From: Kan Liang
This patch fixes a perf diff regression issue which was introduced by:
commit 5baecbcd9c9a ("perf symbols: we can now read separate debug-info
files based on a build ID")
The binary name could be same when perf diff different binaries. Build
id is used to distingui
From: Kan Liang
This patch introduces the concept of NET policy object and policy object
list.
The NET policy object is the instance of CPU/queue mapping. The object
can be shared between different tasks/sockets. So besides CPU and queue
information, the object also maintains a reference
From: Kan Liang
Signed-off-by: Kan Liang
---
Documentation/networking/netpolicy.txt | 157 +
1 file changed, 157 insertions(+)
create mode 100644 Documentation/networking/netpolicy.txt
diff --git a/Documentation/networking/netpolicy.txt
b/Documentation
From: Kan Liang
Current implementation searches the hash table to get assigned object
for each transmit/receive packet. It's not necessory, because the
assigned object usually remain unchanged. This patch store the assigned
queue to speed up the searching process.
But under certain situa
From: Kan Liang
This patches introduces functions to set and remove IRQ affinity
according to cpu and queue mapping.
The functions will not record the previous affinity status. After a
set/remove cycles, it will set the affinity on all online CPU with IRQ
balance enabling.
Signed-off-by: Kan
From: Kan Liang
Users may not want to change the source code to add per task net polic
support. Or they may want to change a running task's net policy. prctl
does not work for both cases.
This patch adds an interface in /proc, which can be used to set and
retrieve policy of already ru
From: Kan Liang
User can write policy name to /proc/net/netpolicy/$DEV/policy to enable
net policy for specific device.
When the policy is enabled, the subsystem automatically disables IRQ
balance and set IRQ affinity. The object list is also generated
accordingly.
It is device driver&#
From: Kan Liang
When the device tries to transmit a packet, netdev_pick_tx is called to
find the available Tx queues. If the net policy is applied, it picks up
the assigned Tx queue from net policy subsystem, and redirect the
traffic to the assigned queue.
Signed-off-by: Kan Liang
---
include
From: Kan Liang
NET policy can not fulfill users request without limit, because of the
security consideration and device limitation. For security
consideration, the attacker may fake millions of per task/socket request
to crash the system. For device limitation, the flow director rules
number is
From: Kan Liang
Usually, application as a whole has specific requirement. Applying the
net policy to all sockets one by one in the application is too complex.
This patch introduces per task net policy to address this case.
Once the per task net policy is applied, all the sockets in the
From: Kan Liang
Introduce three NET policies
CPU policy: configure for higher throughput and lower CPU% (power
saving).
BULK policy: configure for highest throughput.
LATENCY policy: configure for lowest latency.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 3 +++
net/core
From: Kan Liang
For setting Rx queues, this patch configure Rx network flow
classification rules to redirect the packets to the assigned queue.
Since we may not get all the information required for rule until the
first packet arrived, it will add the rule after recvmsg. Also, to
avoid
From: Kan Liang
The socket/task can only be benefited when it register itself with
specific policy. If it's the first time to register, a record will be
created and inserted into RCU hash table. The record includes ptr,
policy and object information. ptr is the socket/task's pointe
From: Kan Liang
For CPU hotplug, the NET policy subsystem will rebuild the sys map and
object list.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 76
1 file changed, 76 insertions(+)
diff --git a/net/core/netpolicy.c b/net/core
From: Kan Liang
This patch introduces functions to enable and disable NET policy.
For enabling, it collects device and CPU information, setup CPU/queue
mapping, and set IRQ affinity accordingly.
For disabling, it removes the IRQ affinity and mapping information.
np_lock should protect the
From: Kan Liang
The network socket is the most basic unit which control the network
traffic. This patch introduces a new socket option SO_NETPOLICY to
set/get net policy for socket. so that the application can set its own
policy on socket to improve the network performance.
Per socket net policy
From: Kan Liang
Current implementation forces CPU and queue 1:1 mapping. This patch
introduces the function netpolicy_update_sys_map to create this mapping.
The result is stored in netpolicy_sys_info.
If the CPU count and queue count are different, the remaining
CPUs/queues are not used for now
From: Kan Liang
To achieve better network performance, the key step is to distribute the
packets to dedicated queues according to policy and system run time
status.
This patch provides an interface which can return the proper dedicated
queue for socket/task. Then the packets of the socket/task
From: Kan Liang
User can uses ethtool to set the channel number. This patch handles the
channel changes by rebuilding the object list.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 8
net/core/ethtool.c| 8 +++-
net/core/netpolicy.c | 1 +
3 files changed
From: Kan Liang
It is a big challenge to get good network performance. First, the network
performance is not good with default system settings. Second, it is too
difficult to do automatic tuning for all possible workloads, since workloads
have different requirements. Some workloads may want high
From: Kan Liang
Support NET device up/down/namechange in the NET policy code.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 66 +---
1 file changed, 58 insertions(+), 8 deletions(-)
diff --git a/net/core/netpolicy.c b/net/core/netpolicy.c
From: Kan Liang
MIX policy is combine of other policies. It allows different queue has
different policy. If MIX policy is applied,
/proc/net/netpolicy/$DEV/policy shows per queue policy.
Usually, the workloads requires either high throughput or low latency.
So for current implementation, MIX
From: Kan Liang
This patch introduce NET policy subsystem. If proc is supported in the
system, it creates netpolicy node in proc system.
Signed-off-by: Kan Liang
---
include/linux/netdevice.h | 7 +++
include/net/net_namespace.h | 3 ++
net/Kconfig | 7 +++
net/core
From: Kan Liang
This patch tries to initialize NET policy for all the devices in the
system. However, not all device drivers have NET policy support. For
those drivers who does not have NET policy support, the node will not be
showed in /proc/net/netpolicy/.
The device driver who has NET policy
From: Kan Liang
Net policy also needs to know CPU information. Currently, online
CPU number is enough.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/net/core/netpolicy.c b/net/core/netpolicy.c
index 7c34c8a..075aaca 100644
--- a/net
From: Kan Liang
Net policy needs to know device information. Currently, it's enough to
only get irq information of rx and tx queues.
This patch introduces ndo ops to do so, not ethtool ops.
Because there are already several ways to get irq information in
userspace. It's not necessory
From: Kan Liang
Some drivers like i40e driver does not support separate Tx and Rx queues
as channels. Using Rx queue to stand for the channels, if queue_pair is
set by driver.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 1 +
net/core/netpolicy.c | 3 +++
2 files changed, 4
From: Kan Liang
Net policy needs to know device information. Currently, it's enough to
only get irq information of rx and tx queues.
This patch introduces ndo ops to do so, not ethtool ops.
Because there are already several ways to get irq information in
userspace. It's not necessory
From: Kan Liang
This patch introduces the concept of NET policy object and policy object
list.
The NET policy object is the instance of CPU/queue mapping. The object
can be shared between different tasks/sockets. So besides CPU and queue
information, the object also maintains a reference
From: Kan Liang
User can uses ethtool to set the channel number. This patch handles the
channel changes by rebuilding the object list.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 8
net/core/ethtool.c| 8 +++-
net/core/netpolicy.c | 1 +
3 files changed
From: Kan Liang
For setting Rx queues, this patch configure Rx network flow
classification rules to redirect the packets to the assigned queue.
Since we may not get all the information required for rule until the
first packet arrived, it will add the rule after recvmsg. Also, to
avoid
From: Kan Liang
Current implementation searches the hash table to get assigned object
for each transmit/receive packet. It's not necessory, because the
assigned object usually remain unchanged. This patch store the assigned
queue to speed up the searching process.
But under certain situa
From: Kan Liang
Signed-off-by: Kan Liang
---
Documentation/networking/netpolicy.txt | 157 +
1 file changed, 157 insertions(+)
create mode 100644 Documentation/networking/netpolicy.txt
diff --git a/Documentation/networking/netpolicy.txt
b/Documentation
From: Kan Liang
Some drivers like i40e driver does not support separate Tx and Rx queues
as channels. Using Rx queue to stand for the channels, if queue_pair is
set by driver.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 1 +
net/core/netpolicy.c | 3 +++
2 files changed, 4
From: Kan Liang
Users may not want to change the source code to add per task net polic
support. Or they may want to change a running task's net policy. prctl
does not work for both cases.
This patch adds an interface in /proc, which can be used to set and
retrieve policy of already ru
From: Kan Liang
NET policy can not fulfill users request without limit, because of the
security consideration and device limitation. For security
consideration, the attacker may fake millions of per task/socket request
to crash the system. For device limitation, the flow director rules
number is
From: Kan Liang
Usually, application as a whole has specific requirement. Applying the
net policy to all sockets one by one in the application is too complex.
This patch introduces per task net policy to address this case.
Once the per task net policy is applied, all the sockets in the
From: Kan Liang
The network socket is the most basic unit which control the network
traffic. This patch introduces a new socket option SO_NETPOLICY to
set/get net policy for socket. so that the application can set its own
policy on socket to improve the network performance.
Per socket net policy
From: Kan Liang
To achieve better network performance, the key step is to distribute the
packets to dedicated queues according to policy and system run time
status.
This patch provides an interface which can return the proper dedicated
queue for socket/task. Then the packets of the socket/task
From: Kan Liang
When the device tries to transmit a packet, netdev_pick_tx is called to
find the available Tx queues. If the net policy is applied, it picks up
the assigned Tx queue from net policy subsystem, and redirect the
traffic to the assigned queue.
Signed-off-by: Kan Liang
---
include
From: Kan Liang
This patches introduces functions to set and remove IRQ affinity
according to cpu and queue mapping.
The functions will not record the previous affinity status. After a
set/remove cycles, it will set the affinity on all online CPU with IRQ
balance enabling.
Signed-off-by: Kan
From: Kan Liang
Introduce three NET policies
CPU policy: configure for higher throughput and lower CPU% (power
saving).
BULK policy: configure for highest throughput.
LATENCY policy: configure for lowest latency.
Signed-off-by: Kan Liang
---
include/linux/netpolicy.h | 3 +++
net/core
From: Kan Liang
This patch introduces functions to enable and disable NET policy.
For enabling, it collects device and CPU information, setup CPU/queue
mapping, and set IRQ affinity accordingly.
For disabling, it removes the IRQ affinity and mapping information.
np_lock should protect the
From: Kan Liang
MIX policy is combine of other policies. It allows different queue has
different policy. If MIX policy is applied,
/proc/net/netpolicy/$DEV/policy shows per queue policy.
Usually, the workloads requires either high throughput or low latency.
So for current implementation, MIX
From: Kan Liang
The socket/task can only be benefited when it register itself with
specific policy. If it's the first time to register, a record will be
created and inserted into RCU hash table. The record includes ptr,
policy and object information. ptr is the socket/task's pointe
From: Kan Liang
For CPU hotplug, the NET policy subsystem will rebuild the sys map and
object list.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 76
1 file changed, 76 insertions(+)
diff --git a/net/core/netpolicy.c b/net/core
From: Kan Liang
Support NET device up/down/namechange in the NET policy code.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 66 +---
1 file changed, 58 insertions(+), 8 deletions(-)
diff --git a/net/core/netpolicy.c b/net/core/netpolicy.c
From: Kan Liang
User can write policy name to /proc/net/netpolicy/$DEV/policy to enable
net policy for specific device.
When the policy is enabled, the subsystem automatically disables IRQ
balance and set IRQ affinity. The object list is also generated
accordingly.
It is device driver&#
From: Kan Liang
This patch introduce NET policy subsystem. If proc is supported in the
system, it creates netpolicy node in proc system.
Signed-off-by: Kan Liang
---
include/linux/netdevice.h | 7 +++
include/net/net_namespace.h | 3 ++
net/Kconfig | 7 +++
net/core
From: Kan Liang
This patch tries to initialize NET policy for all the devices in the
system. However, not all device drivers have NET policy support. For
those drivers who does not have NET policy support, the node will not be
showed in /proc/net/netpolicy/.
The device driver who has NET policy
From: Kan Liang
Current implementation forces CPU and queue 1:1 mapping. This patch
introduces the function netpolicy_update_sys_map to create this mapping.
The result is stored in netpolicy_sys_info.
If the CPU count and queue count are different, the remaining
CPUs/queues are not used for now
From: Kan Liang
(re-send to correct system time issue. Sorry for any inconvenience.)
It is a big challenge to get good network performance. First, the network
performance is not good with default system settings. Second, it is too
difficult to do automatic tuning for all possible workloads
From: Kan Liang
Net policy also needs to know CPU information. Currently, online
CPU number is enough.
Signed-off-by: Kan Liang
---
net/core/netpolicy.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/net/core/netpolicy.c b/net/core/netpolicy.c
index 7c34c8a..075aaca 100644
--- a/net
From: Kan Liang
For understanding how the workload maps to memory channels and hardware
behavior, it's very important to collect address maps with physical
addresses. For example, 3D XPoint access can only be found by filtering
the physical address.
However, perf doesn't collect physic
From: Kan Liang
For understanding how the workload maps to memory channels and hardware
behavior, it's very important to collect address maps with physical
addresses. For example, 3D XPoint access can only be found by filtering
the physical address.
However, perf doesn't collect physic
From: Kan Liang
Some users reported spurious NMI watchdog timeouts.
We now have more and more systems where the Turbo range is wide enough
that the NMI watchdog expires faster than the soft watchdog timer that
updates the interrupt tick the NMI watchdog relies on.
This problem was originally
From: Kan Liang
Some users reported spurious NMI watchdog timeouts.
We now have more and more systems where the Turbo range is wide enough
that the NMI watchdog expires faster than the soft watchdog timer that
updates the interrupt tick the NMI watchdog relies on.
This problem was originally
From: Kan Liang
There could be different types of memory in the system. E.g normal
System Memory, Persistent Memory. To understand how the workload maps to
those memories, it's important to know the I/O statistics on different
type of memorys. Perf can collect address maps with phy
From: Kan Liang
Currently, there is no way to measure the time cost in System management
mode (SMM) by perf.
Intel perfmon supports FREEZE_WHILE_SMM bit in IA32_DEBUGCTL. Once it sets,
the PMU core counters will freeze on SMI handler. But it will not have an
effect on free running counters. E.g
From: Kan Liang
Adding sysfs__write_int function to ease up writing int to sysfs.
New interface is:
int sysfs__write_int(const char *entry, int value);
Also, introducing filename__write_int which is useful for new helpers to
write sysctl values.
Signed-off-by: Kan Liang
---
tools/lib/api
From: Kan Liang
Implementing a new --smi-cost mode in perf stat to measure SMI cost.
During the measurement, the /sys/device/cpu/freeze_on_smi will be set.
The measurement can be done with one counter (unhalted core cycles),
and two free running MSR counters (IA32_APERF and SMI_COUNT).
In
From: Kan Liang
For Nehalem and Westmere, there is only one fixed counter for W-Box.
There is no index which is bigger than UNCORE_PMC_IDX_FIXED.
It is not correct to use >= to check fixed counter.
The code quality issue will bring problem when new counter index is
introduced.
Signed-off
From: Kan Liang
As of Skylake Server, there are a number of free-running counters in
each IIO Box that collect counts for per box IO clocks and per Port
Input/Output x BW/Utilization.
The free running counter is read-only and always active. Counting will
be suspended only when the IIO Box is
From: Kan Liang
Some uncore has custom pmu. For custom pmu, it does not need to
customize everything. For example, it only needs to customize init()
function for client IMC uncore. Other functions like
add()/del()/start()/stop()/read() can use generic code.
Expose the uncore_pmu_event_add/del
From: Kan Liang
There are two free running counters for client IMC uncore. The custom
event_init() function hardcode their index to 'UNCORE_PMC_IDX_FIXED' and
'UNCORE_PMC_IDX_FIXED + 1'. To support the 'UNCORE_PMC_IDX_FIXED + 1'
case, the generic uncore_perf_event_u
From: Kan Liang
The free running counter is read-only and always active. Current generic
uncore code does not support this kind of counters.
The free running counter is read-only. It cannot be enable/disable in
event_start/stop.
The free running counter event and free running counter are 1:1
From: Kan Liang
The counters in client IMC uncore are free running counters, not fixed
counters. It should be corrected. The new infrastructure for free
running counter should be applied.
Introduce free running counter type SNB_PCI_UNCORE_IMC_DATA for data
read and data write counters.
Keep
From: Kan Liang
There are a number of free running counters introduced for uncore, which
provide highly valuable information to a wide array of customers.
For example, Skylake Server has IIO free running counters to collect
Input/Output x BW/Utilization.
The precious generic counters could be
From: Kan Liang
There is no index which is bigger than UNCORE_PMC_IDX_FIXED. The only
exception is client IMC uncore. It has customized function to deal with
the 'UNCORE_PMC_IDX_FIXED + 1' case. It does not touch the generic code.
For generic code, it is not correct to use >= t
From: Kan Liang
There are a number of free running counters introduced for uncore, which
provide highly valuable information to a wide array of customers.
For example, Skylake Server has IIO freerunning counters to collect
Input/Output x BW/Utilization.
The precious generic counters could be
From: Kan Liang
The client IMC uncore obscurely hack the generic
uncore_perf_event_update to support the 'UNCORE_PMC_IDX_FIXED + 1' case.
The code quality issue will bring problem when new counter index is
introduced into generic code. For example, free running counter.
Introduce cust
From: Kan Liang
As of Skylake Server, there are a number of free-running counters in
each IIO Box that collect counts for per box IO clocks and per Port
Input/Output x BW/Utilization.
The free running counter is read-only and always active. Counting will
be suspended only when the IIO Box is
From: Kan Liang
Some uncore has custom pmu. Usually, it is not fully customized.
Most of the counter operation functions can still use the generic code.
For example, it only needs to customize init() function for client
IMC uncore. Other counter operation functions,
add()/del()/start()/stop
From: Kan Liang
The counters in client IMC uncore are free running counters, not fixed
counters. It should be corrected. The new infrastructure for free
running counter should be applied.
Introduce free running counter type SNB_PCI_UNCORE_IMC_DATA for data
read and data write counters.
Keep
From: Kan Liang
There could be different types of memory in the system. E.g normal
System Memory, Persistent Memory. To understand how the workload maps to
those memories, it's important to know the I/O statistics of them.
Perf can collect physical addresses, but those are raw data.
It
From: Kan Liang
Event select bit 7 'Use Occupancy' in PCU Box is not available for
counter 0 on BDX
Add a constraint to fix it.
Reported-by: Stephane Eranian
Signed-off-by: Kan Liang
Tested-by: Stephane Eranian
---
arch/x86/events/intel/uncore_snbep.c | 8
1 file
From: Kan Liang
Add a Intel event file for perf.
Signed-off-by: Kan Liang
---
.../pmu-events/arch/x86/goldmontplus/cache.json| 1453
.../pmu-events/arch/x86/goldmontplus/frontend.json | 62 +
.../pmu-events/arch/x86/goldmontplus/memory.json | 38 +
.../pmu
From: Kan Liang
The event synthesization multithreading is introduced in
("perf top optimization") https://lkml.org/lkml/2017/9/29/269
But it was not enabled for perf record. Because the process function
process_synthesized_event was not multithreading friendly.
The patch series t
From: Kan Liang
The process function process_synthesized_event writes the process
result to perf.data, which is not multithreading friendly.
Create per thread file to temporarily keep the processing result.
Write them to the perf.data at the end of event synthesization.
The new method doesn&#
From: Kan Liang
Using UINT_MAX to indicate the default thread#, which is the number of
online CPU.
Signed-off-by: Kan Liang
---
tools/perf/Documentation/perf-record.txt | 4
tools/perf/builtin-record.c | 13 +++--
2 files changed, 15 insertions(+), 2 deletions
From: Kan Liang
copyfile_offset could be used to merge per thread file to perf.data in
the following patch.
Signed-off-by: Kan Liang
---
tools/perf/util/util.c | 2 +-
tools/perf/util/util.h | 2 ++
2 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/tools/perf/util/util.c b/tools
From: Kan Liang
Pass the thread idx to process function, which is used by the following
patch.
Signed-off-by: Kan Liang
---
tools/perf/builtin-record.c | 4 +-
tools/perf/tests/dwarf-unwind.c | 3 +-
tools/perf/util/event.c | 98 ++---
tools
From: Kan Liang
For multithreading, the process function needs to know the thread
related information. E.g. saving the process result to the buffer or
file which belongs to specific thread.
Add struct thread_info parameter for process function.
Currently, it only includes thread index
From: Kan Liang
Remove the special codes in generic uncore_perf_event_update.
Introduce inline function to check the fixed counter event.
Signed-off-by: Kan Liang
---
Changes since V1:
- New file to address check event->hw.idx >= UNCORE_PMC_IDX_FIXED
arch/x86/events/intel/uncore
From: Kan Liang
As of Skylake Server, there are a number of free-running counters in
each IIO Box that collect counts for per box IO clocks and per Port
Input/Output x BW/Utilization.
Freerunning counters cannot be written by SW. Counting will be suspended
only when the IIO Box is powered down
From: Kan Liang
The clinet IMC uncore is the only one who claims two 'fixed counters'.
To specially handle it, event->hw.idx >= UNCORE_PMC_IDX_FIXED is used to
check fixed counters in the generic uncore_perf_event_update.
It does not have problem in current code. Because there
From: Kan Liang
There are a number of freerunning counters introduced for uncore.
For example, Skylake Server has IIO freerunning counters to collect
Input/Output x BW/Utilization.
The freerunning counter is similar as fixed counter, except it cannot
be written by SW. It needs to be specially
From: Kan Liang
Add backpointer to evlist, so we can easily access env when processing
something where we have a evsel or evlist.
Suggested-by: Arnaldo Carvalho de Melo
Signed-off-by: Kan Liang
---
tools/perf/util/evlist.h | 1 +
tools/perf/util/header.c | 1 +
2 files changed, 2 insertions
From: Kan Liang
This patch introduces generic FEAT for CPU attributes. For the patch
set, we only need cpu max frequency. But it can be easily extented to
support more other CPU attributes.
The cpu max frequency is from the first online cpu.
Signed-off-by: Kan Liang
---
tools/perf/util
From: Arnaldo Carvalho de Melo
So that functions that deal primarily with an evsel to access
information that concerns the whole evlist it is in.
Signed-off-by: Arnaldo Carvalho de Melo
Signed-off-by: Kan Liang
---
tools/perf/util/evlist.c | 2 ++
tools/perf/util/evsel.c | 2 ++
tools/perf
nv and add backpointer to evlist patches
Arnaldo Carvalho de Melo (1):
perf evsel: Add a backpointer to the evlist a evsel is in
Kan Liang (7):
perf,tools: introduce generic FEAT for CPU attributes
perf,tools: read msr pmu type from header.
perf,tools: rename perf_session_env to perf_env
pe
From: Kan Liang
Get msr pmu type when processing pmu_mappings
Signed-off-by: Kan Liang
---
tools/perf/util/header.c | 3 +++
tools/perf/util/header.h | 1 +
2 files changed, 4 insertions(+)
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index df4461a..8df0582 100644
--- a
From: Kan Liang
Rename perf_session_env to perf_env.
Suggested-by: Arnaldo Carvalho de Melo
Signed-off-by: Kan Liang
---
tools/perf/arch/common.c| 4 ++--
tools/perf/arch/common.h| 2 +-
tools/perf/ui/browser.h | 4 ++--
tools/perf/ui/browsers/header.c | 2
From: Kan Liang
Caculate freq/CPU%/CORE_BUSY% in add_entry_cb, and update the value in
he_stat.
Signed-off-by: Kan Liang
---
tools/perf/builtin-report.c | 36
tools/perf/util/sort.h | 3 +++
2 files changed, 39 insertions(+)
diff --git a/tools/perf
From: Kan Liang
Show frequency, CPU Utilization and percent performance for each symbol
in perf report by --stdio --show-freq-perf
In sampling group, only group leader do sampling. So only need to print
group leader's freq in --group.
Here is an example.
$ perf report --stdio --group -
From: Kan Liang
The group read results from cycles/ref-cycles/TSC/ASTATE/MSTATE event
can be used to calculate the frequency, CPU Utilization and percent
performance during each sampling period.
This patch shows them in report -D.
Here is an example:
$ perf record -e
'{cycles,ref-cycle
From: Kan Liang
This patch stores cpu socket_id and core_id in perf.date, and read them
to perf_env in header process.
Signed-off-by: Kan Liang
---
Changes since V1:
- Store core_id and socket_id in perf.date
tools/perf/util/header.c | 97
From: Jiri Olsa
This patch test cpu core_id and socket_id which are stored in perf_env.
Signed-off-by: Jiri Olsa
Signed-off-by: Kan Liang
---
Changes since jirka's original version
- Use pr_debug to replace fprintf
- Add date_size to avoid warning
- Introduce cpu_map, and compare co
From: Kan Liang
This patch stores cpu socket_id and core_id in perf.date, and read them
to perf_env in header process.
Although the changes modify the CPU_TOPOLOGY feature, the old perf still
can correctly read new perf.data. Because the new codes are added at the
end of the cpu_topology section
From: Kan Liang
This patch moves the codes which read core_id and socket_id into
separated functions, and expose them.
Signed-off-by: Kan Liang
---
tools/perf/util/cpumap.c | 51 +++-
tools/perf/util/cpumap.h | 2 ++
2 files changed, 35 insertions
1 - 100 of 2267 matches
Mail list logo
