literature regarding potential projects
  • Previous summer's research paper and an older paper on Embedded Xinu
  • Paper on MiniOS and Papers on the Tock Embedded OS
  • Paper describing Xest
  • Looked at the ARM Cortex-A53 Technical Reference Manual to familiarize myself with ARM assembly and C calling convention
  • Skimmed papers on Computer Science Secondary Education
• Began to familiarize myself with Embedded Xinu by completing the UART homework from COSC 3250
• Briefly explored the Rust programming language
• Finalized a research project

Week 2

• Read through relevant portions of the ACM/IEEE Computer Science Curricula 2013 Report
• Found and read papers relating to computer science education in parallel and distributed computing and its relation to hardware
  • Primarily searched through publications that derived from the Special Interest Group in Computer Science Education
• Furthered my familiarity with the multi-core Pi 3
• Learned more ARM assembly by exploring the factorial homework from COSC 2200
• Learned how not to give a presentation
• Began to build a sandbox environment for XINU PI3 to mirror the arm playground for the Raspberry Pi
  • Successfully able to run an isolated assembly file on the Pi3
  • Environment still needs better structure
  • Need to be able to easily run code on all cores

Week 3

• Sandbox environment easily runs code on all processors
  • Behavior is unpredictable after code terminates, loop forever for now
• Completed Responsible Conduct of Research training including online modules and face-to-face discussion
• Searched for programming tasks used in education
  • Tasks designed for parallel assembly elude me, but some may be able to be modified to fit this purpose
• Serial output needs attention now that up to 4 cores may write to it at once
• A specific teachable task will be chosen next week

Week 4

• Decided on a specific problem, the coupon collector's problem
  • Easy to parallelize, works naturally with shared memory, and fits well with existing curricula
• Implemented single core version in ARM assembly
• Extended functionality to multi-core
• Began to use mutexes and made use of atomic assembly instructions
• Program does not yet work properly; either there exists race conditions or issue with cache coherence
• Continued to debug software and explored solutions presented online

Week 5

• Changed memory attributes to fix existing issues with cache
• Can now demonstrate existence of race conditions
• Continued to read through ARM documentation to better understand the ISA
• Presented first five weeks' work to peers and faculty
• Looked over NEON advanced SIMD documentation
• Continued to adjust the memory model and MMU setup

Week 6

• Isolated memory issues in program
  • Can now demonstrate 2 types of memory/cache problems caused by sharing memory among cores
• Re-wrote MMU initialization function in assembly to follow the ARM bare metal example
  • This makes setting up the memory translation table more transparent
• Redesigned reserved space sharing setup to improve readability
• Improved function termination strategy—Core 0 no longer has to guess if other cores are still running
• Attempted implementing a semaphore for the PRNG use, though this largely negates the purpose of running on multiple cores
  • I may just have each core have its own PRNG
• Next week will focus on curricular material and improving formal documentation