v3.8.0

What's Changed

Version 3.8.0 is a feature and maintenance release

New Features, Changes, and Enhancements

• metrics command now supports Intel Granite Rapids processors on Google Cloud (C4 instances)
• metrics command's TMA metrics for Granite Rapids updated
• Network IRQs table format improved to avoid one long line of data by adding separators that will allow wrapping
• metrics command no longer errors and exits when the PMU is determined to be in use, warning is generated instead
• Intel Clearwater Forest now recognized and identified by report and config commands
• Intel Granite Rapids D now recognized and identified by report and config commands
• Intel Arrow Lake CPUs now recognized and identified by report command
• AWS Graviton 4, ARM Neoverse-V2 CPUs now recognized and identified by report command

Fixes

• power and temperature benchmarks in report command now works on additional architectures by fixing turbostat output parsing
• NIC table fixed in report command
• race condition in config command fixed when setting multiple configuration options at the same time
• memory benchmark in report command fixed when output format changed in newer MLC release
• frequency benchmark in report command fixed when number of cores per die differs per die

Full Changelog: v3.7.0...v3.8.0

v3.7.0

What's Changed

Version 3.7.0 is a feature and maintenance release.

To install, download and extract the pre-built package (perfspect.tgz) from the Assets listed below.

New Features and Enhancements

• the metrics HTML report now supports comparing two sets of metrics
• metrics command can optionally expose a Prometheus compatible metrics endpoint using --prometheus-server and --prometheus-server-addr
• flame command can now target multiple PIDs using --pids
• flame command can now control the depth of the call stack using --max-depth
• eliminated the requirement to have Perl installed on the target for the flame command
• config command can now enable/disable c6 and c1-demotion
• config command can now configure LLC size on SRF and GNR
• config command can now enable/disable LLC prefetcher on SRF
• telemetry command now reports CPU temperature, IPC and C6 residency
• report command now includes vendor and model ID in the NIC table
• logs can now be directed to stdout using --log-stdout; useful when combined with the metrics prometheus server feature
• metrics command "PMU in use" error and exit changed to a warning

Fixes

• address problems found with collecting metrics for cgroups
• fix memory benchmark chart X-axis label from MB/s to GB/s
• fix index out of range error in renderXlsxTableMultiTarget
• fix determination of availability of fixed counters

Full Changelog: v3.6.1...v3.7.0

v3.6.1

Version 3.6.1 fixes a bug found in 3.6.0 when parsing non-padded HEX values for CPU frequencies.

To install, download and extract the pre-built package (perfspect.tgz) from the Assets listed below.

Full Changelog: v3.6.0...v3.6.1

v3.6.0

Version 3.6.0 is a feature and maintenance release.

To install, download and extract the pre-built package (perfspect.tgz) from the Assets listed below.

New Features & Enhancements

• The CPU frequency table from the report command now includes frequencies for SSE, AVX2, AVX512, and AMX, when supported by architecture
• Flamegraphs can now be limited to a specific process (PID)
• Prefetchers can be enabled/disabled with the config command
• A brief system configuration summary table has been added to the metrics, flame, lock, and telemetry reports
• Added preliminary support for the Intel Clearwater Forest CPU architecture
• The lock command can now retrieve a binary perf package that can be used for analysis off the target
• Added support for metrics, including per-transaction metrics, on EC2 m7a (AMD Genoa) and AMD Turin

Fixes

• The config command can now set the max core frequency on SRF and GNR
• The targets.yaml file no longer requires a value for the target name field

Breaking Changes

• Some flags for the config command have been renamed for consistency and readability. See perfspect config -h.

Full Changelog: v3.5.0...v3.6.0

v3.5.2

v3.5.2 is a bug-fix release (Note: v3.5.1 was a bad build/release and has since been deleted)

Two issues were found in 3.5.0 and are now fixed in 3.5.2.

• perfspect will exit with a panic when an incorrect command line argument is presented
• perfspect will exit with an error when falsely identifying the temp directory as being located on a file system mounted with 'noexec'

Full Changelog: v3.5.0...v3.5.2

v3.5.0

Version 3.5.0 is a feature and maintenance release with the following additions/fixes.

Breaking Change

• The --targettemp flag has been removed. Use the --tempdir flag to override the directory where collection scripts are executed.

New Features & Enhancements

• The --txnrate flag used with the metrics command now augments the metrics list with transaction-oriented metrics rather than replacing existing metrics.
• The --syslog flag redirects log output to the local syslog daemon. This is useful when running PerfSpect for long durations and/or running as a CRON job.
• Improved shutdown when PerfSpect receives SIGINT (ctrl-c).
• Added GNR prefetcher settings to report.
• Added clustering mode (SNC, UMA) for GNR and SRF to report.
• Added CPU frequency chart to telemetry report.
• Added table of network-related kernel parameters to report.
• Added TME (total memory encryption) on/off to report.
• Added TMA level 1 over time chart to the metrics HTML report.
• Added configured DIMM speed and DIMM rank to DIMM table.

Fixes

• Addressed incorrect measured CPU frequency chart on GNR when SNC is disabled.
• Addressed missing NIC information in report.
• Addressed error when /tmp is on a file system mounted with 'noexec' (use --tempdir to override).
• Addressed incorrect memory channels listed for SRF-AP.

Full Changelog: v3.4.0...v3.5.0

v3.4.0

Version 3.4.0 is a feature and maintenance release with the following additions/fixes.

New Features & Enhancements

• Gaudi device stats now included in the telemetry command report.
• Metrics command event data can now be re-processed so that a previously unknown transaction rate (--txnrate) can be applied.
• The telemetry command now accepts a duration value of zero (--duration 0) to run until interrupted by SIGINT (ctrl-c).
• The telemetry command HTML report now includes time stamps on the x-axis of charts.
• The config command now allows setting the compute and I/O die frequencies independently (SRF and GNR)
• The branch misprediction metric was added to the metrics report.
• The report command now includes the Speed Select Technology frequency table when it is enabled.
• Added insight entry to report command to warn when ELC is configured in latency-optimized mode and EPB is non-zero.
• The report and config commands now determine which EPB configuration value (OS or BIOS) is active and report and/or change the appropriate entry.
• Report command tables that are not relevant to a given CPU architecture are now not include in the output.

Fixes

• L3 per core reported by the report command was inaccurate on some CPU architectures
• On multi-socket systems where a socket has been disabled via BIOS, the microarchitecture may be reported incorrectly.

What's Changed

• enable post-processing of pre-collected metric events by @harp-intel in #192
• enable indefinite duration for telemetry collection by @harp-intel in #203
• show timestamps in metrics summary and telemetry charts by @harp-intel in #205
• refactor html report generation to reduce duplication by @harp-intel in #206
• add branch mispredict ratio metric by @harp-intel in #207
• use remote target's perf for metrics collection if it is installed and new enough by @harp-intel in #208
• Highlight notes, tips, and warnings in README by @harp-intel in #209
• Bump github.com/spf13/cobra from 1.8.1 to 1.9.1 by @dependabot in #210
• add speed select turbo frequency tables by @harp-intel in #211
• fix report for l3 size per core when l3 instances are used by multipl… by @harp-intel in #213
• Get and set compute and I/O die max/min frequencies independently by @harp-intel in #216
• add example output images to README by @harp-intel in #218
• refactor scripts to use templating by @harp-intel in #222
• use alternate EPB value when configured to do so by @harp-intel in #223
• report tables associated with CPU models by @harp-intel in #225
• fix GNR_X* microarchitecture detection by @harp-intel in #227
• collect and report Gaudi device telemetry by @harp-intel in #217

Full Changelog: v3.3.1...v3.4.0

v3.3.1

Maintenance/Bug Fix Release

What's Changed

• Bump golang.org/x/term from 0.28.0 to 0.29.0 by @dependabot in #193
• Bump golang.org/x/text from 0.21.0 to 0.22.0 by @dependabot in #194
• fix regression in metrics command causing error on some AWS m5 and m6 instance types by @harp-intel in #196
• address race condition in metrics event processing by @harp-intel in #198
• process last set of events consistently by @harp-intel in #199
• send sigkill to child processes when receive sigint by @harp-intel in #201

Full Changelog: v3.3.0...v3.3.1

v3.3.0

Features/Enhancements:

• add instruction mix reporting to telemetry command by @harp-intel in #179
• add storage performance benchmark to 'report' command by @harp-intel in #161
• add 2nd level TMA metrics to chart by @harp-intel in #165
• add metadata tab to metrics summary HTML report by @harp-intel in #171

Maintenance/Bug Fixes:

• accept metric list without including "metric_" prefix by @harp-intel in #155
• add support for customized "no data found" message for any table by @harp-intel in #156
• recognize L3 cache size in GiB by @harp-intel in #158
• add cpu def for Turin zen 5c by @harp-intel in #160
• add benchmark descriptions by @harp-intel in #163
• add pid to ssh control master file name so that it doesn't get reused… by @harp-intel in #169
• if instructions event cannot be collected then assume target is not s… by @harp-intel in #170
• remove parse error from report, add to log by @harp-intel in #175
• provide sudo password to script when needed by @harp-intel in #177
• change flame and lock help to show 'all' is default format option by @harp-intel in #184
• fix bug where higher granularity metrics are not properly printed by @harp-intel in #182
• don't use bc in script as it is not available, by default, on some Linux OS distributions by @harp-intel in #189
• Bump github.com/spf13/pflag from 1.0.5 to 1.0.6 by @dependabot in #190
• fix metrics socket and cpu granularity metric calculations by @harp-intel in #191

Full Changelog: v3.2.0...v3.3.0

v3.2.0

What's Changed

Features/Enhancements:

• support for GCP C4 instances by @harp-intel in #134
• add AMD Turin CPU identifier by @harp-intel in #127
• make sure PMUs are not in use when running the metrics command by @harp-intel in #144
• Enable interruption of PerfSpect with SIGINT (ctrl-c) when collecting data over SSH by @harp-intel in #145
• limit config flags to specific uarchs by @harp-intel in #149

Maintenance/Bug Fixes:

• consider pkg control value when getting and setting EPP by @harp-intel in #142
• clean up benchmark summary table when no min latency collected by @harp-intel in #129
• perfspect report flag 'all' is true by default by @harp-intel in #128
• assume events not supported on failure when loading metadata for metrics command by @harp-intel in #139
• Bump golang.org/x/term from 0.27.0 to 0.28.0 by @dependabot in #140
• fix metrics HTML title field by @harp-intel in #147
• filter out infinite values in metrics summary by @harp-intel in #152
• Update README.md by @HarpPDX in #136

Full Changelog: v3.1.0...v3.2.0