From patchwork Tue Apr 15 13:35:41 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Will Newton X-Patchwork-Id: 28416 Return-Path: X-Original-To: linaro@patches.linaro.org Delivered-To: linaro@patches.linaro.org Received: from mail-vc0-f197.google.com (mail-vc0-f197.google.com [209.85.220.197]) by ip-10-151-82-157.ec2.internal (Postfix) with ESMTPS id D24462036A for ; Tue, 15 Apr 2014 13:36:03 +0000 (UTC) Received: by mail-vc0-f197.google.com with SMTP id if11sf32587903vcb.0 for ; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:mime-version:delivered-to:mailing-list :precedence:list-id:list-unsubscribe:list-subscribe:list-archive :list-post:list-help:sender:delivered-to:from:to:subject:date :message-id:x-original-sender:x-original-authentication-results; bh=QZy+IwsC6nfi1fvXtE+QVHqQ8yc9L8Fv2Fx3gd7g1eE=; b=ZxwPGCNBtRRGT55GldqmOqlvruBG2R+rxL5MatTun/4MQY9NTWVAmEad++kWp+Yhuz z9g+0Beb4n9ZkdM0F2AEsxikDRWB8t/9bYck3PuWHszQhNZCfvyVyjaNuCclGhYiATDa GTstUBFPR7RJDVqJdzV/ZbxbwJGs8q1vhJOyiujkOuJpKztO8kHqmesG7TFnQb72S2sD U2z2dPLS0mPiJEaIFk6WykVxVvlfZXgupCjR5hWE1OXOkX6syARp6LziKtKbQk3g/i5l q20Nzwn7EPbqXjE4wB6mwWQd2j2Ik6EP5MmTZso1nMa7U2hie2fKd9KANiz5VywVDJS6 3HzA== X-Gm-Message-State: ALoCoQmK4ebrEvsTHdR2/btR82Zbnx1cwzvoFSgDku5LsH4Hgv+XHFL6RZmc622ZkJT6j2sd+kym X-Received: by 10.236.138.113 with SMTP id z77mr610967yhi.25.1397568963595; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) MIME-Version: 1.0 X-BeenThere: patchwork-forward@linaro.org Received: by 10.140.91.8 with SMTP id y8ls121347qgd.82.gmail; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) X-Received: by 10.220.7.131 with SMTP id d3mr89522vcd.45.1397568963451; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) Received: from mail-vc0-x235.google.com (mail-vc0-x235.google.com [2607:f8b0:400c:c03::235]) by mx.google.com with ESMTPS id w15si3391418vcz.181.2014.04.15.06.36.03 for (version=TLSv1 cipher=ECDHE-RSA-RC4-SHA bits=128/128); Tue, 15 Apr 2014 06:36:03 -0700 (PDT) Received-SPF: neutral (google.com: 2607:f8b0:400c:c03::235 is neither permitted nor denied by best guess record for domain of patch+caf_=patchwork-forward=linaro.org@linaro.org) client-ip=2607:f8b0:400c:c03::235; Received: by mail-vc0-f181.google.com with SMTP id id10so9357935vcb.12 for ; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) X-Received: by 10.58.243.72 with SMTP id ww8mr85186vec.49.1397568963265; Tue, 15 Apr 2014 06:36:03 -0700 (PDT) X-Forwarded-To: patchwork-forward@linaro.org X-Forwarded-For: patch@linaro.org patchwork-forward@linaro.org Delivered-To: patch@linaro.org Received: by 10.220.221.72 with SMTP id ib8csp232147vcb; Tue, 15 Apr 2014 06:36:02 -0700 (PDT) X-Received: by 10.66.153.80 with SMTP id ve16mr1874721pab.143.1397568962455; Tue, 15 Apr 2014 06:36:02 -0700 (PDT) Received: from sourceware.org (server1.sourceware.org. [209.132.180.131]) by mx.google.com with ESMTPS id zw7si10841198pac.316.2014.04.15.06.36.01 for (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Tue, 15 Apr 2014 06:36:02 -0700 (PDT) Received-SPF: pass (google.com: domain of libc-alpha-return-49059-patch=linaro.org@sourceware.org designates 209.132.180.131 as permitted sender) client-ip=209.132.180.131; Received: (qmail 17737 invoked by alias); 15 Apr 2014 13:35:53 -0000 Mailing-List: list patchwork-forward@linaro.org; contact patchwork-forward+owners@linaro.org Precedence: list List-Id: List-Unsubscribe: , List-Subscribe: List-Archive: List-Post: , List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 17720 invoked by uid 89); 15 Apr 2014 13:35:52 -0000 X-Virus-Found: No X-Spam-SWARE-Status: No, score=-2.1 required=5.0 tests=AWL, BAYES_00, RCVD_IN_DNSWL_LOW, SPF_PASS autolearn=ham version=3.3.2 X-HELO: mail-wi0-f180.google.com X-Received: by 10.194.103.36 with SMTP id ft4mr1706438wjb.66.1397568945320; Tue, 15 Apr 2014 06:35:45 -0700 (PDT) From: Will Newton To: libc-alpha@sourceware.org Subject: [PATCH] benchtests: Add malloc microbenchmark Date: Tue, 15 Apr 2014 14:35:41 +0100 Message-Id: <1397568941-4298-1-git-send-email-will.newton@linaro.org> X-Original-Sender: will.newton@linaro.org X-Original-Authentication-Results: mx.google.com; spf=neutral (google.com: 2607:f8b0:400c:c03::235 is neither permitted nor denied by best guess record for domain of patch+caf_=patchwork-forward=linaro.org@linaro.org) smtp.mail=patch+caf_=patchwork-forward=linaro.org@linaro.org; dkim=pass header.i=@sourceware.org X-Google-Group-Id: 836684582541 Add a microbenchmark for measuring malloc and free performance. The benchmark allocates and frees buffers of random sizes in a random order and measures the overall execution time and RSS. Variants of the benchmark are run with 8, 32 and 64 threads to measure the effect of concurrency on allocator performance. The random block sizes used follow an inverse square distribution which is intended to mimic the behaviour of real applications which tend to allocate many more small blocks than large ones. ChangeLog: 2014-04-15 Will Newton * benchtests/Makefile: (benchset): Add malloc benchmarks. Link threaded malloc benchmarks with libpthread. * benchtests/bench-malloc-threads-32.c: New file. * benchtests/bench-malloc-threads-64.c: Likewise. * benchtests/bench-malloc-threads-8.c: Likewise. * benchtests/bench-malloc.c: Likewise. --- benchtests/Makefile | 7 +- benchtests/bench-malloc-threads-32.c | 20 ++++ benchtests/bench-malloc-threads-64.c | 20 ++++ benchtests/bench-malloc-threads-8.c | 20 ++++ benchtests/bench-malloc.c | 183 +++++++++++++++++++++++++++++++++++ 5 files changed, 249 insertions(+), 1 deletion(-) create mode 100644 benchtests/bench-malloc-threads-32.c create mode 100644 benchtests/bench-malloc-threads-64.c create mode 100644 benchtests/bench-malloc-threads-8.c create mode 100644 benchtests/bench-malloc.c diff --git a/benchtests/Makefile b/benchtests/Makefile index a0954cd..f38380d 100644 --- a/benchtests/Makefile +++ b/benchtests/Makefile @@ -37,9 +37,11 @@ string-bench := bcopy bzero memccpy memchr memcmp memcpy memmem memmove \ strspn strstr strcpy_chk stpcpy_chk memrchr strsep strtok string-bench-all := $(string-bench) +malloc-bench := malloc malloc-threads-8 malloc-threads-32 malloc-threads-64 + stdlib-bench := strtod -benchset := $(string-bench-all) $(stdlib-bench) +benchset := $(string-bench-all) $(stdlib-bench) $(malloc-bench) CFLAGS-bench-ffs.c += -fno-builtin CFLAGS-bench-ffsll.c += -fno-builtin @@ -47,6 +49,9 @@ CFLAGS-bench-ffsll.c += -fno-builtin $(addprefix $(objpfx)bench-,$(bench-math)): $(common-objpfx)math/libm.so $(addprefix $(objpfx)bench-,$(bench-pthread)): \ $(common-objpfx)nptl/libpthread.so +$(objpfx)bench-malloc-threads-8: $(common-objpfx)nptl/libpthread.so +$(objpfx)bench-malloc-threads-32: $(common-objpfx)nptl/libpthread.so +$(objpfx)bench-malloc-threads-64: $(common-objpfx)nptl/libpthread.so diff --git a/benchtests/bench-malloc-threads-32.c b/benchtests/bench-malloc-threads-32.c new file mode 100644 index 0000000..463ceb7 --- /dev/null +++ b/benchtests/bench-malloc-threads-32.c @@ -0,0 +1,20 @@ +/* Measure malloc and free functions with threads. + Copyright (C) 2014 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 + . */ + +#define NUM_THREADS 32 +#include "bench-malloc.c" diff --git a/benchtests/bench-malloc-threads-64.c b/benchtests/bench-malloc-threads-64.c new file mode 100644 index 0000000..61d8c10 --- /dev/null +++ b/benchtests/bench-malloc-threads-64.c @@ -0,0 +1,20 @@ +/* Measure malloc and free functions with threads. + Copyright (C) 2014 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 + . */ + +#define NUM_THREADS 64 +#include "bench-malloc.c" diff --git a/benchtests/bench-malloc-threads-8.c b/benchtests/bench-malloc-threads-8.c new file mode 100644 index 0000000..ac4ff79 --- /dev/null +++ b/benchtests/bench-malloc-threads-8.c @@ -0,0 +1,20 @@ +/* Measure malloc and free functions with threads. + Copyright (C) 2014 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 + . */ + +#define NUM_THREADS 8 +#include "bench-malloc.c" diff --git a/benchtests/bench-malloc.c b/benchtests/bench-malloc.c new file mode 100644 index 0000000..6809bba --- /dev/null +++ b/benchtests/bench-malloc.c @@ -0,0 +1,183 @@ +/* Benchmark malloc and free functions. + Copyright (C) 2013-2014 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 +#include +#include +#include +#include +#include + +#include "bench-timing.h" +#include "json-lib.h" + +#define BENCHMARK_ITERATIONS 10000000 +#define RAND_SEED 88 + +#ifndef NUM_THREADS +#define NUM_THREADS 1 +#endif + +/* Maximum memory that can be allocated at any one time is: + + NUM_THREADS * WORKING_SET_SIZE * MAX_ALLOCATION_SIZE + + However due to the distribution of the random block sizes + the typical amount allocated will be much smaller. */ +#define WORKING_SET_SIZE 1024 + +#define MIN_ALLOCATION_SIZE 4 +#define MAX_ALLOCATION_SIZE 32768 + +/* Get a random block size with an inverse square distribution. */ +static size_t +get_block_size (unsigned int rand_data) +{ + /* Inverse square. */ + float exponent = -2; + /* Minimum value of distribution. */ + float dist_min = MIN_ALLOCATION_SIZE; + /* Maximum value of distribution. */ + float dist_max = MAX_ALLOCATION_SIZE; + + float min_pow = powf (dist_min, exponent + 1); + float max_pow = powf (dist_max, exponent + 1); + + float r = (float) rand_data / RAND_MAX; + + return (size_t) powf ((max_pow - min_pow) * r + min_pow, 1 / (exponent + 1)); +} + +/* Allocate and free blocks in a random order. */ +static void +malloc_benchmark_loop (size_t iters, void **ptr_arr, unsigned int *rand_data) +{ + size_t i; + + for (i = 0; i < iters; i++) + { + unsigned int r = rand_r (rand_data); + size_t next_idx = r % WORKING_SET_SIZE; + size_t next_block = get_block_size (r); + + if (ptr_arr[next_idx]) + free (ptr_arr[next_idx]); + + ptr_arr[next_idx] = malloc (next_block); + } +} + +static void *working_set[NUM_THREADS][WORKING_SET_SIZE]; + +#if NUM_THREADS > 1 +static pthread_t threads[NUM_THREADS]; + +struct thread_args +{ + size_t iters; + void **working_set; +}; + +static void * +benchmark_thread (void *arg) +{ + struct thread_args *args = (struct thread_args *) arg; + size_t iters = args->iters; + void *thread_set = args->working_set; + unsigned int rand_data = RAND_SEED; + + malloc_benchmark_loop (iters, thread_set, &rand_data); + + return NULL; +} +#endif + +int +main (int argc, char **argv) +{ + timing_t start, stop, cur; + size_t iters = BENCHMARK_ITERATIONS; + unsigned long res; + json_ctx_t json_ctx; + double d_total_s, d_total_i; + + json_init (&json_ctx, 0, stdout); + + json_document_begin (&json_ctx); + + json_attr_string (&json_ctx, "timing_type", TIMING_TYPE); + + json_attr_object_begin (&json_ctx, "functions"); + + json_attr_object_begin (&json_ctx, "malloc"); + + json_attr_object_begin (&json_ctx, ""); + + TIMING_INIT (res); + + (void) res; + + TIMING_NOW (start); +#if NUM_THREADS == 1 + unsigned int rand_data = RAND_SEED; + malloc_benchmark_loop (iters, working_set[0], &rand_data); +#else + struct thread_args args[NUM_THREADS]; + + size_t i; + + for (i = 0; i < NUM_THREADS; i++) + { + args[i].iters = iters; + args[i].working_set = working_set[i]; + pthread_create(&threads[i], NULL, benchmark_thread, &args[i]); + } + + for (i = 0; i < NUM_THREADS; i++) + pthread_join(threads[i], NULL); +#endif + TIMING_NOW (stop); + + struct rusage usage; + getrusage(RUSAGE_SELF, &usage); + + TIMING_DIFF (cur, start, stop); + + d_total_s = cur; + d_total_i = iters * NUM_THREADS; + + json_attr_double (&json_ctx, "duration", d_total_s); + json_attr_double (&json_ctx, "iterations", d_total_i); + json_attr_double (&json_ctx, "mean", d_total_s / d_total_i); + json_attr_double (&json_ctx, "max_rss", usage.ru_maxrss); + + json_attr_double (&json_ctx, "threads", NUM_THREADS); + json_attr_double (&json_ctx, "min_size", MIN_ALLOCATION_SIZE); + json_attr_double (&json_ctx, "max_size", MAX_ALLOCATION_SIZE); + json_attr_double (&json_ctx, "random_seed", RAND_SEED); + + json_attr_object_end (&json_ctx); + + json_attr_object_end (&json_ctx); + + json_attr_object_end (&json_ctx); + + json_document_end (&json_ctx); + + return 0; +}