source: benchmark/readyQ/cycle.cpp @ e3282fe

#include "rq_bench.hpp"
#include <libfibre/fibre.h>

class __attribute__((aligned(128))) bench_sem {
        Fibre * volatile ptr = nullptr;
public:
        inline bool wait() {
                static Fibre * const ready  = reinterpret_cast<Fibre * const>(1ull);
                for(;;) {
                        Fibre * expected = this->ptr;
                        if(expected == ready) {
                                if(__atomic_compare_exchange_n(&this->ptr, &expected, nullptr, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) {
                                        return false;
                                }
                        }
                        else {
                                /* paranoid */ assert( expected == nullptr );
                                if(__atomic_compare_exchange_n(&this->ptr, &expected, fibre_self(), false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) {
                                        fibre_park();
                                        return true;
                                }
                        }

                }
        }

        inline bool post() {
                static Fibre * const ready  = reinterpret_cast<Fibre * const>(1ull);
                for(;;) {
                        Fibre * expected = this->ptr;
                        if(expected == ready) return false;
                        if(expected == nullptr) {
                                if(__atomic_compare_exchange_n(&this->ptr, &expected, ready, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) {
                                        return false;
                                }
                        }
                        else {
                                if(__atomic_compare_exchange_n(&this->ptr, &expected, nullptr, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) {
                                        fibre_unpark( expected );
                                        return true;
                                }
                        }
                }
        }
};
struct Partner {
        unsigned long long count  = 0;
        unsigned long long blocks = 0;
        bench_sem self;
        bench_sem * next;
};

void partner_main( Partner * self ) {
        self->count = 0;
        for(;;) {
                self->blocks += self->self.wait();
                self->next->post();
                self->count ++;
                if( clock_mode && stop) break;
                if(!clock_mode && self->count >= stop_count) break;
        }

        __atomic_fetch_add(&threads_left, -1, __ATOMIC_SEQ_CST);
}

int main(int argc, char * argv[]) {
        unsigned ring_size = 2;
        option_t opt[] = {
                BENCH_OPT,
                { 'r', "ringsize", "Number of threads in a cycle", ring_size }
        };
        BENCH_OPT_PARSE("cforall cycle benchmark");

        {
                unsigned long long global_counter = 0;
                unsigned long long global_blocks  = 0;
                unsigned tthreads = nthreads * ring_size;
                uint64_t start, end;
                FibreInit(1, nprocs);
                {
                        threads_left = tthreads;
                        Fibre * threads[tthreads];
                        Partner thddata[tthreads];
                        for(int i = 0; i < tthreads; i++) {
                                unsigned pi = (i + nthreads) % tthreads;
                                thddata[i].next = &thddata[pi].self;
                        }
                        for(int i = 0; i < tthreads; i++) {
                                threads[i] = new Fibre( reinterpret_cast<void (*)(void *)>(partner_main), &thddata[i] );
                        }
                        printf("Starting\n");

                        bool is_tty = isatty(STDOUT_FILENO);
                        start = getTimeNsec();

                        for(int i = 0; i < nthreads; i++) {
                                thddata[i].self.post();
                        }
                        wait<Fibre>(start, is_tty);

                        stop = true;
                        end = getTimeNsec();
                        printf("\nDone\n");

                        for(int i = 0; i < tthreads; i++) {
                                thddata[i].self.post();
                                fibre_join( threads[i], nullptr );
                                global_counter += thddata[i].count;
                                global_blocks  += thddata[i].blocks;
                        }
                }

                printf("Duration (ms)        : %'ld\n", to_miliseconds(end - start));
                printf("Number of processors : %'d\n", nprocs);
                printf("Number of threads    : %'d\n", tthreads);
                printf("Cycle size (# thrds) : %'d\n", ring_size);
                printf("Total Operations(ops): %'15llu\n", global_counter);
                printf("Total blocks         : %'15llu\n", global_blocks);
                printf("Ops per second       : %'18.2lf\n", ((double)global_counter) / to_fseconds(end - start));
                printf("ns per ops           : %'18.2lf\n", ((double)(end - start)) / global_counter);
                printf("Ops per threads      : %'15llu\n", global_counter / tthreads);
                printf("Ops per procs        : %'15llu\n", global_counter / nprocs);
                printf("Ops/sec/procs        : %'18.2lf\n", (((double)global_counter) / nprocs) / to_fseconds(end - start));
                printf("ns per ops/procs     : %'18.2lf\n", ((double)(end - start)) / (global_counter / nprocs));
                fflush(stdout);
        }

        return 0;
}