perf bench: Change the procps visible command-name of invididual benchmark tests...
authorIngo Molnar <mingo@kernel.org>
Wed, 23 Oct 2013 12:37:56 +0000 (14:37 +0200)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 23 Oct 2013 12:57:34 +0000 (09:57 -0300)
Before this patch, looking at 'perf bench sched pipe' behavior over
'top' only told us that something related to perf is running:

      PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
    19934 mingo     20   0 54836 1296  952 R 18.6  0.0   0:00.56 perf
    19935 mingo     20   0 54836  384   36 S 18.6  0.0   0:00.56 perf

After the patch it's clearly visible what's going on:

      PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
    19744 mingo     20   0  125m 3536 2644 R 68.2  0.0   0:01.12 sched-pipe
    19745 mingo     20   0  125m 1172  276 R 68.2  0.0   0:01.12 sched-pipe

The benchmark-subsystem name is concatenated with the individual
testcase name.

Unfortunately 'perf top' does not show the reconfigured name, possibly
because it caches ->comm[] values and does not recognize changes to
them?

Also clean up a few bits in builtin-bench.c while at it and reorganize
the code and the output strings to be consistent.

Use iterators to access the various arrays. Rename 'suites' concept to
'benchmark collection' and the 'bench_suite' to 'benchmark/bench'. The
many repetitions of 'suite' made the code harder to read and understand.

The new output is:

  comet:~/tip/tools/perf> ./perf bench
  Usage:
        perf bench [<common options>] <collection> <benchmark> [<options>]

        # List of all available benchmark collections:

         sched: Scheduler and IPC benchmarks
           mem: Memory access benchmarks
          numa: NUMA scheduling and MM benchmarks
           all: All benchmarks

  comet:~/tip/tools/perf> ./perf bench sched

        # List of available benchmarks for collection 'sched':

     messaging: Benchmark for scheduling and IPC
          pipe: Benchmark for pipe() between two processes
           all: Test all scheduler benchmarks

  comet:~/tip/tools/perf> ./perf bench mem

        # List of available benchmarks for collection 'mem':

        memcpy: Benchmark for memcpy()
        memset: Benchmark for memset() tests
           all: Test all memory benchmarks

  comet:~/tip/tools/perf> ./perf bench numa

        # List of available benchmarks for collection 'numa':

           mem: Benchmark for NUMA workloads
           all: Test all NUMA benchmarks

Individual benchmark modules were not touched.

Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Hitoshi Mitake <h.mitake@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20131023123756.GA17871@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/builtin-bench.c

index 33af80f..e47f90c 100644 (file)
@@ -1,21 +1,18 @@
 /*
- *
  * builtin-bench.c
  *
- * General benchmarking subsystem provided by perf
+ * General benchmarking collections provided by perf
  *
  * Copyright (C) 2009, Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
- *
  */
 
 /*
+ * Available benchmark collection list:
  *
- * Available subsystem list:
- *  sched ... scheduler and IPC mechanism
+ *  sched ... scheduler and IPC performance
  *  mem   ... memory access performance
- *
+ *  numa  ... NUMA scheduling and MM performance
  */
-
 #include "perf.h"
 #include "util/util.h"
 #include "util/parse-options.h"
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <sys/prctl.h>
 
-struct bench_suite {
-       const char *name;
-       const char *summary;
-       int (*fn)(int, const char **, const char *);
+typedef int (*bench_fn_t)(int argc, const char **argv, const char *prefix);
+
+struct bench {
+       const char      *name;
+       const char      *summary;
+       bench_fn_t      fn;
 };
-                                               \
-/* sentinel: easy for help */
-#define suite_all { "all", "Test all benchmark suites", NULL }
 
 #ifdef HAVE_LIBNUMA_SUPPORT
-static struct bench_suite numa_suites[] = {
-       { "mem",
-         "Benchmark for NUMA workloads",
-         bench_numa },
-       suite_all,
-       { NULL,
-         NULL,
-         NULL                  }
+static struct bench numa_benchmarks[] = {
+       { "mem",        "Benchmark for NUMA workloads",                 bench_numa              },
+       { "all",        "Test all NUMA benchmarks",                     NULL                    },
+       { NULL,         NULL,                                           NULL                    }
 };
 #endif
 
-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      },
-       suite_all,
-       { NULL,
-         NULL,
-         NULL                  }
+static struct bench sched_benchmarks[] = {
+       { "messaging",  "Benchmark for scheduling and IPC",             bench_sched_messaging   },
+       { "pipe",       "Benchmark for pipe() between two processes",   bench_sched_pipe        },
+       { "all",        "Test all scheduler benchmarks",                NULL                    },
+       { NULL,         NULL,                                           NULL                    }
 };
 
-static struct bench_suite mem_suites[] = {
-       { "memcpy",
-         "Simple memory copy in various ways",
-         bench_mem_memcpy },
-       { "memset",
-         "Simple memory set in various ways",
-         bench_mem_memset },
-       suite_all,
-       { NULL,
-         NULL,
-         NULL             }
+static struct bench mem_benchmarks[] = {
+       { "memcpy",     "Benchmark for memcpy()",                       bench_mem_memcpy        },
+       { "memset",     "Benchmark for memset() tests",                 bench_mem_memset        },
+       { "all",        "Test all memory benchmarks",                   NULL                    },
+       { NULL,         NULL,                                           NULL                    }
 };
 
-struct bench_subsys {
-       const char *name;
-       const char *summary;
-       struct bench_suite *suites;
+struct collection {
+       const char      *name;
+       const char      *summary;
+       struct bench    *benchmarks;
 };
 
-static struct bench_subsys subsystems[] = {
+static struct collection collections[] = {
+       { "sched",      "Scheduler and IPC benchmarks",         sched_benchmarks        },
+       { "mem",        "Memory access benchmarks",                     mem_benchmarks          },
 #ifdef HAVE_LIBNUMA_SUPPORT
-       { "numa",
-         "NUMA scheduling and MM behavior",
-         numa_suites },
+       { "numa",       "NUMA scheduling and MM benchmarks",            numa_benchmarks         },
 #endif
-       { "sched",
-         "scheduler and IPC mechanism",
-         sched_suites },
-       { "mem",
-         "memory access performance",
-         mem_suites },
-       { "all",                /* sentinel: easy for help */
-         "all benchmark subsystem",
-         NULL },
-       { NULL,
-         NULL,
-         NULL       }
+       { "all",        "All benchmarks",                               NULL                    },
+       { NULL,         NULL,                                           NULL                    }
 };
 
-static void dump_suites(int subsys_index)
+/* Iterate over all benchmark collections: */
+#define for_each_collection(coll) \
+       for (coll = collections; coll->name; coll++)
+
+/* Iterate over all benchmarks within a collection: */
+#define for_each_bench(coll, bench) \
+       for (bench = coll->benchmarks; bench->name; bench++)
+
+static void dump_benchmarks(struct collection *coll)
 {
-       int i;
+       struct bench *bench;
 
-       printf("# List of available suites for %s...\n\n",
-              subsystems[subsys_index].name);
+       printf("\n        # List of available benchmarks for collection '%s':\n\n", coll->name);
 
-       for (i = 0; subsystems[subsys_index].suites[i].name; i++)
-               printf("%14s: %s\n",
-                      subsystems[subsys_index].suites[i].name,
-                      subsystems[subsys_index].suites[i].summary);
+       for_each_bench(coll, bench)
+               printf("%14s: %s\n", bench->name, bench->summary);
 
        printf("\n");
-       return;
 }
 
 static const char *bench_format_str;
+
+/* Output/formatting style, exported to benchmark modules: */
 int bench_format = BENCH_FORMAT_DEFAULT;
 
 static const struct option bench_options[] = {
-       OPT_STRING('f', "format", &bench_format_str, "default",
-                   "Specify format style"),
+       OPT_STRING('f', "format", &bench_format_str, "default", "Specify format style"),
        OPT_END()
 };
 
 static const char * const bench_usage[] = {
-       "perf bench [<common options>] <subsystem> <suite> [<options>]",
+       "perf bench [<common options>] <collection> <benchmark> [<options>]",
        NULL
 };
 
 static void print_usage(void)
 {
+       struct collection *coll;
        int i;
 
        printf("Usage: \n");
@@ -138,11 +115,10 @@ static void print_usage(void)
                printf("\t%s\n", bench_usage[i]);
        printf("\n");
 
-       printf("# List of available subsystems...\n\n");
+       printf("        # List of all available benchmark collections:\n\n");
 
-       for (i = 0; subsystems[i].name; i++)
-               printf("%14s: %s\n",
-                      subsystems[i].name, subsystems[i].summary);
+       for_each_collection(coll)
+               printf("%14s: %s\n", coll->name, coll->summary);
        printf("\n");
 }
 
@@ -159,44 +135,74 @@ static int bench_str2int(const char *str)
        return BENCH_FORMAT_UNKNOWN;
 }
 
-static void all_suite(struct bench_subsys *subsys)       /* FROM HERE */
+/*
+ * Run a specific benchmark but first rename the running task's ->comm[]
+ * to something meaningful:
+ */
+static int run_bench(const char *coll_name, const char *bench_name, bench_fn_t fn,
+                    int argc, const char **argv, const char *prefix)
 {
-       int i;
+       int size;
+       char *name;
+       int ret;
+
+       size = strlen(coll_name) + 1 + strlen(bench_name) + 1;
+
+       name = zalloc(size);
+       BUG_ON(!name);
+
+       scnprintf(name, size, "%s-%s", coll_name, bench_name);
+
+       prctl(PR_SET_NAME, name);
+       argv[0] = name;
+
+       ret = fn(argc, argv, prefix);
+
+       free(name);
+
+       return ret;
+}
+
+static void run_collection(struct collection *coll)
+{
+       struct bench *bench;
        const char *argv[2];
-       struct bench_suite *suites = subsys->suites;
 
        argv[1] = NULL;
        /*
         * TODO:
-        * preparing preset parameters for
+        *
+        * Preparing preset parameters for
         * embedded, ordinary PC, HPC, etc...
-        * will be helpful
+        * would be helpful.
         */
-       for (i = 0; suites[i].fn; i++) {
-               printf("# Running %s/%s benchmark...\n",
-                      subsys->name,
-                      suites[i].name);
+       for_each_bench(coll, bench) {
+               if (!bench->fn)
+                       break;
+               printf("# Running %s/%s benchmark...\n", coll->name, bench->name);
                fflush(stdout);
 
-               argv[1] = suites[i].name;
-               suites[i].fn(1, argv, NULL);
+               argv[1] = bench->name;
+               run_bench(coll->name, bench->name, bench->fn, 1, argv, NULL);
                printf("\n");
        }
 }
 
-static void all_subsystem(void)
+static void run_all_collections(void)
 {
-       int i;
-       for (i = 0; subsystems[i].suites; i++)
-               all_suite(&subsystems[i]);
+       struct collection *coll;
+
+       for_each_collection(coll)
+               run_collection(coll);
 }
 
 int cmd_bench(int argc, const char **argv, const char *prefix __maybe_unused)
 {
-       int i, j, status = 0;
+       struct collection *coll;
+       int ret = 0;
 
        if (argc < 2) {
-               /* No subsystem specified. */
+               /* No collection specified. */
                print_usage();
                goto end;
        }
@@ -206,7 +212,7 @@ int cmd_bench(int argc, const char **argv, const char *prefix __maybe_unused)
 
        bench_format = bench_str2int(bench_format_str);
        if (bench_format == BENCH_FORMAT_UNKNOWN) {
-               printf("Unknown format descriptor:%s\n", bench_format_str);
+               printf("Unknown format descriptor: '%s'\n", bench_format_str);
                goto end;
        }
 
@@ -216,52 +222,51 @@ int cmd_bench(int argc, const char **argv, const char *prefix __maybe_unused)
        }
 
        if (!strcmp(argv[0], "all")) {
-               all_subsystem();
+               run_all_collections();
                goto end;
        }
 
-       for (i = 0; subsystems[i].name; i++) {
-               if (strcmp(subsystems[i].name, argv[0]))
+       for_each_collection(coll) {
+               struct bench *bench;
+
+               if (strcmp(coll->name, argv[0]))
                        continue;
 
                if (argc < 2) {
-                       /* No suite specified. */
-                       dump_suites(i);
+                       /* No bench specified. */
+                       dump_benchmarks(coll);
                        goto end;
                }
 
                if (!strcmp(argv[1], "all")) {
-                       all_suite(&subsystems[i]);
+                       run_collection(coll);
                        goto end;
                }
 
-               for (j = 0; subsystems[i].suites[j].name; j++) {
-                       if (strcmp(subsystems[i].suites[j].name, argv[1]))
+               for_each_bench(coll, bench) {
+                       if (strcmp(bench->name, argv[1]))
                                continue;
 
                        if (bench_format == BENCH_FORMAT_DEFAULT)
-                               printf("# Running %s/%s benchmark...\n",
-                                      subsystems[i].name,
-                                      subsystems[i].suites[j].name);
+                               printf("# Running '%s/%s' benchmark:\n", coll->name, bench->name);
                        fflush(stdout);
-                       status = subsystems[i].suites[j].fn(argc - 1,
-                                                           argv + 1, prefix);
+                       ret = run_bench(coll->name, bench->name, bench->fn, argc-1, argv+1, prefix);
                        goto end;
                }
 
                if (!strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) {
-                       dump_suites(i);
+                       dump_benchmarks(coll);
                        goto end;
                }
 
-               printf("Unknown suite:%s for %s\n", argv[1], argv[0]);
-               status = 1;
+               printf("Unknown benchmark: '%s' for collection '%s'\n", argv[1], argv[0]);
+               ret = 1;
                goto end;
        }
 
-       printf("Unknown subsystem:%s\n", argv[0]);
-       status = 1;
+       printf("Unknown collection: '%s'\n", argv[0]);
+       ret = 1;
 
 end:
-       return status;
+       return ret;
 }