Implement Savina benchmarks in C

[cmnrd]

This issue is to keep track of the current status of the C benchmarks.

• micro benchmarks
  • Big -> Arthur
  • Chameneos -> Marten
  • Count
  • Fork Join
  • Ping Pong
  • Thread Ring
• concurrency benchmarks
  • Banking
    • Implement
    • Resolve Epoch and lfc become unusable for programs with large (>500) banks and multiports #433
    • Merge banking and pseudorandom #624
  • Barber -> Matt
    • Implement
    • Integrate with runner script
    • Address the comments made in Add C Benchmarks for Cigarette Smokers and Sleeping Barber #401
    • Merge C example sleepingbarber #671
  • Bounded Buffer (ProdCons) -> Matt
    • Update to align with C++ version
  • Cigarette Smokers
    • Implement
    • Integrate with runner script
    • Address the comments made in Adding new C benchmarks and updating old ones #445
  • Concurrent Dictionary -> Matt
    • First implementation
    • Check if it aligns with the C++ version
    • Integrate with the runner script
  • Concurrent Sorted List -> Matt
    • First implementation
    • Check if it aligns with the C++ version
    • Integrate with the runner script
  • Logistic Map
  • Philosophers -> Christian
    • Align C and C++ version
    • cog the main reactor and integrate with runner script
• parallelism benchmarks
  • APSP -> Soroush
  • ASTAR -> Soroush
  • Filter Bank
  • N Queens K Solutions -> Arthur
    • Implement
    • Fix warnings regarding duplicated frees (see Big and NQueens C Benchmarks #721)
  • Pi Precise -> Christian
    • Simplify and align with C++ version
    • Fix FIXMEs
  • Radix Sort
    • use cog and multiports/banks to instantiate all workers
    • Investigate performance issues (in C++)
    • Make sure updated C++ version and C version align
  • Rec Matrix Multiplication -> Peter
  • Trapezoid
• other tasks:
  • consistent use of deque
  • C version of MatMul is slower due to caching effects -> use BenchmarkRunner reactor!?
  • What to do with Smokers.lf? (Thee's version of the smoker problem)

[edwardalee]

In the branch c-example-sleepingbarber, I fixed benchmark/C/src/concurrency/SleepingBarber.lf to correctly use the random number generator. It runs much faster.

I have have also implemented example/C/src/SleepingBarber.lf, which has the same structure as the benchmark, but uses logical actions and randomly generated logical times. That one is useful for watching what is going on.

I then created a version of the example that has the same parameters as benchmark/C/src/concurrency/SleepingBarber.lf and includes the same busy waits. Its performance was about the same, however, so it's not worth checking it in.

[cmnrd]

Closing this since most of the tasks are done. There are separate issues for the remaining tasks (#765, #624).