/* Skeleton for benchmark programs.
- Copyright (C) 2013 Free Software Foundation, Inc.
+ Copyright (C) 2013-2024 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
- <http://www.gnu.org/licenses/>. */
+ <https://www.gnu.org/licenses/>. */
#include <string.h>
#include <stdint.h>
+#include <stdbool.h>
#include <stdio.h>
#include <time.h>
#include <inttypes.h>
+#include "bench-timing.h"
+#include "json-lib.h"
+#include "bench-util.h"
+
+#include "bench-util.c"
#define TIMESPEC_AFTER(a, b) \
- (((a).tv_sec == (b).tv_sec) ? \
- ((a).tv_nsec > (b).tv_nsec) : \
- ((a).tv_sec > (b).tv_sec))
+ (((a).tv_sec == (b).tv_sec) \
+ ? ((a).tv_nsec > (b).tv_nsec) \
+ : ((a).tv_sec > (b).tv_sec))
int
main (int argc, char **argv)
{
unsigned long i, k;
- struct timespec start, end, runtime;
+ struct timespec runtime;
+ timing_t start, end;
+ bool detailed = false;
+ json_ctx_t json_ctx;
+
+ if (argc == 2 && !strcmp (argv[1], "-d"))
+ detailed = true;
+
+ bench_start ();
memset (&runtime, 0, sizeof (runtime));
- memset (&start, 0, sizeof (start));
- memset (&end, 0, sizeof (end));
- clock_getres (CLOCK_PROCESS_CPUTIME_ID, &start);
+ unsigned long iters = 2000;
- /* Measure 1000 times the resolution of the clock. So for a 1ns resolution
- clock, we measure 1000 iterations of the function call at a time.
- Measurements close to the minimum clock resolution won't make much sense,
- but it's better than having nothing at all. */
- unsigned long iters = 1000 * start.tv_nsec;
+#ifdef BENCH_INIT
+ BENCH_INIT ();
+#endif
+
+ json_init (&json_ctx, 2, stdout);
+
+ /* Begin function. */
+ json_attr_object_begin (&json_ctx, FUNCNAME);
for (int v = 0; v < NUM_VARIANTS; v++)
{
clock_gettime (CLOCK_MONOTONIC_RAW, &runtime);
runtime.tv_sec += DURATION;
+ bool is_bench = strncmp (VARIANT (v), "workload-", 9) == 0;
double d_total_i = 0;
- uint64_t total = 0, max = 0, min = 0x7fffffffffffffff;
+ timing_t total = 0, max = 0, min = 0x7fffffffffffffff;
+ timing_t throughput = 0, latency = 0;
+ int64_t c = 0;
+ uint64_t cur;
+ BENCH_VARS;
while (1)
{
- for (i = 0; i < NUM_SAMPLES (v); i++)
+ if (is_bench)
{
- clock_gettime (CLOCK_PROCESS_CPUTIME_ID, &start);
+ /* Benchmark a real trace of calls - all samples are iterated
+ over once before repeating. This models actual use more
+ accurately than repeating the same sample many times. */
+ TIMING_NOW (start);
for (k = 0; k < iters; k++)
- BENCH_FUNC (v, i);
- clock_gettime (CLOCK_PROCESS_CPUTIME_ID, &end);
-
- uint64_t cur = (end.tv_nsec - start.tv_nsec
- + ((end.tv_sec - start.tv_sec)
- * (uint64_t) 1000000000));
+ for (i = 0; i < NUM_SAMPLES (v); i++)
+ BENCH_FUNC (v, i);
+ TIMING_NOW (end);
+ TIMING_DIFF (cur, start, end);
+ TIMING_ACCUM (throughput, cur);
- if (cur > max)
- max = cur;
-
- if (cur < min)
- min = cur;
-
- total += cur;
+ TIMING_NOW (start);
+ for (k = 0; k < iters; k++)
+ for (i = 0; i < NUM_SAMPLES (v); i++)
+ BENCH_FUNC_LAT (v, i);
+ TIMING_NOW (end);
+ TIMING_DIFF (cur, start, end);
+ TIMING_ACCUM (latency, cur);
- d_total_i += iters;
+ d_total_i += iters * NUM_SAMPLES (v);
}
+ else
+ for (i = 0; i < NUM_SAMPLES (v); i++)
+ {
+ TIMING_NOW (start);
+ for (k = 0; k < iters; k++)
+ BENCH_FUNC (v, i);
+ TIMING_NOW (end);
+
+ TIMING_DIFF (cur, start, end);
+
+ if (cur > max)
+ max = cur;
+
+ if (cur < min)
+ min = cur;
+
+ TIMING_ACCUM (total, cur);
+ /* Accumulate timings for the value. In the end we will divide
+ by the total iterations. */
+ RESULT_ACCUM (cur, v, i, c * iters, (c + 1) * iters);
+
+ d_total_i += iters;
+ }
+ c++;
struct timespec curtime;
memset (&curtime, 0, sizeof (curtime));
double d_iters;
done:
- d_total_s = total * 1e-9;
+ d_total_s = total;
d_iters = iters;
- printf ("%s: ITERS:%g: TOTAL:%gs, MAX:%gns, MIN:%gns, %g iter/s\n",
- VARIANT (v),
- d_total_i, d_total_s, max / d_iters, min / d_iters,
- d_total_i / d_total_s);
+ /* Begin variant. */
+ json_attr_object_begin (&json_ctx, VARIANT (v));
+
+ if (is_bench)
+ {
+ json_attr_double (&json_ctx, "duration", throughput + latency);
+ json_attr_double (&json_ctx, "iterations", 2 * d_total_i);
+ json_attr_double (&json_ctx, "reciprocal-throughput",
+ throughput / d_total_i);
+ json_attr_double (&json_ctx, "latency", latency / d_total_i);
+ json_attr_double (&json_ctx, "max-throughput",
+ d_total_i / throughput * 1000000000.0);
+ json_attr_double (&json_ctx, "min-throughput",
+ d_total_i / latency * 1000000000.0);
+ }
+ else
+ {
+ json_attr_double (&json_ctx, "duration", d_total_s);
+ json_attr_double (&json_ctx, "iterations", d_total_i);
+ json_attr_double (&json_ctx, "max", max / d_iters);
+ json_attr_double (&json_ctx, "min", min / d_iters);
+ json_attr_double (&json_ctx, "mean", d_total_s / d_total_i);
+ }
+
+ if (detailed && !is_bench)
+ {
+ json_array_begin (&json_ctx, "timings");
+
+ for (int i = 0; i < NUM_SAMPLES (v); i++)
+ json_element_double (&json_ctx, RESULT (v, i));
+
+ json_array_end (&json_ctx);
+ }
+
+ /* End variant. */
+ json_attr_object_end (&json_ctx);
}
+ /* End function. */
+ json_attr_object_end (&json_ctx);
+
return 0;
}
projects / thirdparty / glibc.git / blobdiff