Added Entire Sync Section

[haneefmubarak]

Mutexes

• implemented no-overhead mutexes (init(), lock(), unlock() - without extra memory allocation)
• used direct type (int) to reduce performance lost from indirection
• used native int size to maximize performance
• added p_cpu_relax to reduce load and cycle wastage during lock contention (when possible)
• utilized architecture optimized assembly (when possible)

  Atomics

• implemented using fairly portable extensions (these are as far as I checked, more platforms are actually supported in reality):
  • GCC
  • Intel
    • x86
  • x86_64
  • ARM
    • armv2
  • armv3
  • armv4
  • armv5
  • armv6
  • armv7
  • AArch64
    • armv8
  • IBM
    • PowerPC
    • Power7
    • Power8
  • Clang
  • x86
  • x86_64
  • Intel Compiler Collection
  • x86
  • x86_64
• implemented for all common integer sizes (8, 16, 32, 64)
• supports both signed and unsigned integers

  p_fence

• used an actual memory fence (printf() is not a real memory fence)
• amended function signature (memory fences work across processor - object specificity doesn't do anything & nothing to return)
• utilized highly portable extension (see above list for supporting compiler/platforms)

[haneefmubarak]

cc @olajep @aolofsson