My favorites | Sign in
Logo
                
Search
for
Updated Mar 03, 2010 by jaredhoberock
Labels: Phase-Requirements
Roadmap  
Tentative schedule of Thrust development

Milestone 1.1

  1. Implement lower_bound(), upper_bound(), & binary_search() (equal_range will come for free with those)
  2. Implement pair
  3. Implement all_of(), any_of(), none_of()
  4. Implement segmented scan operations
    • segmented_inclusive_scan() etc.
    • Need to determine form of segment descriptors (head flags vs. indices) or support both such as fn_by_flag() fn_by_key()
    • Challenge: 4/5
    • Experimental implementation added in r282
  5. Implement minmax_element()
  6. Implement fancy iterators

Milestone 1.2

  1. More fancy iterators
  2. Implement pseudo random number generators
  3. Implement set operations on sorted ranges
  4. Implement unique_copy()
    • Challenge: 1/5
  5. Implement unique_by_key() and unique_copy_by_key()
    • Like std::unique() except keys are specified by a separate range
    • Also takes a BinaryFunction to reduce values with the same key
    • Challenge: 2/5
  6. Implement OpenMP device backend

Milestone 1.x

  1. Implement nth_element()
  2. Versions of XXX_partition_XXX which take a sequence of predicates
    • Challenge: 1/5
  3. Implement find(), find_if(), find_if_not() and our variation with predicates?
  4. Implement temporary_buffer
  5. Implement set operations on sorted ranges
  6. Implement merge(), inplace_merge() (this might be hardcore)
  7. Implement lexicographical_compare(), lexicographical_compare_3way()
  8. Implement mismatch()
  9. Implement next_permutation(), prev_permutation()
  10. Implement partial_sort(), partial_sort_copy()
  11. Implement reverse(), reverse_copy()
  12. Implement rotate(), rotate_copy()
  13. Implement is_partitioned()
  14. Implement partition_point()
  15. Flesh out functional.h with more functions
  16. Implement Boost timer
  17. Implement some sort of exception reporting system based on Boost/C++0x std::system_error
  18. Implement a comprehensive solution to CUDA streams and task parallelism based on C++0x concurrency primitives

Comment by Damien.Massart, Dec 10, 2009

Do you have any plan to use openCL instead of Cuda ?

Thx


Sign in to add a comment
Powered by Google Project Hosting