From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 1791) id 0728A383542E; Fri, 22 Jul 2022 15:32:58 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 0728A383542E Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Adhemerval Zanella To: glibc-cvs@sourceware.org Subject: [glibc] benchtests: Add arc4random benchtest X-Act-Checkin: glibc X-Git-Author: Adhemerval Zanella Netto X-Git-Refname: refs/heads/master X-Git-Oldrev: 8dd890d96f1833a58de6e112a14d63ab44e0a402 X-Git-Newrev: 5d765ada01d140d8d1ecf94953a4751593af720d Message-Id: <20220722153258.0728A383542E@sourceware.org> Date: Fri, 22 Jul 2022 15:32:58 +0000 (GMT) X-BeenThere: glibc-cvs@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Glibc-cvs mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 22 Jul 2022 15:32:58 -0000 https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=5d765ada01d140d8d1ecf94953a4751593af720d commit 5d765ada01d140d8d1ecf94953a4751593af720d Author: Adhemerval Zanella Netto Date: Thu Jul 21 10:05:01 2022 -0300 benchtests: Add arc4random benchtest It shows both throughput (total bytes obtained in the test duration) and latecy for both arc4random and arc4random_buf with different sizes. Checked on x86_64-linux-gnu, aarch64-linux, and powerpc64le-linux-gnu. Diff: --- benchtests/Makefile | 5 +- benchtests/bench-arc4random.c | 218 +++++++++++++++++++++++++++++++++++ benchtests/bench-hash-funcs-kernel.h | 1 + benchtests/bench-hash-funcs.c | 2 - benchtests/bench-util.h | 7 ++ 5 files changed, 230 insertions(+), 3 deletions(-) diff --git a/benchtests/Makefile b/benchtests/Makefile index c279041e19..d99771be74 100644 --- a/benchtests/Makefile +++ b/benchtests/Makefile @@ -233,7 +233,10 @@ hash-benchset := \ nss-hash \ # hash-benchset -stdlib-benchset := strtod +stdlib-benchset := \ + arc4random \ + strtod \ + # stdlib-benchset stdio-common-benchset := sprintf diff --git a/benchtests/bench-arc4random.c b/benchtests/bench-arc4random.c new file mode 100644 index 0000000000..d8fd40298e --- /dev/null +++ b/benchtests/bench-arc4random.c @@ -0,0 +1,218 @@ +/* arc4random benchmarks. + Copyright (C) 2022 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 + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "bench-timing.h" +#include "bench-util.h" +#include "json-lib.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static volatile sig_atomic_t timer_finished; + +static void timer_callback (int unused) +{ + timer_finished = 1; +} + +static timer_t timer; + +/* Run for approximately DURATION seconds, and it does not matter who + receive the signal (so not need to mask it on main thread). */ +static void +timer_start (void) +{ + timer_finished = 0; + timer = support_create_timer (DURATION, 0, false, timer_callback); +} +static void +timer_stop (void) +{ + support_delete_timer (timer); +} + +static const uint32_t sizes[] = { 0, 16, 32, 48, 64, 80, 96, 112, 128 }; + +static double +bench_throughput (void) +{ + uint64_t n = 0; + + struct timespec start, end; + clock_gettime (CLOCK_MONOTONIC, &start); + while (1) + { + DO_NOT_OPTIMIZE_OUT (arc4random ()); + n++; + + if (timer_finished == 1) + break; + } + clock_gettime (CLOCK_MONOTONIC, &end); + struct timespec diff = timespec_sub (end, start); + + double total = (double) n * sizeof (uint32_t); + double duration = (double) diff.tv_sec + + (double) diff.tv_nsec / TIMESPEC_HZ; + + return total / duration; +} + +static double +bench_latency (void) +{ + timing_t start, stop, cur; + const size_t iters = 1024; + + TIMING_NOW (start); + for (size_t i = 0; i < iters; i++) + DO_NOT_OPTIMIZE_OUT (arc4random ()); + TIMING_NOW (stop); + + TIMING_DIFF (cur, start, stop); + + return (double) (cur) / (double) iters; +} + +static double +bench_buf_throughput (size_t len) +{ + uint8_t buf[len]; + uint64_t n = 0; + + struct timespec start, end; + clock_gettime (CLOCK_MONOTONIC, &start); + while (1) + { + arc4random_buf (buf, len); + n++; + + if (timer_finished == 1) + break; + } + clock_gettime (CLOCK_MONOTONIC, &end); + struct timespec diff = timespec_sub (end, start); + + double total = (double) n * len; + double duration = (double) diff.tv_sec + + (double) diff.tv_nsec / TIMESPEC_HZ; + + return total / duration; +} + +static double +bench_buf_latency (size_t len) +{ + timing_t start, stop, cur; + const size_t iters = 1024; + + uint8_t buf[len]; + + TIMING_NOW (start); + for (size_t i = 0; i < iters; i++) + arc4random_buf (buf, len); + TIMING_NOW (stop); + + TIMING_DIFF (cur, start, stop); + + return (double) (cur) / (double) iters; +} + +static void +bench_singlethread (json_ctx_t *json_ctx) +{ + json_element_object_begin (json_ctx); + + json_array_begin (json_ctx, "throughput"); + for (int i = 0; i < array_length (sizes); i++) + { + timer_start (); + double r = sizes[i] == 0 + ? bench_throughput () : bench_buf_throughput (sizes[i]); + timer_stop (); + + json_element_double (json_ctx, r); + } + json_array_end (json_ctx); + + json_array_begin (json_ctx, "latency"); + for (int i = 0; i < array_length (sizes); i++) + { + timer_start (); + double r = sizes[i] == 0 + ? bench_latency () : bench_buf_latency (sizes[i]); + timer_stop (); + + json_element_double (json_ctx, r); + } + json_array_end (json_ctx); + + json_element_object_end (json_ctx); +} + +static void +run_bench (json_ctx_t *json_ctx, const char *name, + char *const*fnames, size_t fnameslen, + void (*bench) (json_ctx_t *ctx)) +{ + json_attr_object_begin (json_ctx, name); + json_array_begin (json_ctx, "functions"); + for (int i = 0; i < fnameslen; i++) + json_element_string (json_ctx, fnames[i]); + json_array_end (json_ctx); + + json_array_begin (json_ctx, "results"); + bench (json_ctx); + json_array_end (json_ctx); + json_attr_object_end (json_ctx); +} + +static int +do_test (void) +{ + char *fnames[array_length (sizes)]; + for (int i = 0; i < array_length (sizes); i++) + if (sizes[i] == 0) + fnames[i] = xasprintf ("arc4random"); + else + fnames[i] = xasprintf ("arc4random_buf(%u)", sizes[i]); + + json_ctx_t json_ctx; + json_init (&json_ctx, 0, stdout); + + json_document_begin (&json_ctx); + json_attr_string (&json_ctx, "timing_type", TIMING_TYPE); + + run_bench (&json_ctx, "single-thread", fnames, array_length (fnames), + bench_singlethread); + + json_document_end (&json_ctx); + + for (int i = 0; i < array_length (sizes); i++) + free (fnames[i]); + + return 0; +} + +#include diff --git a/benchtests/bench-hash-funcs-kernel.h b/benchtests/bench-hash-funcs-kernel.h index 83995cc0ae..63034f7e44 100644 --- a/benchtests/bench-hash-funcs-kernel.h +++ b/benchtests/bench-hash-funcs-kernel.h @@ -17,6 +17,7 @@ . */ +#include "bench-util.h" /* We go through the trouble of using macros here because many of the hash functions are meant to be inlined so its not fair to benchmark diff --git a/benchtests/bench-hash-funcs.c b/benchtests/bench-hash-funcs.c index 578c5cbae2..44b349d30c 100644 --- a/benchtests/bench-hash-funcs.c +++ b/benchtests/bench-hash-funcs.c @@ -38,8 +38,6 @@ #include #include -#define DO_NOT_OPTIMIZE_OUT(x) __asm__ volatile("" : : "r,m"(x) : "memory") - enum { NFIXED_ITERS = 1048576, diff --git a/benchtests/bench-util.h b/benchtests/bench-util.h index d0e29423aa..00f78d649f 100644 --- a/benchtests/bench-util.h +++ b/benchtests/bench-util.h @@ -16,6 +16,13 @@ License along with the GNU C Library; if not, see . */ +/* Prevent compiler to optimize away call. */ +#define DO_NOT_OPTIMIZE_OUT(value) \ + ({ \ + __typeof (value) __v = (value); \ + asm volatile ("" : : "r,m" (__v) : "memory"); \ + __v; \ + }) #ifndef START_ITER # define START_ITER (100000000)