perf/core improvements and fixes
From: Arnaldo Carvalho de Melo <[email protected]>
Hi Ingo,
Please consider pulling,
- Arnaldo
The following changes since commit aa30a2e03a453aad9fd96c3f2d4a82c3497674e5:
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2013-10-23 09:45:50 +0200)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux tags/perf-core-for-mingo
for you to fetch changes up to c1fb5651bb40f9efaf32d280f39e06df7e352673:
perf tools: Show progress on histogram collapsing (2013-10-23 15:48:24 -0300)
----------------------------------------------------------------
perf/core improvements and fixes:
. Show progress on histogram collapsing, that can take a long time, from
Namhyung Kim.
. Support "$vars" meta argument syntax for local variables, allowing
asking for all possible variables at a given probe point to be collected when it hits, from Masami Hiramatsu.
. Address the root cause of that 'perf sched' stack initialization build
slowdown, by programmatically setting a big array after moving the global variable back to the stack. Fix from Adrian Hunter.
. Do not repipe attributes to a perf.data file in 'perf inject',
fix from Adrian Hunter
. Change the procps visible command-name of invididual benchmark tests
plus cleanups, from Ingo Molnar.
. Do not accept parse_tag_value() overflow, fix from Adrian Hunter.
. Validate that mmap_pages is not too big. From Adrian Hunter.
. Fix non-debug build, from Adrian Hunter
. Clarify the "sample parsing" test entry.
. Consider PERF_SAMPLE_TRANSACTION in the "sample parsing" test.
----------------------------------------------------------------
Adrian Hunter (7):
perf sched: Make struct perf_sched sched a local variable perf sched: Optimize build time
perf script: Make perf_script a local variable perf inject: Do not repipe attributes to a perf.data file perf tools: Do not accept parse_tag_value() overflow perf evlist: Validate that mmap_pages is not too big perf tools: Fix non-debug build
Arnaldo Carvalho de Melo (5):
perf test: Clarify the "sample parsing" test entry
perf test: Consider PERF_SAMPLE_TRANSACTION in the "sample parsing" test perf tools: Stop using 'self' in some more places
perf ui: Rename ui_progress to ui_progress_ops
perf ui progress: Per progress bar state
Ingo Molnar (1):
perf bench: Change the procps visible command-name of invididual benchmark tests plus cleanups
Masami Hiramatsu (2):
perf probe: Support "$vars" meta argument syntax for local variables perf probe: Find fentry mcount fuzzed parameter location
Namhyung Kim (1):
perf tools: Show progress on histogram collapsing
tools/perf/Makefile.perf | 1 + tools/perf/builtin-annotate.c | 6 +- tools/perf/builtin-bench.c | 239 +++++++++++++++++++------------------- tools/perf/builtin-diff.c | 7 +- tools/perf/builtin-inject.c | 27 +++-- tools/perf/builtin-report.c | 24 ++-- tools/perf/builtin-sched.c | 44 +++---- tools/perf/builtin-script.c | 40 ++++--- tools/perf/builtin-top.c | 4 +- tools/perf/config/Makefile | 4 + tools/perf/tests/hists_link.c | 2 +- tools/perf/tests/sample-parsing.c | 4 +- tools/perf/ui/gtk/gtk.h | 2 +- tools/perf/ui/gtk/progress.c | 20 ++-- tools/perf/ui/gtk/setup.c | 2 +- tools/perf/ui/progress.c | 32 +++-- tools/perf/ui/progress.h | 19 +-- tools/perf/ui/tui/progress.c | 15 +-- tools/perf/ui/tui/setup.c | 3 +- tools/perf/ui/tui/tui.h | 6 + tools/perf/util/build-id.c | 6 +- tools/perf/util/evlist.c | 14 ++- tools/perf/util/hist.c | 23 ++-- tools/perf/util/hist.h | 3 +- tools/perf/util/probe-event.c | 1 - tools/perf/util/probe-finder.c | 133 ++++++++++++++++++--- tools/perf/util/probe-finder.h | 1 + tools/perf/util/session.c | 24 ++-- tools/perf/util/sort.c | 124 ++++++++++---------- tools/perf/util/strfilter.c | 46 ++++---- tools/perf/util/thread.c | 72 ++++++------ tools/perf/util/util.c | 2 + 32 files changed, 560 insertions(+), 390 deletions(-) create mode 100644 tools/perf/ui/tui/tui.h
PATCH 1/16 perf test: Clarify the "sample parsing" test entry
From: Arnaldo Carvalho de Melo <[email protected]>
Before:
[root@sandy ~]# perf test -v 22
22: Test sample parsing :
--- start ---
sample format has changed - test needs updating ---- end ----
Test sample parsing: FAILED! [root@sandy ~]#
After:
[root@sandy ~]# perf test -v 22
22: Test sample parsing :
--- start ---
sample format has changed, some new PERF_SAMPLE_ bit was introduced - test needs updating ---- end ----
Test sample parsing: FAILED!
[root@sandy ~]#
--- tools/perf/tests/sample-parsing.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/tests/sample-parsing.c +1 -1
@@ -276,7 +276,7 @@ int test__sample_parsing(void)* were added. */ if (PERF_SAMPLE_MAX >PERF_SAMPLE_IDENTIFIER <<1 ) { -pr_debug ("sample format has changed - test needs updating\n" ); +pr_debug ("sample format has changed, some new PERF_SAMPLE_ bit was introduced - test needs updating\n" );return -1 ; }
Signature · 1 line
1.8.1.4
PATCH 2/16 perf test: Consider PERF_SAMPLE_TRANSACTION in the "sample parsing" test
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
[root@sandy ~]# perf test -v 22
22: Test sample parsing :
--- start ---
sample format has changed, some new PERF_SAMPLE_ bit was introduced - test needs updating ---- end ----
Test sample parsing: FAILED!
[root@sandy ~]#
--- tools/perf/tests/sample-parsing.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/tests/sample-parsing.c +1 -1
@@ -275,7 +275,7 @@ int test__sample_parsing(void)* Fail the test if it has not been updated when new sample format bits * were added. */ -if (PERF_SAMPLE_MAX >PERF_SAMPLE_IDENTIFIER <<1 ) { +if (PERF_SAMPLE_MAX >PERF_SAMPLE_TRANSACTION <<1 ) {pr_debug ("sample format has changed, some new PERF_SAMPLE_ bit was introduced - test needs updating\n" );return -1 ; }
Signature · 1 line
1.8.1.4
PATCH 3/16 perf tools: Stop using 'self' in some more places
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
As suggested by tglx, 'self' should be replaced by something that is more useful.
--- tools/perf/builtin-annotate.c | 4 +- tools/perf/builtin-diff.c | 5 +- tools/perf/builtin-inject.c | 22 ++++---- tools/perf/builtin-report.c | 14 ++--- tools/perf/util/build-id.c | 6 +- tools/perf/util/hist.c | 18 +++--- tools/perf/util/sort.c | 124 +++++++++++++++++++++--------------------- tools/perf/util/strfilter.c | 46 ++++++++-------- tools/perf/util/thread.c | 72 ++++++++++++------------ 9 files changed, 155 insertions(+), 156 deletions(-)
tools/perf/builtin-annotate.c +2 -2
@@ -118,11 +118,11 @@ static int hist_entry__tty_annotate(struct hist_entry *he, ann->print_line, ann->full_paths, 0, 0); } -static void hists__find_annotations (struct hists *self, +static void hists__find_annotations (struct hists *hists,struct perf_evsel *evsel,struct perf_annotate *ann) { -struct rb_node *nd =rb_first (&self->entries), *next; +struct rb_node *nd =rb_first (&hists->entries), *next;int key =K_RIGHT ;while (nd) {
tools/perf/builtin-diff.c +2 -3
@@ -303,12 +303,11 @@ static int formula_fprintf(struct hist_entry *he, struct hist_entry *pair,return -1 ; } -static int hists__add_entry (struct hists *self, +static int hists__add_entry (struct hists *hists,struct addr_location *al,u64 period,u64 weight,u64 transaction) { -if (__hists__add_entry (self, al,NULL , period, weight, transaction) - !=NULL ) +if (__hists__add_entry (hists, al,NULL , period, weight, transaction) !=NULL )return 0 ;return -ENOMEM ; }
tools/perf/builtin-inject.c +11 -11
@@ -162,38 +162,38 @@ static int perf_event__repipe_tracing_data(struct perf_tool *tool,return err; } -static int dso__read_build_id (struct dso *self) +static int dso__read_build_id (struct dso *dso) { -if (self->has_build_id) +if (dso->has_build_id)return 0 ; -if (filename__read_build_id (self->long_name, self->build_id, -sizeof (self->build_id)) >0 ) { - self->has_build_id = true; +if (filename__read_build_id (dso->long_name, dso->build_id, +sizeof (dso->build_id)) >0 ) { + dso->has_build_id = true;return 0 ; }return -1 ; } -static int dso__inject_build_id (struct dso *self,struct perf_tool *tool, +static int dso__inject_build_id (struct dso *dso,struct perf_tool *tool,struct machine *machine) {u16 misc =PERF_RECORD_MISC_USER ;int err; -if (dso__read_build_id (self) <0 ) { -pr_debug ("no build_id found for %s\n" , self->long_name); +if (dso__read_build_id (dso) <0 ) { +pr_debug ("no build_id found for %s\n" , dso->long_name);return -1 ; } -if (self->kernel) +if (dso->kernel) misc =PERF_RECORD_MISC_KERNEL ; - err =perf_event__synthesize_build_id (tool, self, misc, perf_event__repipe, + err =perf_event__synthesize_build_id (tool, dso, misc, perf_event__repipe, machine);if (err) { -pr_err ("Can't synthesize build_id event for %s\n" , self->long_name); +pr_err ("Can't synthesize build_id event for %s\n" , dso->long_name);return -1 ; }
tools/perf/builtin-report.c +7 -7
@@ -373,9 +373,9 @@ static int process_read_event(struct perf_tool *tool,/* For pipe mode, sample_type is not currently set */ static int perf_report__setup_sample_type (struct perf_report *rep) { -struct perf_session *self = rep->session; -u64 sample_type =perf_evlist__combined_sample_type (self->evlist); -bool is_pipe =perf_data_file__is_pipe (self->file); +struct perf_session *session = rep->session; +u64 sample_type =perf_evlist__combined_sample_type (session->evlist); +bool is_pipe =perf_data_file__is_pipe (session->file);if (!is_pipe && !(sample_type &PERF_SAMPLE_CALLCHAIN )) {if (sort__has_parent) {
@@ -417,14 +417,14 @@ static void sig_handler(int sig __maybe_unused) }static size_t hists__fprintf_nr_sample_events (struct perf_report *rep, -struct hists *self, +struct hists *hists,const char *evname,FILE *fp) {size_t ret;char unit; -unsigned long nr_samples = self->stats.nr_events[PERF_RECORD_SAMPLE ]; -u64 nr_events = self->stats.total_period; -struct perf_evsel *evsel =hists_to_evsel (self); +unsigned long nr_samples = hists->stats.nr_events[PERF_RECORD_SAMPLE ]; +u64 nr_events = hists->stats.total_period; +struct perf_evsel *evsel =hists_to_evsel (hists);char buf[512 ];size_t size =sizeof (buf);
tools/perf/util/build-id.c +3 -3
@@ -89,14 +89,14 @@ int build_id__sprintf(const u8 *build_id, int len, char *bf)return raw - build_id; } -char *dso__build_id_filename (struct dso *self,char *bf,size_t size) +char *dso__build_id_filename (struct dso *dso,char *bf,size_t size) {char build_id_hex[BUILD_ID_SIZE *2 +1 ]; -if (!self->has_build_id) +if (!dso->has_build_id)return NULL ; -build_id__sprintf (self->build_id,sizeof (self->build_id), build_id_hex); +build_id__sprintf (dso->build_id,sizeof (dso->build_id), build_id_hex);if (bf ==NULL ) {if (asprintf (&bf,"%s/.build-id/%.2s/%s" , buildid_dir, build_id_hex, build_id_hex +2 ) <0 )
tools/perf/util/hist.c +9 -9
@@ -406,7 +406,7 @@ out:return he; } -struct hist_entry *__hists__add_mem_entry(struct hists *self, +struct hist_entry *__hists__add_mem_entry(struct hists *hists, struct addr_location *al, struct symbol *sym_parent, struct mem_info *mi,
@@ -429,14 +429,14 @@ struct hist_entry *__hists__add_mem_entry(struct hists *self, .level = al->level, .parent = sym_parent, .filtered = symbol__parent_filter(sym_parent), - .hists = self, + .hists = hists, .mem_info = mi, .branch_info = NULL, }; -return add_hist_entry (self, &entry, al, period, weight); +return add_hist_entry (hists, &entry, al, period, weight); } -struct hist_entry *__hists__add_branch_entry(struct hists *self, +struct hist_entry *__hists__add_branch_entry(struct hists *hists, struct addr_location *al, struct symbol *sym_parent, struct branch_info *bi,
@@ -460,14 +460,14 @@ struct hist_entry *__hists__add_branch_entry(struct hists *self, .parent = sym_parent, .filtered = symbol__parent_filter(sym_parent), .branch_info = bi, - .hists = self, + .hists = hists, .mem_info = NULL, }; -return add_hist_entry (self, &entry, al, period, weight); +return add_hist_entry (hists, &entry, al, period, weight); } -struct hist_entry *__hists__add_entry (struct hists *self, +struct hist_entry *__hists__add_entry (struct hists *hists,struct addr_location *al,struct symbol *sym_parent,u64 period,u64 weight,u64 transaction)
@@ -488,13 +488,13 @@ struct hist_entry *__hists__add_entry(struct hists *self, }, .parent = sym_parent, .filtered = symbol__parent_filter(sym_parent), - .hists = self, + .hists = hists, .branch_info = NULL, .mem_info = NULL, .transaction = transaction, }; -return add_hist_entry (self, &entry, al, period, weight); +return add_hist_entry (hists, &entry, al, period, weight); }int64_t
tools/perf/util/sort.c +62 -62
349 lines not on this page · raw message
tools/perf/util/strfilter.c +23 -23
75 lines not on this page · raw message
tools/perf/util/thread.c +36 -36
117 lines not on this page · raw message
Signature · 1 line
1.8.1.4
PATCH 4/16 perf probe: Support "$vars" meta argument syntax for local variables
In reply to Arnaldo Carvalho de MeloFrom: Masami Hiramatsu <[email protected]>
Support "$vars" meta argument syntax for tracing all local variables at probe point.
Now you can trace all available local variables (including function parameters) at the probe point by passing $vars.
# perf probe --add foo $vars
This automatically finds all local variables at foo() and adds it as probe arguments.
--- tools/perf/util/probe-event.c | 1 - tools/perf/util/probe-finder.c | 96 ++++++++++++++++++++++++++++++++++++++---- tools/perf/util/probe-finder.h | 1 + 3 files changed, 89 insertions(+), 9 deletions(-)
tools/perf/util/probe-event.c +0 -1
@@ -47,7 +47,6 @@#include "session.h" #define MAX_CMDLEN 256 -#define MAX_PROBE_ARGS 128#define PERFPROBE_GROUP "probe"bool probe_event_dry_run;/* Dry run flag */
tools/perf/util/probe-finder.c +88 -8
121 lines not on this page · raw message
tools/perf/util/probe-finder.h +1 -0
@@ -7,6 +7,7 @@#define MAX_PROBE_BUFFER 1024#define MAX_PROBES 128 +#define MAX_PROBE_ARGS 128static inline int is_c_varname (const char *name) {
Signature · 1 line
1.8.1.4
PATCH 5/16 perf probe: Find fentry mcount fuzzed parameter location
In reply to Arnaldo Carvalho de MeloFrom: Masami Hiramatsu <[email protected]>
At this point, --fentry (mcount function entry) option for gcc fuzzes the debuginfo variable locations by skipping the mcount instruction offset (on x86, this is a 5 byte call instruction).
This makes variable searching fail at the entry of functions which are mcount'ed.
e.g.)
Available variables at vfs_read
@<vfs_read+0>
(No matched variables)
This patch adds additional location search at the function entry point to solve this issue, which tries to find the earliest address for the variable location.
Note that this only works with function parameters (formal parameters) because any local variables should not exist on the function entry address (those are not initialized yet).
With this patch, perf probe shows correct parameters if possible;
# perf probe --vars vfs_read
Available variables at vfs_read
@<vfs_read+0>
char* buf
loff_t* pos
size_t count
struct file* file--- tools/perf/util/probe-finder.c | 39 +++++++++++++++++++++++++++++++-------- 1 file changed, 31 insertions(+), 8 deletions(-)
tools/perf/util/probe-finder.c +31 -8
@@ -273,12 +273,15 @@ static struct probe_trace_arg_ref *alloc_trace_arg_ref(long offs)/* * Convert a location into trace_arg. * If tvar == NULL, this just checks variable can be converted. +* If fentry == true and vr_die is a parameter, do huristic search +* for the location fuzzed by function entry mcount. */ static int convert_variable_location (Dwarf_Die *vr_die,Dwarf_Addr addr, -Dwarf_Op *fb_ops, +Dwarf_Op *fb_ops,Dwarf_Die *sp_die,struct probe_trace_arg *tvar) {Dwarf_Attribute attr; +Dwarf_Addr tmp =0 ;Dwarf_Op *op;size_t nops;unsigned int regn;
@@ -291,12 +294,29 @@ static int convert_variable_location(Dwarf_Die *vr_die, Dwarf_Addr addr, goto static_var;/* TODO: handle more than 1 exprs */ -if (dwarf_attr (vr_die, DW_AT_location, &attr) ==NULL || -dwarf_getlocation_addr (&attr, addr, &op, &nops,1 ) <=0 || - nops ==0 ) { -/* TODO: Support const_value */ +if (dwarf_attr (vr_die, DW_AT_location, &attr) ==NULL ) +return -EINVAL ;/* Broken DIE ? */ +if (dwarf_getlocation_addr (&attr, addr, &op, &nops,1 ) <=0 ) { + ret =dwarf_entrypc (sp_die, &tmp); +if (ret || addr != tmp || +dwarf_tag (vr_die) != DW_TAG_formal_parameter || +dwarf_highpc (sp_die, &tmp)) +return -ENOENT ; +/* +* This is fuzzed by fentry mcount. We try to find the +* parameter location at the earliest address. +*/ +for (addr +=1 ; addr <= tmp; addr++) { +if (dwarf_getlocation_addr (&attr, addr, &op, + &nops,1 ) >0 ) + goto found; + }return -ENOENT ; } +found: +if (nops ==0 ) +/* TODO: Support const_value */ +return -ENOENT ;if (op->atom == DW_OP_addr) { static_var:
@@ -600,7 +620,7 @@ static int convert_variable(Dwarf_Die *vr_die, struct probe_finder *pf) dwarf_diename(vr_die)); ret =convert_variable_location (vr_die, pf->addr, pf->fb_ops, - pf->tvar); + &pf->sp_die, pf->tvar);if (ret == -ENOENT )pr_err ("Failed to find the location of %s at this address.\n" " Perhaps, it has been optimized out.\n" , pf->pvar->var);
@@ -1148,13 +1168,15 @@ struct local_vars_finder {static int copy_variables_cb (Dwarf_Die *die_mem,void *data) {struct local_vars_finder *vf = data; +struct probe_finder *pf = vf->pf;int tag; tag =dwarf_tag (die_mem);if (tag == DW_TAG_formal_parameter || tag == DW_TAG_variable) {if (convert_variable_location (die_mem, vf->pf->addr, - vf->pf->fb_ops,NULL ) ==0 ) { + vf->pf->fb_ops, &pf->sp_die, +NULL ) ==0 ) { vf->args[vf->nargs].var = (char *)dwarf_diename (die_mem);if (vf->args[vf->nargs].var ==NULL ) { vf->ret = -ENOMEM ;
@@ -1302,7 +1324,8 @@ static int collect_variables_cb(Dwarf_Die *die_mem, void *data)if (tag == DW_TAG_formal_parameter || tag == DW_TAG_variable) { ret =convert_variable_location (die_mem, af->pf.addr, - af->pf.fb_ops,NULL ); + af->pf.fb_ops, &af->pf.sp_die, +NULL );if (ret ==0 ) { ret =die_get_varname (die_mem, buf,MAX_VAR_LEN );pr_debug2 ("Add new var: %s\n" , buf);
Signature · 1 line
1.8.1.4
PATCH 6/16 perf bench: Change the procps visible command-name of invididual benchmark tests plus cleanups
In reply to Arnaldo Carvalho de MeloFrom: Ingo Molnar <[email protected]>
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 perfAfter 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-pipeThe 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.
--- tools/perf/builtin-bench.c | 239 +++++++++++++++++++++++---------------------- 1 file changed, 122 insertions(+), 117 deletions(-)
tools/perf/builtin-bench.c +122 -117
@@ -1,21 +1,18 @@/* -* * builtin-bench.c * -* General benchmarking subsystem provided by perf +* General benchmarking collections provided by perf * * Copyright (C) 2009, Hitoshi Mitake <[email protected]> -* */ /* +* 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"
@@ -25,112 +22,92 @@#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; }
Signature · 1 line
1.8.1.4
PATCH 7/16 perf sched: Make struct perf_sched sched a local variable
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
Change "struct perf_sched sched" from being global to being local.
The build slowdown cured by f36f83f947ed is dealt with in the following patch, by programatically setting perf_sched.curr_pid.
--- tools/perf/builtin-sched.c | 41 ++++++++++++++++++++--------------------- 1 file changed, 20 insertions(+), 21 deletions(-)
tools/perf/builtin-sched.c +20 -21
@@ -1655,29 +1655,28 @@ static int __cmd_record(int argc, const char **argv)return cmd_record (i, rec_argv,NULL ); } -static const char default_sort_order[] ="avg, max, switch, runtime" ; -static struct perf_sched sched = { - .tool = { - .sample = perf_sched__process_tracepoint_sample, - .comm = perf_event__process_comm, - .lost = perf_event__process_lost, - .fork = perf_sched__process_fork_event, - .ordered_samples = true, - }, - .cmp_pid =LIST_HEAD_INIT (sched.cmp_pid), - .sort_list =LIST_HEAD_INIT (sched.sort_list), - .start_work_mutex =PTHREAD_MUTEX_INITIALIZER , - .work_done_wait_mutex =PTHREAD_MUTEX_INITIALIZER , - .curr_pid = { [0 ...MAX_CPUS -1 ] =-1 }, - .sort_order = default_sort_order, - .replay_repeat =10 , - .profile_cpu =-1 , - .next_shortname1 ='A' , - .next_shortname2 ='0' , -}; -int cmd_sched (int argc,const char **argv,const char *prefix__maybe_unused ) { +const char default_sort_order[] ="avg, max, switch, runtime" ; +struct perf_sched sched = { + .tool = { + .sample = perf_sched__process_tracepoint_sample, + .comm = perf_event__process_comm, + .lost = perf_event__process_lost, + .fork = perf_sched__process_fork_event, + .ordered_samples = true, + }, + .cmp_pid =LIST_HEAD_INIT (sched.cmp_pid), + .sort_list =LIST_HEAD_INIT (sched.sort_list), + .start_work_mutex =PTHREAD_MUTEX_INITIALIZER , + .work_done_wait_mutex =PTHREAD_MUTEX_INITIALIZER , + .curr_pid = { [0 ...MAX_CPUS -1 ] =-1 }, + .sort_order = default_sort_order, + .replay_repeat =10 , + .profile_cpu =-1 , + .next_shortname1 ='A' , + .next_shortname2 ='0' , + };const struct option latency_options[] = {OPT_STRING ('s' ,"sort" , &sched.sort_order,"key[,key2...]" ,"sort by key(s): runtime, switch, avg, max" ),
Signature · 1 line
1.8.1.4
PATCH 8/16 perf sched: Optimize build time
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
builtin-sched.c took a log time to build with -O6 optimization. This turned out to be caused by:
.curr_pid = { [0 ... MAX_CPUS - 1] = -1 },Fix by initializing curr_pid programmatically.
This addresses the problem cured in f36f83f947ed using a smaller hammer.
--- tools/perf/builtin-sched.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-sched.c +4 -1
@@ -1670,7 +1670,6 @@ int cmd_sched(int argc, const char **argv, const char *prefix __maybe_unused) .sort_list = LIST_HEAD_INIT(sched.sort_list), .start_work_mutex = PTHREAD_MUTEX_INITIALIZER, .work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER, - .curr_pid = { [0 ... MAX_CPUS - 1] = -1 }, .sort_order = default_sort_order, .replay_repeat = 10, .profile_cpu = -1,
@@ -1732,6 +1731,10 @@ int cmd_sched(int argc, const char **argv, const char *prefix __maybe_unused) .switch_event = replay_switch_event, .fork_event = replay_fork_event, }; +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (sched.curr_pid); i++) + sched.curr_pid[i] =-1 ; argc =parse_options (argc, argv, sched_options, sched_usage,PARSE_OPT_STOP_AT_NON_OPTION );
Signature · 1 line
1.8.1.4
PATCH 9/16 perf script: Make perf_script a local variable
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
Change perf_script from being global to being local.
--- tools/perf/builtin-script.c | 40 ++++++++++++++++++++++++---------------- 1 file changed, 24 insertions(+), 16 deletions(-)
tools/perf/builtin-script.c +24 -16
@@ -542,18 +542,9 @@ static int process_sample_event(struct perf_tool *tool __maybe_unused,return 0 ; } -static struct perf_tool perf_script = { - .sample = process_sample_event, - .mmap = perf_event__process_mmap, - .mmap2 = perf_event__process_mmap2, - .comm = perf_event__process_comm, - .exit = perf_event__process_exit, - .fork = perf_event__process_fork, - .attr = perf_event__process_attr, - .tracing_data = perf_event__process_tracing_data, - .build_id = perf_event__process_build_id, - .ordered_samples = true, - .ordering_requires_timestamps = true, +struct perf_script { +struct perf_tool tool; +struct perf_session *session; };static void sig_handler (int sig__maybe_unused )
@@ -561,13 +552,13 @@ static void sig_handler(int sig __maybe_unused) session_done =1 ; } -static int __cmd_script (struct perf_session *session) +static int __cmd_script (struct perf_script *script) {int ret;signal (SIGINT , sig_handler); - ret =perf_session__process_events (session, &perf_script); + ret =perf_session__process_events (script->session, &script->tool);if (debug_mode)pr_err ("Misordered timestamps: %" PRIu64"\n" , nr_unordered);
@@ -1273,6 +1264,21 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)char *script_path =NULL ;const char **__argv;int i, j, err; +struct perf_script script = { + .tool = { + .sample = process_sample_event, + .mmap = perf_event__process_mmap, + .mmap2 = perf_event__process_mmap2, + .comm = perf_event__process_comm, + .exit = perf_event__process_exit, + .fork = perf_event__process_fork, + .attr = perf_event__process_attr, + .tracing_data = perf_event__process_tracing_data, + .build_id = perf_event__process_build_id, + .ordered_samples = true, + .ordering_requires_timestamps = true, + }, + };const struct option options[] = {OPT_BOOLEAN ('D' ,"dump-raw-trace" , &dump_trace,"dump raw trace in ASCII" ),
@@ -1498,10 +1504,12 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)if (!script_name)setup_pager (); - session =perf_session__new (&file, false, &perf_script); + session =perf_session__new (&file, false, &script.tool);if (session ==NULL )return -ENOMEM ; + script.session = session; +if (cpu_list) {if (perf_session__cpu_bitmap (session, cpu_list, cpu_bitmap))return -1 ;
@@ -1565,7 +1573,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)if (err <0 ) goto out; - err =__cmd_script (session); + err =__cmd_script (&script);perf_session__delete (session);cleanup_scripting ();
Signature · 1 line
1.8.1.4
PATCH 10/16 perf inject: Do not repipe attributes to a perf.data file
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
perf.data files contain the attributes separately, do not put them in the event stream as well.
--- tools/perf/builtin-inject.c | 5 +++++ 1 file changed, 5 insertions(+)
tools/perf/builtin-inject.c +5 -0
@@ -72,12 +72,17 @@ static int perf_event__repipe_attr(struct perf_tool *tool,union perf_event *event, struct perf_evlist **pevlist) { + struct perf_inject *inject = container_of(tool, struct perf_inject, + tool);int ret; ret =perf_event__process_attr (tool, event, pevlist);if (ret)return ret; +if (!inject->pipe_output) +return 0 ; +return perf_event__repipe_synth (tool, event); }
Signature · 1 line
1.8.1.4
PATCH 11/16 perf tools: Do not accept parse_tag_value() overflow
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
parse_tag_value() accepts an "unsigned long" and multiplies it according to a tag character. Do not accept the value if the multiplication overflows.
--- tools/perf/util/util.c | 2 ++ 1 file changed, 2 insertions(+)
tools/perf/util/util.c +2 -0
@@ -386,6 +386,8 @@ unsigned long parse_tag_value(const char *str, struct parse_tag *tags)if (s != endptr)break ; +if (value >ULONG_MAX / i->mult) +break ; value *= i->mult;return value; }
Signature · 1 line
1.8.1.4
PATCH 12/16 perf evlist: Validate that mmap_pages is not too big
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
Amend perf_evlist__parse_mmap_pages() to check that the mmap_pages entered via the --mmap_pages/-m option is not too big.
--- tools/perf/util/evlist.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)
tools/perf/util/evlist.c +10 -4
@@ -698,7 +698,8 @@ static size_t perf_evlist__mmap_size(unsigned long pages)int perf_evlist__parse_mmap_pages (const struct option *opt,const char *str,int unset__maybe_unused ) { -unsigned int pages, val, *mmap_pages = opt->value; +unsigned int *mmap_pages = opt->value; +unsigned long pages, val;size_t size;static struct parse_tag tags[] = { { .tag ='B' , .mult =1 },
@@ -709,12 +710,12 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str, }; val =parse_tag_value (str, tags); -if (val != (unsigned int )-1 ) { +if (val != (unsigned long )-1 ) {/* we got file size value */ pages =PERF_ALIGN (val, page_size) / page_size; -if (!is_power_of_2 (pages)) { +if (pages < (1UL <<31 ) && !is_power_of_2 (pages)) { pages =next_pow2 (pages); -pr_info ("rounding mmap pages size to %u (%u pages)\n" , +pr_info ("rounding mmap pages size to %lu (%lu pages)\n" , pages * page_size, pages); } }else {
@@ -727,6 +728,11 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str, } } +if (pages >UINT_MAX || pages >SIZE_MAX / page_size) { +pr_err ("--mmap_pages/-m value too big\n" ); +return -1 ; + } + size =perf_evlist__mmap_size (pages);if (!size) {pr_err ("--mmap_pages/-m value must be a power of two." );
Signature · 1 line
1.8.1.4
PATCH 13/16 perf tools: Fix non-debug build
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
In the absence of s DEBUG variable definition on the command line perf tools was building without optimization. Fix by assigning DEBUG if it is not defined.
--- tools/perf/config/Makefile | 4 ++++ 1 file changed, 4 insertions(+)
tools/perf/config/Makefile +4 -0
@@ -66,6 +66,10 @@ ifneq ($(WERROR),0) CFLAGS += -Werror endif +ifndef DEBUG + DEBUG := 0 +endif + ifeq ($(DEBUG),0) CFLAGS += -O6 endif
Signature · 1 line
1.8.1.4
PATCH 14/16 perf ui: Rename ui_progress to ui_progress_ops
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
Reserving 'struct ui_progress' to the per progress instances, not to the particular set of operations used to implmenet a progress bar in the current UI (GTK, TUI, etc).
--- tools/perf/Makefile.perf | 1 + tools/perf/ui/gtk/gtk.h | 2 +- tools/perf/ui/gtk/progress.c | 14 +++++++------- tools/perf/ui/gtk/setup.c | 2 +- tools/perf/ui/progress.c | 18 +++++++++--------- tools/perf/ui/progress.h | 6 ++---- tools/perf/ui/tui/progress.c | 7 ++++--- tools/perf/ui/tui/setup.c | 3 ++- tools/perf/ui/tui/tui.h | 6 ++++++ 9 files changed, 33 insertions(+), 26 deletions(-) create mode 100644 tools/perf/ui/tui/tui.h
tools/perf/Makefile.perf +1 -0
@@ -487,6 +487,7 @@ ifndef NO_SLANG LIB_OBJS += $(OUTPUT)ui/tui/util.o LIB_OBJS += $(OUTPUT)ui/tui/helpline.o LIB_OBJS += $(OUTPUT)ui/tui/progress.o + LIB_H += ui/tui/tui.h LIB_H += ui/browser.h LIB_H += ui/browsers/map.h LIB_H += ui/keysyms.h
tools/perf/ui/gtk/gtk.h +1 -1
@@ -34,7 +34,7 @@ struct perf_gtk_context *perf_gtk__activate_context(GtkWidget *window);int perf_gtk__deactivate_context (struct perf_gtk_context **ctx);void perf_gtk__init_helpline (void ); -void perf_gtk__init_progress (void ); +void gtk_ui_progress__init (void );void perf_gtk__init_hpp (void );void perf_gtk__signal (int sig);
tools/perf/ui/gtk/progress.c +7 -7
@@ -7,7 +7,7 @@static GtkWidget *dialog;static GtkWidget *progress; -static void gtk_progress_update (u64 curr,u64 total,const char *title) +static void gtk_ui_progress__update (u64 curr,u64 total,const char *title) {double fraction = total ?1.0 * curr / total :0.0 ;char buf[1024 ];
@@ -40,7 +40,7 @@ static void gtk_progress_update(u64 curr, u64 total, const char *title)gtk_main_iteration (); } -static void gtk_progress_finish (void ) +static void gtk_ui_progress__finish (void ) {/* this will also destroy all of its children */ gtk_widget_destroy (dialog);
@@ -48,12 +48,12 @@ static void gtk_progress_finish(void) dialog =NULL ; } -static struct ui_progress gtk_progress_fns = { - .update = gtk_progress_update, - .finish = gtk_progress_finish, +static struct ui_progress_ops gtk_ui_progress__ops = { + .update = gtk_ui_progress__update, + .finish = gtk_ui_progress__finish, }; -void perf_gtk__init_progress (void ) +void gtk_ui_progress__init (void ) { - progress_fns = >k_progress_fns; + ui_progress__ops = >k_ui_progress__ops; }
tools/perf/ui/gtk/setup.c +1 -1
@@ -8,7 +8,7 @@ int perf_gtk__init(void) {perf_error__register (&perf_gtk_eops);perf_gtk__init_helpline (); -perf_gtk__init_progress (); +gtk_ui_progress__init ();perf_gtk__init_hpp ();return gtk_init_check (NULL ,NULL ) ?0 :-1 ;
tools/perf/ui/progress.c +9 -9
@@ -1,26 +1,26 @@#include "../cache.h" #include "progress.h" -static void nop_progress_update (u64 curr__maybe_unused , -u64 total__maybe_unused , -const char *title__maybe_unused ) +static void null_progress__update (u64 curr__maybe_unused , +u64 total__maybe_unused , +const char *title__maybe_unused ) { } -static struct ui_progress default_progress_fns = +static struct ui_progress_ops null_progress__ops = { - .update = nop_progress_update, + .update = null_progress__update, }; -struct ui_progress *progress_fns = &default_progress_fns; +struct ui_progress_ops *ui_progress__ops = &null_progress__ops;void ui_progress__update (u64 curr,u64 total,const char *title) { -return progress_fns->update (curr, total, title); +return ui_progress__ops->update (curr, total, title); }void ui_progress__finish (void ) { -if (progress_fns->finish) - progress_fns->finish (); +if (ui_progress__ops->finish) + ui_progress__ops->finish (); }
tools/perf/ui/progress.h +2 -4
@@ -3,14 +3,12 @@#include <../types.h> -struct ui_progress { +struct ui_progress_ops {void (*update)(u64 ,u64 ,const char *);void (*finish)(void ); }; -extern struct ui_progress *progress_fns; - -void ui_progress__init (void ); +extern struct ui_progress_ops *ui_progress__ops;void ui_progress__update (u64 curr,u64 total,const char *title);void ui_progress__finish (void );
tools/perf/ui/tui/progress.c +4 -3
@@ -2,6 +2,7 @@#include "../progress.h" #include "../libslang.h" #include "../ui.h" +#include "tui.h" #include "../browser.h" static void tui_progress__update (u64 curr,u64 total,const char *title)
@@ -31,12 +32,12 @@ static void tui_progress__update(u64 curr, u64 total, const char *title)pthread_mutex_unlock (&ui__lock); } -static struct ui_progress tui_progress_fns = +static struct ui_progress_ops tui_progress__ops = { .update = tui_progress__update, }; -void ui_progress__init (void ) +void tui_progress__init (void ) { - progress_fns = &tui_progress_fns; + ui_progress__ops = &tui_progress__ops; }
tools/perf/ui/tui/setup.c +2 -1
@@ -9,6 +9,7 @@#include "../util.h" #include "../libslang.h" #include "../keysyms.h" +#include "tui.h" static volatile int ui__need_resize;
@@ -119,7 +120,7 @@ int ui__init(void)ui_helpline__init ();ui_browser__init (); -ui_progress__init (); +tui_progress__init ();signal (SIGSEGV , ui__signal);signal (SIGFPE , ui__signal);
tools/perf/ui/tui/tui.h new file +6 -0
@@ -0,0 +1,6 @@ +#ifndef _PERF_TUI_H_ +#define _PERF_TUI_H_ 1 + +void tui_progress__init (void ); + +#endif /* _PERF_TUI_H_ */
Signature · 1 line
1.8.1.4
PATCH 15/16 perf ui progress: Per progress bar state
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
That will ease using a progress bar across multiple functions, like in the upcoming patches that will present a progress bar when collapsing histograms.
Based on a previous patch by Namhyung Kim.
--- tools/perf/ui/gtk/progress.c | 8 ++++---- tools/perf/ui/progress.c | 22 +++++++++++++++++----- tools/perf/ui/progress.h | 15 +++++++++++---- tools/perf/ui/tui/progress.c | 8 ++++---- tools/perf/util/session.c | 24 +++++++++++------------- 5 files changed, 47 insertions(+), 30 deletions(-)
tools/perf/ui/gtk/progress.c +4 -4
@@ -7,14 +7,14 @@static GtkWidget *dialog;static GtkWidget *progress; -static void gtk_ui_progress__update (u64 curr,u64 total,const char *title) +static void gtk_ui_progress__update (struct ui_progress *p) { -double fraction = total ?1.0 * curr / total :0.0 ; +double fraction = p->total ?1.0 * p->curr / p->total :0.0 ;char buf[1024 ];if (dialog ==NULL ) {GtkWidget *vbox =gtk_vbox_new (TRUE,5 ); -GtkWidget *label =gtk_label_new (title); +GtkWidget *label =gtk_label_new (p->title); dialog =gtk_window_new (GTK_WINDOW_TOPLEVEL ); progress =gtk_progress_bar_new ();
@@ -32,7 +32,7 @@ static void gtk_ui_progress__update(u64 curr, u64 total, const char *title) }gtk_progress_bar_set_fraction (GTK_PROGRESS_BAR (progress), fraction); -snprintf (buf,sizeof (buf),"%" PRIu64" / %" PRIu64, curr, total); +snprintf (buf,sizeof (buf),"%" PRIu64" / %" PRIu64, p->curr, p->total);gtk_progress_bar_set_text (GTK_PROGRESS_BAR (progress), buf);/* we didn't call gtk_main yet, so do it manually */
tools/perf/ui/progress.c +17 -5
@@ -1,9 +1,7 @@#include "../cache.h" #include "progress.h" -static void null_progress__update (u64 curr__maybe_unused , -u64 total__maybe_unused , -const char *title__maybe_unused ) +static void null_progress__update (struct ui_progress *p__maybe_unused ) { }
@@ -14,9 +12,23 @@ static struct ui_progress_ops null_progress__ops =struct ui_progress_ops *ui_progress__ops = &null_progress__ops; -void ui_progress__update (u64 curr,u64 total,const char *title) +void ui_progress__update (struct ui_progress *p,u64 adv) { -return ui_progress__ops->update (curr, total, title); + p->curr += adv; + +if (p->curr >= p->next) { + p->next += p->step; + ui_progress__ops->update (p); + } +} + +void ui_progress__init (struct ui_progress *p,u64 total,const char *title) +{ + p->curr =0 ; + p->next = p->step = total /16 ; + p->total = total; + p->title = title; + }void ui_progress__finish (void )
tools/perf/ui/progress.h +11 -4
@@ -3,14 +3,21 @@#include <../types.h> +void ui_progress__finish (void ); + +struct ui_progress { +const char *title; +u64 curr, next, step, total; +}; + +void ui_progress__init (struct ui_progress *p,u64 total,const char *title); +void ui_progress__update (struct ui_progress *p,u64 adv); +struct ui_progress_ops { -void (*update)(u64 ,u64 ,const char *); +void (*update)(struct ui_progress *p);void (*finish)(void ); };extern struct ui_progress_ops *ui_progress__ops; -void ui_progress__update (u64 curr,u64 total,const char *title); -void ui_progress__finish (void ); -#endif
tools/perf/ui/tui/progress.c +4 -4
@@ -5,7 +5,7 @@#include "tui.h" #include "../browser.h" -static void tui_progress__update (u64 curr,u64 total,const char *title) +static void tui_progress__update (struct ui_progress *p) {int bar, y;/*
@@ -15,7 +15,7 @@ static void tui_progress__update(u64 curr, u64 total, const char *title)if (use_browser <=0 )return ; -if (total ==0 ) +if (p->total ==0 )return ;ui__refresh_dimensions (true);
@@ -24,9 +24,9 @@ static void tui_progress__update(u64 curr, u64 total, const char *title)SLsmg_set_color (0 );SLsmg_draw_box (y,0 ,3 , SLtt_Screen_Cols);SLsmg_gotorc (y++,1 ); -SLsmg_write_string ((char *)title); +SLsmg_write_string ((char *)p->title);SLsmg_set_color (HE_COLORSET_SELECTED ); - bar = ((SLtt_Screen_Cols -2 ) * curr) / total; + bar = ((SLtt_Screen_Cols -2 ) * p->curr) / p->total;SLsmg_fill_region (y,1 ,1 , bar,' ' );SLsmg_refresh ();pthread_mutex_unlock (&ui__lock);
tools/perf/util/session.c +11 -13
@@ -503,13 +503,16 @@ static int flush_sample_queue(struct perf_session *s,struct perf_sample sample;u64 limit = os->next_flush;u64 last_ts = os->last_sample ? os->last_sample->timestamp :0ULL ; -unsigned idx =0 , progress_next = os->nr_samples /16 ;bool show_progress = limit ==ULLONG_MAX ; +struct ui_progress prog;int ret;if (!tool->ordered_samples || !limit)return 0 ; +if (show_progress) +ui_progress__init (&prog, os->nr_samples,"Processing time ordered events..." ); +list_for_each_entry_safe (iter, tmp, head, list) {if (session_done ())return 0 ;
@@ -530,11 +533,9 @@ static int flush_sample_queue(struct perf_session *s, os->last_flush = iter->timestamp;list_del (&iter->list);list_add (&iter->list, &os->sample_cache); -if (show_progress && (++idx >= progress_next)) { - progress_next += os->nr_samples /16 ; -ui_progress__update (idx, os->nr_samples, -"Processing time ordered events..." ); - } + +if (show_progress) +ui_progress__update (&prog,1 ); }if (list_empty (head)) {
@@ -1285,12 +1286,13 @@ int __perf_session__process_events(struct perf_session *session, u64 file_size, struct perf_tool *tool) { int fd = perf_data_file__fd(session->file); -u64 head, page_offset, file_offset, file_pos, progress_next; +u64 head, page_offset, file_offset, file_pos;int err, mmap_prot, mmap_flags, map_idx =0 ;size_t mmap_size;char *buf, *mmaps[NUM_MMAPS ];union perf_event *event;uint32_t size; +struct ui_progress prog;perf_tool__fill_defaults (tool);
@@ -1301,7 +1303,7 @@ int __perf_session__process_events(struct perf_session *session,if (data_size && (data_offset + data_size < file_size)) file_size = data_offset + data_size; - progress_next = file_size /16 ; +ui_progress__init (&prog, file_size,"Processing events..." ); mmap_size =MMAP_SIZE ;if (mmap_size > file_size)
@@ -1356,11 +1358,7 @@ more: head += size; file_pos += size; -if (file_pos >= progress_next) { - progress_next += file_size /16 ; -ui_progress__update (file_pos, file_size, -"Processing events..." ); - } +ui_progress__update (&prog, size);if (session_done ()) goto out;
Signature · 1 line
1.8.1.4
PATCH 16/16 perf tools: Show progress on histogram collapsing
In reply to Arnaldo Carvalho de MeloFrom: Namhyung Kim <[email protected]>
It can take quite amount of time so add progress bar UI to inform user.
--- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-diff.c | 2 +- tools/perf/builtin-report.c | 10 +++++++++- tools/perf/builtin-top.c | 4 ++-- tools/perf/tests/hists_link.c | 2 +- tools/perf/util/hist.c | 5 ++++- tools/perf/util/hist.h | 3 ++- 7 files changed, 20 insertions(+), 8 deletions(-)
tools/perf/builtin-annotate.c +1 -1
@@ -247,7 +247,7 @@ static int __cmd_annotate(struct perf_annotate *ann)if (nr_samples >0 ) { total_nr_samples += nr_samples; -hists__collapse_resort (hists); +hists__collapse_resort (hists,NULL );hists__output_resort (hists); if (symbol_conf.event_group &&
tools/perf/builtin-diff.c +1 -1
@@ -369,7 +369,7 @@ static void perf_evlist__collapse_resort(struct perf_evlist *evlist)list_for_each_entry (evsel, &evlist->entries, node) {struct hists *hists = &evsel->hists; -hists__collapse_resort (hists); +hists__collapse_resort (hists,NULL ); } }
tools/perf/builtin-report.c +9 -1
@@ -496,6 +496,7 @@ static int __cmd_report(struct perf_report *rep)struct map *kernel_map;struct kmap *kernel_kmap;const char *help ="For a higher level overview, try: perf report --sort comm,dso" ; +struct ui_progress prog;struct perf_data_file *file = session->file;signal (SIGINT , sig_handler);
@@ -558,13 +559,19 @@ static int __cmd_report(struct perf_report *rep) } nr_samples =0 ; +list_for_each_entry (pos, &session->evlist->entries, node) + nr_samples += pos->hists.nr_entries; + +ui_progress__init (&prog, nr_samples,"Merging related events..." ); + + nr_samples =0 ;list_for_each_entry (pos, &session->evlist->entries, node) {struct hists *hists = &pos->hists;if (pos->idx ==0 ) hists->symbol_filter_str = rep->symbol_filter_str; -hists__collapse_resort (hists); +hists__collapse_resort (hists, &prog); nr_samples += hists->stats.nr_events[PERF_RECORD_SAMPLE ];/* Non-group events are considered as leader */
@@ -576,6 +583,7 @@ static int __cmd_report(struct perf_report *rep)hists__link (leader_hists, hists); } } +ui_progress__finish ();if (session_done ())return 0 ;
tools/perf/builtin-top.c +2 -2
@@ -286,7 +286,7 @@ static void perf_top__print_sym_table(struct perf_top *top)return ; } -hists__collapse_resort (&top->sym_evsel->hists); +hists__collapse_resort (&top->sym_evsel->hists,NULL );hists__output_resort (&top->sym_evsel->hists); hists__decay_entries(&top->sym_evsel->hists, top->hide_user_symbols,
@@ -552,7 +552,7 @@ static void perf_top__sort_new_samples(void *arg)if (t->evlist->selected !=NULL ) t->sym_evsel = t->evlist->selected; -hists__collapse_resort (&t->sym_evsel->hists); +hists__collapse_resort (&t->sym_evsel->hists,NULL );hists__output_resort (&t->sym_evsel->hists); hists__decay_entries(&t->sym_evsel->hists, t->hide_user_symbols,
tools/perf/tests/hists_link.c +1 -1
@@ -467,7 +467,7 @@ int test__hists_link(void) goto out;list_for_each_entry (evsel, &evlist->entries, node) { -hists__collapse_resort (&evsel->hists); +hists__collapse_resort (&evsel->hists,NULL );if (verbose >2 )print_hists (&evsel->hists);
tools/perf/util/hist.c +4 -1
@@ -399,6 +399,7 @@ static struct hist_entry *add_hist_entry(struct hists *hists,if (!he)return NULL ; + hists->nr_entries++;rb_link_node (&he->rb_node_in, parent, p);rb_insert_color (&he->rb_node_in, hists->entries_in); out:
@@ -604,7 +605,7 @@ static void hists__apply_filters(struct hists *hists, struct hist_entry *he)hists__filter_entry_by_symbol (hists, he); } -void hists__collapse_resort (struct hists *hists) +void hists__collapse_resort (struct hists *hists,struct ui_progress *prog) {struct rb_root *root;struct rb_node *next;
@@ -631,6 +632,8 @@ void hists__collapse_resort(struct hists *hists)*/ hists__apply_filters (hists, n); } +if (prog) +ui_progress__update (prog,1 ); } }
tools/perf/util/hist.h +2 -1
@@ -5,6 +5,7 @@#include <pthread.h> #include "callchain.h" #include "header.h" +#include "ui/progress.h" extern struct callchain_param callchain_param;
@@ -108,7 +109,7 @@ struct hist_entry *__hists__add_mem_entry(struct hists *self, u64 weight);void hists__output_resort (struct hists *self); -void hists__collapse_resort (struct hists *self); +void hists__collapse_resort (struct hists *self,struct ui_progress *prog);void hists__decay_entries (struct hists *hists,bool zap_user,bool zap_kernel);void hists__output_recalc_col_len (struct hists *hists,int max_rows);
Signature · 1 line
1.8.1.4
Quoted 106 lines · Arnaldo Carvalho de Melo
From: Arnaldo Carvalho de Melo <[email protected]>
Hi Ingo,
Please consider pulling,
- Arnaldo
The following changes since commit
aa30a2e03a453aad9fd96c3f2d4a82c3497674e5:Merge tag 'perf-core-for-mingo' of
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linuxinto perf/core (2013-10-23 09:45:50 +0200)are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linuxtags/perf-core-for-mingofor you to fetch changes up to
c1fb5651bb40f9efaf32d280f39e06df7e352673:perf tools: Show progress on histogram collapsing (2013-10-23 15:48:24 -0300)
----------------------------------------------------------------
perf/core improvements and fixes:. Show progress on histogram collapsing, that can take a long time, from
Namhyung Kim.. Support "$vars" meta argument syntax for local variables, allowing
asking for all possible variables at a given probe point to be collected when it hits, from Masami Hiramatsu.. Address the root cause of that 'perf sched' stack initialization build
slowdown, by programmatically setting a big array after moving the global variable back to the stack. Fix from Adrian Hunter.. Do not repipe attributes to a perf.data file in 'perf inject',
fix from Adrian Hunter. Change the procps visible command-name of invididual benchmark tests
plus cleanups, from Ingo Molnar.. Do not accept
parse_tag_value()overflow, fix from Adrian Hunter.. Validate that
mmap_pagesis not too big. From Adrian Hunter.. Fix non-debug build, from Adrian Hunter
. Clarify the "sample parsing" test entry.
. Consider
PERF_SAMPLE_TRANSACTIONin the "sample parsing" test.Signed-off-by Arnaldo Carvalho de Melo <[email protected]>----------------------------------------------------------------
Adrian Hunter (7):
perf sched: Make structperf_schedsched a local variable perf sched: Optimize build time
perf script: Makeperf_scripta local variable perf inject: Do not repipe attributes to a perf.data file perf tools: Do not acceptparse_tag_value()overflow perf evlist: Validate thatmmap_pagesis not too big perf tools: Fix non-debug buildArnaldo Carvalho de Melo (5):
perf test: Clarify the "sample parsing" test entry
perf test: ConsiderPERF_SAMPLE_TRANSACTIONin the "sample parsing" test perf tools: Stop using 'self' in some more places
perf ui: Renameui_progresstoui_progress_ops
perf ui progress: Per progress bar stateIngo Molnar (1):
perf bench: Change the procps visible command-name of invididual benchmark tests plus cleanupsMasami Hiramatsu (2):
perf probe: Support "$vars" meta argument syntax for local variables perf probe: Find fentry mcount fuzzed parameter locationNamhyung Kim (1):
perf tools: Show progress on histogram collapsingtools/perf/Makefile.perf | 1 + tools/perf/builtin-annotate.c | 6 +- tools/perf/builtin-bench.c | 239 +++++++++++++++++++------------------- tools/perf/builtin-diff.c | 7 +- tools/perf/builtin-inject.c | 27 +++-- tools/perf/builtin-report.c | 24 ++-- tools/perf/builtin-sched.c | 44 +++---- tools/perf/builtin-script.c | 40 ++++--- tools/perf/builtin-top.c | 4 +- tools/perf/config/Makefile | 4 + tools/perf/tests/hists_link.c | 2 +- tools/perf/tests/sample-parsing.c | 4 +- tools/perf/ui/gtk/gtk.h | 2 +- tools/perf/ui/gtk/progress.c | 20 ++-- tools/perf/ui/gtk/setup.c | 2 +- tools/perf/ui/progress.c | 32 +++-- tools/perf/ui/progress.h | 19 +-- tools/perf/ui/tui/progress.c | 15 +-- tools/perf/ui/tui/setup.c | 3 +- tools/perf/ui/tui/tui.h | 6 + tools/perf/util/build-id.c | 6 +- tools/perf/util/evlist.c | 14 ++- tools/perf/util/hist.c | 23 ++-- tools/perf/util/hist.h | 3 +- tools/perf/util/probe-event.c | 1 - tools/perf/util/probe-finder.c | 133 ++++++++++++++++++---
Quoted 8 lines
tools/perf/util/probe-finder.h | 1 + tools/perf/util/session.c | 24 ++-- tools/perf/util/sort.c | 124 ++++++++++---------- tools/perf/util/strfilter.c | 46 ++++---- tools/perf/util/thread.c | 72 ++++++------ tools/perf/util/util.c | 2 + 32 files changed, 560 insertions(+), 390 deletions(-) create mode 100644 tools/perf/ui/tui/tui.h
Pulled, thanks Arnaldo!
Ingo