README.md

wfqueue.h

c/c++ FIFO wait-free queue, easy built cross platform(no extra dependencies needed)

Guarantee thread safety memory management, and it's all in one header only, as fast as never wait.

All Platform tests

GCC/CLANG/G++/CLANG++ |

VS x64/x86 |

support MPMC, MPSC and MCSP

Example

if c++

#include "wfqueue.h"


tWaitFree::Queue<MyVal> myqueue(sz);

// wrap in to thread
new ClassVal s(...);
tWaitFree::WfqEnqCtx<MyVal> enqCtx; // init 1 time in a thread only
tWaitFree::WfqDeqCtx<MyVal> deqCtx; // init 1 time in a thread only


// wrap in to thread
// please use enq to guarantee enqueue.
if(myqueue.tryEnq(s, enqCtx)) {
	printf("%s\n", "Enq Done");
} else {
	printf("%s\n", "queue is full, please try again to re-enqueue ");
}

// wrap in to thread
// please use deq to guarantee dequeue.
if(myqueue.tryDeq(s, deqCtx) {
  s->do_op();
}

if c

#include "wfqueue.h"

// Fixed size of queue
wfqueue_t *q = wfq_create(sz); 
wfq_enq_ctx_t enq_ctx = wfq_init_enq_ctx(); // init 1 time in a thread only
wfq_deq_ctx_t deq_ctx = wfq_init_deq_ctx(); // init 1 time in a thread only
// wrap in to thread
ClassVal *s = malloc(sizeof(ClassVal);

// wfq_enq_must for guarantee enqueue
wfq_enq_must(q, s, &enq_ctx);

// wrap in to thread
// wfq_deq_must for guarantee dequeue
s = (ClassVal*)wfq_deq(q, &deq_ctx); // return NULL if no val consuming

if(s) {
  s->do_op();

  free(s);
}

wfq_destroy(q);

Build

include header file in your project

Comparison with MoodyCamel/ConcurrentQueue

4 Concurrent Threads

Type	WFQUEUE (c++) ms	WFQUEUE (c)ms	MOODYCAMEL (ms)	INPUTS PER THREAD
MPMC	389.0	370	393.9	1,000,000 * 4
MPSC	494.50	600	684.0	1,000,000 * 4
MCSP	252.70	290.5	175.7	1,000,000 * 4

8 Concurrent Threads

Type	WFQUEUE (c++) ms	WFQUEUE (c)ms	MOODYCAMEL (ms)	INPUTS PER THREAD
MPMC	657.20	642.2	517.90	1,000,000 * 8
MPSC	1140.50	1420	1471.8	1,000,000 * 8
MCSP	247.80	245.2	253.2	1,000,000 * 8

8 Concurrent Threads with wfqueue(2 capacity only), create a small queue

Type	WFQUEUE (c++) ms	WFQUEUE (c)ms	MOODYCAMEL (ms)	INPUTS PER THREAD
MPMC	499	496	517.90	1,000,000 * 8
MPSC	1718	1700	1471.8	1,000,000 * 8
MCSP	403	360	253.2	1,000,000 * 8

Next feature target

No pre allocated size needed, it will expand itself (expandable size)

You may also like lock free queue FIFO

lfqueue