From 629cc356653719c206a05f4dee5c5e242edb6546 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:34 +0900 Subject: perf bench: Add builtin-bench.c: General framework for benchmark suites This patch adds builtin-bench.c builtin-bench.c is a general framework for benchmark suites. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-5-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 128 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 128 insertions(+) create mode 100644 tools/perf/builtin-bench.c (limited to 'tools/perf/builtin-bench.c') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c new file mode 100644 index 00000000000..31f41643b0c --- /dev/null +++ b/tools/perf/builtin-bench.c @@ -0,0 +1,128 @@ +/* + * + * builtin-bench.c + * + * General benchmarking subsystem provided by perf + * + * Copyright (C) 2009, Hitoshi Mitake + * + */ + +/* + * + * Available subsystem list: + * sched ... scheduler and IPC mechanism + * + */ + +#include "perf.h" +#include "util/util.h" +#include "util/parse-options.h" +#include "builtin.h" +#include "bench/bench.h" + +#include +#include +#include + +struct bench_suite { + const char *name; + const char *summary; + int (*fn)(int, const char **, const char *); +}; + +static struct bench_suite sched_suites[] = { + { "messaging", + "Benchmark for scheduler and IPC mechanisms", + bench_sched_messaging }, + { "pipe", + "Flood of communication over pipe() between two processes", + bench_sched_pipe }, + { NULL, + NULL, + NULL } +}; + +struct bench_subsys { + const char *name; + const char *summary; + struct bench_suite *suites; +}; + +static struct bench_subsys subsystems[] = { + { "sched", + "scheduler and IPC mechanism", + sched_suites }, + { NULL, + NULL, + NULL } +}; + +static void dump_suites(int subsys_index) +{ + int i; + + printf("List of available suites for %s...\n\n", + subsystems[subsys_index].name); + + for (i = 0; subsystems[subsys_index].suites[i].name; i++) + printf("\t%s: %s\n", + subsystems[subsys_index].suites[i].name, + subsystems[subsys_index].suites[i].summary); + + printf("\n"); + return; +} + +int cmd_bench(int argc, const char **argv, const char *prefix __used) +{ + int i, j, status = 0; + + if (argc < 2) { + /* No subsystem specified. */ + printf("Usage: perf bench []\n\n"); + printf("List of available subsystems...\n\n"); + + for (i = 0; subsystems[i].name; i++) + printf("\t%s: %s\n", + subsystems[i].name, subsystems[i].summary); + printf("\n"); + + goto end; + } + + for (i = 0; subsystems[i].name; i++) { + if (strcmp(subsystems[i].name, argv[1])) + continue; + + if (argc < 3) { + /* No suite specified. */ + dump_suites(i); + goto end; + } + + for (j = 0; subsystems[i].suites[j].name; j++) { + if (strcmp(subsystems[i].suites[j].name, argv[2])) + continue; + + status = subsystems[i].suites[j].fn(argc - 2, + argv + 2, prefix); + goto end; + } + + if (!strcmp(argv[2], "-h") || !strcmp(argv[2], "--help")) { + dump_suites(i); + goto end; + } + + printf("Unknown suite:%s for %s\n", argv[2], argv[1]); + status = 1; + goto end; + } + + printf("Unknown subsystem:%s\n", argv[1]); + status = 1; + +end: + return status; +} -- cgit v1.2.3-70-g09d2 From 386d7e9e542c2115d5d300747e57f503458a1617 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 08:20:00 +0900 Subject: perf bench: Modify builtin-bench.c for processing common options This patch modifies builtin-bench.c for processing common options. The first option added is "--format". Users of perf bench will be able to specify output style by --format. Usage example: % ./perf bench sched messaging # with no style specify (20 sender and receiver processes per group) (10 groups == 400 processes run) Total time:1.431 sec % ./perf bench --format=simple sched messaging # specified simple 1.431 Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257808802-9420-3-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 79 ++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 65 insertions(+), 14 deletions(-) (limited to 'tools/perf/builtin-bench.c') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index 31f41643b0c..c7505eaff84 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -74,53 +74,104 @@ static void dump_suites(int subsys_index) return; } +static char *bench_format_str; +int bench_format = BENCH_FORMAT_DEFAULT; + +static const struct option bench_options[] = { + OPT_STRING('f', "format", &bench_format_str, "default", + "Specify format style"), + OPT_END() +}; + +static const char * const bench_usage[] = { + "perf bench [] []", + NULL +}; + +static void print_usage(void) +{ + int i; + + printf("Usage: \n"); + for (i = 0; bench_usage[i]; i++) + printf("\t%s\n", bench_usage[i]); + printf("\n"); + + printf("List of available subsystems...\n\n"); + + for (i = 0; subsystems[i].name; i++) + printf("\t%s: %s\n", + subsystems[i].name, subsystems[i].summary); + printf("\n"); +} + +static int bench_str2int(char *str) +{ + if (!str) + return BENCH_FORMAT_DEFAULT; + + if (!strcmp(str, BENCH_FORMAT_DEFAULT_STR)) + return BENCH_FORMAT_DEFAULT; + else if (!strcmp(str, BENCH_FORMAT_SIMPLE_STR)) + return BENCH_FORMAT_SIMPLE; + + return BENCH_FORMAT_UNKNOWN; +} + int cmd_bench(int argc, const char **argv, const char *prefix __used) { int i, j, status = 0; if (argc < 2) { /* No subsystem specified. */ - printf("Usage: perf bench []\n\n"); - printf("List of available subsystems...\n\n"); + print_usage(); + goto end; + } - for (i = 0; subsystems[i].name; i++) - printf("\t%s: %s\n", - subsystems[i].name, subsystems[i].summary); - printf("\n"); + argc = parse_options(argc, argv, bench_options, bench_usage, + PARSE_OPT_STOP_AT_NON_OPTION); + + bench_format = bench_str2int(bench_format_str); + if (bench_format == BENCH_FORMAT_UNKNOWN) { + printf("Unknown format descriptor:%s\n", bench_format_str); + goto end; + } + if (argc < 1) { + print_usage(); goto end; } for (i = 0; subsystems[i].name; i++) { - if (strcmp(subsystems[i].name, argv[1])) + if (strcmp(subsystems[i].name, argv[0])) continue; - if (argc < 3) { + if (argc < 2) { /* No suite specified. */ dump_suites(i); goto end; } for (j = 0; subsystems[i].suites[j].name; j++) { - if (strcmp(subsystems[i].suites[j].name, argv[2])) + if (strcmp(subsystems[i].suites[j].name, argv[1])) continue; - status = subsystems[i].suites[j].fn(argc - 2, - argv + 2, prefix); + status = subsystems[i].suites[j].fn(argc - 1, + argv + 1, prefix); goto end; } - if (!strcmp(argv[2], "-h") || !strcmp(argv[2], "--help")) { + if (!strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) { dump_suites(i); goto end; } - printf("Unknown suite:%s for %s\n", argv[2], argv[1]); + printf("Unknown suite:%s for %s\n", argv[1], argv[0]); status = 1; goto end; } - printf("Unknown subsystem:%s\n", argv[1]); + printf("Unknown subsystem:%s\n", argv[0]); status = 1; end: -- cgit v1.2.3-70-g09d2 From 79e295d4bd0f524257299e7c4e42f643f21abcc2 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 11 Nov 2009 00:04:00 +0900 Subject: perf bench: Improve builtin-bench.c for more friendly output This patch makes output of perf bench more friendly. Current style of putput, keeping user wait and printing everything suddenly when we finish, may confuse users. So I improved it: | % perf bench sched messaging | # Running sched/messaging benchmark... <- printed right after invocation | # 20 sender and receiver processes per group | # 10 groups == 400 processes run | | Total time: 1.476 [sec] Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257865442-20252-2-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'tools/perf/builtin-bench.c') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index c7505eaff84..90c39baae0d 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -156,6 +156,10 @@ int cmd_bench(int argc, const char **argv, const char *prefix __used) if (strcmp(subsystems[i].suites[j].name, argv[1])) continue; + if (bench_format == BENCH_FORMAT_DEFAULT) + printf("# Running %s/%s benchmark...\n", + subsystems[i].name, + subsystems[i].suites[j].name); status = subsystems[i].suites[j].fn(argc - 1, argv + 1, prefix); goto end; -- cgit v1.2.3-70-g09d2 From 827f3b4974c5db2968d4979fe6a0ae00ab37bdd8 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 18 Nov 2009 00:20:09 +0900 Subject: perf bench: Add memcpy() benchmark 'perf bench mem memcpy' is a benchmark suite for measuring memcpy() performance. Example on a Intel(R) Core(TM)2 Duo CPU E6850 @ 3.00GHz: | % perf bench mem memcpy -l 1GB | # Running mem/memcpy benchmark... | # Copying 1MB Bytes from 0xb7d98008 to 0xb7e99008 ... | | 726.216412 MB/Sec Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1258471212-30281-1-git-send-email-mitake@dcl.info.waseda.ac.jp> [ v2: updated changelog, clarified history of builtin-bench.c ] Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/bench/bench.h | 1 + tools/perf/bench/mem-memcpy.c | 186 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-bench.c | 15 +++- 4 files changed, 202 insertions(+), 1 deletion(-) create mode 100644 tools/perf/bench/mem-memcpy.c (limited to 'tools/perf/builtin-bench.c') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3f0666af93d..53e663a5fa2 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -432,6 +432,7 @@ BUILTIN_OBJS += builtin-bench.o # Benchmark modules BUILTIN_OBJS += bench/sched-messaging.o BUILTIN_OBJS += bench/sched-pipe.o +BUILTIN_OBJS += bench/mem-memcpy.o BUILTIN_OBJS += builtin-help.o BUILTIN_OBJS += builtin-sched.o diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index 9fbd8d745fa..f7781c6267c 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -3,6 +3,7 @@ extern int bench_sched_messaging(int argc, const char **argv, const char *prefix); extern int bench_sched_pipe(int argc, const char **argv, const char *prefix); +extern int bench_mem_memcpy(int argc, const char **argv, const char *prefix __used); #define BENCH_FORMAT_DEFAULT_STR "default" #define BENCH_FORMAT_DEFAULT 0 diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c new file mode 100644 index 00000000000..d4f4f9806ae --- /dev/null +++ b/tools/perf/bench/mem-memcpy.c @@ -0,0 +1,186 @@ +/* + * mem-memcpy.c + * + * memcpy: Simple memory copy in various ways + * + * Written by Hitoshi Mitake + */ +#include + +#include "../perf.h" +#include "../util/util.h" +#include "../util/parse-options.h" +#include "../util/string.h" +#include "../util/header.h" +#include "bench.h" + +#include +#include +#include +#include +#include + +#define K 1024 + +static const char *length_str = "1MB"; +static const char *routine = "default"; +static int use_clock = 0; + +static const struct option options[] = { + OPT_STRING('l', "length", &length_str, "1MB", + "Specify length of memory to copy. " + "available unit: B, MB, GB (upper and lower)"), + OPT_STRING('r', "routine", &routine, "default", + "Specify routine to copy"), + OPT_BOOLEAN('c', "clock", &use_clock, + "Use CPU clock for measuring"), + OPT_END() +}; + +struct routine { + const char *name; + const char *desc; + void * (*fn)(void *dst, const void *src, size_t len); +}; + +struct routine routines[] = { + { "default", + "Default memcpy() provided by glibc", + memcpy }, + { NULL, + NULL, + NULL } +}; + +static const char * const bench_mem_memcpy_usage[] = { + "perf bench mem memcpy ", + NULL +}; + +static int clock_fd; + +static struct perf_event_attr clock_attr = { + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES +}; + +static void init_clock(void) +{ + clock_fd = sys_perf_event_open(&clock_attr, getpid(), -1, -1, 0); + BUG_ON(clock_fd < 0); +} + +static u64 get_clock(void) +{ + int ret; + u64 clk; + + ret = read(clock_fd, &clk, sizeof(u64)); + BUG_ON(ret != sizeof(u64)); + + return clk; +} + +static double timeval2double(struct timeval *ts) +{ + return (double)ts->tv_sec + + (double)ts->tv_usec / (double)1000000; +} + +int bench_mem_memcpy(int argc, const char **argv, + const char *prefix __used) +{ + int i; + void *dst, *src; + size_t length; + double bps = 0.0; + struct timeval tv_start, tv_end, tv_diff; + u64 clock_start, clock_end, clock_diff; + + clock_start = clock_end = clock_diff = 0ULL; + argc = parse_options(argc, argv, options, + bench_mem_memcpy_usage, 0); + + tv_diff.tv_sec = 0; + tv_diff.tv_usec = 0; + length = (size_t)perf_atoll((char *)length_str); + if ((long long int)length <= 0) { + fprintf(stderr, "Invalid length:%s\n", length_str); + return 1; + } + + for (i = 0; routines[i].name; i++) { + if (!strcmp(routines[i].name, routine)) + break; + } + if (!routines[i].name) { + printf("Unknown routine:%s\n", routine); + printf("Available routines...\n"); + for (i = 0; routines[i].name; i++) { + printf("\t%s ... %s\n", + routines[i].name, routines[i].desc); + } + return 1; + } + + dst = calloc(length, sizeof(char)); + assert(dst); + src = calloc(length, sizeof(char)); + assert(src); + + if (bench_format == BENCH_FORMAT_DEFAULT) { + printf("# Copying %s Bytes from %p to %p ...\n\n", + length_str, src, dst); + } + + if (use_clock) { + init_clock(); + clock_start = get_clock(); + } else + BUG_ON(gettimeofday(&tv_start, NULL)); + + routines[i].fn(dst, src, length); + + if (use_clock) { + clock_end = get_clock(); + clock_diff = clock_end - clock_start; + } else { + BUG_ON(gettimeofday(&tv_end, NULL)); + timersub(&tv_end, &tv_start, &tv_diff); + bps = (double)((double)length / timeval2double(&tv_diff)); + } + + switch (bench_format) { + case BENCH_FORMAT_DEFAULT: + if (use_clock) { + printf(" %14lf Clock/Byte\n", + (double)clock_diff / (double)length); + } else { + if (bps < K) + printf(" %14lf B/Sec\n", bps); + else if (bps < K * K) + printf(" %14lfd KB/Sec\n", bps / 1024); + else if (bps < K * K * K) + printf(" %14lf MB/Sec\n", bps / 1024 / 1024); + else { + printf(" %14lf GB/Sec\n", + bps / 1024 / 1024 / 1024); + } + } + break; + case BENCH_FORMAT_SIMPLE: + if (use_clock) { + printf("%14lf\n", + (double)clock_diff / (double)length); + } else + printf("%lf\n", bps); + break; + default: + /* reaching here is something disaster */ + fprintf(stderr, "Unknown format:%d\n", bench_format); + exit(1); + break; + } + + return 0; +} diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index 90c39baae0d..e043eb83092 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -12,6 +12,7 @@ * * Available subsystem list: * sched ... scheduler and IPC mechanism + * mem ... memory access performance * */ @@ -43,6 +44,15 @@ static struct bench_suite sched_suites[] = { NULL } }; +static struct bench_suite mem_suites[] = { + { "memcpy", + "Simple memory copy in various ways", + bench_mem_memcpy }, + { NULL, + NULL, + NULL } +}; + struct bench_subsys { const char *name; const char *summary; @@ -53,9 +63,12 @@ static struct bench_subsys subsystems[] = { { "sched", "scheduler and IPC mechanism", sched_suites }, + { "mem", + "memory access performance", + mem_suites }, { NULL, NULL, - NULL } + NULL } }; static void dump_suites(int subsys_index) -- cgit v1.2.3-70-g09d2