powerpc/perf: Add json file metric support for the hv_24x7 socket/chip level events
Subsystems (2)
v2 cover powerpc/perf: Add json file metric support for the hv_24x7 socket/chip level events
The hv_24×7 feature in IBM® POWER9™ processor-based servers provide the facility to continuously collect large numbers of hardware performance metrics efficiently and accurately.
First patch of the patchset fix inconsistent results we are getting when we run multiple 24x7 events.
Patchset adds json file metric support for the hv_24x7 socket/chip level events. "hv_24x7" pmu interface events needs system dependent parameter like socket/chip/core. For example, hv_24x7 chip level events needs specific chip-id to which the data is requested should be added as part of pmu events.
So to enable JSON file support to "hv_24x7" interface, patchset expose total number of sockets and chips per-socket details in sysfs files (sockets, chips) under "/sys/devices/hv_24x7/interface/".
To get sockets and number of chips per sockets, patchset adds a rtas call with token "PROCESSOR_MODULE_INFO" to get these details. Patchset also handles partition migration case to re-init these system depended parameters by adding proper calls in post_mobility_fixup() (mobility.c).
Patch 6 & 8 of the patchset handles perf tool plumbing needed to replace the "?" character in the metric expression to proper value and hv_24x7 json metric file for different Socket/chip resources.
Patch set also enable Hz/hz prinitg for --metric-only option to print metric data for bus frequency.
Changelog:
v1 -> v2
- Rename hv-24x7 metric json file as nest_metrics.json
Kajol Jain (8):
powerpc/perf/hv-24x7: Fix inconsistent output values incase multiple
hv-24x7 events run
powerpc/hv-24x7: Add rtas call in hv-24x7 driver to get processor
details
powerpc/hv-24x7: Add sysfs files inside hv-24x7 device to show
processor details
Documentation/ABI: Add ABI documentation for chips and sockets powerpc/hv-24x7: Update post_mobility_fixup() to handle migration perf/tools: Enhance JSON/metric infrastructure to handle "?" tools/perf: Enable Hz/hz prinitg for --metric-only option perf/tools/pmu-events/powerpc: Add hv_24x7 socket/chip level metric
events
.../sysfs-bus-event_source-devices-hv_24x7 | 14 +++ arch/powerpc/perf/hv-24x7.c | 96 ++++++++++++++- arch/powerpc/platforms/pseries/mobility.c | 12 ++ arch/powerpc/platforms/pseries/pseries.h | 3 + tools/perf/arch/powerpc/util/header.c | 40 +++++++ .../arch/powerpc/power9/nest_metrics.json | 19 +++ tools/perf/util/expr.h | 1 + tools/perf/util/expr.y | 17 ++- tools/perf/util/metricgroup.c | 112 +++++++++++------- tools/perf/util/metricgroup.h | 1 + tools/perf/util/stat-display.c | 2 - tools/perf/util/stat-shadow.c | 5 + 12 files changed, 277 insertions(+), 45 deletions(-) create mode 100644 tools/perf/pmu-events/arch/powerpc/power9/nest_metrics.json
Signature · 1 line
2.18.1
PATCH v2 1/8 powerpc/perf/hv-24x7: Fix inconsistent output values incase multiple hv-24x7 events run
Commit 2b206ee6b0df ("powerpc/perf/hv-24x7: Display change in counter values")' added to print _change_ in the counter value rather then raw value for 24x7 counters. Incase of transactions, the event count is set to 0 at the beginning of the transaction. It also sets the event's prev_count to the raw value at the time of initialization. Because of setting event count to 0, we are seeing some weird behaviour, whenever we run multiple 24x7 events at a time.
For example:
command#: ./perf stat -e "{hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/,
hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/}"
-C 0 -I 1000 sleep 100
1.000121704 120 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
1.000121704 5 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
2.000357733 8 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
2.000357733 10 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
3.000495215 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
3.000495215 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.000641884 56 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
4.000641884 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
5.000791887 18,446,744,073,709,551,616 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/Getting these large values in case we do -I.
As we are setting event_count to 0, for interval case, overall event_count is not coming in incremental order. As we may can get new delta lesser then previous count. Because of which when we print intervals, we are getting negative value which create these large values.
This patch rather then setting event_count to 0, it change local64_set to local64_add in function 'h_24x7_event_read'.
With this patch
In power9 platform
command#: ./perf stat -e "{hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/,
hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/}"
-C 0 -I 1000 sleep 100
1.000117685 93 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
1.000117685 1 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
2.000349331 98 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
2.000349331 2 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
3.000495900 131 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
3.000495900 4 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.000645920 204 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/
4.000645920 61 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=1/
4.284169997 22 hv_24x7/PM_MCS01_128B_RD_DISP_PORT01,chip=0/--- arch/powerpc/perf/hv-24x7.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/perf/hv-24x7.c +1 -1
@@ -1409,7 +1409,7 @@ static void h_24x7_event_read(struct perf_event *event)* that would require issuing a hcall, which would then * defeat the purpose of using the txn interface. */ -local64_set (&event->count,0 ); +local64_add (0 , &event->count); }put_cpu_var (hv_24x7_reqb);
Signature · 1 line
2.18.1
Re: [PATCH v2 1/8] powerpc/perf/hv-24x7: Fix inconsistent output values incase multiple hv-24x7 events run
Quoted 7 lines · Kajol Jain [[email protected]] wrote:
Commit
2b206ee6b0df("powerpc/perf/hv-24x7: Display change in counter values")' added to print _change_ in the counter value rather then raw value for 24x7 counters. Incase of transactions, the event count is set to 0 at the beginning of the transaction. It also sets the event'sprev_countto the raw value at the time of initialization. Because of setting event count to 0, we are seeing some weird behaviour, whenever we run multiple 24x7 events at a time.
Interesting. Are we taking delta of a delta and ending up with large negative values in the -I case? However...
<snip>
Quoted 11 lines
Signed-off-by Kajol Jain <[email protected]>--- arch/powerpc/perf/hv-24x7.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/powerpc/perf/hv-24x7.c +0
-0
arch/powerpc/perf/hv-24x7.c static void h_24x7_event_read(struct perf_event *event) · line 1412
Quoted 5 lines
arch/powerpc/perf/hv-24x7.c +1
-1@@ -1409,7 +1409,7 @@ static void h_24x7_event_read(struct perf_event *event)* that would require issuing a hcall, which would then * defeat the purpose of using the txn interface. */ -local64_set (&event->count,0 ); +local64_add (0 , &event->count);
... not sure, how adding zero to the count helps. Should we just remove the line (and the comment block above it)? Or does it help to clear the event count in ->start_txn() rather than on read()?
How does the change impact the counts when run without the -I?
Thanks for chasing this down.
Sukadev
Re: [PATCH v2 1/8] powerpc/perf/hv-24x7: Fix inconsistent output values incase multiple hv-24x7 events run
Quoted 9 lines · Sukadev Bhattiprolu
Quoted 7 lines · Kajol Jain [[email protected]] wrote:
Commit
2b206ee6b0df("powerpc/perf/hv-24x7: Display change in counter values")' added to print _change_ in the counter value rather then raw value for 24x7 counters. Incase of transactions, the event count is set to 0 at the beginning of the transaction. It also sets the event'sprev_countto the raw value at the time of initialization. Because of setting event count to 0, we are seeing some weird behaviour, whenever we run multiple 24x7 events at a time.
Quoted 3 lines
Interesting. Are we taking delta of a delta and ending up with large negative values in the -I case? However...
Hi Sukadev,
That's right, we are ending up in calculating delta of delta which may give us negative values
because of which we are getting these large values in -I case.
Quoted 8 lines
<snip>
Quoted 6 lines
Signed-off-by Kajol Jain <[email protected]>--- arch/powerpc/perf/hv-24x7.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 14 lines
arch/powerpc/perf/hv-24x7.c static void h_24x7_event_read(struct perf_event *event) · line 1412
Quoted 10 lines
arch/powerpc/perf/hv-24x7.c +1
-1@@ -1409,7 +1409,7 @@ static void h_24x7_event_read(struct perf_event *event)* that would require issuing a hcall, which would then * defeat the purpose of using the txn interface. */ -local64_set (&event->count,0 ); +local64_add (0 , &event->count);... not sure, how adding zero to the count helps. Should we just remove the line (and the comment block above it)? Or does it help to clear the event count in ->start_txn() rather than on
read()?
Its not impacting much as we are just adding and not setting event count.I think we can remove that line with the comment added. Will it be ok if I remove that whole part?
Quoted 2 lines
How does the change impact the counts when run without the -I?
There won't be much impact because from my understanding you did add (void)local64_xchg(&event->hw.prev_count, ct);, to print change value in your Commit 2b206ee6b0df in function 'h_24x7_event_init()'.
And we will go through this path in case of 'without -I'. Please let me know if my understanding is correct.
Thanks,
Kajol
Quoted 5 lines
Thanks for chasing this down.
Sukadev
PATCH v2 2/8 powerpc/hv-24x7: Add rtas call in hv-24x7 driver to get processor details
In reply to Kajol JainFor hv_24x7 socket/chip level events, specific chip-id to which the data requested should be added as part of pmu events. But number of chips/socket in the system details are not exposed.
Patch implements read_sys_info_pseries() to get system parameter values like number of sockets and chips per socket. Rtas_call with token "PROCESSOR_MODULE_INFO"
is used to get these values.
Sub-sequent patch exports these values via sysfs.
Patch also make these parameters default to 1.
--- arch/powerpc/perf/hv-24x7.c | 72 ++++++++++++++++++++++++ arch/powerpc/platforms/pseries/pseries.h | 3 + 2 files changed, 75 insertions(+)
arch/powerpc/perf/hv-24x7.c +72 -0
@@ -20,6 +20,11 @@#include <asm/io.h> #include <linux/byteorder/generic.h> +#ifdef CONFIG_PPC_RTAS +#include <asm/rtas.h> +#include <../../platforms/pseries/pseries.h> +#endif +#include "hv-24x7.h" #include "hv-24x7-catalog.h" #include "hv-common.h"
@@ -57,6 +62,69 @@ static bool is_physical_domain(unsigned domain) } } +#ifdef CONFIG_PPC_RTAS +#define PROCESSOR_MODULE_INFO 43 +#define PROCESSOR_MAX_LENGTH (8 * 1024) + +static int strbe16toh (const char *buf,int offset) +{ +return (buf[offset] <<8 ) + buf[offset +1 ]; +} + +static u32 physsockets;/* Physical sockets */ +static u32 physchips;/* Physical chips */ + +/* +* Function read_sys_info_pseries() make a rtas_call which require +* data buffer of size 8K. As standard 'rtas_data_buf' is of size +* 4K, we are adding new local buffer 'rtas_local_data_buf'. +*/ +char rtas_local_data_buf[PROCESSOR_MAX_LENGTH] __cacheline_aligned; + +/* +* read_sys_info_pseries() +* Retrieve the number of sockets and chips per socket details +* through the get-system-parameter rtas call. +*/ +void read_sys_info_pseries (void ) +{ +int call_status, len, ntypes; + +/* +* Making system parameter: chips and sockets default to 1. +*/ + physsockets =1 ; + physchips =1 ; +memset (rtas_local_data_buf,0 ,PROCESSOR_MAX_LENGTH ); +spin_lock (&rtas_data_buf_lock); + + call_status =rtas_call (rtas_token ("ibm,get-system-parameter" ),3 ,1 , +NULL , +PROCESSOR_MODULE_INFO , +__pa (rtas_local_data_buf), +PROCESSOR_MAX_LENGTH ); + +spin_unlock (&rtas_data_buf_lock); + +if (call_status !=0 ) { +pr_info ("%s %s Error calling get-system-parameter (0x%x)\n" , + __FILE__, __func__, call_status); + }else { + rtas_local_data_buf[PROCESSOR_MAX_LENGTH -1 ] ='\0' ; + len =strbe16toh (rtas_local_data_buf,0 ); +if (len <6 ) +return ; + + ntypes =strbe16toh (rtas_local_data_buf,2 ); + +if (!ntypes) +return ; + physsockets =strbe16toh (rtas_local_data_buf,4 ); + physchips =strbe16toh (rtas_local_data_buf,6 ); + } +} +#endif /* CONFIG_PPC_RTAS */ +/* Domains for which more than one result element are returned for each event. */ static bool domain_needs_aggregation (unsigned int domain) {
@@ -1615,6 +1683,10 @@ static int hv_24x7_init(void)if (r)return r; +#ifdef CONFIG_PPC_RTAS +read_sys_info_pseries (); +#endif +return 0 ; }
arch/powerpc/platforms/pseries/pseries.h +3 -0
@@ -19,6 +19,9 @@ extern void request_event_sources_irqs(struct device_node *np,struct pt_regs ;extern int pSeries_system_reset_exception (struct pt_regs *regs); +#ifdef CONFIG_PPC_RTAS +extern void read_sys_info_pseries (void ); +#endif extern int pSeries_machine_check_exception (struct pt_regs *regs);extern long pseries_machine_check_realmode (struct pt_regs *regs);
Signature · 1 line
2.18.1
PATCH v2 3/8 powerpc/hv-24x7: Add sysfs files inside hv-24x7 device to show processor details
In reply to Kajol JainTo expose the system dependent parameter like total number of sockets and numbers of chips per socket, patch adds two sysfs files. "sockets" and "chips" are added to /sys/devices/hv_24x7/interface/ of the "hv_24x7" pmu.
--- arch/powerpc/perf/hv-24x7.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
arch/powerpc/perf/hv-24x7.c +22 -0
@@ -454,6 +454,20 @@ static ssize_t device_show_string(struct device *dev,return sprintf (buf,"%s\n" , (char *)d->var); } +#ifdef CONFIG_PPC_RTAS +static ssize_t sockets_show (struct device *dev, +struct device_attribute *attr,char *buf) +{ +return sprintf (buf,"%d\n" , physsockets); +} + +static ssize_t chips_show (struct device *dev,struct device_attribute *attr, +char *buf) +{ +return sprintf (buf,"%d\n" , physchips); +} +#endif +static struct attribute *device_str_attr_create_ (char *name,char *str) {struct dev_ext_attribute *attr =kzalloc (sizeof (*attr),GFP_KERNEL );
@@ -1100,6 +1114,10 @@ PAGE_0_ATTR(catalog_len, "%lld\n", (unsigned long long)be32_to_cpu(page_0->length) * 4096); static BIN_ATTR_RO(catalog, 0/* real length varies */);static DEVICE_ATTR_RO (domains); +#ifdef CONFIG_PPC_RTAS +static DEVICE_ATTR_RO (sockets); +static DEVICE_ATTR_RO (chips); +#endif static struct bin_attribute *if_bin_attrs[] = { &bin_attr_catalog,
@@ -1110,6 +1128,10 @@ static struct attribute *if_attrs[] = { &dev_attr_catalog_len.attr, &dev_attr_catalog_version.attr, &dev_attr_domains.attr, +#ifdef CONFIG_PPC_RTAS + &dev_attr_sockets.attr, + &dev_attr_chips.attr, +#endif NULL, };
Signature · 1 line
2.18.1
PATCH v2 4/8 Documentation/ABI: Add ABI documentation for chips and sockets
In reply to Kajol JainAdd documentation for the following sysfs files:/sys/devices/hv_24x7/interface/chips,/sys/devices/hv_24x7/interface/sockets
--- .../testing/sysfs-bus-event_source-devices-hv_24x7 | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_24x7 +14 -0
@@ -22,6 +22,20 @@ Description: Exposes the "version" field of the 24x7 catalog. This is also extractable from the provided binary "catalog" sysfs entry. +What: /sys/devices/hv_24x7/interface/sockets +Date: December 2019 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: read only + This sysfs interface exposes the number of sockets present in the + system. + +What: /sys/devices/hv_24x7/interface/chips +Date: December 2019 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: read only + This sysfs interface exposes the number of chips per socket + present in the system. + What: /sys/bus/event_source/devices/hv_24x7/event_descs/<event-name> Date: February 2014 Contact: Linux on PowerPC Developer List <[email protected]>
Signature · 1 line
2.18.1
PATCH v2 5/8 powerpc/hv-24x7: Update post_mobility_fixup() to handle migration
In reply to Kajol JainFunction 'read_sys_info_pseries()' is added to get system parameter values like number of sockets and chips per socket.
and it gets these details via rtas_call with token "PROCESSOR_MODULE_INFO".
Incase lpar migrate from one system to another, system parameter details like chips per sockets or number of sockets might change. So, it needs to be re-initialized otherwise, these values corresponds to previous system values.
This patch adds a call to 'read_sys_info_pseries()' from 'post-mobility_fixup()' to re-init the physsockets and physchips values.
--- arch/powerpc/platforms/pseries/mobility.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
arch/powerpc/platforms/pseries/mobility.c +12 -0
@@ -371,6 +371,18 @@ void post_mobility_fixup(void)/* Possibly switch to a new RFI flush type */ pseries_setup_rfi_flush (); +/* +* Incase lpar migrate from one system to another, system +* parameter details like chips per sockets and number of sockets +* might change. So, it needs to be re-initialized otherwise these +* values corresponds to previous system. +* Here, adding a call to read_sys_info_pseries() declared in +* platforms/pseries/pseries.h to re-init the physsockets and +* physchips value. +*/ +if (IS_ENABLED (CONFIG_HV_PERF_CTRS ) &&IS_ENABLED (CONFIG_PPC_RTAS )) +read_sys_info_pseries (); +return ; }
Signature · 1 line
2.18.1
PATCH v2 6/8 perf/tools: Enhance JSON/metric infrastructure to handle "?"
In reply to Kajol JainPatch enhances current metric infrastructure to handle "?" in the metric expression. The "?" can be use for parameters whose value not known while creating metric events and which can be replace later at runtime to the proper value. It also add flexibility to create multiple events out of single metric event added in json file.
Patch adds function 'arch_get_runtimeparam' which is a arch specific function, returns the count of metric events need to be created. By default it return 1.
One loop is added in function 'metricgroup__add_metric', which create multiple events at run time depend on return value of 'arch_get_runtimeparam' and merge that event in 'group_list'.
This infrastructure needed for hv_24x7 socket/chip level events. "hv_24x7" chip level events needs specific chip-id to which the data is requested. Function 'arch_get_runtimeparam' implemented in header.c which extract number of sockets from sysfs file "sockets" under "/sys/devices/hv_24x7/interface/".
--- tools/perf/arch/powerpc/util/header.c | 40 +++++++++ tools/perf/util/expr.h | 1 + tools/perf/util/expr.y | 17 +++- tools/perf/util/metricgroup.c | 112 ++++++++++++++++---------- tools/perf/util/metricgroup.h | 1 + tools/perf/util/stat-shadow.c | 5 ++ 6 files changed, 134 insertions(+), 42 deletions(-)
tools/perf/arch/powerpc/util/header.c +40 -0
@@ -7,6 +7,11 @@#include <string.h> #include <linux/stringify.h> #include "header.h" +#include "metricgroup.h" +#include "evlist.h" +#include <dirent.h> +#include "pmu.h" +#include <api/fs/fs.h> #define mfspr (rn) ({unsigned long rval; \ asm volatile("mfspr %0," __stringify(rn) \
@@ -16,6 +21,8 @@#define PVR_VER (pvr) (((pvr) >> 16) & 0xFFFF)/* Version field */ #define PVR_REV (pvr) (((pvr) >> 0) & 0xFFFF)/* Revison field */ +#define SOCKETS_INFO_FILE_PATH "/devices/hv_24x7/interface/" +int get_cpuid (char *buffer,size_t sz) {
@@ -44,3 +51,36 @@ get_cpuid_str(struct perf_pmu *pmu __maybe_unused)return bufp; } + +int arch_get_runtimeparam (void ) +{ +int count =0 ; +DIR *dir; +char path[PATH_MAX ]; +const char *sysfs =sysfs__mountpoint (); +char filename[] ="sockets" ; +FILE *file; +char buf[16 ], *num; +int data; + +if (!sysfs) + goto out; +snprintf (path,PATH_MAX , +"%s" SOCKETS_INFO_FILE_PATH , sysfs); + dir =opendir (path); +if (!dir) + goto out; +strcat (path, filename); + file =fopen (path,"r" ); +if (!file) + goto out; + + data =fread (buf,1 ,sizeof (buf), file); +if (data ==0 ) + goto out; + count =strtol (buf, &num,10 ); +out: +if (!count) + count =1 ; +return count; +}
tools/perf/util/expr.h +1 -0
@@ -15,6 +15,7 @@ struct parse_ctx {struct parse_id ids[MAX_PARSE_ID ]; }; +extern int expr__runtimeparam;void expr__ctx_init (struct parse_ctx *ctx);void expr__add_id (struct parse_ctx *ctx,const char *id,double val);#ifndef IN_EXPR_Y
tools/perf/util/expr.y +16 -1
@@ -37,6 +37,8 @@ %type <num> expr if_expr %{ +int expr__runtimeparam; + static int expr__lex(YYSTYPE *res, const char **pp); static void expr__error(double *final_val __maybe_unused,
@@ -102,7 +104,7 @@ static int expr__symbol(YYSTYPE *res, const char *p, const char **pp) if (*p == '#') *dst++ = *p++; - while (isalnum(*p) || *p == '_' || *p == '.' || *p == ':' || *p == '@' || *p == '\\') { + while (isalnum(*p) || *p == '_' || *p == '.' || *p == ':' || *p == '@' || *p == '\\' || *p == '?') { if (p - s >= MAXIDLEN) return -1; /*
@@ -113,6 +115,19 @@ static int expr__symbol(YYSTYPE *res, const char *p, const char **pp) *dst++ = '/'; else if (*p == '\\') *dst++ = *++p; + else if (*p == '?') { + int size = snprintf(NULL, 0, "%d", expr__runtimeparam); + char * paramval = (char *)malloc(size); + int i = 0; + if(!paramval) + *dst++ = '0'; + else { + sprintf(paramval, "%d", expr__runtimeparam); + while(i < size) + *dst++ = paramval[i++]; + free(paramval); + } + } else *dst++ = *p; p++;
tools/perf/util/metricgroup.c +71 -41
129 lines not on this page · raw message
tools/perf/util/metricgroup.h +1 -0
@@ -34,4 +34,5 @@ int metricgroup__parse_groups(const struct option *opt,void metricgroup__print (bool metrics,bool groups,char *filter,bool raw,bool details);bool metricgroup__has_metric (const char *metric); +int arch_get_runtimeparam (void );#endif
tools/perf/util/stat-shadow.c +5 -0
@@ -780,7 +780,12 @@ static void generic_metric(struct perf_stat_config *config,if (!metric_events[i]) {const char *p = metric_expr; +if (strstr (p,"?" )) { +char *tmp =strrchr (metric_name,'_' ); + tmp++; + expr__runtimeparam =strtol (tmp, &tmp,10 ); + }if (expr__parse (&ratio, &pctx, &p) ==0 ) {char *unit;char metric_bf[64 ];
Signature · 1 line
2.18.1
PATCH v2 7/8 tools/perf: Enable Hz/hz prinitg for --metric-only option
In reply to Kajol JainCommit 54b5091606c18 ("perf stat: Implement --metric-only mode") added function 'valid_only_metric()' which drops "Hz" or "hz", if it is part of "ScaleUnit". This patch enable it since hv_24x7 supports couple of frequency events.
--- tools/perf/util/stat-display.c | 2 -- 1 file changed, 2 deletions(-)
tools/perf/util/stat-display.c +0 -2
@@ -236,8 +236,6 @@ static bool valid_only_metric(const char *unit)if (!unit)return false;if (strstr (unit,"/sec" ) || -strstr (unit,"hz" ) || -strstr (unit,"Hz" ) ||strstr (unit,"CPUs utilized" ))return false;return true;
Signature · 1 line
2.18.1
PATCH v2 8/8 perf/tools/pmu-events/powerpc: Add hv_24x7 socket/chip level metric events
In reply to Kajol JainThe hv_24×7 feature in IBM® POWER9™ processor-based servers provide the facility to continuously collect large numbers of hardware performance metrics efficiently and accurately.
This patch adds hv_24x7 metric file for different Socket/chip resources.
Result:
power9 platform:
command:# ./perf stat --metric-only -M Memory_RD_BW_Chip -C 0
-I 1000 sleep 1
time MB Memory_RD_BW_Chip_0 MB Memory_RD_BW_Chip_1 MB
1.000192635 0.4 0.0 1.001695883 0.0 0.0
--- .../arch/powerpc/power9/nest_metrics.json | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) create mode 100644 tools/perf/pmu-events/arch/powerpc/power9/nest_metrics.json
tools/perf/pmu-events/arch/powerpc/power9/nest_metrics.json new file +19 -0
@@ -0,0 +1,19 @@ +[ + { + "MetricExpr": "(hv_24x7@PM_MCS01_128B_RD_DISP_PORT01\\,chip\\=?@ + hv_24x7@PM_MCS01_128B_RD_DISP_PORT23\\,chip\\=?@ + hv_24x7@PM_MCS23_128B_RD_DISP_PORT01\\,chip\\=?@ + hv_24x7@PM_MCS23_128B_RD_DISP_PORT23\\,chip\\=?@)", + "MetricName": "Memory_RD_BW_Chip", + "MetricGroup": "Memory_BW", + "ScaleUnit": "1.6e-2MB" + }, + { + "MetricExpr": "(hv_24x7@PM_MCS01_128B_WR_DISP_PORT01\\,chip\\=?@ + hv_24x7@PM_MCS01_128B_WR_DISP_PORT23\\,chip\\=?@ + hv_24x7@PM_MCS23_128B_WR_DISP_PORT01\\,chip\\=?@ + hv_24x7@PM_MCS23_128B_WR_DISP_PORT23\\,chip\\=?@ )", + "MetricName": "Memory_WR_BW_Chip", + "MetricGroup": "Memory_BW", + "ScaleUnit": "1.6e-2MB" + }, + { + "MetricExpr": "(hv_24x7@PM_PB_CYC\\,chip\\=?@ )", + "MetricName": "PowerBUS_Frequency", + "ScaleUnit": "2.5e-7GHz" + } +]
Signature · 1 line
2.18.1