bench: replace wall-clock timer with per-process CPU timer

README.md

@@ -159,6 +159,8 @@ Benchmark
 ------------
 If configured with `--enable-benchmark` (which is the default), binaries for benchmarking the libsecp256k1 functions will be present in the root directory after the build.
 
+For reliable benchmarks, it is strongly recommended to pin the process to a dedicated CPU core and to disable CPU frequency scaling.
+
 To print the benchmark result to the command line:
 
     $ ./bench_name

src/bench.c

@@ -4,6 +4,8 @@
  * file COPYING or https://www.opensource.org/licenses/mit-license.php.*
  ***********************************************************************/
 
+#define _POSIX_C_SOURCE 200112L /* for clock_gettime() */
+
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -252,6 +254,7 @@ int main(int argc, char** argv) {
     data.pubkeylen = 33;
     CHECK(secp256k1_ec_pubkey_serialize(data.ctx, data.pubkey, &data.pubkeylen, &pubkey, SECP256K1_EC_COMPRESSED) == 1);
 
+    print_clock_info();
     print_output_table_header_row();
     if (d || have_flag(argc, argv, "ecdsa") || have_flag(argc, argv, "verify") || have_flag(argc, argv, "ecdsa_verify")) run_benchmark("ecdsa_verify", bench_verify, NULL, NULL, &data, 10, iters);
 

src/bench.h

@@ -7,30 +7,44 @@
 #ifndef SECP256K1_BENCH_H
 #define SECP256K1_BENCH_H
 
+#if defined(_WIN32)
+# include <windows.h>
+#else /* POSIX */
+# include <time.h>
+#endif
+
 #include <stdlib.h>
 #include <stdint.h>
 #include <stdio.h>
 #include <string.h>
 
-#if (defined(_MSC_VER) && _MSC_VER >= 1900)
-#  include <time.h>
-#else
-#  include <sys/time.h>
-#endif
+static int64_t gettime_us(void) {
+#if defined(_WIN32)
+
+    LARGE_INTEGER freq, counter;
+    QueryPerformanceFrequency(&freq);
+    QueryPerformanceCounter(&counter);
+    return (int64_t)(counter.QuadPart * 1000000 / freq.QuadPart);
+
+#else /* POSIX */
+
+# if defined(CLOCK_PROCESS_CPUTIME_ID)
+    /* In theory, CLOCK_PROCESS_CPUTIME_ID is only useful if the process is locked to a core,
+     * see `man clock_gettime` on Linux. In practice, modern CPUs have synchronized TSCs which
+     * address this issue, see https://docs.amd.com/r/en-US/ug1586-onload-user/Timer-TSC-Stability . */
+    const clockid_t clock_type = CLOCK_PROCESS_CPUTIME_ID;
+# elif defined(CLOCK_MONOTONIC)
+    /* fallback to wall-clock timer */
+    const clockid_t clock_type = CLOCK_MONOTONIC;
+# else
+    /* fallback to less precise wall-clock timer */
+    const clockid_t clock_type = CLOCK_REALTIME;
+# endif
+
+    struct timespec ts;
+    clock_gettime(clock_type, &ts);
+    return (int64_t)ts.tv_sec * 1000000 + ts.tv_nsec / 1000;
 
-static int64_t gettime_i64(void) {
-#if (defined(_MSC_VER) && _MSC_VER >= 1900)
-    /* C11 way to get wallclock time */
-    struct timespec tv;
-    if (!timespec_get(&tv, TIME_UTC)) {
-        fputs("timespec_get failed!", stderr);
-        exit(EXIT_FAILURE);
-    }
-    return (int64_t)tv.tv_nsec / 1000 + (int64_t)tv.tv_sec * 1000000LL;
-#else
-    struct timeval tv;
-    gettimeofday(&tv, NULL);
-    return (int64_t)tv.tv_usec + (int64_t)tv.tv_sec * 1000000LL;
 #endif
 }
 
@@ -105,9 +119,9 @@ static void run_benchmark(char *name, void (*benchmark)(void*, int), void (*setu
         if (setup != NULL) {
             setup(data);
         }
-        begin = gettime_i64();
+        begin = gettime_us();
         benchmark(data, iter);
-        total = gettime_i64() - begin;
+        total = gettime_us() - begin;
         if (teardown != NULL) {
             teardown(data, iter);
         }
@@ -176,6 +190,14 @@ static int get_iters(int default_iters) {
     }
 }
 
+static void print_clock_info(void) {
+#if defined(CLOCK_PROCESS_CPUTIME_ID)
+    printf("INFO: Using per-process CPU timer\n\n");
+#else
+    printf("WARN: using Wall-Clock timer, results are highly influenced by other running processes\n\n");
+#endif
+}
+
 static void print_output_table_header_row(void) {
     char* bench_str = "Benchmark";     /* left justified */
     char* min_str = "    Min(us)    "; /* center alignment */

src/bench_ecmult.c

@@ -3,6 +3,9 @@
  * Distributed under the MIT software license, see the accompanying    *
  * file COPYING or https://www.opensource.org/licenses/mit-license.php.*
  ***********************************************************************/
+
+#define _POSIX_C_SOURCE 200112L /* for clock_gettime() */
+
 #include <stdio.h>
 #include <stdlib.h>
 
@@ -362,7 +365,7 @@ int main(int argc, char **argv) {
     }
     secp256k1_ge_set_all_gej_var(data.pubkeys, data.pubkeys_gej, POINTS);
 
-
+    print_clock_info();
     print_output_table_header_row();
     /* Initialize offset1 and offset2 */
     hash_into_offset(&data, 0);

src/bench_internal.c

@@ -3,6 +3,9 @@
  * Distributed under the MIT software license, see the accompanying    *
  * file COPYING or https://www.opensource.org/licenses/mit-license.php.*
  ***********************************************************************/
+
+#define _POSIX_C_SOURCE 200112L /* for clock_gettime() */
+
 #include <stdio.h>
 #include <stdlib.h>
 
@@ -398,6 +401,7 @@ int main(int argc, char **argv) {
         }
     }
 
+    print_clock_info();
     print_output_table_header_row();
 
     if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "half")) run_benchmark("scalar_half", bench_scalar_half, bench_setup, NULL, &data, 10, iters*100);