C11 <threads.h> and <stdatomic.h> compatibility headers for MSVC++.

• C11 <threads.h> emulation library is distributed under the Boost Software License, Version 1.0.
• C11 <stdatomic.h> emulation library is distributed under the PLEASE LICENSE.

the heeader currently supports a subset of the atomic types.

• atomic_flag, atomic_bool
• atomic_char, atomic_uchar
• atomic_short, atomic_ushort
• atomic_int, atomic_uint
• atomic_long, atomic_ulong
• atomic_llong, atomic_ullong
• atomic_ptr (non-standard)

the header currently uses strong ordering for all primitives.

• atomic_load, atomic_load_explicit
• atomic_store, atomic_store_explicit
• atomic_exchange, atomic_exchange_explicit
• atomic_compare_exchange_weak, atomic_compare_exchange_weak_explicit
• atomic_compare_exchange_strong, atomic_compare_exchange_strong_explicit
• atomic_fetch_add, atomic_fetch_add_explicit
• atomic_fetch_sub, atomic_fetch_sub_explicit
• atomic_fetch_and, atomic_fetch_and_explicit
• atomic_fetch_or, atomic_fetch_or_explicit
• atomic_fetch_xor, atomic_fetch_xor_explicit
• atomic_flag_test_and_set, atomic_flag_test_and_set_explicit
• atomic_flag_clear, atomic_flag_clear_explicit
• atomic_thread_fence, atomic_signal_fence
• atomic_is_lock_free

custom pointer types are not supported due to limitations with _Generic dispatch as all types need to be specified in advance. GCC and Clang implement atomics using polymorphic compiler builtin functions so they are not exposed to the limitations of _Generic dispatch. as a work around atomic_ptr has been added which is a void* pointer type that can be used with casts to support atomic operations on pointers to custom types.

typedef void * _Atomic atomic_ptr;

simple compile, link and code generation tests. this is not intended to be a full test suite.

• test_threads - tests <threads.h> thread creation.
• test_atomics - tests <stdatomic.h> interlocked operations.
• codegen_atomics - is for inspection of <stdatomic.h> code generation.

two threads exchange counter bumps to measure inter-thread latency using the memory system.

• C11 <threads.h> and <stdatomic.h> tests:

  • latency_atomic_acquire_release_c - sender and receiver loop using acquire and release memory ops
  • latency_atomic_relaxed_barrier_c - sender and receiver loop using relaxed memory ops and barriers

• C++11 <threads>, <atomic>, and <chrono> tests:

  • latency_atomic_acquire_release_cxx - sender and receiver loop using acquire and release memory ops
  • latency_atomic_relaxed_barrier_cxx - sender and receiver loop using relaxed memory ops and barriers

mumule is a simple thread pool implementation using the C11 thread support library and C11 atomics.

• test_mumule - minimal thread pool dispatcher using queue counters

mule_submit: queue-start
mule_start: starting-threads
mule_sync: quench-queue
mule_thread-0: worker-started
mule_thread-1: worker-started
arg=(nil) thr_idx=1 item_idx=2
arg=(nil) thr_idx=1 item_idx=3
arg=(nil) thr_idx=1 item_idx=4
arg=(nil) thr_idx=1 item_idx=5
arg=(nil) thr_idx=1 item_idx=6
arg=(nil) thr_idx=1 item_idx=7
arg=(nil) thr_idx=1 item_idx=8
mule_sync: queue-processing (t=1970-01-01T00:00:01.249553791Z)
arg=(nil) thr_idx=0 item_idx=1
mule_thread-0: queue-complete
mule_sync: dispatcher-woke
mule_sync: queue-complete
mule_stop: stopping-threads
mule_thread-1: queue-empty (t=1970-01-01T00:00:01.258748034Z)
mule_thread-1: worker-exiting
mule_thread-0: queue-empty (t=1970-01-01T00:00:01.258818636Z)
mule_thread-0: worker-exiting

portable replacement for clock_gettime using platform specific clock APIs on Windows, Linux, and macOS, plus x86 TSC using rdtsc, and a fallback to the lower resolution C Standard Library clock API.

• test_muclock - clock busy loop overshoot, jitter and latency histogram

# test_meta
clock    : clock-std        freq     : 1000000
clock    : clock-posix      freq     : 1000000000
clock    : clock-x86        freq     : 2600000000
clk: clock-x86

# test_busy
wait_1ms :    2600034 cycles (elapsed 1000013.077 ns, overshoot   13.077 ns)

# test_jitter
dmin     :         48 cycles (  18.462 ns)
dmax     :        122 cycles (  46.923 ns)
drange   :         74 cycles (  28.462 ns)

# histogram
      48 (  18.462 ns)     82 ███████████▉
      52 (  20.000 ns)    345 ██████████████████████████████████████████████████
      54 (  20.769 ns)     41 █████▉
      56 (  21.538 ns)    125 ██████████████████
      58 (  22.308 ns)     82 ███████████▉
      60 (  23.077 ns)    143 ████████████████████▋
      62 (  23.846 ns)     41 █████▉
      64 (  24.615 ns)    163 ███████████████████████▌
      72 (  27.692 ns)      1 ▏
     122 (  46.923 ns)      1 ▏

interface to x86 cpuid on Windows, Linux, and macOS to query processor topology (cpuid leaf 0x1f or 0x0b) and report SMT thread placement.

• test_mucpu - reports x86 processor model name, apic_id, core_id and smt_id

Intel(R) Core(TM) i9-7980XE CPU @ 2.60GHz (apic_id:17 core_id:8 smt_id:1)