New function and literals for metrics
Ian Rogers
reviewercover New function and literals for metrics
Add a test for hyphenated events that are common with Intel's icelake topdown events.
Try to tidy the existing cputopo code and make it more consistent with the sysfs ABI. Refactor the #smt_on literal parsing into a general literal token that we can use for more literals. This is then used for num_cpus, num_cores, num_dies and num_packages literals that use the existing topology code. Finally, a source_count function is added which is used to determine the number of events contributing to an aggregate event. The intent for these new literals and function is for them to be used in upcoming metrics.
Ian Rogers (7):
perf test: Add expr test for events with hyphens perf cputopo: Update to use pakage_cpus
perf cputopo: Match die_siblings to topology ABI name perf cputopo: Match thread_siblings to topology ABI name perf expr: Add literal values starting with # perf expr: Add metric literals for topology. perf expr: Add source_count for aggregating events
tools/perf/tests/expr.c | 34 ++++++++++++++- tools/perf/util/cputopo.c | 78 ++++++++++++++++++----------------- tools/perf/util/cputopo.h | 33 ++++++++++++--- tools/perf/util/evsel.c | 12 ++++++ tools/perf/util/evsel.h | 1 + tools/perf/util/expr.c | 65 ++++++++++++++++++++++++++--- tools/perf/util/expr.h | 4 ++ tools/perf/util/expr.l | 16 ++++++- tools/perf/util/expr.y | 72 ++++++++++++++++++-------------- tools/perf/util/header.c | 20 ++++----- tools/perf/util/stat-shadow.c | 7 +++- 11 files changed, 250 insertions(+), 92 deletions(-)
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 1/7 perf test: Add expr test for events with hyphens
An example of such an event is topdown-fe-bound.
--- tools/perf/tests/expr.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
tools/perf/tests/expr.c +10 -0
@@ -134,6 +134,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)TEST_ASSERT_VAL ("find ids" ,hashmap__find (ctx->ids,"EVENT2,param=3@" , (void **)&val_ptr)); +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("find ids" , +expr__find_ids ("dash\\-event1 - dash\\-event2" , +NULL , ctx) ==0 ); +TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==2 ); +TEST_ASSERT_VAL ("find ids" ,hashmap__find (ctx->ids,"dash-event1" , + (void **)&val_ptr)); +TEST_ASSERT_VAL ("find ids" ,hashmap__find (ctx->ids,"dash-event2" , + (void **)&val_ptr)); +/* Only EVENT1 or EVENT2 need be measured depending on the value of smt_on. */ expr__ctx_clear (ctx); TEST_ASSERT_VAL("find ids",
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 2/7 perf cputopo: Update to use pakage_cpus
In reply to Ian Rogerscore_siblings_list is the deprecated topology name for package_cpus_list, update the code to try the non-deprecated path first. Adjust variable names to match topology name.
--- tools/perf/util/cputopo.c | 26 ++++++++++++++++---------- tools/perf/util/cputopo.h | 11 +++++++++-- tools/perf/util/header.c | 6 +++--- 3 files changed, 28 insertions(+), 15 deletions(-)
tools/perf/util/cputopo.c +16 -10
@@ -14,7 +14,9 @@#include "env.h" #include "pmu-hybrid.h" -#define CORE_SIB_FMT \ +#define PACKAGE_CPUS_FMT \ + "%s/devices/system/cpu/cpu%d/topology/package_cpus_list" +#define PACKAGE_CPUS_FMT_OLD \ "%s/devices/system/cpu/cpu%d/topology/core_siblings_list"#define DIE_SIB_FMT \ "%s/devices/system/cpu/cpu%d/topology/die_cpus_list"
@@ -39,8 +41,12 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu)u32 i =0 ;int ret =-1 ; -scnprintf (filename,MAXPATHLEN ,CORE_SIB_FMT , +scnprintf (filename,MAXPATHLEN ,PACKAGE_CPUS_FMT ,sysfs__mountpoint (), cpu); +if (access (filename,F_OK ) ==-1 ) { +scnprintf (filename,MAXPATHLEN ,PACKAGE_CPUS_FMT_OLD , +sysfs__mountpoint (), cpu); + } fp =fopen (filename,"r" );if (!fp) goto try_dies;
@@ -54,13 +60,13 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu)if (p) *p ='\0' ; -for (i =0 ; i < tp->core_sib; i++) { -if (!strcmp (buf, tp->core_siblings[i])) +for (i =0 ; i < tp->package_cpus_lists; i++) { +if (!strcmp (buf, tp->package_cpus_list[i]))break ; } -if (i == tp->core_sib) { - tp->core_siblings[i] = buf; - tp->core_sib++; +if (i == tp->package_cpus_lists) { + tp->package_cpus_list[i] = buf; + tp->package_cpus_lists++; buf =NULL ; len =0 ; }
@@ -139,8 +145,8 @@ void cpu_topology__delete(struct cpu_topology *tp)if (!tp)return ; -for (i =0 ; i < tp->core_sib; i++) -zfree (&tp->core_siblings[i]); +for (i =0 ; i < tp->package_cpus_lists; i++) +zfree (&tp->package_cpus_list[i]);if (tp->die_sib) {for (i =0 ; i < tp->die_sib; i++)
@@ -205,7 +211,7 @@ struct cpu_topology *cpu_topology__new(void) tp = addr; addr +=sizeof (*tp); - tp->core_siblings = addr; + tp->package_cpus_list = addr; addr += sz;if (has_die) { tp->die_siblings = addr;
tools/perf/util/cputopo.h +9 -2
@@ -5,10 +5,17 @@#include <linux/types.h> struct cpu_topology { -u32 core_sib; +/* The number of unique package_cpus_lists below. */ +u32 package_cpus_lists;u32 die_sib;u32 thread_sib; -char **core_siblings; +/* +* An array of strings where each string is unique and read from +* /sys/devices/system/cpu/cpuX/topology/package_cpus_list. From the ABI +* each of these is a human-readable list of CPUs sharing the same +* physical_package_id. The format is like 0-3, 8-11, 14,17. +*/ +const char **package_cpus_list;char **die_siblings;char **thread_siblings; };
tools/perf/util/header.c +3 -3
@@ -582,12 +582,12 @@ static int write_cpu_topology(struct feat_fd *ff,if (!tp)return -1 ; - ret =do_write (ff, &tp->core_sib,sizeof (tp->core_sib)); + ret =do_write (ff, &tp->package_cpus_lists,sizeof (tp->package_cpus_lists));if (ret <0 ) goto done; -for (i =0 ; i < tp->core_sib; i++) { - ret =do_write_string (ff, tp->core_siblings[i]); +for (i =0 ; i < tp->package_cpus_lists; i++) { + ret =do_write_string (ff, tp->package_cpus_list[i]);if (ret <0 ) goto done; }
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 3/7 perf cputopo: Match die_siblings to topology ABI name
In reply to Ian RogersThe topology name for die_siblings is die_cpus_list, use this for consistency and add documentation.
--- tools/perf/util/cputopo.c | 26 ++++++++++++-------------- tools/perf/util/cputopo.h | 11 +++++++++-- tools/perf/util/header.c | 8 ++++---- 3 files changed, 25 insertions(+), 20 deletions(-)
tools/perf/util/cputopo.c +12 -14
@@ -18,7 +18,7 @@ "%s/devices/system/cpu/cpu%d/topology/package_cpus_list"#define PACKAGE_CPUS_FMT_OLD \ "%s/devices/system/cpu/cpu%d/topology/core_siblings_list" -#define DIE_SIB_FMT \ +#define DIE_CPUS_FMT \ "%s/devices/system/cpu/cpu%d/topology/die_cpus_list"#define THRD_SIB_FMT \ "%s/devices/system/cpu/cpu%d/topology/thread_siblings_list"
@@ -73,10 +73,10 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu) ret =0 ; try_dies: -if (!tp->die_siblings) +if (!tp->die_cpus_list) goto try_threads; -scnprintf (filename,MAXPATHLEN ,DIE_SIB_FMT , +scnprintf (filename,MAXPATHLEN ,DIE_CPUS_FMT ,sysfs__mountpoint (), cpu); fp =fopen (filename,"r" );if (!fp)
@@ -91,13 +91,13 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu)if (p) *p ='\0' ; -for (i =0 ; i < tp->die_sib; i++) { -if (!strcmp (buf, tp->die_siblings[i])) +for (i =0 ; i < tp->die_cpus_lists; i++) { +if (!strcmp (buf, tp->die_cpus_list[i]))break ; } -if (i == tp->die_sib) { - tp->die_siblings[i] = buf; - tp->die_sib++; +if (i == tp->die_cpus_lists) { + tp->die_cpus_list[i] = buf; + tp->die_cpus_lists++; buf =NULL ; len =0 ; }
@@ -148,10 +148,8 @@ void cpu_topology__delete(struct cpu_topology *tp)for (i =0 ; i < tp->package_cpus_lists; i++)zfree (&tp->package_cpus_list[i]); -if (tp->die_sib) { -for (i =0 ; i < tp->die_sib; i++) -zfree (&tp->die_siblings[i]); - } +for (i =0 ; i < tp->die_cpus_lists; i++) +zfree (&tp->die_cpus_list[i]);for (i =0 ; i < tp->thread_sib; i++)zfree (&tp->thread_siblings[i]);
@@ -170,7 +168,7 @@ static bool has_die_topology(void)if (strncmp (uts.machine,"x86_64" ,6 ))return false; -scnprintf (filename,MAXPATHLEN ,DIE_SIB_FMT , +scnprintf (filename,MAXPATHLEN ,DIE_CPUS_FMT ,sysfs__mountpoint (),0 );if (access (filename,F_OK ) ==-1 )return false;
@@ -214,7 +212,7 @@ struct cpu_topology *cpu_topology__new(void) tp->package_cpus_list = addr; addr += sz;if (has_die) { - tp->die_siblings = addr; + tp->die_cpus_list = addr; addr += sz; } tp->thread_siblings = addr;
tools/perf/util/cputopo.h +9 -2
@@ -7,7 +7,8 @@struct cpu_topology {/* The number of unique package_cpus_lists below. */ u32 package_cpus_lists; -u32 die_sib; +/* The number of unique die_cpu_lists below. */ +u32 die_cpus_lists;u32 thread_sib;/* * An array of strings where each string is unique and read from
@@ -16,7 +17,13 @@ struct cpu_topology {* physical_package_id. The format is like 0-3, 8-11, 14,17. */ const char **package_cpus_list; -char **die_siblings; +/* +* An array of string where each string is unique and from +* /sys/devices/system/cpu/cpuX/topology/die_cpus_list. From the ABI +* each of these is a human-readable list of CPUs within the same die. +* The format is like 0-3, 8-11, 14,17. +*/ +const char **die_cpus_list;char **thread_siblings; };
tools/perf/util/header.c +4 -4
@@ -616,15 +616,15 @@ static int write_cpu_topology(struct feat_fd *ff,return ret; } -if (!tp->die_sib) +if (!tp->die_cpus_lists) goto done; - ret =do_write (ff, &tp->die_sib,sizeof (tp->die_sib)); + ret =do_write (ff, &tp->die_cpus_lists,sizeof (tp->die_cpus_lists));if (ret <0 ) goto done; -for (i =0 ; i < tp->die_sib; i++) { - ret =do_write_string (ff, tp->die_siblings[i]); +for (i =0 ; i < tp->die_cpus_lists; i++) { + ret =do_write_string (ff, tp->die_cpus_list[i]);if (ret <0 ) goto done; }
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 4/7 perf cputopo: Match thread_siblings to topology ABI name
In reply to Ian RogersThe topology name for thread_siblings is core_cpus_list, use this for consistency and add documentation.
--- tools/perf/util/cputopo.c | 26 +++++++++++++------------- tools/perf/util/cputopo.h | 11 +++++++++-- tools/perf/util/header.c | 6 +++--- 3 files changed, 25 insertions(+), 18 deletions(-)
tools/perf/util/cputopo.c +13 -13
@@ -20,10 +20,10 @@ "%s/devices/system/cpu/cpu%d/topology/core_siblings_list"#define DIE_CPUS_FMT \ "%s/devices/system/cpu/cpu%d/topology/die_cpus_list" -#define THRD_SIB_FMT \ - "%s/devices/system/cpu/cpu%d/topology/thread_siblings_list" -#define THRD_SIB_FMT_NEW \ +#define CORE_CPUS_FMT \ "%s/devices/system/cpu/cpu%d/topology/core_cpus_list" +#define CORE_CPUS_FMT_OLD \ + "%s/devices/system/cpu/cpu%d/topology/thread_siblings_list"#define NODE_ONLINE_FMT \ "%s/devices/system/node/online"#define NODE_MEMINFO_FMT \
@@ -104,10 +104,10 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu) ret =0 ; try_threads: -scnprintf (filename,MAXPATHLEN ,THRD_SIB_FMT_NEW , +scnprintf (filename,MAXPATHLEN ,CORE_CPUS_FMT ,sysfs__mountpoint (), cpu);if (access (filename,F_OK ) ==-1 ) { -scnprintf (filename,MAXPATHLEN ,THRD_SIB_FMT , +scnprintf (filename,MAXPATHLEN ,CORE_CPUS_FMT_OLD ,sysfs__mountpoint (), cpu); } fp =fopen (filename,"r" );
@@ -121,13 +121,13 @@ static int build_cpu_topology(struct cpu_topology *tp, int cpu)if (p) *p ='\0' ; -for (i =0 ; i < tp->thread_sib; i++) { -if (!strcmp (buf, tp->thread_siblings[i])) +for (i =0 ; i < tp->core_cpus_lists; i++) { +if (!strcmp (buf, tp->core_cpus_list[i]))break ; } -if (i == tp->thread_sib) { - tp->thread_siblings[i] = buf; - tp->thread_sib++; +if (i == tp->core_cpus_lists) { + tp->core_cpus_list[i] = buf; + tp->core_cpus_lists++; buf =NULL ; } ret =0 ;
@@ -151,8 +151,8 @@ void cpu_topology__delete(struct cpu_topology *tp)for (i =0 ; i < tp->die_cpus_lists; i++)zfree (&tp->die_cpus_list[i]); -for (i =0 ; i < tp->thread_sib; i++) -zfree (&tp->thread_siblings[i]); +for (i =0 ; i < tp->core_cpus_lists; i++) +zfree (&tp->core_cpus_list[i]);free (tp); }
@@ -215,7 +215,7 @@ struct cpu_topology *cpu_topology__new(void) tp->die_cpus_list = addr; addr += sz; } - tp->thread_siblings = addr; + tp->core_cpus_list = addr;for (i =0 ; i < nr; i++) {if (!cpu_map__has (map, i))
tools/perf/util/cputopo.h +9 -2
@@ -9,7 +9,8 @@ struct cpu_topology {u32 package_cpus_lists;/* The number of unique die_cpu_lists below. */ u32 die_cpus_lists; -u32 thread_sib; +/* The number of unique core_cpu_lists below. */ +u32 core_cpus_lists;/* * An array of strings where each string is unique and read from * /sys/devices/system/cpu/cpuX/topology/package_cpus_list. From the ABI
@@ -24,7 +25,13 @@ struct cpu_topology {* The format is like 0-3, 8-11, 14,17. */ const char **die_cpus_list; -char **thread_siblings; +/* +* An array of string where each string is unique and from +* /sys/devices/system/cpu/cpuX/topology/core_cpus_list. From the ABI +* each of these is a human-readable list of CPUs within the same +* core. The format is like 0-3, 8-11, 14,17. +*/ +const char **core_cpus_list; };struct numa_topology_node {
tools/perf/util/header.c +3 -3
@@ -591,12 +591,12 @@ static int write_cpu_topology(struct feat_fd *ff,if (ret <0 ) goto done; } - ret =do_write (ff, &tp->thread_sib,sizeof (tp->thread_sib)); + ret =do_write (ff, &tp->core_cpus_lists,sizeof (tp->core_cpus_lists));if (ret <0 ) goto done; -for (i =0 ; i < tp->thread_sib; i++) { - ret =do_write_string (ff, tp->thread_siblings[i]); +for (i =0 ; i < tp->core_cpus_lists; i++) { + ret =do_write_string (ff, tp->core_cpus_list[i]);if (ret <0 )break ; }
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 5/7 perf expr: Add literal values starting with #
In reply to Ian RogersIt is useful to have literal values for constants relating to topologies, SMT, etc. Make the parsing of literals shared code and add a lookup function. Move #smt_on to this function.
--- tools/perf/util/expr.c | 11 +++++++++++ tools/perf/util/expr.h | 1 + tools/perf/util/expr.l | 15 ++++++++++++++- tools/perf/util/expr.y | 9 ++++----- 4 files changed, 30 insertions(+), 6 deletions(-)
tools/perf/util/expr.c +11 -0
@@ -9,9 +9,11 @@#include "expr.h" #include "expr-bison.h" #include "expr-flex.h" +#include "smt.h" #include <linux/kernel.h> #include <linux/zalloc.h> #include <ctype.h> +#include <math.h> #ifdef PARSER_DEBUG extern int expr_debug;
@@ -370,3 +372,12 @@ double expr_id_data__value(const struct expr_id_data *data)assert (data->kind ==EXPR_ID_DATA__REF_VALUE );return data->ref.val; } + +double expr__get_literal (const char *literal) +{ +if (!strcmp ("#smt_on" , literal)) +return smt_on () >0 ?1.0 :0.0 ; + +pr_err ("Unrecognized literal '%s'" , literal); +return NAN ; +}
tools/perf/util/expr.h +1 -0
@@ -55,5 +55,6 @@ int expr__find_ids(const char *expr, const char *one, struct expr_parse_ctx *ids);double expr_id_data__value (const struct expr_id_data *data); +double expr__get_literal (const char *literal);#endif
tools/perf/util/expr.l +14 -1
@@ -6,6 +6,7 @@ #include <linux/compiler.h> #include "expr.h" #include "expr-bison.h" +#include <math.h> char *expr_get_text(yyscan_t yyscanner); YYSTYPE *expr_get_lval(yyscan_t yyscanner);
@@ -77,6 +78,17 @@ static int str(yyscan_t scanner, int token, int runtime) yylval->str = normalize(yylval->str, runtime); return token; } + +static int literal(yyscan_t scanner) +{ + YYSTYPE *yylval = expr_get_lval(scanner); + + yylval->num = expr__get_literal(expr_get_text(scanner)); + if (isnan(yylval->num)) + return EXPR_ERROR; + + return LITERAL; +} %} number ([0-9]+\.?[0-9]*|[0-9]*\.?[0-9]+)
@@ -85,6 +97,7 @@ sch [-,=] spec \\{sch} sym [0-9a-zA-Z_\.:@?]+ symbol ({spec}|{sym})+ +literal #[0-9a-zA-Z_\.\-]+ %% struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);
@@ -94,7 +107,7 @@ max { return MAX; } min { return MIN; } if { return IF; } else { return ELSE; } -#smt_on { return SMT_ON; } +{literal} { return literal(yyscanner); } {number} { return value(yyscanner); } {symbol} { return str(yyscanner, ID, sctx->runtime); } "|" { return '|'; }
tools/perf/util/expr.y +4 -5
@@ -4,7 +4,6 @@ #include <assert.h> #include <math.h> #include "util/debug.h" -#include "smt.h" #define IN_EXPR_Y 1 #include "expr.h" %}
@@ -37,7 +36,7 @@ } ids; } -%token ID NUMBER MIN MAX IF ELSE SMT_ON D_RATIO EXPR_ERROR +%token ID NUMBER MIN MAX IF ELSE LITERAL D_RATIO EXPR_ERROR %left MIN MAX IF %left '|' %left '^'
@@ -46,7 +45,7 @@ %left '-' '+' %left '*' '/' '%' %left NEG NOT -%type <num> NUMBER +%type <num> NUMBER LITERAL %type <str> ID %destructor { free ($$); } <str> %type <ids> expr if_expr
@@ -280,9 +279,9 @@ expr: NUMBER $$ = union_expr($3, $5); } } -| SMT_ON +| LITERAL { - $$.val = smt_on() > 0 ? 1.0 : 0.0; + $$.val = $1; $$.ids = NULL; } ;
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerPATCH 6/7 perf expr: Add metric literals for topology.
In reply to Ian RogersAllow the number of cpus, cores, dies and packages to be queried by a metric expression.
--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)
tools/perf/tests/expr.c +11 -1
@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;
@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;
tools/perf/util/expr.c +27 -0
@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h"
@@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ; +if (!strcmp ("#num_cpus" , literal)) +return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new (); +if (!topology) { +pr_err ("Error creating CPU topology" ); +return NAN ; + } + } +if (!strcmp ("#num_packages" , literal)) +return topology->package_cpus_lists; +if (!strcmp ("#num_dies" , literal)) +return topology->die_cpus_lists; +if (!strcmp ("#num_cores" , literal)) +return topology->core_cpus_lists; +pr_err ("Unrecognized literal '%s'" , literal);return NAN ; }
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Re: [PATCH 6/7] perf expr: Add metric literals for topology.
Quoted 62 lines · Ian Rogers
Allow the number of cpus, cores, dies and packages to be queried by a metric expression.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)tools/perf/tests/expr.c +11
-1@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;tools/perf/util/expr.c +5
-0@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h" @@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ; +if (!strcmp ("#num_cpus" , literal))
tools/perf/util/expr.c double expr_id_data__value(const struct expr_id_data *data) · line 395
Quoted 10 lines
tools/perf/util/expr.c +10
-0+return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new ();
any chance we could propagate expr_scanner_ctx in here and store topology to it and release it at the end? I think we have several places like this, so it'd be nice not to make more if it's possible ;-)
thanks,
jirka
Quoted 18 lines
tools/perf/util/expr.c +12
-0+if (!topology) { +pr_err ("Error creating CPU topology" ); +return NAN ; + } + } +if (!strcmp ("#num_packages" , literal)) +return topology->package_cpus_lists; +if (!strcmp ("#num_dies" , literal)) +return topology->die_cpus_lists; +if (!strcmp ("#num_cores" , literal)) +return topology->core_cpus_lists; +pr_err ("Unrecognized literal '%s'" , literal);return NAN ; }Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerRe: [PATCH 6/7] perf expr: Add metric literals for topology.
Quoted 63 lines · Jiri Olsa
Quoted 61 lines · Ian Rogers
Allow the number of cpus, cores, dies and packages to be queried by a metric expression.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)tools/perf/tests/expr.c +11
-1@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;tools/perf/util/expr.c +4
-0@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h" @@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ;
Quoted 15 lines
tools/perf/util/expr.c double expr_id_data__value(const struct expr_id_data *data) · line 395
Quoted 11 lines
tools/perf/util/expr.c +11
-0+if (!strcmp ("#num_cpus" , literal)) +return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new ();any chance we could propagate
expr_scanner_ctxin here and store topology to it and release it at the end? I think we have several places like this, so it'd be nice not to make more if it's possible ;-)
The topology here is static and so will only get computed once per execution rather than once pre expression parse. I was worried about the cost of recomputing the topology for something like 'perf stat -I 1000 -M ...' in which case the static will do less recomputation.
Thanks,
Ian
Quoted 22 lines
thanks,
jirkaQuoted 18 lines
tools/perf/util/expr.c +12
-1+if (!topology) { +pr_err ("Error creating CPU topology" ); +return NAN ; + } + } +if (!strcmp ("#num_packages" , literal)) +return topology->package_cpus_lists; +if (!strcmp ("#num_dies" , literal)) +return topology->die_cpus_lists; +if (!strcmp ("#num_cores" , literal)) +return topology->core_cpus_lists; +pr_err ("Unrecognized literal '%s'" , literal);return NAN ; } --2.34.0.rc0.344.g81b53c2807-goog
Re: [PATCH 6/7] perf expr: Add metric literals for topology.
Quoted 80 lines · Ian Rogers
Quoted 63 lines · Jiri Olsa
Quoted 61 lines · Ian Rogers
Allow the number of cpus, cores, dies and packages to be queried by a metric expression.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)tools/perf/tests/expr.c +11
-1@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;tools/perf/util/expr.c +4
-0@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h" @@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ;Quoted 15 lines
tools/perf/util/expr.c double expr_id_data__value(const struct expr_id_data *data) · line 395
Quoted 11 lines
tools/perf/util/expr.c +11
-0+if (!strcmp ("#num_cpus" , literal)) +return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new ();any chance we could propagate
expr_scanner_ctxin here and store topology to it and release it at the end? I think we have several places like this, so it'd be nice not to make more if it's possible ;-)
Quoted 4 lines
The topology here is static and so will only get computed once per execution rather than once pre expression parse. I was worried about the cost of recomputing the topology for something like 'perf stat -I 1000 -M ...' in which case the static will do less recomputation.
can't we have the topology created/release on the top fo the parsing and released after all expressions are parsed?
or we could come up with some generic way to handle this kind of release
jirka
Quoted 27 lines
Thanks,
IanQuoted 22 lines
thanks,
jirkaQuoted 18 lines
tools/perf/util/expr.c +12
-1+if (!topology) { +pr_err ("Error creating CPU topology" ); +return NAN ; + } + } +if (!strcmp ("#num_packages" , literal)) +return topology->package_cpus_lists; +if (!strcmp ("#num_dies" , literal)) +return topology->die_cpus_lists; +if (!strcmp ("#num_cores" , literal)) +return topology->core_cpus_lists; +pr_err ("Unrecognized literal '%s'" , literal);return NAN ; } --2.34.0.rc0.344.g81b53c2807-goog
Ian Rogers
reviewerRe: [PATCH 6/7] perf expr: Add metric literals for topology.
Quoted 87 lines · Jiri Olsa
Quoted 80 lines · Ian Rogers
Quoted 63 lines · Jiri Olsa
Quoted 61 lines · Ian Rogers
Allow the number of cpus, cores, dies and packages to be queried by a metric expression.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)tools/perf/tests/expr.c +11
-1@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;tools/perf/util/expr.c +4
-0@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h" @@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ;Quoted 15 lines
tools/perf/util/expr.c double expr_id_data__value(const struct expr_id_data *data) · line 395
Quoted 11 lines
tools/perf/util/expr.c +11
-0+if (!strcmp ("#num_cpus" , literal)) +return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new ();any chance we could propagate
expr_scanner_ctxin here and store topology to it and release it at the end? I think we have several places like this, so it'd be nice not to make more if it's possible ;-)Quoted 4 lines
The topology here is static and so will only get computed once per execution rather than once pre expression parse. I was worried about the cost of recomputing the topology for something like 'perf stat -I 1000 -M ...' in which case the static will do less recomputation.
Quoted 4 lines
can't we have the topology created/release on the top fo the parsing and released after all expressions are parsed?
or we could come up with some generic way to handle this kind of release
Creating the topology at the top of metric parsing would incur a sysfs parsing cost on metrics regardless of whether they used the topology. I feel a lazy approach is better to avoid this cost. We could make this part of the expression context and lazily initialize it there. It'd be good to keep the expression context around in 'perf stat' in that case. I quite like this approach as really the topology should be part of the perf environment and part of the header. It is interesting to think how can we get metrics to work with 'perf stat record'. Currently you will get metrics for something like 'perf stat record -e instructions,cycles -a sleep 2; perf stat report', but none baked in metrics don't work and we don't even have the metric-id to make this work (even using '-M IPC' doesn't work). Putting the environment in the session, making the topology part of it, writing/reading data based on that seems like worthwhile clean up but beyond the scope of what I was trying to do here.
In terms of releasing for the sake of memory leaks, leak sanitizer doesn't treat memory allocated and reachable from a global as a leak. We can't precompute the topology and so this style of approach is what I'm used to. Putting the topology into the parsing context and lazily initializing would be the second best option imo, and I can do that if this is a blocker.
Thanks,
Ian
Quoted 30 lines
jirka
Quoted 27 lines
Thanks,
IanQuoted 22 lines
thanks,
jirkaQuoted 18 lines
tools/perf/util/expr.c +12
-1+if (!topology) { +pr_err ("Error creating CPU topology" ); +return NAN ; + } + } +if (!strcmp ("#num_packages" , literal)) +return topology->package_cpus_lists; +if (!strcmp ("#num_dies" , literal)) +return topology->die_cpus_lists; +if (!strcmp ("#num_cores" , literal)) +return topology->core_cpus_lists; +pr_err ("Unrecognized literal '%s'" , literal);return NAN ; } --2.34.0.rc0.344.g81b53c2807-goog
Re: [PATCH 6/7] perf expr: Add metric literals for topology.
Quoted 109 lines · Ian Rogers
Quoted 87 lines · Jiri Olsa
Quoted 80 lines · Ian Rogers
Quoted 63 lines · Jiri Olsa
Quoted 61 lines · Ian Rogers
Allow the number of cpus, cores, dies and packages to be queried by a metric expression.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++++++- tools/perf/util/expr.c | 27 +++++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 1 deletion(-)tools/perf/tests/expr.c +11
-1@@ -66,7 +66,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) {struct expr_id_data *val_ptr;const char *p; -double val; +double val, num_cpus, num_cores, num_dies, num_packages;int ret;struct expr_parse_ctx *ctx;@@ -161,6 +161,16 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) NULL, ctx) == 0);TEST_ASSERT_VAL ("find ids" ,hashmap__size (ctx->ids) ==0 ); +/* Test toplogy constants appear well ordered. */ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("#num_cpus" ,expr__parse (&num_cpus, ctx,"#num_cpus" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores" ,expr__parse (&num_cores, ctx,"#num_cores" ) ==0 ); +TEST_ASSERT_VAL ("#num_cpus >= #num_cores" , num_cpus >= num_cores); +TEST_ASSERT_VAL ("#num_dies" ,expr__parse (&num_dies, ctx,"#num_dies" ) ==0 ); +TEST_ASSERT_VAL ("#num_cores >= #num_dies" , num_cores >= num_dies); +TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 ); +TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +expr__ctx_free (ctx);return 0 ;tools/perf/util/expr.c +4
-0@@ -5,6 +5,8 @@#include <stdlib.h> #include <string.h> #include "metricgroup.h" +#include "cpumap.h" +#include "cputopo.h" #include "debug.h" #include "expr.h" #include "expr-bison.h" @@ -375,9 +377,34 @@ double expr_id_data__value(const struct expr_id_data *data)double expr__get_literal (const char *literal) { +static struct cpu_topology *topology; +if (!strcmp ("#smt_on" , literal))return smt_on () >0 ?1.0 :0.0 ;Quoted 15 lines
tools/perf/util/expr.c double expr_id_data__value(const struct expr_id_data *data) · line 395
Quoted 11 lines
tools/perf/util/expr.c +11
-0+if (!strcmp ("#num_cpus" , literal)) +return cpu__max_present_cpu (); + +/* +* Assume that topology strings are consistent, such as CPUs "0-1" +* wouldn't be listed as "0,1", and so after deduplication the number of +* these strings gives an indication of the number of packages, dies, +* etc. +*/ +if (!topology) { + topology =cpu_topology__new ();any chance we could propagate
expr_scanner_ctxin here and store topology to it and release it at the end? I think we have several places like this, so it'd be nice not to make more if it's possible ;-)Quoted 4 lines
The topology here is static and so will only get computed once per execution rather than once pre expression parse. I was worried about the cost of recomputing the topology for something like 'perf stat -I 1000 -M ...' in which case the static will do less recomputation.
Quoted 4 lines
can't we have the topology created/release on the top fo the parsing and released after all expressions are parsed?
or we could come up with some generic way to handle this kind of release
Creating the topology at the top of metric parsing would incur a sysfs parsing cost on metrics regardless of whether they used the topology. I feel a lazy approach is better to avoid this cost. We could make this part of the expression context and lazily initialize it there. It'd be good to keep the expression context around in 'perf stat' in that case. I quite like this approach as really the topology should be part of the perf environment and part of the header. It is interesting to think how can we get metrics to work with 'perf stat record'. Currently you will get metrics for something like 'perf stat record -e instructions,cycles -a sleep 2; perf stat report', but none baked in metrics don't work and we don't even have the metric-id to make this work (even using '-M IPC' doesn't work). Putting the environment in the session, making the topology part of it, writing/reading data based on that seems like worthwhile clean up but beyond the scope of what I was trying to do here.
Quoted 6 lines
In terms of releasing for the sake of memory leaks, leak sanitizer doesn't treat memory allocated and reachable from a global as a leak. We can't precompute the topology and so this style of approach is what I'm used to. Putting the topology into the parsing context and lazily initializing would be the second best option imo, and I can do that if this is a blocker.
ok, let's leave it as it is.. thanks for the details
jirka
Ian Rogers
reviewerPATCH 7/7 perf expr: Add source_count for aggregating events
In reply to Ian RogersEvents like uncore_imc/cas_count_read/ on Skylake open multiple events and then aggregate in the metric leader. To determine the average value per event the number of these events is needed. Add a source_count function that returns this value by counting the number of events with the given metric leader. For most events the value is 1 but for uncore_imc/cas_count_read/ it can yield values like 6.
Add a generic test, but manually tested with a test metric that uses the function.
--- tools/perf/tests/expr.c | 12 +++++++ tools/perf/util/evsel.c | 12 +++++++ tools/perf/util/evsel.h | 1 + tools/perf/util/expr.c | 27 ++++++++++++--- tools/perf/util/expr.h | 3 ++ tools/perf/util/expr.l | 1 + tools/perf/util/expr.y | 65 ++++++++++++++++++++--------------- tools/perf/util/stat-shadow.c | 7 +++- 8 files changed, 95 insertions(+), 33 deletions(-)
tools/perf/tests/expr.c +12 -0
@@ -171,6 +171,18 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 );TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +/* +* Source count returns the number of events aggregating in a leader +* event including the leader. Check parsing yields an id. +*/ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("source count" , +expr__find_ids ("source_count(EVENT1)" , +NULL , ctx) ==0 ); +TEST_ASSERT_VAL ("source count" ,hashmap__size (ctx->ids) ==1 ); +TEST_ASSERT_VAL ("source count" ,hashmap__find (ctx->ids,"EVENT1" , + (void **)&val_ptr)); +expr__ctx_free (ctx);return 0 ;
tools/perf/util/evsel.c +12 -0
@@ -3018,3 +3018,15 @@ void evsel__set_leader(struct evsel *evsel, struct evsel *leader) { evsel->core.leader = &leader->core; } + +int evsel__source_count (const struct evsel *evsel) +{ +struct evsel *pos; +int count =0 ; + +evlist__for_each_entry (evsel->evlist, pos) { +if (pos->metric_leader == evsel) + count++; + } +return count; +}
tools/perf/util/evsel.h +1 -0
@@ -484,6 +484,7 @@ struct evsel *evsel__leader(struct evsel *evsel);bool evsel__has_leader (struct evsel *evsel,struct evsel *leader);bool evsel__is_leader (struct evsel *evsel);void evsel__set_leader (struct evsel *evsel,struct evsel *leader); +int evsel__source_count (const struct evsel *evsel);/* * Macro to swap the bit-field postition and size.
tools/perf/util/expr.c +22 -5
@@ -23,7 +23,10 @@ extern int expr_debug; struct expr_id_data { union { - double val; + struct { + double val; + int source_count; + } val; struct { double val; const char *metric_name;
@@ -140,6 +143,13 @@ int expr__add_id(struct expr_parse_ctx *ctx, const char *id)/* Caller must make sure id is allocated */ int expr__add_id_val (struct expr_parse_ctx *ctx,const char *id,double val) +{ +return expr__add_id_val_source_count (ctx, id, val,/*source_count=*/ 1 ); +} + +/* Caller must make sure id is allocated */ +int expr__add_id_val_source_count (struct expr_parse_ctx *ctx,const char *id, +double val,int source_count) {struct expr_id_data *data_ptr =NULL , *old_data =NULL ;char *old_key =NULL ;
@@ -148,7 +158,8 @@ int expr__add_id_val(struct expr_parse_ctx *ctx, const char *id, double val) data_ptr =malloc (sizeof (*data_ptr));if (!data_ptr)return -ENOMEM ; - data_ptr->val = val; + data_ptr->val.val = val; + data_ptr->val.source_count = source_count; data_ptr->kind =EXPR_ID_DATA__VALUE ; ret = hashmap__set(ctx->ids, id, data_ptr,
@@ -244,7 +255,7 @@ int expr__resolve_id(struct expr_parse_ctx *ctx, const char *id,switch (data->kind) {case EXPR_ID_DATA__VALUE : -pr_debug2 ("lookup(%s): val %f\n" , id, data->val); +pr_debug2 ("lookup(%s): val %f\n" , id, data->val.val);break ;case EXPR_ID_DATA__REF : pr_debug2("lookup(%s): ref metric name %s\n", id,
@@ -255,7 +266,7 @@ int expr__resolve_id(struct expr_parse_ctx *ctx, const char *id,pr_debug ("%s failed to count\n" , id);return -1 ; } -pr_debug ("processing metric: %s EXIT: %f\n" , id, data->val); +pr_debug ("processing metric: %s EXIT: %f\n" , id, data->ref.val);break ;case EXPR_ID_DATA__REF_VALUE : pr_debug2("lookup(%s): ref val %f metric name %s\n", id,
@@ -370,11 +381,17 @@ int expr__find_ids(const char *expr, const char *one,double expr_id_data__value (const struct expr_id_data *data) {if (data->kind ==EXPR_ID_DATA__VALUE ) -return data->val; +return data->val.val;assert (data->kind ==EXPR_ID_DATA__REF_VALUE );return data->ref.val; } +double expr_id_data__source_count (const struct expr_id_data *data) +{ +assert (data->kind ==EXPR_ID_DATA__VALUE ); +return data->val.source_count; +} +double expr__get_literal (const char *literal) {static struct cpu_topology *topology;
tools/perf/util/expr.h +3 -0
@@ -40,6 +40,8 @@ void expr__ctx_free(struct expr_parse_ctx *ctx);void expr__del_id (struct expr_parse_ctx *ctx,const char *id);int expr__add_id (struct expr_parse_ctx *ctx,const char *id);int expr__add_id_val (struct expr_parse_ctx *ctx,const char *id,double val); +int expr__add_id_val_source_count (struct expr_parse_ctx *ctx,const char *id, +double val,int source_count);int expr__add_ref (struct expr_parse_ctx *ctx,struct metric_ref *ref);int expr__get_id (struct expr_parse_ctx *ctx,const char *id,struct expr_id_data **data);
@@ -55,6 +57,7 @@ int expr__find_ids(const char *expr, const char *one, struct expr_parse_ctx *ids);double expr_id_data__value (const struct expr_id_data *data); +double expr_id_data__source_count (const struct expr_id_data *data);double expr__get_literal (const char *literal);#endif
tools/perf/util/expr.l +1 -0
@@ -107,6 +107,7 @@ max { return MAX; } min { return MIN; } if { return IF; } else { return ELSE; } +source_count { return SOURCE_COUNT; } {literal} { return literal(yyscanner); } {number} { return value(yyscanner); } {symbol} { return str(yyscanner, ID, sctx->runtime); }
tools/perf/util/expr.y +38 -27
89 lines not on this page · raw message
tools/perf/util/stat-shadow.c +6 -1
@@ -828,10 +828,12 @@ static int prepare_metric(struct evsel **metric_events,struct saved_value *v;struct stats *stats;u64 metric_total =0 ; +int source_count;if (!strcmp (metric_events[i]->name,"duration_time" )) { stats = &walltime_nsecs_stats; scale =1e-9 ; + source_count =1 ; }else { v = saved_value_lookup(metric_events[i], cpu, false, STAT_NONE, 0, st,
@@ -840,6 +842,7 @@ static int prepare_metric(struct evsel **metric_events,break ; stats = &v->stats; scale =1.0 ; + source_count =evsel__source_count (metric_events[i]);if (v->metric_other) metric_total = v->metric_total;
@@ -848,7 +851,9 @@ static int prepare_metric(struct evsel **metric_events,if (!n)return -ENOMEM ; -expr__add_id_val (pctx, n, metric_total ? :avg_stats (stats) * scale); +expr__add_id_val_source_count (pctx, n, + metric_total ? :avg_stats (stats) * scale, + source_count); }for (j =0 ; metric_refs && metric_refs[j].metric_name; j++) {
Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog
Re: [PATCH 7/7] perf expr: Add source_count for aggregating events
Quoted 233 lines · Ian Rogers
Events like uncore_imc/cas_count_read/ on Skylake open multiple events and then aggregate in the metric leader. To determine the average value per event the number of these events is needed. Add a
source_countfunction that returns this value by counting the number of events with the given metric leader. For most events the value is 1 but for uncore_imc/cas_count_read/ it can yield values like 6.Add a generic test, but manually tested with a test metric that uses the function.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/tests/expr.c | 12 +++++++ tools/perf/util/evsel.c | 12 +++++++ tools/perf/util/evsel.h | 1 + tools/perf/util/expr.c | 27 ++++++++++++--- tools/perf/util/expr.h | 3 ++ tools/perf/util/expr.l | 1 + tools/perf/util/expr.y | 65 ++++++++++++++++++++--------------- tools/perf/util/stat-shadow.c | 7 +++- 8 files changed, 95 insertions(+), 33 deletions(-)tools/perf/tests/expr.c +12
-0@@ -171,6 +171,18 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)TEST_ASSERT_VAL ("#num_packages" ,expr__parse (&num_packages, ctx,"#num_packages" ) ==0 );TEST_ASSERT_VAL ("#num_dies >= #num_packages" , num_dies >= num_packages); +/* +* Source count returns the number of events aggregating in a leader +* event including the leader. Check parsing yields an id. +*/ +expr__ctx_clear (ctx); +TEST_ASSERT_VAL ("source count" , +expr__find_ids ("source_count(EVENT1)" , +NULL , ctx) ==0 ); +TEST_ASSERT_VAL ("source count" ,hashmap__size (ctx->ids) ==1 ); +TEST_ASSERT_VAL ("source count" ,hashmap__find (ctx->ids,"EVENT1" , + (void **)&val_ptr)); +expr__ctx_free (ctx);return 0 ;tools/perf/util/evsel.c +12
-0@@ -3018,3 +3018,15 @@ void evsel__set_leader(struct evsel *evsel, struct evsel *leader) { evsel->core.leader = &leader->core; } + +int evsel__source_count (const struct evsel *evsel) +{ +struct evsel *pos; +int count =0 ; + +evlist__for_each_entry (evsel->evlist, pos) { +if (pos->metric_leader == evsel) + count++; + } +return count; +}tools/perf/util/evsel.h +1
-0@@ -484,6 +484,7 @@ struct evsel *evsel__leader(struct evsel *evsel);bool evsel__has_leader (struct evsel *evsel,struct evsel *leader);bool evsel__is_leader (struct evsel *evsel);void evsel__set_leader (struct evsel *evsel,struct evsel *leader); +int evsel__source_count (const struct evsel *evsel);/* * Macro to swap the bit-field postition and size. tools/perf/util/expr.c +22
-5@@ -23,7 +23,10 @@ extern int expr_debug; struct expr_id_data { union { - double val; + struct { + double val; + int source_count; + } val; struct { double val; const char *metric_name;@@ -140,6 +143,13 @@ int expr__add_id(struct expr_parse_ctx *ctx, const char *id)/* Caller must make sure id is allocated */ int expr__add_id_val (struct expr_parse_ctx *ctx,const char *id,double val) +{ +return expr__add_id_val_source_count (ctx, id, val,/*source_count=*/ 1 ); +} + +/* Caller must make sure id is allocated */ +int expr__add_id_val_source_count (struct expr_parse_ctx *ctx,const char *id, +double val,int source_count) {struct expr_id_data *data_ptr =NULL , *old_data =NULL ;char *old_key =NULL ;@@ -148,7 +158,8 @@ int expr__add_id_val(struct expr_parse_ctx *ctx, const char *id, double val) data_ptr =malloc (sizeof (*data_ptr));if (!data_ptr)return -ENOMEM ; - data_ptr->val = val; + data_ptr->val.val = val; + data_ptr->val.source_count = source_count; data_ptr->kind =EXPR_ID_DATA__VALUE ; ret = hashmap__set(ctx->ids, id, data_ptr,@@ -244,7 +255,7 @@ int expr__resolve_id(struct expr_parse_ctx *ctx, const char *id,switch (data->kind) {case EXPR_ID_DATA__VALUE : -pr_debug2 ("lookup(%s): val %f\n" , id, data->val); +pr_debug2 ("lookup(%s): val %f\n" , id, data->val.val);break ;case EXPR_ID_DATA__REF : pr_debug2("lookup(%s): ref metric name %s\n", id,@@ -255,7 +266,7 @@ int expr__resolve_id(struct expr_parse_ctx *ctx, const char *id,pr_debug ("%s failed to count\n" , id);return -1 ; } -pr_debug ("processing metric: %s EXIT: %f\n" , id, data->val); +pr_debug ("processing metric: %s EXIT: %f\n" , id, data->ref.val);break ;case EXPR_ID_DATA__REF_VALUE : pr_debug2("lookup(%s): ref val %f metric name %s\n", id,@@ -370,11 +381,17 @@ int expr__find_ids(const char *expr, const char *one,double expr_id_data__value (const struct expr_id_data *data) {if (data->kind ==EXPR_ID_DATA__VALUE ) -return data->val; +return data->val.val;assert (data->kind ==EXPR_ID_DATA__REF_VALUE );return data->ref.val; } +double expr_id_data__source_count (const struct expr_id_data *data) +{ +assert (data->kind ==EXPR_ID_DATA__VALUE ); +return data->val.source_count; +} +double expr__get_literal (const char *literal) {static struct cpu_topology *topology;tools/perf/util/expr.h +3
-0@@ -40,6 +40,8 @@ void expr__ctx_free(struct expr_parse_ctx *ctx);void expr__del_id (struct expr_parse_ctx *ctx,const char *id);int expr__add_id (struct expr_parse_ctx *ctx,const char *id);int expr__add_id_val (struct expr_parse_ctx *ctx,const char *id,double val); +int expr__add_id_val_source_count (struct expr_parse_ctx *ctx,const char *id, +double val,int source_count);int expr__add_ref (struct expr_parse_ctx *ctx,struct metric_ref *ref);int expr__get_id (struct expr_parse_ctx *ctx,const char *id,struct expr_id_data **data);@@ -55,6 +57,7 @@ int expr__find_ids(const char *expr, const char *one, struct expr_parse_ctx *ids);double expr_id_data__value (const struct expr_id_data *data); +double expr_id_data__source_count (const struct expr_id_data *data);double expr__get_literal (const char *literal);#endif tools/perf/util/expr.l +1
-0@@ -107,6 +107,7 @@ max { return MAX; } min { return MIN; } if { return IF; } else { return ELSE; } +source_count { return SOURCE_COUNT; } {literal} { return literal(yyscanner); } {number} { return value(yyscanner); } {symbol} { return str(yyscanner, ID, sctx->runtime); }tools/perf/util/expr.y +23
-1@@ -3,6 +3,7 @@ #define YYDEBUG 1 #include <assert.h> #include <math.h> +#include <stdlib.h> #include "util/debug.h" #define IN_EXPR_Y 1 #include "expr.h"@@ -36,7 +37,7 @@ } ids; } -%token ID NUMBER MIN MAX IF ELSE LITERAL D_RATIO EXPR_ERROR +%token ID NUMBER MIN MAX IF ELSE LITERAL D_RATIO SOURCE_COUNT EXPR_ERROR %left MIN MAX IF %left '|' %left '^'@@ -82,6 +83,40 @@ static struct ids union_expr(struct ids ids1, struct ids ids2) return result; } +static struct ids handle_id(struct expr_parse_ctx *ctx, char *id, + bool compute_ids, bool source_count) +{ + struct ids result; + if (!compute_ids) { + /* + * Compute the event's value from ID. If the ID isn't known then + * it isn't used to compute the formula so set to NAN. + */ + struct expr_id_data *data; + + result.val = NAN; + if (expr__resolve_id(ctx, id, &data) == 0) { + result.val = source_count + ? expr_id_data__source_count(data) + : expr_id_data__value(data); + } + result.ids = NULL; + free(id); + } else { + /*
tools/perf/util/expr.y static struct ids union_expr(struct ids ids1, struct ids ids2) · line 116
Quoted 10 lines
tools/perf/util/expr.y +10
-0+ * Set the value to BOTTOM to show that any value is possible + * when the event is computed. Create a set of just the ID. + */ + result.val = BOTTOM; + result.ids = ids__new(); + if (!result.ids || ids__insert(result.ids, id)) { + pr_err("Error creating IDs for '%s'", id); + free(id); + } + }
could you please put the move of handle_id into separate patch? it''ll be easier to review the actual changes
thanks,
jirka
Quoted 80 lines
tools/perf/util/expr.y +5
-26+ return result; +} + /* * If we're not computing ids or $1 and $3 are constants, compute the new * constant value using OP. Its invariant that there are no ids. If computing@@ -167,32 +202,8 @@ expr: NUMBER $$.val = $1; $$.ids = NULL; } -| ID -{ - if (!compute_ids) { - /* - * Compute the event's value from ID. If the ID isn't known then - * it isn't used to compute the formula so set to NAN. - */ - struct expr_id_data *data; - - $$.val = NAN; - if (expr__resolve_id(ctx, $1, &data) == 0) - $$.val = expr_id_data__value(data); - - $$.ids = NULL; - free($1); - } else { - /* - * Set the value to BOTTOM to show that any value is possible - * when the event is computed. Create a set of just the ID. - */ - $$.val = BOTTOM; - $$.ids = ids__new(); - if (!$$.ids || ids__insert($$.ids, $1)) - YYABORT; - } -} +| ID { $$ = handle_id(ctx, $1, compute_ids, /*source_count=*/false); } +| SOURCE_COUNT '(' ID ')' { $$ = handle_id(ctx, $3, compute_ids, /*source_count=*/true); } | expr '|' expr { BINARY_LONG_OP($$, |, $1, $3); } | expr '&' expr { BINARY_LONG_OP($$, &, $1, $3); } | expr '^' expr { BINARY_LONG_OP($$, ^, $1, $3); }tools/perf/util/stat-shadow.c +6
-1@@ -828,10 +828,12 @@ static int prepare_metric(struct evsel **metric_events,struct saved_value *v;struct stats *stats;u64 metric_total =0 ; +int source_count;if (!strcmp (metric_events[i]->name,"duration_time" )) { stats = &walltime_nsecs_stats; scale =1e-9 ; + source_count =1 ; }else { v = saved_value_lookup(metric_events[i], cpu, false, STAT_NONE, 0, st,@@ -840,6 +842,7 @@ static int prepare_metric(struct evsel **metric_events,break ; stats = &v->stats; scale =1.0 ; + source_count =evsel__source_count (metric_events[i]);if (v->metric_other) metric_total = v->metric_total;@@ -848,7 +851,9 @@ static int prepare_metric(struct evsel **metric_events,if (!n)return -ENOMEM ; -expr__add_id_val (pctx, n, metric_total ? :avg_stats (stats) * scale); +expr__add_id_val_source_count (pctx, n, + metric_total ? :avg_stats (stats) * scale, + source_count); }for (j =0 ; metric_refs && metric_refs[j].metric_name; j++) {Signature · 1 line
2.34.0.rc0.344.g81b53c2807-goog