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: 339277 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 902E6140080 for ; Tue, 15 Apr 2014 23:35:58 +1000 (EST) DomainKey-Signature: a=rsa-sha1; c=nofws; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:from:to:subject:date:message-id; q=dns; s= default; b=fvIdbLVymsd5hPFU/ZOqMlhpGooXlDyG3WEKRBzK0O2nVxFZfxN4s X5BgaDL3PeOBZRwbc6MnVid3Gw5HGnnnBFUiT1rgd9n92oYMYO4yAIRHVZYk8Cbs /A+jMKFuDFC3iMBNuSVqEJ+TgEo0rhP+swnz05QRVmXDLRG9r7OqMo= DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:from:to:subject:date:message-id; s=default; bh=VmAnsV1KBmIXMwy68BhA77ZoaTg=; b=BVs9P6epVyrz3x4Mj9+VY6elxgPZ bYqd273aRixDlY5vNU5ssv64iPo4bxB0NG0tFEQscY+NUbVOVPxRvf3DVraVhMGR h8N4C7OPUKLdLcuA97uzwc0UcaVLoDpYgZwPrBfmVvFWNXlZk5PIPuto3bCVErTL 2kbysVi0HInlpx0= Received: (qmail 17737 invoked by alias); 15 Apr 2014 13:35:53 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk 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 Authentication-Results: sourceware.org; auth=none 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-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:from:to:subject:date:message-id; bh=QZy+IwsC6nfi1fvXtE+QVHqQ8yc9L8Fv2Fx3gd7g1eE=; b=ANKBFBzCE8Lzft5OXjAKqOt61WvrcaGfRinrNDsJyo+b805+1R4xwj+KZseDeNQ/No ur0NeEtd6o7BuPQ4UWzzAyJ0wPG74YPdp5Og69Ums0Rej2EwDUb20iTjAMCfRgYvK11u 09YMlaGeGL4oFoy0OUv+/OLykuSnIIuFYEs27Kc6cVzdYv4lWE3m5+XvrDJkS4m3oeGT jhawnzbO435b6nrZJrrIKmx92/nNT8Mj8lRFZKOGCZgbk/ii0bspoXDoBVmPpOhsqCga 2lp+MqwIqZ52T/rWCEYjuXhWDO740wJ7yENptnIft2h8sY5goJXe95RZ9wQb2i++zB90 QCSw== X-Gm-Message-State: ALoCoQlUs2XEpQUHgpKcaLfmpz28b6GkZJZ28E7nFvTiNtOc3qwNAgBtUKiONADig7Lt7qrIjqvO 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> 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; +}