^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) // SPDX-License-Identifier: GPL-2.0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) #ifdef HAVE_EVENTFD_SUPPORT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) * Copyright (C) 2018 Davidlohr Bueso.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * This program benchmarks concurrent epoll_wait(2) monitoring multiple
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * file descriptors under one or two load balancing models. The first,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * and default, is the single/combined queueing (which refers to a single
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) * epoll instance for N worker threads):
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) * |---> [worker A]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * |---> [worker B]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) * [combined queue] .---> [worker C]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * |---> [worker D]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) * |---> [worker E]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) * While the second model, enabled via --multiq option, uses multiple
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) * queueing (which refers to one epoll instance per worker). For example,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) * short lived tcp connections in a high throughput httpd server will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) * ditribute the accept()'ing connections across CPUs. In this case each
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) * worker does a limited amount of processing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) * [queue A] ---> [worker]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) * [queue B] ---> [worker]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) * [queue C] ---> [worker]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) * [queue D] ---> [worker]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) * [queue E] ---> [worker]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) * Naturally, the single queue will enforce more concurrency on the epoll
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) * instance, and can therefore scale poorly compared to multiple queues.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) * However, this is a benchmark raw data and must be taken with a grain of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) * salt when choosing how to make use of sys_epoll.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) * Each thread has a number of private, nonblocking file descriptors,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) * referred to as fdmap. A writer thread will constantly be writing to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) * the fdmaps of all threads, minimizing each threads's chances of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) * epoll_wait not finding any ready read events and blocking as this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) * is not what we want to stress. The size of the fdmap can be adjusted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) * by the user; enlarging the value will increase the chances of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) * epoll_wait(2) blocking as the lineal writer thread will take "longer",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) * at least at a high level.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) * Note that because fds are private to each thread, this workload does
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) * not stress scenarios where multiple tasks are awoken per ready IO; ie:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) * EPOLLEXCLUSIVE semantics.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) * The end result/metric is throughput: number of ops/second where an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) * operation consists of:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) * epoll_wait(2) + [others]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) * ... where [others] is the cost of re-adding the fd (EPOLLET),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) * or rearming it (EPOLLONESHOT).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) * The purpose of this is program is that it be useful for measuring
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) * kernel related changes to the sys_epoll, and not comparing different
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) * IO polling methods, for example. Hence everything is very adhoc and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) * outputs raw microbenchmark numbers. Also this uses eventfd, similar
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) * tools tend to use pipes or sockets, but the result is the same.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) /* For the CLR_() macros */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) #include <string.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) #include <pthread.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) #include <unistd.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) #include <errno.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) #include <inttypes.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) #include <signal.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) #include <stdlib.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) #include <linux/compiler.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) #include <linux/kernel.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) #include <sys/time.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) #include <sys/resource.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) #include <sys/epoll.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) #include <sys/eventfd.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) #include <sys/types.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) #include <internal/cpumap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) #include <perf/cpumap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) #include "../util/stat.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) #include <subcmd/parse-options.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) #include "bench.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) #include <err.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) #define printinfo(fmt, arg...) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) do { if (__verbose) { printf(fmt, ## arg); fflush(stdout); } } while (0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) static unsigned int nthreads = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) static unsigned int nsecs = 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) static bool wdone, done, __verbose, randomize, nonblocking;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) * epoll related shared variables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) /* Maximum number of nesting allowed inside epoll sets */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) #define EPOLL_MAXNESTS 4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) static int epollfd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) static int *epollfdp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) static bool noaffinity;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) static unsigned int nested = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) static bool et; /* edge-trigger */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) static bool oneshot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) static bool multiq; /* use an epoll instance per thread */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) /* amount of fds to monitor, per thread */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) static unsigned int nfds = 64;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) static pthread_mutex_t thread_lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) static unsigned int threads_starting;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) static struct stats throughput_stats;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) static pthread_cond_t thread_parent, thread_worker;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) struct worker {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) int tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) int epollfd; /* for --multiq */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) pthread_t thread;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) unsigned long ops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) int *fdmap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) static const struct option options[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) /* general benchmark options */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) OPT_UINTEGER('t', "threads", &nthreads, "Specify amount of threads"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) OPT_UINTEGER('r', "runtime", &nsecs, "Specify runtime (in seconds)"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) OPT_UINTEGER('f', "nfds", &nfds, "Specify amount of file descriptors to monitor for each thread"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) OPT_BOOLEAN( 'n', "noaffinity", &noaffinity, "Disables CPU affinity"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) OPT_BOOLEAN('R', "randomize", &randomize, "Enable random write behaviour (default is lineal)"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) OPT_BOOLEAN( 'v', "verbose", &__verbose, "Verbose mode"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) /* epoll specific options */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) OPT_BOOLEAN( 'm', "multiq", &multiq, "Use multiple epoll instances (one per thread)"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) OPT_BOOLEAN( 'B', "nonblocking", &nonblocking, "Nonblocking epoll_wait(2) behaviour"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) OPT_UINTEGER( 'N', "nested", &nested, "Nesting level epoll hierarchy (default is 0, no nesting)"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) OPT_BOOLEAN( 'S', "oneshot", &oneshot, "Use EPOLLONESHOT semantics"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) OPT_BOOLEAN( 'E', "edge", &et, "Use Edge-triggered interface (default is LT)"),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) OPT_END()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) static const char * const bench_epoll_wait_usage[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) "perf bench epoll wait <options>",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) NULL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) * Arrange the N elements of ARRAY in random order.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) * Only effective if N is much smaller than RAND_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) * if this may not be the case, use a better random
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) * number generator. -- Ben Pfaff.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) static void shuffle(void *array, size_t n, size_t size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) char *carray = array;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) void *aux;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) size_t i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) if (n <= 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) aux = calloc(1, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) if (!aux)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) err(EXIT_FAILURE, "calloc");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) for (i = 1; i < n; ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) size_t j = i + rand() / (RAND_MAX / (n - i) + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) j *= size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) memcpy(aux, &carray[j], size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) memcpy(&carray[j], &carray[i*size], size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) memcpy(&carray[i*size], aux, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) free(aux);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) static void *workerfn(void *arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) int fd, ret, r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) struct worker *w = (struct worker *) arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) unsigned long ops = w->ops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) struct epoll_event ev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) uint64_t val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) int to = nonblocking? 0 : -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) int efd = multiq ? w->epollfd : epollfd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) pthread_mutex_lock(&thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) threads_starting--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) if (!threads_starting)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) pthread_cond_signal(&thread_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) pthread_cond_wait(&thread_worker, &thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) pthread_mutex_unlock(&thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) * Block undefinitely waiting for the IN event.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) * In order to stress the epoll_wait(2) syscall,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) * call it event per event, instead of a larger
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) * batch (max)limit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) ret = epoll_wait(efd, &ev, 1, to);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) } while (ret < 0 && errno == EINTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) err(EXIT_FAILURE, "epoll_wait");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) fd = ev.data.fd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216) r = read(fd, &val, sizeof(val));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) } while (!done && (r < 0 && errno == EAGAIN));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) if (et) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220) ev.events = EPOLLIN | EPOLLET;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) ret = epoll_ctl(efd, EPOLL_CTL_ADD, fd, &ev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) if (oneshot) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) /* rearm the file descriptor with a new event mask */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) ev.events |= EPOLLIN | EPOLLONESHOT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) ret = epoll_ctl(efd, EPOLL_CTL_MOD, fd, &ev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) ops++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) } while (!done);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) if (multiq)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234) close(w->epollfd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) w->ops = ops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) static void nest_epollfd(struct worker *w)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) unsigned int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) struct epoll_event ev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) int efd = multiq ? w->epollfd : epollfd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) if (nested > EPOLL_MAXNESTS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) nested = EPOLL_MAXNESTS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) epollfdp = calloc(nested, sizeof(*epollfdp));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) if (!epollfdp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) err(EXIT_FAILURE, "calloc");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253) for (i = 0; i < nested; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) epollfdp[i] = epoll_create(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) if (epollfdp[i] < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) err(EXIT_FAILURE, "epoll_create");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) ev.events = EPOLLHUP; /* anything */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) ev.data.u64 = i; /* any number */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) for (i = nested - 1; i; i--) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) if (epoll_ctl(epollfdp[i - 1], EPOLL_CTL_ADD,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264) epollfdp[i], &ev) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265) err(EXIT_FAILURE, "epoll_ctl");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) if (epoll_ctl(efd, EPOLL_CTL_ADD, *epollfdp, &ev) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) err(EXIT_FAILURE, "epoll_ctl");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) static void toggle_done(int sig __maybe_unused,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) siginfo_t *info __maybe_unused,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274) void *uc __maybe_unused)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) /* inform all threads that we're done for the day */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277) done = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) gettimeofday(&bench__end, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) timersub(&bench__end, &bench__start, &bench__runtime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282) static void print_summary(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) unsigned long avg = avg_stats(&throughput_stats);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) double stddev = stddev_stats(&throughput_stats);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287) printf("\nAveraged %ld operations/sec (+- %.2f%%), total secs = %d\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288) avg, rel_stddev_stats(stddev, avg),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) (int)bench__runtime.tv_sec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) static int do_threads(struct worker *worker, struct perf_cpu_map *cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) pthread_attr_t thread_attr, *attrp = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295) cpu_set_t cpuset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296) unsigned int i, j;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297) int ret = 0, events = EPOLLIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299) if (oneshot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) events |= EPOLLONESHOT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301) if (et)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) events |= EPOLLET;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304) printinfo("starting worker/consumer %sthreads%s\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305) noaffinity ? "":"CPU affinity ",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) nonblocking ? " (nonblocking)":"");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) if (!noaffinity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) pthread_attr_init(&thread_attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310) for (i = 0; i < nthreads; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) struct worker *w = &worker[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) if (multiq) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) w->epollfd = epoll_create(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315) if (w->epollfd < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) err(EXIT_FAILURE, "epoll_create");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) if (nested)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319) nest_epollfd(w);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) w->tid = i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) w->fdmap = calloc(nfds, sizeof(int));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324) if (!w->fdmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327) for (j = 0; j < nfds; j++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) int efd = multiq ? w->epollfd : epollfd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329) struct epoll_event ev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) w->fdmap[j] = eventfd(0, EFD_NONBLOCK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) if (w->fdmap[j] < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) err(EXIT_FAILURE, "eventfd");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) ev.data.fd = w->fdmap[j];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) ev.events = events;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338) ret = epoll_ctl(efd, EPOLL_CTL_ADD,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) w->fdmap[j], &ev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341) err(EXIT_FAILURE, "epoll_ctl");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344) if (!noaffinity) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345) CPU_ZERO(&cpuset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) CPU_SET(cpu->map[i % cpu->nr], &cpuset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) ret = pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352) attrp = &thread_attr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) ret = pthread_create(&w->thread, attrp, workerfn,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356) (void *)(struct worker *) w);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) err(EXIT_FAILURE, "pthread_create");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361) if (!noaffinity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) pthread_attr_destroy(&thread_attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) static void *writerfn(void *p)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369) struct worker *worker = p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) size_t i, j, iter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) const uint64_t val = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372) ssize_t sz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) struct timespec ts = { .tv_sec = 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) .tv_nsec = 500 };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376) printinfo("starting writer-thread: doing %s writes ...\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) randomize? "random":"lineal");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) for (iter = 0; !wdone; iter++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) if (randomize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) shuffle((void *)worker, nthreads, sizeof(*worker));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384) for (i = 0; i < nthreads; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) struct worker *w = &worker[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) if (randomize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) shuffle((void *)w->fdmap, nfds, sizeof(int));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391) for (j = 0; j < nfds; j++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393) sz = write(w->fdmap[j], &val, sizeof(val));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394) } while (!wdone && (sz < 0 && errno == EAGAIN));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) nanosleep(&ts, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401) printinfo("exiting writer-thread (total full-loops: %zd)\n", iter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405) static int cmpworker(const void *p1, const void *p2)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) struct worker *w1 = (struct worker *) p1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409) struct worker *w2 = (struct worker *) p2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) return w1->tid > w2->tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) int bench_epoll_wait(int argc, const char **argv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416) struct sigaction act;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) unsigned int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418) struct worker *worker = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419) struct perf_cpu_map *cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) pthread_t wthread;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421) struct rlimit rl, prevrl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423) argc = parse_options(argc, argv, options, bench_epoll_wait_usage, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) if (argc) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425) usage_with_options(bench_epoll_wait_usage, options);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426) exit(EXIT_FAILURE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429) memset(&act, 0, sizeof(act));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430) sigfillset(&act.sa_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) act.sa_sigaction = toggle_done;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) sigaction(SIGINT, &act, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) cpu = perf_cpu_map__new(NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435) if (!cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436) goto errmem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) /* a single, main epoll instance */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) if (!multiq) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440) epollfd = epoll_create(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441) if (epollfd < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) err(EXIT_FAILURE, "epoll_create");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445) * Deal with nested epolls, if any.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447) if (nested)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) nest_epollfd(NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) printinfo("Using %s queue model\n", multiq ? "multi" : "single");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) printinfo("Nesting level(s): %d\n", nested);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) /* default to the number of CPUs and leave one for the writer pthread */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) if (!nthreads)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 456) nthreads = cpu->nr - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 458) worker = calloc(nthreads, sizeof(*worker));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) if (!worker) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) goto errmem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) if (getrlimit(RLIMIT_NOFILE, &prevrl))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) err(EXIT_FAILURE, "getrlimit");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465) rl.rlim_cur = rl.rlim_max = nfds * nthreads * 2 + 50;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466) printinfo("Setting RLIMIT_NOFILE rlimit from %" PRIu64 " to: %" PRIu64 "\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) (uint64_t)prevrl.rlim_max, (uint64_t)rl.rlim_max);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) if (setrlimit(RLIMIT_NOFILE, &rl) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469) err(EXIT_FAILURE, "setrlimit");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471) printf("Run summary [PID %d]: %d threads monitoring%s on "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) "%d file-descriptors for %d secs.\n\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) getpid(), nthreads, oneshot ? " (EPOLLONESHOT semantics)": "", nfds, nsecs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475) init_stats(&throughput_stats);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) pthread_mutex_init(&thread_lock, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477) pthread_cond_init(&thread_parent, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478) pthread_cond_init(&thread_worker, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480) threads_starting = nthreads;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482) gettimeofday(&bench__start, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) do_threads(worker, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) pthread_mutex_lock(&thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487) while (threads_starting)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488) pthread_cond_wait(&thread_parent, &thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) pthread_cond_broadcast(&thread_worker);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) pthread_mutex_unlock(&thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493) * At this point the workers should be blocked waiting for read events
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) * to become ready. Launch the writer which will constantly be writing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) * to each thread's fdmap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) ret = pthread_create(&wthread, NULL, writerfn,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498) (void *)(struct worker *) worker);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500) err(EXIT_FAILURE, "pthread_create");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) sleep(nsecs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503) toggle_done(0, NULL, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) printinfo("main thread: toggling done\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) sleep(1); /* meh */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) wdone = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) ret = pthread_join(wthread, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510) err(EXIT_FAILURE, "pthread_join");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512) /* cleanup & report results */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513) pthread_cond_destroy(&thread_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514) pthread_cond_destroy(&thread_worker);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) pthread_mutex_destroy(&thread_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517) /* sort the array back before reporting */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518) if (randomize)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) qsort(worker, nthreads, sizeof(struct worker), cmpworker);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) for (i = 0; i < nthreads; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) unsigned long t = bench__runtime.tv_sec > 0 ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523) worker[i].ops / bench__runtime.tv_sec : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) update_stats(&throughput_stats, t);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 527) if (nfds == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 528) printf("[thread %2d] fdmap: %p [ %04ld ops/sec ]\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529) worker[i].tid, &worker[i].fdmap[0], t);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531) printf("[thread %2d] fdmap: %p ... %p [ %04ld ops/sec ]\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532) worker[i].tid, &worker[i].fdmap[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) &worker[i].fdmap[nfds-1], t);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) print_summary();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) close(epollfd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) errmem:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541) err(EXIT_FAILURE, "calloc");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) #endif // HAVE_EVENTFD_SUPPORT
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags