perf tools: Add support for user defined metric
Jiri Olsa
reviewerv3 cover perf tools: Add support for user defined metric
hi,
Joe asked for possibility to add user defined metrics. Given that we already have metrics support, I added --metrics-file option that allows to specify custom metrics.
$ cat metrics # IPC mine1 = instructions / cycles; /* DECODED_ICACHE_UOPS% */ mine2 = 100 * (idq.dsb_uops / \ (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops));
$ sudo perf stat --metrics-file ./metrics -M mine1,mine2 --metric-only -a -I 1000
# time insn per cycle mine1 mine2
1.000536263 0.71 0.7 41.4
2.002069025 0.31 0.3 14.1
3.003427684 0.27 0.3 14.8
4.004807132 0.25 0.2 12.1
...v3 changes:
- added doc for metrics file in perf stat man page
- reporting error line number now
- changed '#' style comment to C way with '//'
v2 changes:
- add new --metrics-file option
- rebased on current perf/core expression bison/flex enhancements
Also available in:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/metric
thanks,
jirka
---
Jiri Olsa (4):
perf expr: Add parsing support for multiple expressions perf expr: Allow comments in custom metric file perf stat: Add --metrics-file option
perf expr: Report line number with error
tools/perf/Documentation/perf-stat.txt | 77 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 7 +++++-- tools/perf/tests/expr.c | 18 ++++++++++++++++++ tools/perf/util/expr.c | 6 ++++++ tools/perf/util/expr.h | 21 +++++++++++++++++++-- tools/perf/util/expr.l | 34 ++++++++++++++++++++++++++++++++++ tools/perf/util/expr.y | 21 +++++++++++++++++---- tools/perf/util/metricgroup.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------- tools/perf/util/metricgroup.h | 3 ++- tools/perf/util/stat.h | 1 + 10 files changed, 242 insertions(+), 16 deletions(-)
Jiri Olsa
reviewerPATCH v3 1/4 perf expr: Add parsing support for multiple expressions
Adding support to parse metric difinitions in following form:
NAME = EXPRESSION ; NAME = EXPRESSION ; ...
The parsed NAME and EXPRESSION will be used in following changes to feed the metric code with definitions from custom file.
--- tools/perf/tests/expr.c | 13 +++++++++++++ tools/perf/util/expr.c | 6 ++++++ tools/perf/util/expr.h | 19 +++++++++++++++++-- tools/perf/util/expr.l | 12 ++++++++++++ tools/perf/util/expr.y | 13 ++++++++++++- 5 files changed, 60 insertions(+), 3 deletions(-)
tools/perf/tests/expr.c +13 -0
@@ -71,5 +71,18 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)zfree (&other[i]);free ((void *)other); +expr__ctx_init (&ctx); + ret =expr__parse_custom (&ctx,"IPC=INSTRUCTIONS / CYCLES; CPI=CYCLES / INSTRUCTIONS;" ); +TEST_ASSERT_VAL ("parse custom failed" , ret ==0 ); +TEST_ASSERT_VAL ("parse custom count" , ctx.num_custom ==2 ); +TEST_ASSERT_VAL ("parse custom name" , !strcmp (ctx.custom[0 ].name,"IPC" )); +TEST_ASSERT_VAL ("parse custom name" , !strcmp (ctx.custom[1 ].name,"CPI" )); +TEST_ASSERT_VAL ("parse custom expr" , !strcmp (ctx.custom[0 ].expr,"INSTRUCTIONS / CYCLES" )); +TEST_ASSERT_VAL ("parse custom expr" , !strcmp (ctx.custom[1 ].expr,"CYCLES / INSTRUCTIONS" )); + +for (i =0 ; i < ctx.num_custom; i++) { +zfree (&ctx.custom[i].name); +zfree (&ctx.custom[i].expr); + }return 0 ; }
tools/perf/util/expr.c +6 -0
@@ -23,6 +23,7 @@ void expr__add_id(struct expr_parse_ctx *ctx, const char *name, double val)void expr__ctx_init (struct expr_parse_ctx *ctx) { ctx->num_ids =0 ; + ctx->num_custom =0 ; } static int
@@ -61,6 +62,11 @@ int expr__parse(double *final_val, struct expr_parse_ctx *ctx, const char *expr,return __expr__parse (final_val, ctx, expr,EXPR_PARSE , runtime) ?-1 :0 ; } +int expr__parse_custom (struct expr_parse_ctx *ctx,const char *expr) +{ +return __expr__parse (NULL , ctx, expr,EXPR_CUSTOM ,0 ); +} +static bool already_seen (const char *val,const char *one,const char **other,int num_other)
tools/perf/util/expr.h +17 -2
@@ -4,15 +4,29 @@#define EXPR_MAX_OTHER 64#define MAX_PARSE_ID EXPR_MAX_OTHER +#define EXPR_MAX 20struct expr_parse_id {const char *name;double val; }; +struct expr_parse_custom { +const char *name; +const char *expr; +}; +struct expr_parse_ctx { -int num_ids; -struct expr_parse_id ids[MAX_PARSE_ID ]; +union { +struct { +int num_ids; +struct expr_parse_id ids[MAX_PARSE_ID ]; + }; +struct { +int num_custom; +struct expr_parse_custom custom[EXPR_MAX ]; + }; + }; };struct expr_scanner_ctx {
@@ -23,6 +37,7 @@ struct expr_scanner_ctx {void expr__ctx_init (struct expr_parse_ctx *ctx);void expr__add_id (struct expr_parse_ctx *ctx,const char *id,double val);int expr__parse (double *final_val,struct expr_parse_ctx *ctx,const char *expr,int runtime); +int expr__parse_custom (struct expr_parse_ctx *ctx,const char *expr);int expr__find_other (const char *expr,const char *one,const char ***other,int *num_other,int runtime);
tools/perf/util/expr.l +12 -0
@@ -81,12 +81,15 @@ static int str(yyscan_t scanner, int token, int runtime) } %} +%x custom + number [0-9]*\.?[0-9]+ sch [-,=] spec \\{sch} sym [0-9a-zA-Z_\.:@?]+ symbol ({spec}|{sym})+ +all [^;]+ %% struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);
@@ -100,6 +103,12 @@ symbol ({spec}|{sym})+ } } +<custom>{ + +{all} { BEGIN(INITIAL); return str(yyscanner, ALL, sctx->runtime); } + +} + max { return MAX; } min { return MIN; } if { return IF; }
@@ -118,6 +127,9 @@ else { return ELSE; } "(" { return '('; } ")" { return ')'; } "," { return ','; } +";" { return ';'; } +"=" { BEGIN(custom); return '='; } +\n { } . { } %%
tools/perf/util/expr.y +12 -1
@@ -24,9 +24,10 @@ char *str; } -%token EXPR_PARSE EXPR_OTHER EXPR_ERROR +%token EXPR_PARSE EXPR_OTHER EXPR_CUSTOM EXPR_ERROR %token <num> NUMBER %token <str> ID +%token <str> ALL %token MIN MAX IF ELSE SMT_ON %left MIN MAX IF %left '|'
@@ -66,6 +67,16 @@ start: EXPR_PARSE all_expr | EXPR_OTHER all_other +| +EXPR_CUSTOM all_custom + +all_custom: all_custom ID '=' ALL ';' +{ + ctx->custom[ctx->num_custom].name = $2; + ctx->custom[ctx->num_custom].expr = $4; + ctx->num_custom++; +} +| all_other: all_other other |
Signature · 1 line
2.25.4
Jiri Olsa
reviewerPATCH v3 2/4 perf expr: Allow comments in custom metric file
In reply to Jiri OlsaAdding support to use comments within the custom metric file with both the shell '# ... ' and C '/* ... */ styles.
--- tools/perf/util/expr.l | 12 ++++++++++++ 1 file changed, 12 insertions(+)
tools/perf/util/expr.l +12 -0
@@ -82,6 +82,8 @@ static int str(yyscan_t scanner, int token, int runtime) %} %x custom +%x comment_single +%x comment_multi number [0-9]*\.?[0-9]+
@@ -116,6 +118,16 @@ else { return ELSE; } #smt_on { return SMT_ON; } {number} { return value(yyscanner); } {symbol} { return str(yyscanner, ID, sctx->runtime); } + +"//" { BEGIN(comment_single); } +<comment_single>\n { BEGIN(INITIAL); } +<comment_single>. { } + +"/*" { BEGIN(comment_multi); } +<comment_multi>"*/" { BEGIN(INITIAL); } +<comment_multi>\n { } +<comment_multi>. { } + "|" { return '|'; } "^" { return '^'; } "&" { return '&'; }
Signature · 1 line
2.25.4
Jiri Olsa
reviewerPATCH v3 3/4 perf stat: Add --metrics-file option
In reply to Jiri OlsaAdding --metrics-file option that allows to specify metrics in the file.
It's now possible to define metrics in file and use them like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;
/* DECODED_ICACHE_UOPS% */
mine2 = 100 * (idq.dsb_uops / (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops));
$ sudo perf stat --metrics-file ./metrics -M mine1,mine2 -a -I 1000 --metric-only
# time mine1 mine2
1.000997184 0.41 18.47
2.002479737 0.57 22.46
3.003932935 0.40 17.52
...--- tools/perf/Documentation/perf-stat.txt | 77 ++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 7 ++- tools/perf/util/metricgroup.c | 69 ++++++++++++++++++++--- tools/perf/util/metricgroup.h | 3 +- tools/perf/util/stat.h | 1 + 5 files changed, 147 insertions(+), 10 deletions(-)
tools/perf/Documentation/perf-stat.txt +77 -0
@@ -266,6 +266,10 @@ For a group all metrics from the group are added. The events from the metrics are automatically measured. See perf list output for the possble metrics and metricgroups. +--metrics-file file:: +Read metrics definitions from file in addition to compiled in metrics. +Please check the file's syntax details in METRICS FILE section below. + -A:: --no-aggr:: Do not aggregate counts across all monitored CPUs.
@@ -404,6 +408,79 @@ The fields are in this order: Additional metrics may be printed with all earlier fields being empty. +METRICS FILE +------------ +The file with metrics has following syntax: + + NAME = EXPRESSION ; + NAME = EXPRESSION ; + ... + +where NAME is unique identifier of the metric, which is later used in +perf stat as -M option argument (see below). + +The EXPRESSION is the metric's formula with following grammar: + + EXPR: EVENT + EXPR: EXPR if EXPR else EXPR + EXPR: NUMBER + EXPR: EXPR | EXPR + EXPR: EXPR & EXPR + EXPR: EXPR ^ EXPR + EXPR: EXPR + EXPR + EXPR: EXPR - EXPR + EXPR: EXPR * EXPR + EXPR: EXPR / EXPR + EXPR: EXPR % EXPR + EXPR: - EXPR + EXPR: ( EXPR ) + EXPR: min( EXPR, EXPR ) + EXPR: max( EXPR, EXPR ) + EXPR: #smt_on + +where: + + EVENT is monitored event name + min returns smaller of 2 expressions + max returns bigger of 2 expressions + #smt_on returns true if SMT is enabled + +The metric's definition can be spread across multiple lines and it's finished +with ';' character. + +The syntax allows for C style comments: + + // single line + /* multiple + lines */ + +Metrics file example with 2 custom metrics mine1 and mine2: + + $ cat metrics + // IPC + mine1 = inst_retired.any / cpu_clk_unhalted.thread; + + /* DECODED_ICACHE_UOPS% */ + mine2 = 100 * (idq.dsb_uops / (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops)); + + $ sudo perf stat --metrics-file ./metrics -M mine1,mine2 -a -I 1000 --metric-only + # time mine1 mine2 + 1.000997184 0.41 18.47 + 2.002479737 0.57 22.46 + 3.003932935 0.40 17.52 + ... + + +It's possible to mix custom metrics with compiled-in metrics in one -M argument: + + $ sudo perf stat --metrics-file ./metrics -M mine1,mine2,IPC -a -I 1000 --metric-only + # time mine1 mine2 IPC + 1.001142007 0.85 22.33 0.85 + 2.002460174 0.86 23.37 0.86 + 3.003969795 1.03 23.93 1.03 + ... + + SEE ALSO -------- linkperf:perf-top[1], linkperf:perf-list[1]
tools/perf/builtin-stat.c +5 -2
@@ -840,7 +840,8 @@ static int parse_metric_groups(const struct option *opt,const char *str, int unset __maybe_unused) { - return metricgroup__parse_groups(opt, str, &stat_config.metric_events); + return metricgroup__parse_groups(opt, str, &stat_config.metric_events, + stat_config.metrics_file); } static struct option stat_options[] = {
@@ -925,6 +926,8 @@ static struct option stat_options[] = {OPT_CALLBACK ('M' ,"metrics" , &evsel_list,"metric/metric group list" ,"monitor specified metrics or metric groups (separated by ,)" , parse_metric_groups), +OPT_STRING (0 ,"metrics-file" , &stat_config.metrics_file,"file path" , +"file with metrics definitions" ),OPT_BOOLEAN_FLAG (0 ,"all-kernel" , &stat_config.all_kernel,"Configure all used events to run in kernel space." , PARSE_OPT_EXCLUSIVE),
@@ -1442,7 +1445,7 @@ static int add_default_attributes(void)struct option opt = { .value = &evsel_list };return metricgroup__parse_groups (&opt,"transaction" , - &stat_config.metric_events); + &stat_config.metric_events,NULL ); } if (pmu_have_event("cpu", "cycles-ct") &&
tools/perf/util/metricgroup.c +62 -7
121 lines not on this page · raw message
tools/perf/util/metricgroup.h +2 -1
@@ -30,7 +30,8 @@ struct metric_event *metricgroup__lookup(struct rblist *metric_events, bool create);int metricgroup__parse_groups (const struct option *opt,const char *str, -struct rblist *metric_events); +struct rblist *metric_events, +const char *metrics_file);void metricgroup__print (bool metrics,bool groups,char *filter,bool raw,bool details);
tools/perf/util/stat.h +1 -0
@@ -123,6 +123,7 @@ struct perf_stat_config {struct runtime_stat *stats;int stats_num;const char *csv_sep; +const char *metrics_file;struct stats *walltime_nsecs_stats;struct rusage ru_data;struct perf_cpu_map *aggr_map;
Signature · 1 line
2.25.4
Jiri Olsa
reviewerPATCH v3 4/4 perf expr: Report line number with error
In reply to Jiri OlsaDisplay line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;
krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4
Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3
--- tools/perf/tests/expr.c | 5 +++++ tools/perf/util/expr.h | 2 ++ tools/perf/util/expr.l | 10 ++++++++++ tools/perf/util/expr.y | 8 +++++--- tools/perf/util/metricgroup.c | 3 ++- 5 files changed, 24 insertions(+), 4 deletions(-)
tools/perf/tests/expr.c +5 -0
@@ -84,5 +84,10 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)zfree (&ctx.custom[i].name);zfree (&ctx.custom[i].expr); } + +expr__ctx_init (&ctx); + ret =expr__parse_custom (&ctx,"IPC=INSTRUCTIONS / CYCLES;\n error" ); +TEST_ASSERT_VAL ("parse custom failed" , ret !=0 ); +TEST_ASSERT_VAL ("parse custom count" , ctx.lineno ==2 );return 0 ; }
tools/perf/util/expr.h +2 -0
@@ -27,6 +27,8 @@ struct expr_parse_ctx {struct expr_parse_custom custom[EXPR_MAX ]; }; }; +/* Set on error for custom metrics. */ +int lineno; };struct expr_scanner_ctx {
tools/perf/util/expr.l +10 -0
@@ -1,6 +1,8 @@ %option prefix="expr_" %option reentrant %option bison-bridge +%option bison-locations +%option yylineno %{ #include <linux/compiler.h>
@@ -79,6 +81,13 @@ static int str(yyscan_t scanner, int token, int runtime) yylval->str = normalize(yylval->str, runtime); return token; } + +#define YY_USER_ACTION \ +do { \ + yylloc->last_line = yylloc->first_line; \ + yylloc->first_line = yylineno; \ +} while (0); + %} %x custom
@@ -101,6 +110,7 @@ all [^;]+ if (sctx->start_token) { sctx->start_token = 0; + yylineno = 1; return start_token; } }
tools/perf/util/expr.y +5 -3
@@ -18,6 +18,7 @@ %parse-param { struct expr_parse_ctx *ctx } %parse-param {void *scanner} %lex-param {void* scanner} +%locations %union { double num;
@@ -39,12 +40,13 @@ %type <num> expr if_expr %{ -static void expr_error(double *final_val __maybe_unused, - struct expr_parse_ctx *ctx __maybe_unused, +static void expr_error(YYLTYPE *loc, double *final_val __maybe_unused, + struct expr_parse_ctx *ctx, void *scanner, const char *s) { - pr_debug("%s\n", s); + pr_debug("%s, line %d\n", s, loc->last_line); + ctx->lineno = loc->last_line; } static int lookup_id(struct expr_parse_ctx *ctx, char *id, double *val)
tools/perf/util/metricgroup.c +2 -1
@@ -615,7 +615,8 @@ static int metricgroup__add_metric_list(const char *list, struct strbuf *events, ret =expr__parse_custom (&ctx, buf);free (buf);if (ret) { -pr_err ("failed to parse metrics file: %s\n" , metrics_file); +pr_err ("failed to parse metrics file: %s:%d\n" , + metrics_file, ctx.lineno);return -1 ; } }
Signature · 1 line
2.25.4
Ian Rogers
reviewerRe: [PATCH 1/4] perf expr: Add parsing support for multiple expressions
In reply to Jiri OlsaQuoted 2 lines · Jiri Olsa
Adding support to parse metric difinitions in following form:
Typo on definitions.
Quoted 69 lines
NAME = EXPRESSION ; NAME = EXPRESSION ; ...The parsed NAME and EXPRESSION will be used in following changes to feed the metric code with definitions from custom file.
Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/tests/expr.c | 13 +++++++++++++ tools/perf/util/expr.c | 6 ++++++ tools/perf/util/expr.h | 19 +++++++++++++++++-- tools/perf/util/expr.l | 12 ++++++++++++ tools/perf/util/expr.y | 13 ++++++++++++- 5 files changed, 60 insertions(+), 3 deletions(-)tools/perf/tests/expr.c +13
-0@@ -71,5 +71,18 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)zfree (&other[i]);free ((void *)other); +expr__ctx_init (&ctx); + ret =expr__parse_custom (&ctx,"IPC=INSTRUCTIONS / CYCLES; CPI=CYCLES / INSTRUCTIONS;" ); +TEST_ASSERT_VAL ("parse custom failed" , ret ==0 ); +TEST_ASSERT_VAL ("parse custom count" , ctx.num_custom ==2 ); +TEST_ASSERT_VAL ("parse custom name" , !strcmp (ctx.custom[0 ].name,"IPC" )); +TEST_ASSERT_VAL ("parse custom name" , !strcmp (ctx.custom[1 ].name,"CPI" )); +TEST_ASSERT_VAL ("parse custom expr" , !strcmp (ctx.custom[0 ].expr,"INSTRUCTIONS / CYCLES" )); +TEST_ASSERT_VAL ("parse custom expr" , !strcmp (ctx.custom[1 ].expr,"CYCLES / INSTRUCTIONS" )); + +for (i =0 ; i < ctx.num_custom; i++) { +zfree (&ctx.custom[i].name); +zfree (&ctx.custom[i].expr); + }return 0 ; }tools/perf/util/expr.c +6
-0@@ -23,6 +23,7 @@ void expr__add_id(struct expr_parse_ctx *ctx, const char *name, double val)void expr__ctx_init (struct expr_parse_ctx *ctx) { ctx->num_ids =0 ; + ctx->num_custom =0 ; } static int@@ -61,6 +62,11 @@ int expr__parse(double *final_val, struct expr_parse_ctx *ctx, const char *expr,return __expr__parse (final_val, ctx, expr,EXPR_PARSE , runtime) ?-1 :0 ; } +int expr__parse_custom (struct expr_parse_ctx *ctx,const char *expr) +{ +return __expr__parse (NULL , ctx, expr,EXPR_CUSTOM ,0 ); +} +static bool already_seen (const char *val,const char *one,const char **other,int num_other)tools/perf/util/expr.h +0
-0
tools/perf/util/expr.h line 7
Quoted 4 lines
tools/perf/util/expr.h +1
-0@@ -4,15 +4,29 @@#define EXPR_MAX_OTHER 64#define MAX_PARSE_ID EXPR_MAX_OTHER +#define EXPR_MAX 20
Currently deduplication of ids is done after rather than during expression pasing, meaning hitting these limits is quite easy. This is fixed in:
https://lore.kernel.org/lkml/[email protected]/
But not for custom expressions being added here. I plan to rebase that work and clone hashmap from libbpf into libapi to workaround the dependency issue.
That patch also adds expr__ctx_clear as a convenience for cleaning up the context, and passes the context around inside of metricgroup rather than ids.
Quoted 68 lines
tools/perf/util/expr.h +16
-2struct expr_parse_id {const char *name;double val; }; +struct expr_parse_custom { +const char *name; +const char *expr; +}; +struct expr_parse_ctx { -int num_ids; -struct expr_parse_id ids[MAX_PARSE_ID ]; +union { +struct { +int num_ids; +struct expr_parse_id ids[MAX_PARSE_ID ]; + }; +struct { +int num_custom; +struct expr_parse_custom custom[EXPR_MAX ]; + }; + }; };struct expr_scanner_ctx {@@ -23,6 +37,7 @@ struct expr_scanner_ctx {void expr__ctx_init (struct expr_parse_ctx *ctx);void expr__add_id (struct expr_parse_ctx *ctx,const char *id,double val);int expr__parse (double *final_val,struct expr_parse_ctx *ctx,const char *expr,int runtime); +int expr__parse_custom (struct expr_parse_ctx *ctx,const char *expr);int expr__find_other (const char *expr,const char *one,const char ***other,int *num_other,int runtime);tools/perf/util/expr.l +9
-0@@ -81,12 +81,15 @@ static int str(yyscan_t scanner, int token, int runtime) } %} +%x custom + number [0-9]*\.?[0-9]+ sch [-,=] spec \\{sch} sym [0-9a-zA-Z_\.:@?]+ symbol ({spec}|{sym})+ +all [^;]+ %% struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);@@ -100,6 +103,12 @@ symbol ({spec}|{sym})+ } } +<custom>{ + +{all} { BEGIN(INITIAL); return str(yyscanner, ALL, sctx->runtime); } + +} + max { return MAX; } min { return MIN; } if { return IF; }
tools/perf/util/expr.l else { return ELSE; } · line 131
Quoted 5 lines
tools/perf/util/expr.l +2
-0@@ -118,6 +127,9 @@ else { return ELSE; } "(" { return '('; } ")" { return ')'; } "," { return ','; } +";" { return ';'; } +"=" { BEGIN(custom); return '='; }
Will this interfere with the \\= encoded in MetricExpr? Could you test with:
https://lore.kernel.org/lkml/[email protected]/
Quoted 9 lines
tools/perf/util/expr.l +1
-0+\n { } . { } %%tools/perf/util/expr.y +0
-0
tools/perf/util/expr.y line 30
Quoted 8 lines
tools/perf/util/expr.y +2
-1@@ -24,9 +24,10 @@ char *str; } -%token EXPR_PARSE EXPR_OTHER EXPR_ERROR +%token EXPR_PARSE EXPR_OTHER EXPR_CUSTOM EXPR_ERROR %token <num> NUMBER %token <str> ID +%token <str> ALL
Missing %destructor, fix is here:
https://lore.kernel.org/lkml/[email protected]/
Thanks,
Ian
Quoted 23 lines
tools/perf/util/expr.y +10
-1%token MIN MAX IF ELSE SMT_ON %left MIN MAX IF %left '|'@@ -66,6 +67,16 @@ start: EXPR_PARSE all_expr | EXPR_OTHER all_other +| +EXPR_CUSTOM all_custom + +all_custom: all_custom ID '=' ALL ';' +{ + ctx->custom[ctx->num_custom].name = $2; + ctx->custom[ctx->num_custom].expr = $4; + ctx->num_custom++; +} +| all_other: all_other other | --2.25.4
Ian Rogers
reviewerRe: [PATCH 3/4] perf stat: Add --metrics-file option
In reply to Jiri OlsaQuoted 53 lines · Jiri Olsa
Adding --metrics-file option that allows to specify metrics in the file.
It's now possible to define metrics in file and use them like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;/* DECODED_ICACHE_UOPS% */ mine2 = 100 * (idq.dsb_uops / (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops)); $ sudo perf stat --metrics-file ./metrics -M mine1,mine2 -a -I 1000 --metric-only # time mine1 mine2 1.000997184 0.41 18.47 2.002479737 0.57 22.46 3.003932935 0.40 17.52 ...Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/Documentation/perf-stat.txt | 77 ++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 7 ++- tools/perf/util/metricgroup.c | 69 ++++++++++++++++++++--- tools/perf/util/metricgroup.h | 3 +- tools/perf/util/stat.h | 1 + 5 files changed, 147 insertions(+), 10 deletions(-)tools/perf/Documentation/perf-stat.txt +9
-0@@ -266,6 +266,10 @@ For a group all metrics from the group are added. The events from the metrics are automatically measured. See perf list output for the possble metrics and metricgroups. +--metrics-file file:: +Read metrics definitions from file in addition to compiled in metrics. +Please check the file's syntax details in METRICS FILE section below. + -A:: --no-aggr:: Do not aggregate counts across all monitored CPUs.@@ -404,6 +408,79 @@ The fields are in this order: Additional metrics may be printed with all earlier fields being empty. +METRICS FILE +------------ +The file with metrics has following syntax: + + NAME = EXPRESSION ;
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 425
Quoted 10 lines
tools/perf/Documentation/perf-stat.txt +10
-0+ NAME = EXPRESSION ; + ... + +where NAME is unique identifier of the metric, which is later used in +perf stat as -M option argument (see below). + +The EXPRESSION is the metric's formula with following grammar: + + EXPR: EVENT + EXPR: EXPR if EXPR else EXPR
Not introduced by this patch, but this patch is exposing it as an API. This notion of if-else is really weird. For one thing there are no comparison operators. The unit test doesn't really help:
ret |= test(&ctx, "1+1 if 3*4 else 0", 2);
What kind of comparison is "3*4"? If 0.0 causes the else clause then will -0.0? A typical expression I see written in C is to give a ratio such:
value = denom == 0 ? 0 : nom / denom;
I've worked around encoding this by extending expr.y locally.
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 429
Quoted 4 lines
tools/perf/Documentation/perf-stat.txt +4
-0+ EXPR: NUMBER + EXPR: EXPR | EXPR + EXPR: EXPR & EXPR + EXPR: EXPR ^ EXPR
Again, it's odd that these cast the double to a long and then assign the result back to a double.
Quoted 42 lines
tools/perf/Documentation/perf-stat.txt +42
-0+ EXPR: EXPR + EXPR + EXPR: EXPR - EXPR + EXPR: EXPR * EXPR + EXPR: EXPR / EXPR + EXPR: EXPR % EXPR + EXPR: - EXPR + EXPR: ( EXPR ) + EXPR: min( EXPR, EXPR ) + EXPR: max( EXPR, EXPR ) + EXPR: #smt_on + +where: + + EVENT is monitored event name + min returns smaller of 2 expressions + max returns bigger of 2 expressions + #smt_on returns true if SMT is enabled + +The metric's definition can be spread across multiple lines and it's finished +with ';' character. + +The syntax allows for C style comments: + + // single line + /* multiple + lines */ + +Metrics file example with 2 custom metrics mine1 and mine2: + + $ cat metrics + // IPC + mine1 = inst_retired.any / cpu_clk_unhalted.thread; + + /* DECODED_ICACHE_UOPS% */ + mine2 = 100 * (idq.dsb_uops / (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops)); + + $ sudo perf stat --metrics-file ./metrics -M mine1,mine2 -a -I 1000 --metric-only + # time mine1 mine2 + 1.000997184 0.41 18.47 + 2.002479737 0.57 22.46 + 3.003932935 0.40 17.52 + ...
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 481
Quoted 10 lines
tools/perf/Documentation/perf-stat.txt +10
-0+ + +It's possible to mix custom metrics with compiled-in metrics in one -M argument: + + $ sudo perf stat --metrics-file ./metrics -M mine1,mine2,IPC -a -I 1000 --metric-only + # time mine1 mine2 IPC + 1.001142007 0.85 22.33 0.85 + 2.002460174 0.86 23.37 0.86 + 3.003969795 1.03 23.93 1.03 + ...
A feature request would be to allow metrics in terms of other metrics, not just events :-) For example, it is common to sum all cache hit/miss events. It is laborious to copy that expression for hit rate, miss rate, etc.
Perhaps the expression parsing code should be folded into the event parsing code.
Thanks,
Ian
Quoted 196 lines
tools/perf/Documentation/perf-stat.txt +1
-0+ SEE ALSO -------- linkperf:perf-top[1], linkperf:perf-list[1]tools/perf/builtin-stat.c +5
-2@@ -840,7 +840,8 @@ static int parse_metric_groups(const struct option *opt,const char *str, int unset __maybe_unused) { - return metricgroup__parse_groups(opt, str, &stat_config.metric_events); + return metricgroup__parse_groups(opt, str, &stat_config.metric_events, + stat_config.metrics_file); } static struct option stat_options[] = {@@ -925,6 +926,8 @@ static struct option stat_options[] = {OPT_CALLBACK ('M' ,"metrics" , &evsel_list,"metric/metric group list" ,"monitor specified metrics or metric groups (separated by ,)" , parse_metric_groups), +OPT_STRING (0 ,"metrics-file" , &stat_config.metrics_file,"file path" , +"file with metrics definitions" ),OPT_BOOLEAN_FLAG (0 ,"all-kernel" , &stat_config.all_kernel,"Configure all used events to run in kernel space." , PARSE_OPT_EXCLUSIVE),@@ -1442,7 +1445,7 @@ static int add_default_attributes(void)struct option opt = { .value = &evsel_list };return metricgroup__parse_groups (&opt,"transaction" , - &stat_config.metric_events); + &stat_config.metric_events,NULL ); } if (pmu_have_event("cpu", "cycles-ct") &&tools/perf/util/metricgroup.c +62
-7@@ -527,17 +527,17 @@ static int __metricgroup__add_metric(struct strbuf *events, }static int metricgroup__add_metric (const char *metric,struct strbuf *events, -struct list_head *group_list) +struct list_head *group_list, +struct expr_parse_ctx *ctx) {struct pmu_events_map *map =perf_pmu__find_map (NULL ); -struct pmu_event *pe;int i, ret = -EINVAL ;if (!map)return 0 ;for (i =0 ; ; i++) { - pe = &map->table[i]; +struct pmu_event *pe = &map->table[i];if (!pe->name && !pe->metric_group && !pe->metric_name)break ;@@ -567,15 +567,59 @@ static int metricgroup__add_metric(const char *metric, struct strbuf *events,break ; } } + +if (!ctx->num_custom || ret == -ENOMEM ) +return ret; + +for (i =0 ; i < ctx->num_custom; i++) { +struct pmu_event pe = {0 }; + +if (!match_metric (ctx->custom[i].name, metric)) +continue ; + + pe.metric_name =strdup (ctx->custom[i].name); + pe.metric_expr =strdup (ctx->custom[i].expr); + +if (!pe.metric_name && !pe.metric_expr) +return -ENOMEM ; + + ret =__metricgroup__add_metric (events, group_list, &pe,1 ); +if (ret) { +free ((char *) pe.metric_name); +free ((char *) pe.metric_expr); +break ; + } + } +return ret; }static int metricgroup__add_metric_list (const char *list,struct strbuf *events, -struct list_head *group_list) +struct list_head *group_list, +const char *metrics_file) { +struct expr_parse_ctx ctx = {0 };char *llist, *nlist, *p;int ret = -EINVAL ; +if (metrics_file) { +size_t size; +char *buf; + +if (filename__read_str (metrics_file, &buf, &size)) { +pr_err ("failed to read metrics file: %s\n" , metrics_file); +return -1 ; + } + +expr__ctx_init (&ctx); + ret =expr__parse_custom (&ctx, buf); +free (buf); +if (ret) { +pr_err ("failed to parse metrics file: %s\n" , metrics_file); +return -1 ; + } + } + nlist =strdup (list);if (!nlist)return -ENOMEM ;@@ -585,7 +629,7 @@ static int metricgroup__add_metric_list(const char *list, struct strbuf *events,strbuf_addf (events,"%s" ,"" );while ((p =strsep (&llist,"," )) !=NULL ) { - ret =metricgroup__add_metric (p, events, group_list); + ret =metricgroup__add_metric (p, events, group_list, &ctx);if (ret == -EINVAL ) {fprintf (stderr,"Cannot find metric or group `%s'\n" , p);@@ -597,6 +641,15 @@ static int metricgroup__add_metric_list(const char *list, struct strbuf *events,if (!ret)metricgroup___watchdog_constraint_hint (NULL , true); +if (metrics_file) { +int i; + +for (i =0 ; i < ctx.num_custom; i++) { +zfree (&ctx.custom[i].name); +zfree (&ctx.custom[i].expr); + } + } +return ret; }@@ -616,7 +669,8 @@ static void metricgroup__free_egroups(struct list_head *group_list)int metricgroup__parse_groups (const struct option *opt,const char *str, -struct rblist *metric_events) +struct rblist *metric_events, +const char *metrics_file) {struct parse_events_error parse_error;struct evlist *perf_evlist = *(struct evlist **)opt->value;@@ -626,7 +680,8 @@ int metricgroup__parse_groups(const struct option *opt,if (metric_events->nr_entries ==0 )metricgroup__rblist_init (metric_events); - ret =metricgroup__add_metric_list (str, &extra_events, &group_list); + ret =metricgroup__add_metric_list (str, &extra_events, &group_list, + metrics_file);if (ret)return ret;pr_debug ("adding %s\n" , extra_events.buf);tools/perf/util/metricgroup.h +2
-1@@ -30,7 +30,8 @@ struct metric_event *metricgroup__lookup(struct rblist *metric_events, bool create);int metricgroup__parse_groups (const struct option *opt,const char *str, -struct rblist *metric_events); +struct rblist *metric_events, +const char *metrics_file);void metricgroup__print (bool metrics,bool groups,char *filter,bool raw,bool details);tools/perf/util/stat.h +1
-1@@ -123,6 +123,7 @@ struct perf_stat_config {struct runtime_stat *stats;int stats_num;const char *csv_sep; +const char *metrics_file;struct stats *walltime_nsecs_stats;struct rusage ru_data;struct perf_cpu_map *aggr_map; --2.25.4
Ian Rogers
reviewerRe: [PATCH 4/4] perf expr: Report line number with error
In reply to Jiri OlsaQuoted 15 lines · Jiri Olsa
Display line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
Quoted 6 lines
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3
A line number is better than nothing :-) It'd be nice to be told about broken events and more information about what's broken in the line. A common failure is @ vs / encoding and also no-use or misuse of \\. Perhaps expand the test coverage.
Thanks,
Ian
Quoted 118 lines
Signed-off-by Jiri Olsa <[email protected]>--- tools/perf/tests/expr.c | 5 +++++ tools/perf/util/expr.h | 2 ++ tools/perf/util/expr.l | 10 ++++++++++ tools/perf/util/expr.y | 8 +++++--- tools/perf/util/metricgroup.c | 3 ++- 5 files changed, 24 insertions(+), 4 deletions(-)tools/perf/tests/expr.c +5
-0@@ -84,5 +84,10 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)zfree (&ctx.custom[i].name);zfree (&ctx.custom[i].expr); } + +expr__ctx_init (&ctx); + ret =expr__parse_custom (&ctx,"IPC=INSTRUCTIONS / CYCLES;\n error" ); +TEST_ASSERT_VAL ("parse custom failed" , ret !=0 ); +TEST_ASSERT_VAL ("parse custom count" , ctx.lineno ==2 );return 0 ; }tools/perf/util/expr.h +2
-0@@ -27,6 +27,8 @@ struct expr_parse_ctx {struct expr_parse_custom custom[EXPR_MAX ]; }; }; +/* Set on error for custom metrics. */ +int lineno; };struct expr_scanner_ctx {tools/perf/util/expr.l +10
-0@@ -1,6 +1,8 @@ %option prefix="expr_" %option reentrant %option bison-bridge +%option bison-locations +%option yylineno %{ #include <linux/compiler.h>@@ -79,6 +81,13 @@ static int str(yyscan_t scanner, int token, int runtime) yylval->str = normalize(yylval->str, runtime); return token; } + +#define YY_USER_ACTION \ +do { \ + yylloc->last_line = yylloc->first_line; \ + yylloc->first_line = yylineno; \ +} while (0); + %} %x custom@@ -101,6 +110,7 @@ all [^;]+ if (sctx->start_token) { sctx->start_token = 0; + yylineno = 1; return start_token; } }tools/perf/util/expr.y +5
-3@@ -18,6 +18,7 @@ %parse-param { struct expr_parse_ctx *ctx } %parse-param {void *scanner} %lex-param {void* scanner} +%locations %union { double num;@@ -39,12 +40,13 @@ %type <num> expr if_expr %{ -static void expr_error(double *final_val __maybe_unused, - struct expr_parse_ctx *ctx __maybe_unused, +static void expr_error(YYLTYPE *loc, double *final_val __maybe_unused, + struct expr_parse_ctx *ctx, void *scanner, const char *s) { - pr_debug("%s\n", s); + pr_debug("%s, line %d\n", s, loc->last_line); + ctx->lineno = loc->last_line; } static int lookup_id(struct expr_parse_ctx *ctx, char *id, double *val)tools/perf/util/metricgroup.c +2
-2@@ -615,7 +615,8 @@ static int metricgroup__add_metric_list(const char *list, struct strbuf *events, ret =expr__parse_custom (&ctx, buf);free (buf);if (ret) { -pr_err ("failed to parse metrics file: %s\n" , metrics_file); +pr_err ("failed to parse metrics file: %s:%d\n" , + metrics_file, ctx.lineno);return -1 ; } } --2.25.4
Re: [PATCH 1/4] perf expr: Add parsing support for multiple expressions
In reply to Ian RogersQuoted 5 lines · Ian Rogers
Quoted 2 lines · Jiri Olsa
Adding support to parse metric difinitions in following form:
Typo on definitions.
right
SNIP
Quoted 14 lines
Quoted 14 lines
tools/perf/util/expr.c +5
-0+int expr__parse_custom (struct expr_parse_ctx *ctx,const char *expr) +{ +return __expr__parse (NULL , ctx, expr,EXPR_CUSTOM ,0 ); +} +static bool already_seen (const char *val,const char *one,const char **other,int num_other)tools/perf/util/expr.h +0
-0@@ -4,15 +4,29 @@
Quoted 14 lines
tools/perf/util/expr.h line 7
Quoted 3 lines
tools/perf/util/expr.h +1
-0#define EXPR_MAX_OTHER 64#define MAX_PARSE_ID EXPR_MAX_OTHER +#define EXPR_MAX 20Currently deduplication of ids is done after rather than during expression pasing, meaning hitting these limits is quite easy. This is fixed in:
https://lore.kernel.org/lkml/[email protected]/
But not for custom expressions being added here. I plan to rebase that work and clone hashmap from libbpf into libapi to workaround the dependency issue.
That patch also addsexpr__ctx_clearas a convenience for cleaning up the context, and passes the context around inside of metricgroup rather than ids.
ok
SNIP
Quoted 19 lines
Quoted 19 lines
tools/perf/util/expr.h +7
-0symbol ({spec}|{sym})+ +all [^;]+ %% struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);@@ -100,6 +103,12 @@ symbol ({spec}|{sym})+ } } +<custom>{ + +{all} { BEGIN(INITIAL); return str(yyscanner, ALL, sctx->runtime); } + +} + max { return MAX; } min { return MIN; } if { return IF; }@@ -118,6 +127,9 @@ else { return ELSE; }
Quoted 8 lines
tools/perf/util/expr.h else { return ELSE; } · line 131
Quoted 5 lines
tools/perf/util/expr.h +2
-0"(" { return '('; } ")" { return ')'; } "," { return ','; } +";" { return ';'; } +"=" { BEGIN(custom); return '='; }Will this interfere with the \\= encoded in MetricExpr? Could you test with:
https://lore.kernel.org/lkml/[email protected]/
it shouldn't, the escape is matched first.. but I'll put it on top of the new tests to be sure
Quoted 13 lines
Quoted 13 lines
tools/perf/util/expr.h +1
-0+\n { } . { } %%tools/perf/util/expr.y +0
-0@@ -24,9 +24,10 @@ char *str; }
Quoted 8 lines
tools/perf/util/expr.y line 30
Quoted 5 lines
tools/perf/util/expr.y +2
-1-%token EXPR_PARSE EXPR_OTHER EXPR_ERROR +%token EXPR_PARSE EXPR_OTHER EXPR_CUSTOM EXPR_ERROR %token <num> NUMBER %token <str> ID +%token <str> ALLMissing %destructor, fix is here:
https://lore.kernel.org/lkml/[email protected]/
oops, right.. thanks
jirka
Re: [PATCH 3/4] perf stat: Add --metrics-file option
In reply to Ian RogersOn Wed, May 13, 2020 at 12:04:55AM -0700, Ian Rogers wrote:
SNIP
Quoted 9 lines
Quoted 9 lines
tools/perf/Documentation/perf-stat.txt +9
-0+METRICS FILE +------------ +The file with metrics has following syntax: + + NAME = EXPRESSION ; + NAME = EXPRESSION ; + ... + +where NAME is unique identifier of the metric, which is later used in
Quoted 8 lines
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 425
Quoted 6 lines
tools/perf/Documentation/perf-stat.txt +6
-0+perf stat as -M option argument (see below). + +The EXPRESSION is the metric's formula with following grammar: + + EXPR: EVENT + EXPR: EXPR if EXPR else EXPRNot introduced by this patch, but this patch is exposing it as an API.
yea, I was thinking about this and I think we will put a disclaimer in here that this is not an API and the interface can change.. it's really mostly intended to help out with running a custom metric which is not compiled in ... I don't want to be commited to support old API
Quoted 7 lines
This notion of if-else is really weird. For one thing there are no comparison operators. The unit test doesn't really help:
ret |= test(&ctx, "1+1 if 3*4 else 0", 2);What kind of comparison is "3*4"? If 0.0 causes the else clause then will -0.0? A typical expression I see written in C is to give a ratio such:
value = denom == 0 ? 0 : nom / denom;
I've worked around encoding this by extending expr.y locally.
AFAICS it's used only with #SMT_on in the condition, aybe we could limit the condition only for #SMT_on term?
Quoted 8 lines
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 429
Quoted 5 lines
tools/perf/Documentation/perf-stat.txt +4
-0+ EXPR: NUMBER + EXPR: EXPR | EXPR + EXPR: EXPR & EXPR + EXPR: EXPR ^ EXPRAgain, it's odd that these cast the double to a long and then assign the result back to a double.
is this even used anywhere? perhaps it was added just to be complete
SNIP
Quoted 11 lines
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 481
Quoted 3 lines
tools/perf/Documentation/perf-stat.txt +3
-0+ 2.002460174 0.86 23.37 0.86 + 3.003969795 1.03 23.93 1.03 + ...A feature request would be to allow metrics in terms of other metrics, not just events :-) For example, it is common to sum all cache hit/miss events. It is laborious to copy that expression for hit rate, miss rate, etc.
Perhaps the expression parsing code should be folded into the event parsing code.
nice idea, but let's finish straighten up what we have first ;-)
I'll try to go through all the fixes/tests you posted and let's get it in first
thanks,
jirka
Re: [PATCH 4/4] perf expr: Report line number with error
In reply to Ian RogersQuoted 23 lines · Ian Rogers
Quoted 15 lines · Jiri Olsa
Display line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
Quoted 6 lines
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3
Quoted 4 lines
A line number is better than nothing :-) It'd be nice to be told about broken events and more information about what's broken in the line. A common failure is @ vs / encoding and also no-use or misuse of \\. Perhaps expand the test coverage.
yep, error reporting needs more changes.. but the line is crucial ;-)
jirka
Re: [PATCH 4/4] perf expr: Report line number with error
Em Wed, May 13, 2020 at 01:34:24PM +0200, Jiri Olsa escreveu:
Quoted 24 lines · Ian Rogers
Quoted 23 lines · Jiri Olsa
Quoted 21 lines
Display line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3
Quoted 6 lines
Quoted 4 lines
A line number is better than nothing :-) It'd be nice to be told about broken events and more information about what's broken in the line. A common failure is @ vs / encoding and also no-use or misuse of \\. Perhaps expand the test coverage.
yep, error reporting needs more changes.. but the line is crucial ;-)
So I had started processing this patchkit, I assume you will send a v2 and I should drop what I had processed, is that ok?
- Arnaldo
Re: [PATCH 4/4] perf expr: Report line number with error
Quoted 30 lines · Arnaldo Carvalho de Melo
Em Wed, May 13, 2020 at 01:34:24PM +0200, Jiri Olsa escreveu:
Quoted 29 lines · Ian Rogers
Quoted 27 lines · Jiri Olsa
Quoted 15 lines
Display line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
Quoted 6 lines
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3A line number is better than nothing :-) It'd be nice to be told about broken events and more information about what's broken in the line. A common failure is @ vs / encoding and also no-use or misuse of \\. Perhaps expand the test coverage.
Quoted 4 lines
Quoted 1 line
yep, error reporting needs more changes.. but the line is crucial ;-)
So I had started processing this patchkit, I assume you will send a v2 and I should drop what I had processed, is that ok?
yes, I will resubmit on top of the other expr changes we have now pending
jirka
Re: [PATCH 4/4] perf expr: Report line number with error
Em Wed, May 13, 2020 at 04:46:10PM +0200, Jiri Olsa escreveu:
Quoted 31 lines · Arnaldo Carvalho de Melo
Quoted 30 lines
Em Wed, May 13, 2020 at 01:34:24PM +0200, Jiri Olsa escreveu:
Quoted 29 lines · Ian Rogers
Quoted 27 lines · Jiri Olsa
Quoted 15 lines
Display line number on when parsing custom metrics file, like:
$ cat metrics // IPC mine1 = inst_retired.any / cpu_clk_unhalted.thread;krava
$ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:4Please note that because the grammar is flexible on new lines, the syntax could be broken on the next 'not fitting' item and not the first wrong word, like:
Quoted 6 lines
$ cat metrics // IPC krava mine1 = inst_retired.any / cpu_clk_unhalted.thread; $ sudo perf stat --metrics-file ./metrics -M mine1 -a -I 1000 --metric-only failed to parse metrics file: ./metrics:3A line number is better than nothing :-) It'd be nice to be told about broken events and more information about what's broken in the line. A common failure is @ vs / encoding and also no-use or misuse of \\. Perhaps expand the test coverage.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
yep, error reporting needs more changes.. but the line is crucial ;-)
So I had started processing this patchkit, I assume you will send a v2 and I should drop what I had processed, is that ok?
yes, I will resubmit on top of the other expr changes we have now pending
Ok, I removed those from my local branch, and pushed what I have, which includes a few more fixes from Ian and others, continuing to process other remaining patches now.
- Arnaldo
Ian Rogers
reviewerRe: [PATCH 3/4] perf stat: Add --metrics-file option
In reply to Jiri OlsaQuoted 46 lines · Jiri Olsa
On Wed, May 13, 2020 at 12:04:55AM -0700, Ian Rogers wrote:
SNIP
Quoted 9 lines
Quoted 9 lines
tools/perf/Documentation/perf-stat.txt +9
-0+METRICS FILE +------------ +The file with metrics has following syntax: + + NAME = EXPRESSION ; + NAME = EXPRESSION ; + ... + +where NAME is unique identifier of the metric, which is later used inQuoted 8 lines
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 425
Quoted 6 lines
tools/perf/Documentation/perf-stat.txt +6
-0+perf stat as -M option argument (see below). + +The EXPRESSION is the metric's formula with following grammar: + + EXPR: EVENT + EXPR: EXPR if EXPR else EXPRNot introduced by this patch, but this patch is exposing it as an API.
yea, I was thinking about this and I think we will put a disclaimer in here that this is not an API and the interface can change.. it's really mostly intended to help out with running a custom metric which is not compiled in ... I don't want to be commited to support old API
Quoted 7 lines
This notion of if-else is really weird. For one thing there are no comparison operators. The unit test doesn't really help:
ret |= test(&ctx, "1+1 if 3*4 else 0", 2);What kind of comparison is "3*4"? If 0.0 causes the else clause then will -0.0? A typical expression I see written in C is to give a ratio such:
value = denom == 0 ? 0 : nom / denom;
I've worked around encoding this by extending expr.y locally.AFAICS it's used only with #SMT_on in the condition, aybe we could limit the condition only for #SMT_on term?
Quoted 6 lines
Quoted 5 lines
tools/perf/Documentation/perf-stat.txt +4
-0+ EXPR: NUMBER + EXPR: EXPR | EXPR + EXPR: EXPR & EXPR + EXPR: EXPR ^ EXPR
Quoted 4 lines
Quoted 2 lines
Again, it's odd that these cast the double to a long and then assign the result back to a double.
is this even used anywhere? perhaps it was added just to be complete
I don't believe they are used and checked with the pmu parsing test that removing them doesn't cause any x86 expressions to break. I'd prefer it if we could remove the unused operators and avoid advertising here.
Thanks,
Ian
Quoted 22 lines
SNIP
Quoted 11 lines
tools/perf/Documentation/perf-stat.txt The fields are in this order: · line 481
Quoted 3 lines
tools/perf/Documentation/perf-stat.txt +3
-0+ 2.002460174 0.86 23.37 0.86 + 3.003969795 1.03 23.93 1.03 + ...A feature request would be to allow metrics in terms of other metrics, not just events :-) For example, it is common to sum all cache hit/miss events. It is laborious to copy that expression for hit rate, miss rate, etc.
Perhaps the expression parsing code should be folded into the event parsing code.
nice idea, but let's finish straighten up what we have first ;-)
I'll try to go through all the fixes/tests you posted and let's get it in first
thanks,
jirka
On 11/05/2020 21:53, Jiri Olsa wrote:
+
Hi jirka,
Did you a plan to continue to work? I don't think that this support was ever merged.
We have a requirement to be able to tune parameters of metrics, and support here seems suitable.
Thanks,
John
Quoted 57 lines
hi,
Joe asked for possibility to add user defined metrics. Given that we already have metrics support, I added --metrics-file option that allows to specify custom metrics.$ cat metrics # IPC mine1 = instructions / cycles; /* DECODED_ICACHE_UOPS% */ mine2 = 100 * (idq.dsb_uops / \ (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops));$ sudo perf stat --metrics-file ./metrics -M mine1,mine2 --metric-only -a -I 1000 # time insn per cycle mine1 mine2 1.000536263 0.71 0.7 41.4 2.002069025 0.31 0.3 14.1 3.003427684 0.27 0.3 14.8 4.004807132 0.25 0.2 12.1 ...v3 changes:
- added doc for metrics file in perf stat man page
- reporting error line number now
- changed '#' style comment to C way with '//'v2 changes:
- add new --metrics-file option
- rebased on current perf/core expression bison/flex enhancementsAlso available in:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/metricthanks,
jirka---
Jiri Olsa (4):
perf expr: Add parsing support for multiple expressions perf expr: Allow comments in custom metric file perf stat: Add --metrics-file option
perf expr: Report line number with errortools/perf/Documentation/perf-stat.txt | 77 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 7 +++++-- tools/perf/tests/expr.c | 18 ++++++++++++++++++ tools/perf/util/expr.c | 6 ++++++ tools/perf/util/expr.h | 21 +++++++++++++++++++-- tools/perf/util/expr.l | 34 ++++++++++++++++++++++++++++++++++ tools/perf/util/expr.y | 21 +++++++++++++++++---- tools/perf/util/metricgroup.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------- tools/perf/util/metricgroup.h | 3 ++- tools/perf/util/stat.h | 1 + 10 files changed, 242 insertions(+), 16 deletions(-).
Quoted 11 lines · John Garry
On 11/05/2020 21:53, Jiri Olsa wrote:
+
Hi jirka,
Did you a plan to continue to work? I don't think that this support was ever merged.
We have a requirement to be able to tune parameters of metrics, and support here seems suitable.
John,
Have you tried to apply that series to see if it still applies?
- Arnaldo
Quoted 62 lines
Thanks,
JohnQuoted 57 lines
hi,
Joe asked for possibility to add user defined metrics. Given that we already have metrics support, I added --metrics-file option that allows to specify custom metrics.$ cat metrics # IPC mine1 = instructions / cycles; /* DECODED_ICACHE_UOPS% */ mine2 = 100 * (idq.dsb_uops / \ (idq.ms_uops + idq.mite_uops + idq.dsb_uops + lsd.uops));$ sudo perf stat --metrics-file ./metrics -M mine1,mine2 --metric-only -a -I 1000 # time insn per cycle mine1 mine2 1.000536263 0.71 0.7 41.4 2.002069025 0.31 0.3 14.1 3.003427684 0.27 0.3 14.8 4.004807132 0.25 0.2 12.1 ...v3 changes:
- added doc for metrics file in perf stat man page
- reporting error line number now
- changed '#' style comment to C way with '//'v2 changes:
- add new --metrics-file option
- rebased on current perf/core expression bison/flex enhancementsAlso available in:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/metricthanks,
jirka---
Jiri Olsa (4):
perf expr: Add parsing support for multiple expressions perf expr: Allow comments in custom metric file perf stat: Add --metrics-file option
perf expr: Report line number with errortools/perf/Documentation/perf-stat.txt | 77 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 7 +++++-- tools/perf/tests/expr.c | 18 ++++++++++++++++++ tools/perf/util/expr.c | 6 ++++++ tools/perf/util/expr.h | 21 +++++++++++++++++++-- tools/perf/util/expr.l | 34 ++++++++++++++++++++++++++++++++++ tools/perf/util/expr.y | 21 +++++++++++++++++---- tools/perf/util/metricgroup.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------- tools/perf/util/metricgroup.h | 3 ++- tools/perf/util/stat.h | 1 + 10 files changed, 242 insertions(+), 16 deletions(-).
Quoted 4 lines · Arnaldo Carvalho de Melo
Quoted 4 lines · Jiri Olsa
+
Quoted 10 lines
Quoted 7 lines
Hi jirka,
Did you a plan to continue to work? I don't think that this support was ever merged.
We have a requirement to be able to tune parameters of metrics, and support here seems suitable.
John,
Hi Arnaldo,
Quoted 1 line
Have you tried to apply that series to see if it still applies?
It doesn't apply on v5.17-rc1 (which is same as you acme perf/core branch).
Anyway, from checking this old thread, it seems that further changes were going to be added.
Thanks,
John