X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=ee495ffd88dc5dae82f92717deda18e59ef8bd57;hp=19b36961aa4854f0e3fb27a692b04ed5f52cb9e9;hb=fb390e242cac8868e01cdc2046c63e04e3d53d70;hpb=72c27ff8f2eaf945ae34776dab4da3fee753707d diff --git a/stat.c b/stat.c index 19b36961..ee495ffd 100644 --- a/stat.c +++ b/stat.c @@ -63,12 +63,12 @@ static unsigned int plat_val_to_idx(unsigned int val) /* * Discard the error bits and apply the mask to find the - * index for the buckets in the group + * index for the buckets in the group */ offset = (FIO_IO_U_PLAT_VAL - 1) & (val >> error_bits); /* Make sure the index does not exceed (array size - 1) */ - idx = (base + offset) < (FIO_IO_U_PLAT_NR - 1)? + idx = (base + offset) < (FIO_IO_U_PLAT_NR - 1) ? (base + offset) : (FIO_IO_U_PLAT_NR - 1); return idx; @@ -86,11 +86,11 @@ static unsigned int plat_idx_to_val(unsigned int idx) /* MSB <= (FIO_IO_U_PLAT_BITS-1), cannot be rounded off. Use * all bits of the sample as index */ - if (idx < (FIO_IO_U_PLAT_VAL << 1) ) + if (idx < (FIO_IO_U_PLAT_VAL << 1)) return idx; /* Find the group and compute the minimum value of that group */ - error_bits = (idx >> FIO_IO_U_PLAT_BITS) -1; + error_bits = (idx >> FIO_IO_U_PLAT_BITS) - 1; base = 1 << (error_bits + FIO_IO_U_PLAT_BITS); /* Find its bucket number of the group */ @@ -114,11 +114,9 @@ static int double_cmp(const void *a, const void *b) return cmp; } -static unsigned int calc_clat_percentiles(unsigned int *io_u_plat, - unsigned long nr, fio_fp64_t *plist, - unsigned int **output, - unsigned int *maxv, - unsigned int *minv) +unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, + fio_fp64_t *plist, unsigned int **output, + unsigned int *maxv, unsigned int *minv) { unsigned long sum = 0; unsigned int len, i, j = 0; @@ -142,7 +140,7 @@ static unsigned int calc_clat_percentiles(unsigned int *io_u_plat, * isn't a worry. Also note that this does not work for NaN values. */ if (len > 1) - qsort((void*)plist, len, sizeof(plist[0]), double_cmp); + qsort((void *)plist, len, sizeof(plist[0]), double_cmp); /* * Calculate bucket values, note down max and min values @@ -234,8 +232,8 @@ out: free(ovals); } -static int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max, - double *mean, double *dev) +int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max, + double *mean, double *dev) { double n = is->samples; @@ -287,11 +285,7 @@ void show_group_stats(struct group_run_stats *rs) } } -#define ts_total_io_u(ts) \ - ((ts)->total_io_u[0] + (ts)->total_io_u[1]) - -static void stat_calc_dist(unsigned int *map, unsigned long total, - double *io_u_dist) +void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist) { int i; @@ -329,28 +323,33 @@ static void stat_calc_lat(struct thread_stat *ts, double *dst, } } -static void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat) +void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat) { stat_calc_lat(ts, io_u_lat, ts->io_u_lat_u, FIO_IO_U_LAT_U_NR); } -static void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat) +void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat) { stat_calc_lat(ts, io_u_lat, ts->io_u_lat_m, FIO_IO_U_LAT_M_NR); } -static int usec_to_msec(unsigned long *min, unsigned long *max, double *mean, - double *dev) +static void display_lat(const char *name, unsigned long min, unsigned long max, + double mean, double dev) { - if (*min > 1000 && *max > 1000 && *mean > 1000.0 && *dev > 1000.0) { - *min /= 1000; - *max /= 1000; - *mean /= 1000.0; - *dev /= 1000.0; - return 0; - } + const char *base = "(usec)"; + char *minp, *maxp; - return 1; + if (!usec_to_msec(&min, &max, &mean, &dev)) + base = "(msec)"; + + minp = num2str(min, 6, 1, 0); + maxp = num2str(max, 6, 1, 0); + + log_info(" %s %s: min=%s, max=%s, avg=%5.02f," + " stdev=%5.02f\n", name, base, minp, maxp, mean, dev); + + free(minp); + free(maxp); } static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, @@ -386,66 +385,27 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, free(bw_p); free(iops_p); - if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" slat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); - - free(minp); - free(maxp); - } - if (calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" clat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); - - free(minp); - free(maxp); - } - if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" lat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); + if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) + display_lat("slat", min, max, mean, dev); + if (calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) + display_lat("clat", min, max, mean, dev); + if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) + display_lat(" lat", min, max, mean, dev); - free(minp); - free(maxp); - } if (ts->clat_percentiles) { show_clat_percentiles(ts->io_u_plat[ddir], ts->clat_stat[ddir].samples, ts->percentile_list); } if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { - double p_of_agg; + double p_of_agg = 100.0; const char *bw_str = "KB"; - p_of_agg = mean * 100 / (double) rs->agg[ddir]; - if (p_of_agg > 100.0) - p_of_agg = 100.0; + if (rs->agg[ddir]) { + p_of_agg = mean * 100 / (double) rs->agg[ddir]; + if (p_of_agg > 100.0) + p_of_agg = 100.0; + } if (mean > 999999.9) { min /= 1000.0; @@ -508,8 +468,14 @@ static void show_lat_m(double *io_u_lat_m) show_lat(io_u_lat_m, FIO_IO_U_LAT_M_NR, ranges, "msec"); } -static void show_latencies(double *io_u_lat_u, double *io_u_lat_m) +static void show_latencies(struct thread_stat *ts) { + double io_u_lat_u[FIO_IO_U_LAT_U_NR]; + double io_u_lat_m[FIO_IO_U_LAT_M_NR]; + + stat_calc_lat_u(ts, io_u_lat_u); + stat_calc_lat_m(ts, io_u_lat_m); + show_lat_u(io_u_lat_u); show_lat_m(io_u_lat_m); } @@ -519,8 +485,6 @@ void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) double usr_cpu, sys_cpu; unsigned long runtime; double io_u_dist[FIO_IO_U_MAP_NR]; - double io_u_lat_u[FIO_IO_U_LAT_U_NR]; - double io_u_lat_m[FIO_IO_U_LAT_M_NR]; if (!(ts->io_bytes[0] + ts->io_bytes[1]) && !(ts->total_io_u[0] + ts->total_io_u[1])) @@ -544,9 +508,7 @@ void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) if (ts->io_bytes[DDIR_WRITE]) show_ddir_status(rs, ts, DDIR_WRITE); - stat_calc_lat_u(ts, io_u_lat_u); - stat_calc_lat_m(ts, io_u_lat_m); - show_latencies(io_u_lat_u, io_u_lat_m); + show_latencies(ts); runtime = ts->total_run_time; if (runtime) { @@ -611,7 +573,7 @@ static void show_ddir_status_terse(struct thread_stat *ts, if (ts->runtime[ddir]) { uint64_t runt = ts->runtime[ddir]; - bw = ts->io_bytes[ddir] / runt; + bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024; iops = (1000 * (uint64_t) ts->total_io_u[ddir]) / runt; } @@ -628,11 +590,6 @@ static void show_ddir_status_terse(struct thread_stat *ts, else log_info(";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0); - if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) - log_info(";%lu;%lu;%f;%f", min, max, mean, dev); - else - log_info(";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0); - if (ts->clat_percentiles) { len = calc_clat_percentiles(ts->io_u_plat[ddir], ts->clat_stat[ddir].samples, @@ -648,22 +605,91 @@ static void show_ddir_status_terse(struct thread_stat *ts, } log_info(";%2.2f%%=%u", ts->percentile_list[i].u.f, ovals[i]); } + + if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) + log_info(";%lu;%lu;%f;%f", min, max, mean, dev); + else + log_info(";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0); + if (ovals) free(ovals); if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { - double p_of_agg; + double p_of_agg = 100.0; + + if (rs->agg[ddir]) { + p_of_agg = mean * 100 / (double) rs->agg[ddir]; + if (p_of_agg > 100.0) + p_of_agg = 100.0; + } - p_of_agg = mean * 100 / (double) rs->agg[ddir]; log_info(";%lu;%lu;%f%%;%f;%f", min, max, p_of_agg, mean, dev); } else log_info(";%lu;%lu;%f%%;%f;%f", 0UL, 0UL, 0.0, 0.0, 0.0); } +static void show_thread_status_terse_v2(struct thread_stat *ts, + struct group_run_stats *rs) +{ + double io_u_dist[FIO_IO_U_MAP_NR]; + double io_u_lat_u[FIO_IO_U_LAT_U_NR]; + double io_u_lat_m[FIO_IO_U_LAT_M_NR]; + double usr_cpu, sys_cpu; + int i; + + /* General Info */ + log_info("2;%s;%d;%d", ts->name, ts->groupid, ts->error); + /* Log Read Status */ + show_ddir_status_terse(ts, rs, 0); + /* Log Write Status */ + show_ddir_status_terse(ts, rs, 1); + + /* CPU Usage */ + if (ts->total_run_time) { + double runt = (double) ts->total_run_time; + + usr_cpu = (double) ts->usr_time * 100 / runt; + sys_cpu = (double) ts->sys_time * 100 / runt; + } else { + usr_cpu = 0; + sys_cpu = 0; + } + + log_info(";%f%%;%f%%;%lu;%lu;%lu", usr_cpu, sys_cpu, ts->ctx, ts->majf, + ts->minf); + + /* Calc % distribution of IO depths, usecond, msecond latency */ + stat_calc_dist(ts->io_u_map, ts_total_io_u(ts), io_u_dist); + stat_calc_lat_u(ts, io_u_lat_u); + stat_calc_lat_m(ts, io_u_lat_m); + + /* Only show fixed 7 I/O depth levels*/ + log_info(";%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%", + io_u_dist[0], io_u_dist[1], io_u_dist[2], io_u_dist[3], + io_u_dist[4], io_u_dist[5], io_u_dist[6]); + + /* Microsecond latency */ + for (i = 0; i < FIO_IO_U_LAT_U_NR; i++) + log_info(";%3.2f%%", io_u_lat_u[i]); + /* Millisecond latency */ + for (i = 0; i < FIO_IO_U_LAT_M_NR; i++) + log_info(";%3.2f%%", io_u_lat_m[i]); + /* Additional output if continue_on_error set - default off*/ + if (ts->continue_on_error) + log_info(";%lu;%d", ts->total_err_count, ts->first_error); + log_info("\n"); + + /* Additional output if description is set */ + if (ts->description) + log_info(";%s", ts->description); + + log_info("\n"); +} + #define FIO_TERSE_VERSION "3" -static void show_thread_status_terse(struct thread_stat *ts, - struct group_run_stats *rs) +static void show_thread_status_terse_v3(struct thread_stat *ts, + struct group_run_stats *rs) { double io_u_dist[FIO_IO_U_MAP_NR]; double io_u_lat_u[FIO_IO_U_LAT_U_NR]; @@ -716,11 +742,23 @@ static void show_thread_status_terse(struct thread_stat *ts, /* Additional output if continue_on_error set - default off*/ if (ts->continue_on_error) log_info(";%lu;%d", ts->total_err_count, ts->first_error); - log_info("\n"); /* Additional output if description is set */ if (strlen(ts->description)) log_info(";%s", ts->description); + + log_info("\n"); +} + +static void show_thread_status_terse(struct thread_stat *ts, + struct group_run_stats *rs) +{ + if (terse_version == 2) + show_thread_status_terse_v2(ts, rs); + else if (terse_version == 3) + show_thread_status_terse_v3(ts, rs); + else + log_err("fio: bad terse version!? %d\n", terse_version); } static void sum_stat(struct io_stat *dst, struct io_stat *src, int nr) @@ -919,6 +957,11 @@ void show_run_stats(void) else memset(ts->description, 0, FIO_JOBNAME_SIZE); + /* + * If multiple entries in this group, this is + * the first member. + */ + ts->thread_number = td->thread_number; ts->groupid = td->groupid; /* @@ -967,10 +1010,11 @@ void show_run_stats(void) bw = 0; if (ts->runtime[j]) { - unsigned long runt; + unsigned long runt = ts->runtime[j]; + unsigned long long kb; - runt = ts->runtime[j]; - bw = ts->io_bytes[j] / runt; + kb = ts->io_bytes[j] / rs->kb_base; + bw = kb * 1000 / runt; } if (bw < rs->min_bw[j]) rs->min_bw[j] = bw; @@ -982,16 +1026,12 @@ void show_run_stats(void) } for (i = 0; i < groupid + 1; i++) { - unsigned long max_run[2]; - rs = &runstats[i]; - max_run[0] = rs->max_run[0]; - max_run[1] = rs->max_run[1]; if (rs->max_run[0]) - rs->agg[0] = (rs->io_kb[0] * 1000) / max_run[0]; + rs->agg[0] = (rs->io_kb[0] * 1000) / rs->max_run[0]; if (rs->max_run[1]) - rs->agg[1] = (rs->io_kb[1] * 1000) / max_run[1]; + rs->agg[1] = (rs->io_kb[1] * 1000) / rs->max_run[1]; } /* @@ -1027,12 +1067,59 @@ void show_run_stats(void) else if (!terse_output) show_disk_util(0); - free_disk_util(); - free(runstats); free(threadstats); } +static void *__show_running_run_stats(void *arg) +{ + struct thread_data *td; + unsigned long long *rt; + struct timeval tv; + int i; + + rt = malloc(thread_number * sizeof(unsigned long long)); + fio_gettime(&tv, NULL); + + for_each_td(td, i) { + rt[i] = mtime_since(&td->start, &tv); + if (td_read(td) && td->io_bytes[DDIR_READ]) + td->ts.runtime[DDIR_READ] += rt[i]; + if (td_write(td) && td->io_bytes[DDIR_WRITE]) + td->ts.runtime[DDIR_WRITE] += rt[i]; + + update_rusage_stat(td); + td->ts.io_bytes[0] = td->io_bytes[0]; + td->ts.io_bytes[1] = td->io_bytes[1]; + td->ts.total_run_time = mtime_since(&td->epoch, &tv); + } + + show_run_stats(); + + for_each_td(td, i) { + if (td_read(td) && td->io_bytes[DDIR_READ]) + td->ts.runtime[DDIR_READ] -= rt[i]; + if (td_write(td) && td->io_bytes[DDIR_WRITE]) + td->ts.runtime[DDIR_WRITE] -= rt[i]; + } + + free(rt); + return NULL; +} + +/* + * Called from signal handler. It _should_ be safe to just run this inline + * in the sig handler, but we should be disturbing the system less by just + * creating a thread to do it. + */ +void show_running_run_stats(void) +{ + pthread_t thread; + + pthread_create(&thread, NULL, __show_running_run_stats, NULL); + pthread_detach(thread); +} + static inline void add_stat_sample(struct io_stat *is, unsigned long data) { double val = data; @@ -1058,6 +1145,9 @@ static void __add_log_sample(struct io_log *iolog, unsigned long val, { const int nr_samples = iolog->nr_samples; + if (!iolog->nr_samples) + iolog->avg_last = t; + if (iolog->nr_samples == iolog->max_samples) { int new_size = sizeof(struct io_sample) * iolog->max_samples*2; @@ -1072,14 +1162,66 @@ static void __add_log_sample(struct io_log *iolog, unsigned long val, iolog->nr_samples++; } +static inline void reset_io_stat(struct io_stat *ios) +{ + ios->max_val = ios->min_val = ios->samples = 0; + ios->mean.u.f = ios->S.u.f = 0; +} + static void add_log_sample(struct thread_data *td, struct io_log *iolog, unsigned long val, enum fio_ddir ddir, unsigned int bs) { + unsigned long elapsed, this_window; + if (!ddir_rw(ddir)) return; - __add_log_sample(iolog, val, ddir, bs, mtime_since_now(&td->epoch)); + elapsed = mtime_since_now(&td->epoch); + + /* + * If no time averaging, just add the log sample. + */ + if (!iolog->avg_msec) { + __add_log_sample(iolog, val, ddir, bs, elapsed); + return; + } + + /* + * Add the sample. If the time period has passed, then + * add that entry to the log and clear. + */ + add_stat_sample(&iolog->avg_window[ddir], val); + + /* + * If period hasn't passed, adding the above sample is all we + * need to do. + */ + this_window = elapsed - iolog->avg_last; + if (this_window < iolog->avg_msec) + return; + + /* + * Note an entry in the log. Use the mean from the logged samples, + * making sure to properly round up. Only write a log entry if we + * had actual samples done. + */ + if (iolog->avg_window[DDIR_READ].samples) { + unsigned long mr; + + mr = iolog->avg_window[DDIR_READ].mean.u.f + 0.50; + __add_log_sample(iolog, mr, DDIR_READ, 0, elapsed); + } + if (iolog->avg_window[DDIR_WRITE].samples) { + unsigned long mw; + + mw = iolog->avg_window[DDIR_WRITE].mean.u.f + 0.50; + __add_log_sample(iolog, mw, DDIR_WRITE, 0, elapsed); + } + + reset_io_stat(&iolog->avg_window[DDIR_READ]); + reset_io_stat(&iolog->avg_window[DDIR_WRITE]); + iolog->avg_last = elapsed; } void add_agg_sample(unsigned long val, enum fio_ddir ddir, unsigned int bs) @@ -1160,15 +1302,26 @@ void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs, if (spent < td->o.bw_avg_time) return; - rate = (td->this_io_bytes[ddir] - td->stat_io_bytes[ddir]) * - 1000 / spent / 1024; - add_stat_sample(&ts->bw_stat[ddir], rate); + /* + * Compute both read and write rates for the interval. + */ + for (ddir = DDIR_READ; ddir <= DDIR_WRITE; ddir++) { + uint64_t delta; + + delta = td->this_io_bytes[ddir] - td->stat_io_bytes[ddir]; + if (!delta) + continue; /* No entries for interval */ + + rate = delta * 1000 / spent / 1024; + add_stat_sample(&ts->bw_stat[ddir], rate); + + if (td->bw_log) + add_log_sample(td, td->bw_log, rate, ddir, bs); - if (td->bw_log) - add_log_sample(td, td->bw_log, rate, ddir, bs); + td->stat_io_bytes[ddir] = td->this_io_bytes[ddir]; + } fio_gettime(&td->bw_sample_time, NULL); - td->stat_io_bytes[ddir] = td->this_io_bytes[ddir]; } void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, @@ -1184,15 +1337,24 @@ void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, if (spent < td->o.iops_avg_time) return; - iops = ((td->this_io_blocks[ddir] - td->stat_io_blocks[ddir]) * 1000) / spent; + /* + * Compute both read and write rates for the interval. + */ + for (ddir = DDIR_READ; ddir <= DDIR_WRITE; ddir++) { + uint64_t delta; + + delta = td->this_io_blocks[ddir] - td->stat_io_blocks[ddir]; + if (!delta) + continue; /* No entries for interval */ + + iops = (delta * 1000) / spent; + add_stat_sample(&ts->iops_stat[ddir], iops); - add_stat_sample(&ts->iops_stat[ddir], iops); + if (td->iops_log) + add_log_sample(td, td->iops_log, iops, ddir, 0); - if (td->iops_log) { - assert(iops); - add_log_sample(td, td->iops_log, iops, ddir, 0); + td->stat_io_blocks[ddir] = td->this_io_blocks[ddir]; } fio_gettime(&td->iops_sample_time, NULL); - td->stat_io_blocks[ddir] = td->this_io_blocks[ddir]; }