Category: Diary

Research:

• Implement KroEngine in Unity

LeetCode:

• 19. Remove Nth Node From End of List
• 17. Letter Combinations of a Phone Number
• 27. Remove Element

Parallel Computing Study:

• Lesson3

Guitar:

• practice

Others:

Presentation next Monday:

• Prepare the PPT

Research:

• Read paper: Real-time Rendering on a Power Budget

LeetCode:

• 12. Integer to Roman
  • Initialiazation of the unordered map affects timing performance
• 11. Container With Most Water
  • squeeze, O(n)
• 42. Trapping Rain Water
  • Scan three times

Parallel Computing Study:

• Finish watching video of Lesson3: Fundamental GPU Algorithms
  • Reduce
    • Reduction Operator a) binary (two arguments)     b) associative
  • Scan
    • Input:
      • input array
      • binary associative operator
      • identity element [I op a = a]

Guitar:

• practice

Others:

• O(1) independent of size of inputs

Presentation next Monday:

• Prepare the PPT

Research:

• Implement KFR with Optical Flow in OpenGL
• Cannot Output the model from Unity (only part of the model can be exported), but we can try to test FPS on KroEngine.
  • Origin: 90 – 100 FPS
  • GausBlur: 62 -69 FPS
  • KFR: 58 – 65 FPS

LeetCode:

• 6. ZigZag Conversion
  • Usage of sring.append() is much more flexible than s.push_back() and s+=***
• 7. Reverse Integer
  • Overflow, INT_MAX, INT_MIN
• 12. Integer to Roman
  • Initialiazation of the unordered map affects timing performance

Parallel Computing Study:

• Finish watching video of Lesson2
  • coalesced: access contiguous memory
  • strided: access non-contiguous memory
  • for floats, (a + b) + c  != a + (b + c), e.g. a = 1, b = 10^99, c = -10 ^ 99
  • atomic functions are costly

Guitar:

• practice.

Others:

Presentation next Monday:

• Prepare the PPT

Research:

• Measure time of the three scenes (KFR, Origin, GausOptFlow).
• Origin: 32.2 fps
• KFR: 20.4 fps
• GausOptFlow: 20.2 fps

LeetCode:

• 4. Median of Two Sorted Arrays
  • Cannot use concatinate + sort because the time complexity goes beyond log(m+n).
  • If time complexity is log(*), consider about dichotomy.
• 5. Longest Palindromic Substring
  • Manacher
  • DP
  • s.substr(start, length) (can also use this to read the ith element in string in string format, you know, s[i] is char)

Parallel Computing Study:

• Finish watching video of Lesson2

Guitar:

• practice.

Others:

• About time complexity of sort algorithms

Presentation next Monday:

• Prepare the PPT

Research:

• Merge the three scenes (KFR, Origin, GausOptFLow) into one program.
• Tried to measure time: have problem (the time are all the same)
• Tried to export mesh but failed.

LeetCode:

• 2. Add Two Numbers
  • Cannot use list2num/num2list (overflow), should use full adder algorithm.
• 3. Longest Substring Without Repeating Characters
  • “dvdf”

Parallel Computing Study:

• Finish watching video of Lesson2

733Code->C++:

• Panorama Maker

Guitar:

• practice.

Others:

• About time complexity

Presentation next Monday:

• Prepare the PPT

Try to finish the projects of CMSC733 by using CPP+opencv.

An tutorial about opencv:

http://blog.csdn.net/column/details/opencv-tutorial.html

Research:

• no research task today~~~lol~~~! Where is varshney?????

Leetcode:

Finished 4 leetcode problems.

• 228. Summary Ranges (easy)

• 209. Minimum Size Subarray Sum
  • Add at the tail and subtract at the head.
• 18. 4Sum
  • sort, and squeeze from both sizes. Actually it will be faster to store the sum of two numbers by using hashmap.
• 373. Find K Pairs with Smallest Sums
  • should use priority queue

Parallel Computing Study:

• Finish hw1.
• Finish watching video of Lesson2

Website:

• Installed plugin Crayon Syntax Highlighter, but actually I don’t know how to use it…

Guitar:

• Watch a new tutorial and practice.

Research:

• no research task today~~~lol~~~! Where is varshney?????

Leetcode:

Finished 5 leetcode problems.

• Know the difference between vec.emplace_back() and vec.push_back().
  • https://stackoverflow.com/questions/4303513/push-back-vs-emplace-back

Parallel Computing Study:

• Watch video of class2
  • https://classroom.udacity.com/courses/cs344/lessons/67054969/concepts/673560200923
  • Bill Dally mentioned that CPU is not getting faster…we use more parallelism now.
  • Four communication patterns:
    • map: all the inputs do the same calculation (efficient in GPU).
    • gather: calc average of several pixels.
    • scatter: write several memory units with one value.
    • stencil: tasks read input from a fixed neighborhood in an array.
  • Transpose:
    • reorder data elements in memory
    • Transpose: AoS [array of structures] vs.  SoA [structure of arrays]
  • Transpose: AoS [array of structures] vs.  SoA [structure of arrays]
• install CUDA on my PC.

Website:

• no

Guitar:

• Watch a new tutorial.
  • https://www.youtube.com/watch?v=c3VcnvVZ1dA&index=8&list=PLf1ZeiBxYLUYUbCgiztit-Z8AkvqwCaHZ