AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Sync staging with mainline (#524)

Browse Source 
* External CI: rename pipeline to rocprofiler-compute (#463)

Signed-off-by: Daniel Su <danielsu@amd.com>

* Update webui branding (#459)

* Update name and icon for browser tab to rocprofiler-compute.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Update name and icon for browser tab to rocprofiler-compute.

Signed-off-by: xuchen-amd <xuchen@amd.com>

---------

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Update branding in documentation (#442)

* find/replace Omniperf to ROCm Compute Profiler

Signed-off-by: Peter Park <peter.park@amd.com>

* update name in Sphinx conf

Signed-off-by: Peter Park <peter.park@amd.com>

* mv what-is-omniperf.rst -> what-is-rocprof-compute.rst

Signed-off-by: Peter Park <peter.park@amd.com>

* update Tutorials section

Signed-off-by: Peter Park <peter.park@amd.com>

* add Omniperf as keyword to Conceptual section for internal search

Signed-off-by: Peter Park <peter.park@amd.com>

* update Reference section

Signed-off-by: Peter Park <peter.park@amd.com>

* black fmt conf.py

Signed-off-by: Peter Park <peter.park@amd.com>

* update profile mode and basic usage subsections

Signed-off-by: Peter Park <peter.park@amd.com>

* update how to use analyze mode subsection

Signed-off-by: Peter Park <peter.park@amd.com>

* update install section

Signed-off-by: Peter Park <peter.park@amd.com>

* fix sphinx warnings

Signed-off-by: Peter Park <peter.park@amd.com>

* fix cmd line examples in profile/mode.rst

Signed-off-by: Peter Park <peter.park@amd.com>

* update install decision tree image

Signed-off-by: Peter Park <peter.park@amd.com>

* fix TOC and index

Signed-off-by: Peter Park <peter.park@amd.com>

fix weird wording

* fix cli text: deriving rocprofiler-compute metrics...

Signed-off-by: Peter Park <peter.park@amd.com>

* update standalone-gui.rst

Signed-off-by: Peter Park <peter.park@amd.com>

* restore removed doc updates from #428

Signed-off-by: Peter Park <peter.park@amd.com>

* update ref to Omniperf in index.rst

Signed-off-by: Peter Park <peter.park@amd.com>

* fix grafana connection name to match image

Signed-off-by: Peter Park <peter.park@amd.com>

* update cmds in tutorials

Signed-off-by: Peter Park <peter.park@amd.com>

---------

Signed-off-by: Peter Park <peter.park@amd.com>

* MI300 roofline enablement in rocprofiler-compute (#470)

* MI300 roofline enablement in rocprofiler-compute

requirements.txt
- running some modules complained about numpy version too new, adding extra requirement that numpy be 1.x
pmc_roof_perf.txt
- adding TCC_BUBBLE_sum counter to profile
soc_gfx940.py
soc_gfx941.py
soc_gfx942.py
- remove console logs reading that roofline is temporarily disabled, uncommenting blocks that check for roofline csv and run roofline post-processing
roofline_calc.py
- add mi300 to supported soc
- add new calculation for hbm_data for MI300 using tcc_bubble_sum, checks if counter > 0 to use
- add to a few comments
roofline-ubuntu-20_04-mi300-rocm6
- binary for the ubuntu systems to enable mi300 roofline calculations from rocm-amdgpu-bench

Note- other distros will get roofline bins to enable mi300, but need to be further tested before putting into branch.

Signed-off-by: Carrie Fallows <carrie.fallows@amd.com>

* Reformatting roofline_calc.py

Signed-off-by: Carrie Fallows <carrie.fallows@amd.com>

---------

Signed-off-by: Carrie Fallows <carrie.fallows@amd.com>

* Update Python format checker (#471)

* Add pre commit hook for Python formatting

Signed-off-by: coleramos425 <colramos@amd.com>

* Update formatting workflow to run on latest Python and add isort formatter

Signed-off-by: coleramos425 <colramos@amd.com>

* Fix caught yaml formatting issues

* Update pyproject file

* Add pre-commit hook instruction to CONTRIBUTING guide

* Remove target-version from black pyproject.toml

* Fixed formatting errors found with black and isort

Signed-off-by: David Galiffi <David.Galiffi@amd.com>

* Run hook: Whitespaces, fix end of file spaces

---------

Signed-off-by: coleramos425 <colramos@amd.com>
Signed-off-by: David Galiffi <David.Galiffi@amd.com>
Co-authored-by: David Galiffi <David.Galiffi@amd.com>

* Bump cryptography from 43.0.0 to 43.0.1 in /docs/sphinx (#473)

Bumps [cryptography](https://github.com/pyca/cryptography) from 43.0.0 to 43.0.1.
- [Changelog](https://github.com/pyca/cryptography/blob/main/CHANGELOG.rst)
- [Commits](https://github.com/pyca/cryptography/compare/43.0.0...43.0.1)

---
updated-dependencies:
- dependency-name: cryptography
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Fix file permission on MI300 roofline binary (#477)

Signed-off-by: David Galiffi <David.Galiffi@amd.com>

* Removing numpy requirements of <2 (#478)

Checks are failing if version too high and no need for lower version

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Fix crash when loading web UI roofline for gfx942 (#479)

* Fix crash when loading web UI roofline for gfx942

* Fix formatting

Signed-off-by: benrichard-amd <ben.richard@amd.com>

* Make same changs for gfx940, gfx942.

Signed-off-by: benrichard-amd <ben.richard@amd.com>

* Fix formatting in soc_gfx940 and soc_gfx941.

Signed-off-by: benrichard-amd <ben.richard@amd.com>

---------

Signed-off-by: benrichard-amd <ben.richard@amd.com>

* Rebranding name change patch (#469)

* Patch in missed name change for rebranding.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Patch in missed name change for rebranding.

Signed-off-by: xuchen-amd <xuchen@amd.com>

---------

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Move dependabot.yml to .github/ and bump rocm-docs-core (#481)

* Move dependabot.yml to .github/

* Bump rocm-docs-core to 1.8.5

* Bump rocm-docs-core to 1.9.0

* Fix packaging for upgrading (#486)

Specify that "rocprofiler-compute" replaces / obsoletes the "omniperf" package.

* Renamed extension path from omniperf to rocprofiler_compute (#487)

Signed-off-by: Tim Gu <Tim.Gu@amd.com>

* MI300 rhel and sles roofline binaries (#480)

* Roofline bins for MI300 on rhel and sles distributions
Built from rocm-amdgpu-bench, tested on respective distro systems with MI300 hardware.

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Minor modifications removing hardcoded variables in roofline files.

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Modify test_profile_general.py ctest to include MI300 enablement (#498)

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* part 1 to support rocprofv3 (#492)

* rocprofv3 support initial commit

-Can run rocprofv3 but ultimately fails. rocprofv3 says the counter capacity
is exceeded and the output CSV file format is different from v1/v2.

* Add rocprofv3 detection so v2 can still be used

It's hacky but it'll do for now.

* Add code path to convert rocprofv3 JSON output into CSV

* Grab correct value for Queue ID

* Use _sum suffix to sum TCC counters

Previously we were specifying each channel for TCC counters. rocprofv3 does
not support specifing each TCC channel, and instead will auto sum given
the TCC counter name. The counter name with the _sum suffix is also
supported and is also supported in v1 and v2. So we will use the TCC
counter name with the _sum suffix.

* Fix incorrect counter outputs when using rocprofv3

In the JSON output some counters appear multime times and must be
summed to get the correct value. These summed values match the
rocprofv3 output in CSV mode and also match the rocprofv2
output.

* Remove duplicate Correlation_ID and Wave_Size in output

* Handle json output that does not contain any dispatches

Omniperf was assuming each JSON output from rocprofv3 would always contain
dispatches. This is not the case. For example, in a multi-process
workload where one of the processes does not dispatch any kernels. A JSON
file will still be output for this process but it will not contain any dispatches.

* Code cleanup

* Update search path for rocprofv3 results

Rocprofv3 was updated to include the hostname in the path where
it outputs results.

* Handle accumulate counters

In v1/v2 rocprof uses the SQ_ACCUM_PREV_HIRES counter for the accumualte
counters. v3 does not have this. So we need to define our own counters
in counter_defs.yaml. For this we use the counter name + _ACCUM, for
example SQ_INSTR_LEVEL_SMEM_ACCUM.

To use rocprofv3 you will need to update counter_defs.yaml to include
these new counter definitions.

* Use correct GPU ID

When converting JSON -> CSV we were assigning node_id to GPU_ID. Since
the JSON contains non-GPU devices, the node_id for GPUs might not
start at 0 as expected.

This commit maps the agent ID to the appropriate GPU ID.

* Parse scratch memory per work item from JSON

* Support rocprofv3 CSV parsing

JSON decoding is very slow for large files. Include support for parsing
rocprofv3 CSV output and make that the default.

CSV/JSON can be toggled via the ROCPROF_OUTPUT_FORMAT environment
variable e.g. ROCPROF_OUTPUT_FORMAT=csv or ROCPROF_OUTPUT_FORMAT=json

* black format after merge

* format isort

* change return of rocprof_cmd to try to resolve test's error

* hack to pick last part of rocminfo's name

* debug log of hacks

* Modify test_profile_general.py ctest to include MI300 enablement. Currently failing because of explicitly excluded roofline files for the soc and autofailed asserts for roof-only tests- originally in place because roofline was not enabled on mi300 yet.

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* black and isort formated

* corrected line of copyright

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
Co-authored-by: benrichard-amd <ben.richard@amd.com>
Co-authored-by: YANG WANG <ywang@ywang-ubuntu.amd.com>
Co-authored-by: Carrie Fallows <Carrie.Fallows@amd.com>

* fix for crash of timestamp of part 1 for rocprofv3 (#499)

* fix the error caused by ignoring the lack of counter csv file from rocprofv3 for timestamp

* isort and black formated

* quick fix for gfx906 roofline (#505)

* Multi node support (#503)

* [CTest] Pipeline failures for MI300 (#483)

* Propagate new chip_id logic to testing workflow

Signed-off-by: coleramos425 <colramos@amd.com>

* Add a debug line to tests

Signed-off-by: coleramos425 <colramos@amd.com>

* Trying to set rocprofv2 generally in CTest module

Signed-off-by: coleramos425 <colramos@amd.com>

* Remove temp debugging lines from CI

Signed-off-by: coleramos425 <colramos@amd.com>

* Add roofline entry for MI300 expected files in CI tests

Signed-off-by: coleramos425 <colramos@amd.com>

* Make num_devices modifier global in scope

Signed-off-by: coleramos425 <colramos@amd.com>

* Change kernel name in PyTest to confirm rocprofv2 bug

Related to https://ontrack-internal.amd.com/browse/SWDEV-503453

Signed-off-by: coleramos425 <colramos@amd.com>

---------

Signed-off-by: coleramos425 <colramos@amd.com>

* Spatial-multiplexing: part 1 profiling stage (#465)

* rocprofv3 support initial commit

-Can run rocprofv3 but ultimately fails. rocprofv3 says the counter capacity
is exceeded and the output CSV file format is different from v1/v2.

* Add rocprofv3 detection so v2 can still be used

It's hacky but it'll do for now.

* Add code path to convert rocprofv3 JSON output into CSV

* Grab correct value for Queue ID

* Use _sum suffix to sum TCC counters

Previously we were specifying each channel for TCC counters. rocprofv3 does
not support specifing each TCC channel, and instead will auto sum given
the TCC counter name. The counter name with the _sum suffix is also
supported and is also supported in v1 and v2. So we will use the TCC
counter name with the _sum suffix.

* Fix incorrect counter outputs when using rocprofv3

In the JSON output some counters appear multime times and must be
summed to get the correct value. These summed values match the
rocprofv3 output in CSV mode and also match the rocprofv2
output.

* Remove duplicate Correlation_ID and Wave_Size in output

* Handle json output that does not contain any dispatches

Omniperf was assuming each JSON output from rocprofv3 would always contain
dispatches. This is not the case. For example, in a multi-process
workload where one of the processes does not dispatch any kernels. A JSON
file will still be output for this process but it will not contain any dispatches.

* Code cleanup

* Update search path for rocprofv3 results

Rocprofv3 was updated to include the hostname in the path where
it outputs results.

* Handle accumulate counters

In v1/v2 rocprof uses the SQ_ACCUM_PREV_HIRES counter for the accumualte
counters. v3 does not have this. So we need to define our own counters
in counter_defs.yaml. For this we use the counter name + _ACCUM, for
example SQ_INSTR_LEVEL_SMEM_ACCUM.

To use rocprofv3 you will need to update counter_defs.yaml to include
these new counter definitions.

* debug code

* add logic code for multiplexing

* minor fix

* more fixes

* rocprofv3 support initial commit

-Can run rocprofv3 but ultimately fails. rocprofv3 says the counter capacity
is exceeded and the output CSV file format is different from v1/v2.

* Add rocprofv3 detection so v2 can still be used

It's hacky but it'll do for now.

* Add code path to convert rocprofv3 JSON output into CSV

* Grab correct value for Queue ID

* Use _sum suffix to sum TCC counters

Previously we were specifying each channel for TCC counters. rocprofv3 does
not support specifing each TCC channel, and instead will auto sum given
the TCC counter name. The counter name with the _sum suffix is also
supported and is also supported in v1 and v2. So we will use the TCC
counter name with the _sum suffix.

* Fix incorrect counter outputs when using rocprofv3

In the JSON output some counters appear multime times and must be
summed to get the correct value. These summed values match the
rocprofv3 output in CSV mode and also match the rocprofv2
output.

* Remove duplicate Correlation_ID and Wave_Size in output

* Handle json output that does not contain any dispatches

Omniperf was assuming each JSON output from rocprofv3 would always contain
dispatches. This is not the case. For example, in a multi-process
workload where one of the processes does not dispatch any kernels. A JSON
file will still be output for this process but it will not contain any dispatches.

* Code cleanup

* Update search path for rocprofv3 results

Rocprofv3 was updated to include the hostname in the path where
it outputs results.

* Handle accumulate counters

In v1/v2 rocprof uses the SQ_ACCUM_PREV_HIRES counter for the accumualte
counters. v3 does not have this. So we need to define our own counters
in counter_defs.yaml. For this we use the counter name + _ACCUM, for
example SQ_INSTR_LEVEL_SMEM_ACCUM.

To use rocprofv3 you will need to update counter_defs.yaml to include
these new counter definitions.

* count accu files as well

* Use correct GPU ID

When converting JSON -> CSV we were assigning node_id to GPU_ID. Since
the JSON contains non-GPU devices, the node_id for GPUs might not
start at 0 as expected.

This commit maps the agent ID to the appropriate GPU ID.

* fix error with csv file parse from json and merge during post-processing

* implemented parsing of csv files from v3 output for optimization

* Parse scratch memory per work item from JSON

* Support rocprofv3 CSV parsing

JSON decoding is very slow for large files. Include support for parsing
rocprofv3 CSV output and make that the default.

CSV/JSON can be toggled via the ROCPROF_OUTPUT_FORMAT environment
variable e.g. ROCPROF_OUTPUT_FORMAT=csv or ROCPROF_OUTPUT_FORMAT=json

* black format after merge

* format isort

* change return of rocprof_cmd to try to resolve test's error

* hack to pick last part of rocminfo's name

* debug log of hacks

* Modify test_profile_general.py ctest to include MI300 enablement. Currently failing because of explicitly excluded roofline files for the soc and autofailed asserts for roof-only tests- originally in place because roofline was not enabled on mi300 yet.

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* black and isort formated

* formated by isort and black

* change default rocprof's output to csv

* repaired crash caused by missing csv counter file when running for timestamp

* change name to spatial-multiplexing from multiplexing

* make necessary modification for review

* set the value of spatial_multiplexing in argument defautly to None

* repair the part that blocks regular pmc files' generation

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
Co-authored-by: benrichard-amd <ben.richard@amd.com>
Co-authored-by: fei.zheng <fei.zheng@amd.com>
Co-authored-by: YANG WANG <ywang@ywang-ubuntu.amd.com>
Co-authored-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Simple fix for gpu model value. (#508)

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Add FP64 to plot adhering to pdf name (#507)

* Replacing FP32-only plot with an FP32&FP64 combo plot. Results will likely be negligible but the plot name indicates both should be graphed.

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Remove duplicate AI plot to clean up fp32 fp64 graph

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Add gpu series for roofline (#510)

* Add gpu_series for roofline.

* Use gpu_series in path names for roofline.

* Fix  TCC on MI200 when introduce rocprofv3 (#509)

* quick fix for v2

* one more fix

* revert a bit

---------

Co-authored-by: ywang103-amd <ywang103@amd.com>

* Bump rocm-docs-core from 1.9.0 to 1.12.0 in /docs/sphinx (#511)

Bumps [rocm-docs-core](https://github.com/ROCm/rocm-docs-core) from 1.9.0 to 1.12.0.
- [Release notes](https://github.com/ROCm/rocm-docs-core/releases)
- [Changelog](https://github.com/ROCm/rocm-docs-core/blob/develop/CHANGELOG.md)
- [Commits](https://github.com/ROCm/rocm-docs-core/compare/v1.9.0...v1.12.0)

---
updated-dependencies:
- dependency-name: rocm-docs-core
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Update sample roofline plot img (#516)

* Modify path to use gpu_model instead of gpu_series to match other workload directory path creation/search points. Affects manual testing, does not seem to affect ctests. (#513)

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

* Improve formatting when displaying rocprof command. (#476)

* Improve formatting when displaying rocprof command.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Fix python formatting.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Strip unwanted characters (rocprofv1 specific) from rocprof commands.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Strip unwanted characters (rocprofv1 specific) from rocprof commands.

Signed-off-by: xuchen-amd <xuchen@amd.com>

* Save the unmodified arguments for rocprof for debug message display.

Signed-off-by: xuchen-amd <xuchen@amd.com>

---------

Signed-off-by: xuchen-amd <xuchen@amd.com>

* quick fix for mpi_support (#518)

* Pass accumulate counters to rocprofv3 using -E option (#522)

rocprofv3 has a new -E option where extra counters can be passed (see accum_counters.yaml) instead
of defining them in counter_defs.yaml.

* Unify all file handling with pathlib (#512)

* Replace occurences of os.path functions with equivalent functions from
  pathlib library

* Remove unwanted imports of os.path and os

* Add coding guidelines for using pathlib instead of os.path

* Auto sync staging and mainline on a weekly cadence (#517)

Signed-off-by: coleramos425 <colramos@amd.com>

---------

Signed-off-by: Daniel Su <danielsu@amd.com>
Signed-off-by: xuchen-amd <xuchen@amd.com>
Signed-off-by: Peter Park <peter.park@amd.com>
Signed-off-by: Carrie Fallows <carrie.fallows@amd.com>
Signed-off-by: coleramos425 <colramos@amd.com>
Signed-off-by: David Galiffi <David.Galiffi@amd.com>
Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
Signed-off-by: benrichard-amd <ben.richard@amd.com>
Signed-off-by: Tim Gu <Tim.Gu@amd.com>
Co-authored-by: Daniel Su <danielsu@amd.com>
Co-authored-by: xuchen-amd <xuchen@amd.com>
Co-authored-by: Peter Park <peter.park@amd.com>
Co-authored-by: cfallows-amd <Carrie.Fallows@amd.com>
Co-authored-by: David Galiffi <David.Galiffi@amd.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Ben Richard <143630488+benrichard-amd@users.noreply.github.com>
Co-authored-by: Tim Gu <Tim.Gu@amd.com>
Co-authored-by: ywang103-amd <ywang103@amd.com>
Co-authored-by: benrichard-amd <ben.richard@amd.com>
Co-authored-by: YANG WANG <ywang@ywang-ubuntu.amd.com>
Co-authored-by: Fei Zheng <44449748+feizheng10@users.noreply.github.com>
Co-authored-by: fei.zheng <fei.zheng@amd.com>
Co-authored-by: vedithal-amd <Vignesh.Edithal@amd.com>
This commit is contained in:
Cole Ramos
2025-01-02 13:29:47 -08:00
committed by GitHub
parent 71bc463921
commit 272e5b6e32
444 changed files with 9941 additions and 9233 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/how-to/analyze/cli.rst
+39 -39
View File
@@ -1,14 +1,14 @@
.. meta::
:description: Omniperf analysis: CLI analysis
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, command line, analyze, filtering, metrics, baseline, comparison
:description: ROCm Compute Profiler analysis: CLI analysis
:keywords: ROCm Compute Profiler, ROCm, profiler, tool, Instinct, accelerator, command line, analyze, filtering, metrics, baseline, comparison
************
CLI analysis
************
This section provides an overview of Omniperf's CLI analysis features.
This section provides an overview of ROCm Compute Profiler's CLI analysis features.
* :ref:`Derived metrics <cli-list-metrics>`: All of Omniperf's built-in metrics.
* :ref:`Derived metrics <cli-list-metrics>`: All of ROCm Compute Profiler's built-in metrics.
* :ref:`Baseline comparison <analysis-baseline-comparison>`: Compare multiple
runs in a side-by-side manner.
@@ -19,28 +19,28 @@ This section provides an overview of Omniperf's CLI analysis features.
* :ref:`Filtering <cli-analysis-options>`: Hone in on a particular kernel,
GPU ID, or dispatch ID via post-process filtering.
Run ``omniperf analyze -h`` for more details.
Run ``rocprof-compute analyze -h`` for more details.
.. _cli-walkthrough:
Walkthrough
===========
1. To begin, generate a high-level analysis report using Omniperf's ``-b`` (or ``--block``) flag.
1. To begin, generate a high-level analysis report using ROCm Compute Profiler's ``-b`` (or ``--block``) flag.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf analyze -p workloads/vcopy/MI200/ -b 2
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -b 2
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Analysis mode = cli
[analysis] deriving Omniperf metrics...
[analysis] deriving rocprofiler-compute metrics...
--------------------------------------------------------------------------------
0. Top Stats
@@ -134,19 +134,19 @@ Walkthrough
2. Use ``--list-metrics`` to generate a list of available metrics for inspection.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf analyze -p workloads/vcopy/MI200/ --list-metrics gfx90a
$ rocprof-compute analyze -p workloads/vcopy/MI200/ --list-metrics gfx90a
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Analysis mode = cli
[analysis] deriving Omniperf metrics...
[analysis] deriving rocprofiler-compute metrics...
0 -> Top Stats
1 -> System Info
2 -> System Speed-of-Light
@@ -186,13 +186,13 @@ Walkthrough
3. Choose your own customized subset of metrics with the ``-b`` (or ``--block``)
option. Or, build your own configuration following
`config_template <https://github.com/ROCm/omniperf/blob/amd-mainline/src/rocprof_compute_soc/analysis_configs/panel_config_template.yaml>`_.
`config_template <https://github.com/ROCm/rocprofiler-compute/blob/amd-mainline/src/rocprof_compute_soc/analysis_configs/panel_config_template.yaml>`_.
The following snippet shows how to generate a report containing only metric 2
(:doc:`System Speed-of-Light </conceptual/system-speed-of-light>`).
.. code-block:: shell
.. code-block:: shell-session
$ omniperf analyze -p workloads/vcopy/MI200/ -b 2
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -b 2
--------
Analyze
@@ -280,7 +280,7 @@ Walkthrough
4. Optimize the application, iterate, and re-profile to inspect performance
changes.
5. Redo a comprehensive analysis with Omniperf CLI at any optimization
5. Redo a comprehensive analysis with ROCm Compute Profiler CLI at any optimization
milestone.
.. _cli-analysis-options:
@@ -291,22 +291,22 @@ More analysis options
Single run
.. code-block:: shell
$ omniperf analyze -p workloads/vcopy/MI200/
$ rocprof-compute analyze -p workloads/vcopy/MI200/
List top kernels and dispatches
.. code-block:: shell
$ omniperf analyze -p workloads/vcopy/MI200/ --list-stats
$ rocprof-compute analyze -p workloads/vcopy/MI200/ --list-stats
List metrics
.. code-block:: shell
$ omniperf analyze -p workloads/vcopy/MI200/ --list-metrics gfx90a
$ rocprof-compute analyze -p workloads/vcopy/MI200/ --list-metrics gfx90a
Show System Speed-of-Light and CS_Busy blocks only
.. code-block:: shell
$ omniperf analyze -p workloads/vcopy/MI200/ -b 2 5.1.0
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -b 2 5.1.0
.. note::
@@ -319,10 +319,10 @@ Filter kernels
.. code-block::
$ omniperf analyze -p workloads/vcopy/MI200/ --list-stats
$ rocprof-compute analyze -p workloads/vcopy/MI200/ --list-stats
Analysis mode = cli
[analysis] deriving Omniperf metrics...
[analysis] deriving rocprofiler-compute metrics...
--------------------------------------------------------------------------------
Detected Kernels (sorted descending by duration)
@@ -344,12 +344,12 @@ Filter kernels
``vecCopy(double*, double*, double*, int, int) [clone .kd]`` at index ``0``.
Then, use this index to apply the filter via ``-k`` or ``--kernels``.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf analyze -p workloads/vcopy/MI200/ -k 0
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -k 0
Analysis mode = cli
[analysis] deriving Omniperf metrics...
[analysis] deriving rocprofiler-compute metrics...
--------------------------------------------------------------------------------
0. Top Stats
@@ -369,10 +369,10 @@ Filter kernels
Baseline comparison
.. code-block:: shell
omniperf analyze -p workload1/path/ -p workload2/path/
rocprof-compute analyze -p workload1/path/ -p workload2/path/
OR
.. code-block:: shell
omniperf analyze -p workload1/path/ -k 0 -p workload2/path/ -k 1
rocprof-compute analyze -p workload1/path/ -k 0 -p workload2/path/ -k 1
docs/how-to/analyze/grafana-gui.rst
+81 -80
View File
@@ -1,6 +1,7 @@
.. meta::
:description: Omniperf analysis: Grafana GUI
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, Grafana, panels, GUI, import
:description: ROCm Compute Profiler analysis: Grafana GUI
:keywords: Omniperf, ROCm Compute Profiler, ROCm, profiler, tool,
Instinct, accelerator, Grafana, panels, GUI, import
********************
Grafana GUI analysis
@@ -8,7 +9,7 @@ Grafana GUI analysis
Find setup instructions in :doc:`../../install/grafana-setup`.
The Omniperf Grafana analysis dashboard GUI supports the following features to
The ROCm Compute Profiler Grafana analysis dashboard GUI supports the following features to
facilitate MI accelerator performance profiling and analysis:
* System and hardware component (hardware block)
@@ -40,7 +41,7 @@ facilitate MI accelerator performance profiling and analysis:
* L2 Cache (TCC) (both aggregated and per-channel perf info)
See the full list of :ref:`Omniperf's analysis panels <panels>`.
See the full list of :ref:`ROCm Compute Profiler's analysis panels <panels>`.
.. _analysis-sol:
@@ -70,14 +71,14 @@ normalizations are available.
* ``per_second``
See :ref:`normalization-units` to learn more about Omniperf normalizations.
See :ref:`normalization-units` to learn more about ROCm Compute Profiler normalizations.
.. _analysis-baseline-comparison:
Baseline comparison
-------------------
Omniperf enables baseline comparison to allow checking A/B effect. Currently
ROCm Compute Profiler enables baseline comparison to allow checking A/B effect. Currently
baseline comparison is limited to the same :ref:`SoC <def-soc>`. Cross
comparison between SoCs is in development.
@@ -92,14 +93,14 @@ setup the following filters to allow fine grained comparisons:
* Dispatch ID filtering (regex filtering)
* Omniperf Panels (multi-selection)
* ROCm Compute Profiler Panels (multi-selection)
.. _analysis-regex-dispatch-id:
Regex-based dispatch ID filtering
---------------------------------
Omniperf allows filtering via Regular Expressions (regex), a standard Linux
ROCm Compute Profiler allows filtering via Regular Expressions (regex), a standard Linux
string matching syntax, based dispatch ID filtering to flexibly choose the
kernel invocations.
@@ -116,7 +117,7 @@ corresponding regex is : ``(1[7-9]|[23]\d|4[0-8])``.
Incremental profiling
---------------------
Omniperf supports incremental profiling to speed up performance analysis.
ROCm Compute Profiler supports incremental profiling to speed up performance analysis.
Refer to the :ref:`profiling-hw-component-filtering` section for this command.
@@ -145,7 +146,7 @@ Global variables and configurations
.. image:: ../../data/analyze/global_variables.png
:align: center
:alt: Omniperf global variables and configurations
:alt: ROCm Compute Profiler global variables and configurations
:width: 800
.. _grafana-gui-import:
@@ -153,7 +154,7 @@ Global variables and configurations
Grafana GUI import
------------------
The Omniperf database ``--import`` option imports the raw profiling data to
The ROCm Compute Profiler database ``--import`` option imports the raw profiling data to
Grafana's backend MongoDB database. This step is only required for Grafana
GUI-based performance analysis.
@@ -169,13 +170,13 @@ convention:
.. code-block:: shell
omniperf_<team>_<database>_<soc>
rocprofiler-compute_<team>_<database>_<soc>
For example:
.. code-block:: shell
omniperf_asw_vcopy_mi200
rocprofiler-compute_asw_vcopy_mi200
When using :ref:`database mode <modes-database>`, be sure to tailor the
connection options to the machine hosting your
@@ -187,10 +188,10 @@ called ``dummybox``.
.. code-block:: shell-session
$ omniperf database --help
$ rocprof-compute database --help
usage:
omniperf database <interaction type> [connection options]
rocprof-compute database <interaction type> [connection options]
@@ -198,9 +199,9 @@ called ``dummybox``.
Examples:
omniperf database --import -H pavii1 -u temp -t asw -w workloads/vcopy/mi200/
rocprof-compute database --import -H pavii1 -u temp -t asw -w workloads/vcopy/mi200/
omniperf database --remove -H pavii1 -u temp -w omniperf_asw_sample_mi200
rocprof-compute database --remove -H pavii1 -u temp -w rocprofiler-compute_asw_sample_mi200
-------------------------------------------------------------------------------
@@ -215,8 +216,8 @@ called ``dummybox``.
-s, --specs Print system specs.
Interaction Type:
-i, --import Import workload to Omniperf DB
-r, --remove Remove a workload from Omniperf DB
-i, --import Import workload to ROCm Compute Profiler DB
-r, --remove Remove a workload from ROCm Compute Profiler DB
Connection Options:
-H , --host Name or IP address of the server host.
@@ -228,22 +229,22 @@ called ``dummybox``.
--kernel-verbose Specify Kernel Name verbose level 1-5. Lower the level, shorter the kernel name. (DEFAULT: 5) (DISABLE: 5)
Omniperf import for vcopy:
^^^^^^^^^^^^^^^^^^^^^^^^^^
ROCm Compute Profiler import for vcopy:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.. code-block:: shell
.. code-block:: shell-session
$ omniperf database --import -H dummybox -u temp -t asw -w workloads/vcopy/mi200/
$ rocprof-compute database --import -H dummybox -u temp -t asw -w workloads/vcopy/mi200/
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Pulling data from /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Pulling data from /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
The directory exists
Found sysinfo file
KernelName shortening enabled
@@ -251,15 +252,15 @@ Omniperf import for vcopy:
Password:
Password received
-- Conversion & Upload in Progress --
0%| | 0/11 [00:00<?, ?it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_IFETCH_LEVEL.csv
9%|█████████████████▉ | 1/11 [00:00<00:01, 8.53it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/pmc_perf.csv
18%|███████████████████████████████████▊ | 2/11 [00:00<00:01, 6.99it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_SMEM.csv
27%|█████████████████████████████████████████████████████▋ | 3/11 [00:00<00:01, 7.90it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_LEVEL_WAVES.csv
36%|███████████████████████████████████████████████████████████████████████▋ | 4/11 [00:00<00:00, 8.56it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_LDS.csv
45%|█████████████████████████████████████████████████████████████████████████████████████████▌ | 5/11 [00:00<00:00, 9.00it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_VMEM.csv
55%|███████████████████████████████████████████████████████████████████████████████████████████████████████████▍ | 6/11 [00:00<00:00, 9.24it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/sysinfo.csv
64%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▎ | 7/11 [00:00<00:00, 9.37it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/roofline.csv
82%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏ | 9/11 [00:00<00:00, 12.60it/s]/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/timestamps.csv
0%| | 0/11 [00:00<?, ?it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_IFETCH_LEVEL.csv
9%|█████████████████▉ | 1/11 [00:00<00:01, 8.53it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/pmc_perf.csv
18%|███████████████████████████████████▊ | 2/11 [00:00<00:01, 6.99it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_SMEM.csv
27%|█████████████████████████████████████████████████████▋ | 3/11 [00:00<00:01, 7.90it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_LEVEL_WAVES.csv
36%|███████████████████████████████████████████████████████████████████████▋ | 4/11 [00:00<00:00, 8.56it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_LDS.csv
45%|█████████████████████████████████████████████████████████████████████████████████████████▌ | 5/11 [00:00<00:00, 9.00it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_INST_LEVEL_VMEM.csv
55%|███████████████████████████████████████████████████████████████████████████████████████████████████████████▍ | 6/11 [00:00<00:00, 9.24it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/sysinfo.csv
64%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▎ | 7/11 [00:00<00:00, 9.37it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/roofline.csv
82%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏ | 9/11 [00:00<00:00, 12.60it/s]/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/timestamps.csv
100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 11/11 [00:00<00:00, 11.05it/s]
9 collections added.
Workload name uploaded
@@ -267,8 +268,8 @@ Omniperf import for vcopy:
.. _panels:
Omniperf panels
---------------
ROCm Compute Profiler panels
----------------------------
There are currently 18 main panel categories available for analyzing the compute
workload performance. Each category contains several panels for close inspection
@@ -446,7 +447,7 @@ Kernel Time Histogram
.. figure:: ../../data/analyze/grafana/Kernel_time_histogram.png
:align: center
:alt: Kernel time histogram panel in Omniperf Grafana
:alt: Kernel time histogram panel in ROCm Compute Profiler Grafana
:width: 800
Mapping application kernel launches to execution duration.
@@ -456,7 +457,7 @@ Top Bottleneck Kernels
.. figure:: ../../data/analyze/grafana/top-stat_panel.png
:align: center
:alt: Top bottleneck kernels panel in Omniperf Grafana
:alt: Top bottleneck kernels panel in ROCm Compute Profiler Grafana
:width: 800
Top N kernels and relevant statistics. Sorted by total duration.
@@ -466,7 +467,7 @@ Top Bottleneck Dispatches
.. figure:: ../../data/analyze/grafana/Top_bottleneck_dispatches.png
:align: center
:alt: Top bottleneck dispatches panel in Omniperf Grafana
:alt: Top bottleneck dispatches panel in ROCm Compute Profiler Grafana
:width: 800
Top N kernel dispatches and relevant statistics. Sorted by total duration.
@@ -476,7 +477,7 @@ Current and Baseline Dispatch IDs (Filtered)
.. figure:: ../../data/analyze/grafana/Current_and_baseline_dispatch_ids.png
:align: center
:alt: Current and baseline dispatch IDs panel in Omniperf Grafana
:alt: Current and baseline dispatch IDs panel in ROCm Compute Profiler Grafana
:width: 800
List of all kernel dispatches.
@@ -488,10 +489,10 @@ System Speed-of-Light
.. figure:: ../../data/analyze/grafana/sol_panel.png
:align: center
:alt: System Speed-of-Light panel in Omniperf Grafana
:alt: System Speed-of-Light panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics from various sections of Omniperf’s profiling report.
Key metrics from various sections of ROCm Compute Profiler’s profiling report.
.. tip::
@@ -510,7 +511,7 @@ Memory Chart Analysis
.. figure:: ../../data/analyze/grafana/memory-chart_panel.png
:align: center
:alt: Memory Chart Analysis panel in Omniperf Grafana
:alt: Memory Chart Analysis panel in ROCm Compute Profiler Grafana
:width: 800
A graphical representation of performance data for memory blocks on the GPU.
@@ -523,7 +524,7 @@ Empirical Roofline Analysis
.. figure:: ../../data/analyze/grafana/roofline_panel.png
:align: center
:alt: Roofline Analysis panel in Omniperf Grafana
:alt: Roofline Analysis panel in ROCm Compute Profiler Grafana
:width: 800
Visualize achieved performance relative to a benchmarked peak performance.
@@ -543,7 +544,7 @@ Command Processor Fetcher
.. figure:: ../../data/analyze/grafana/cpc_panel.png
:align: center
:alt: Command Processor Fetcher panel in Omniperf Grafana
:alt: Command Processor Fetcher panel in ROCm Compute Profiler Grafana
:width: 800
Fetches commands out of memory to hand them over to the Command Processor
@@ -554,7 +555,7 @@ Command Processor Compute
.. figure:: ../../data/analyze/grafana/cpf_panel.png
:align: center
:alt: Command Processor Compute panel in Omniperf Grafana
:alt: Command Processor Compute panel in ROCm Compute Profiler Grafana
:width: 800
The micro-controller running the command processing firmware that decodes the
@@ -575,7 +576,7 @@ SPI Stats
.. figure:: ../../data/analyze/grafana/spi-stats_panel.png
:align: center
:alt: SPI Stats panel in Omniperf Grafana
:alt: SPI Stats panel in ROCm Compute Profiler Grafana
:width: 800
..
@@ -586,7 +587,7 @@ SPI Resource Allocation
.. figure:: ../../data/analyze/grafana/spi-resource-allocation_panel.png
:align: center
:alt: SPI Resource Allocation panel in Omniperf Grafana
:alt: SPI Resource Allocation panel in ROCm Compute Profiler Grafana
:width: 800
..
@@ -602,7 +603,7 @@ Wavefront Launch Stats
.. figure:: ../../data/analyze/grafana/wavefront-launch-stats_panel.png
:align: center
:alt: Wavefront Launch Stats panel in Omniperf Grafana
:alt: Wavefront Launch Stats panel in ROCm Compute Profiler Grafana
:width: 800
General information about the kernel launch.
@@ -616,7 +617,7 @@ Wavefront Runtime Stats
.. figure:: ../../data/analyze/grafana/wavefront-runtime-stats_panel.png
:align: center
:alt: Wavefront Runtime Stats panel in Omniperf Grafana.
:alt: Wavefront Runtime Stats panel in ROCm Compute Profiler Grafana.
:width: 800
High-level overview of the execution of wavefronts in a kernel.
@@ -635,7 +636,7 @@ Instruction Mix
.. figure:: ../../data/analyze/grafana/cu-inst-mix_panel.png
:align: center
:alt: Instruction Mix panel in Omniperf Grafana
:alt: Instruction Mix panel in ROCm Compute Profiler Grafana
:width: 800
Breakdown of the various types of instructions executed by the user’s kernel,
@@ -650,7 +651,7 @@ VALU Arithmetic Instruction Mix
.. figure:: ../../data/analyze/grafana/cu-value-arith-instr-mix_panel.png
:align: center
:alt: VALU Arithmetic Instruction Mix panel in Omniperf Grafana
:alt: VALU Arithmetic Instruction Mix panel in ROCm Compute Profiler Grafana
:width: 800
The various types of vector instructions that were issued to the vector
@@ -665,7 +666,7 @@ MFMA Arithmetic Instruction Mix
.. figure:: ../../data/analyze/grafana/cu-mafma-arith-instr-mix_panel.png
:align: center
:alt: MFMA Arithmetic Instruction Mix panel in Omniperf Grafana
:alt: MFMA Arithmetic Instruction Mix panel in ROCm Compute Profiler Grafana
:width: 800
The types of Matrix Fused Multiply-Add (MFMA) instructions that were issued.
@@ -679,7 +680,7 @@ VMEM Arithmetic Instruction Mix
.. figure:: ../../data/analyze/grafana/cu-vmem-instr-mix_panel.png
:align: center
:alt: VMEM Arithmetic Instruction Mix panel in Omniperf Grafana
:alt: VMEM Arithmetic Instruction Mix panel in ROCm Compute Profiler Grafana
:width: 800
The types of vector memory (VMEM) instructions that were issued.
@@ -698,7 +699,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/cu-sol_panel.png
:align: center
:alt: Speed-of-Light (CU) panel in Omniperf Grafana
:alt: Speed-of-Light (CU) panel in ROCm Compute Profiler Grafana
:width: 800
The number of floating-point and integer operations executed on the vector
@@ -714,7 +715,7 @@ Pipeline Stats
.. figure:: ../../data/analyze/grafana/cu-pipeline-stats_panel.png
:align: center
:alt: Pipeline Stats panel in Omniperf Grafana
:alt: Pipeline Stats panel in ROCm Compute Profiler Grafana
:width: 800
More detailed metrics to analyze the several independent pipelines found in
@@ -729,7 +730,7 @@ Arithmetic Operations
.. figure:: ../../data/analyze/grafana/cu-arith-ops_panel.png
:align: center
:alt: Arithmetic Operations panel in Omniperf Grafana
:alt: Arithmetic Operations panel in ROCm Compute Profiler Grafana
:width: 800
The total number of floating-point and integer operations executed in various
@@ -749,7 +750,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/lds-sol_panel.png
:align: center
:alt: Speed-of-Light (LDS) panel in Omniperf Grafana
:alt: Speed-of-Light (LDS) panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics for the Local Data Share (LDS) as a comparison with the peak
@@ -764,7 +765,7 @@ LDS Stats
.. figure:: ../../data/analyze/grafana/lds-stats_panel.png
:align: center
:alt: LDS Stats panel in Omniperf Grafana
:alt: LDS Stats panel in ROCm Compute Profiler Grafana
:width: 800
More detailed view of the Local Data Share (LDS) performance.
@@ -783,7 +784,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/instr-cache-sol_panel.png
:align: center
:alt: Speed-of-Light (instruction cache) panel in Omniperf Grafana
:alt: Speed-of-Light (instruction cache) panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics of the L1 Instruction (L1I) cache as a comparison with the peak
@@ -798,7 +799,7 @@ Instruction Cache Stats
.. figure:: ../../data/analyze/grafana/instr-cache-accesses_panel.png
:align: center
:alt: Instruction Cache Stats panel in Omniperf Grafana
:alt: Instruction Cache Stats panel in ROCm Compute Profiler Grafana
:width: 800
More detail on the hit/miss statistics of the L1 Instruction (L1I) cache.
@@ -821,7 +822,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/sl1d-sol_panel.png
:align: center
:alt: Speed-of-Light (SL1D) panel in Omniperf Grafana
:alt: Speed-of-Light (SL1D) panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics of the Scalar L1 Data (sL1D) cache as a comparison with the peak
@@ -836,7 +837,7 @@ Scalar L1D Cache Accesses
.. figure:: ../../data/analyze/grafana/sl1d-cache-accesses_panel.png
:align: center
:alt: Scalar L1D Cache Accesses panel in Omniperf Grafana
:alt: Scalar L1D Cache Accesses panel in ROCm Compute Profiler Grafana
:width: 800
More detail on the types of accesses made to the Scalar L1 Data (sL1D) cache,
@@ -851,7 +852,7 @@ Scalar L1D Cache - L2 Interface
.. figure:: ../../data/analyze/grafana/sl1d-l12-interface_panel.png
:align: center
:alt: Scalar L1D Cache - L2 Interface panel in Omniperf Grafana
:alt: Scalar L1D Cache - L2 Interface panel in ROCm Compute Profiler Grafana
:width: 800
More detail on the data requested across the Scalar L1 Data (sL1D) cache <->
@@ -871,7 +872,7 @@ Texture Addresser
.. figure:: ../../data/analyze/grafana/ta_panel.png
:align: center
:alt: Texture Addresser in Omniperf Grafana
:alt: Texture Addresser in ROCm Compute Profiler Grafana
:width: 800
Metric specific to texture addresser (TA) which receives commands (e.g.,
@@ -889,7 +890,7 @@ Texture Data
.. figure:: ../../data/analyze/grafana/td_panel.png
:align: center
:alt: Texture Data panel in Omniperf Grafana
:alt: Texture Data panel in ROCm Compute Profiler Grafana
:width: 800
Metrics specific to texture data (TD) which routes data back to the
@@ -909,7 +910,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/vl1d-sol_panel.png
:align: center
:alt: Speed-of-Light (VL1D) panel in Omniperf Grafana
:alt: Speed-of-Light (VL1D) panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics of the vector L1 data (vL1D) cache as a comparison with the peak
@@ -924,7 +925,7 @@ L1D Cache Stalls
.. figure:: ../../data/analyze/grafana/vl1d-cache-stalls_panel.png
:align: center
:alt: L1D Cache Stalls panel in Omniperf Grafana
:alt: L1D Cache Stalls panel in ROCm Compute Profiler Grafana
:width: 800
More detail on where vector L1 data (vL1D) cache is stalled in the pipeline,
@@ -955,7 +956,7 @@ L1D - L2 Transactions
.. figure:: ../../data/analyze/grafana/vl1d-l2-transactions_panel.png
:align: center
:alt: L1D - L2 Transactions in Omniperf Grafana
:alt: L1D - L2 Transactions in ROCm Compute Profiler Grafana
:width: 800
A more granular look at the types of requests made to the L2 cache.
@@ -969,7 +970,7 @@ L1D Addr Translation
.. figure:: ../../data/analyze/grafana/vl1d-addr-translation_panel.png
:align: center
:alt: L1D Addr Translation panel in Omniperf Grafana
:alt: L1D Addr Translation panel in ROCm Compute Profiler Grafana
:width: 800
After a vector memory instruction has been processed/coalesced by the address
@@ -995,7 +996,7 @@ Speed-of-Light
.. figure:: ../../data/analyze/grafana/l2-sol_panel.png
:align: center
:alt: Speed-of-Light (L2 cache) panel in Omniperf Grafana
:alt: Speed-of-Light (L2 cache) panel in ROCm Compute Profiler Grafana
:width: 800
Key metrics about the performance of the L2 cache, aggregated over all the
@@ -1011,7 +1012,7 @@ L2 Cache Accesses
.. figure:: ../../data/analyze/grafana/l2-accesses_panel.png
:align: center
:alt: L2 Cache Accesses panel in Omniperf Grafana
:alt: L2 Cache Accesses panel in ROCm Compute Profiler Grafana
:width: 800
Incoming requests to the L2 cache from the vector L1 data (vL1D) cache and
@@ -1026,7 +1027,7 @@ L2 - Fabric Transactions
.. figure:: ../../data/analyze/grafana/l2-fabric-transactions_panel.png
:align: center
:alt: L2 - Fabric Transactions panel in Omniperf Grafana
:alt: L2 - Fabric Transactions panel in ROCm Compute Profiler Grafana
:width: 800
More detail on the flow of requests through Infinity Fabric™.
@@ -1040,7 +1041,7 @@ L2 - Fabric Interface Stalls
.. figure:: ../../data/analyze/grafana/l2-fabric-interface-stalls_panel.png
:align: center
:alt: L2 - Fabric Interface Stalls panel in Omniperf Grafana
:alt: L2 - Fabric Interface Stalls panel in ROCm Compute Profiler Grafana
:width: 800
A breakdown of what types of requests in a kernel caused a stall
@@ -1065,7 +1066,7 @@ Aggregate Stats
.. figure:: ../../data/analyze/grafana/l2-per-channel-agg-stats_panel.png
:align: center
:alt: Aggregate Stats (L2 cache per channel) panel in Omniperf Grafana
:alt: Aggregate Stats (L2 cache per channel) panel in ROCm Compute Profiler Grafana
:width: 800
L2 Cache per channel performance at a glance. Metrics are aggregated over all available channels.
docs/how-to/analyze/mode.rst
+8 -8
View File
@@ -1,22 +1,22 @@
.. meta::
:description: How to use Omniperf's analyze mode
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, AMD,
:description: How to use ROCm Compute Profiler's analyze mode
:keywords: ROCm Compute Profiler, ROCm, profiler, tool, Instinct, accelerator, AMD,
Grafana, analysis, analyze mode
************
Analyze mode
************
Omniperf offers several ways to interact with the metrics it generates from
ROCm Compute Profiler offers several ways to interact with the metrics it generates from
profiling. Your level of familiarity with the profiled application, computing
environment, and experience with Omniperf should inform the analysis method you
environment, and experience with ROCm Compute Profiler should inform the analysis method you
choose.
While analyzing with the CLI offers quick and straightforward access to Omniperf
While analyzing with the CLI offers quick and straightforward access to ROCm Compute Profiler
metrics from the terminal, Grafana's dashboard GUI adds an extra layer of
readability and interactivity you might prefer.
See the following sections to explore Omniperf's analysis and visualization
See the following sections to explore ROCm Compute Profiler's analysis and visualization
options.
* :doc:`cli`
@@ -32,5 +32,5 @@ options.
Unless otherwise noted, the performance analysis is done on the
:ref:`MI200 platform <def-soc>`.
Learn about profiling with Omniperf in :doc:`../profile/mode`. For an overview of
Omniperf's other modes, see :ref:`modes`.
Learn about profiling with ROCm Compute Profiler in :doc:`../profile/mode`. For an overview of
ROCm Compute Profiler's other modes, see :ref:`modes`.
docs/how-to/analyze/standalone-gui.rst
+15 -15
View File
@@ -1,15 +1,15 @@
.. meta::
:description: Omniperf analysis: Standalone GUI
:description: ROCm Compute Profiler analysis: Standalone GUI
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, GUI, standalone, filter
***********************
Standalone GUI analysis
***********************
Omniperf's standalone analysis GUI is a lightweight web page that you can
ROCm Compute Profiler's standalone analysis GUI is a lightweight web page that you can
generate straight from the command line. The standalone analysis GUI is an
alternative to the CLI if you want to explore profiling results visually, but
without the additional setup requirements or server-side overhead of Omniperf's
without the additional setup requirements or server-side overhead of ROCm Compute Profiler's
detailed :doc:`Grafana interface <grafana-gui>` option. This analysis
option is implemented as a simple `Flask <https://flask.palletsprojects.com>`_
application that lets you view results from your preferred web browser.
@@ -29,22 +29,22 @@ application that lets you view results from your preferred web browser.
Launch the standalone GUI analyzer
----------------------------------
To launch the Omniperf GUI analyzer, include the ``--gui`` flag with your
To launch the ROCm Compute Profiler GUI analyzer, include the ``--gui`` flag with your
desired analysis command. For example:
.. code-block:: shell
.. code-block:: shell-session
$ omniperf analyze -p workloads/vcopy/MI200/ --gui
$ rocprof-compute analyze -p workloads/vcopy/MI200/ --gui
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Analysis mode = web_ui
[analysis] deriving Omniperf metrics...
[analysis] deriving rocprofiler-compute metrics...
Dash is running on http://0.0.0.0:8050/
* Serving Flask app 'rocprof_compute_analyze.analysis_webui' (lazy loading)
@@ -62,7 +62,7 @@ At this point, you can launch your web browser of choice and navigate to
.. image:: ../../data/analyze/standalone_gui.png
:align: center
:alt: Omniperf standalone GUI home screen
:alt: ROCm Compute Profiler standalone GUI home screen
:width: 800
.. tip::
@@ -85,5 +85,5 @@ metrics specific to your selected filters.
Once a filter is applied, you'll see several additional sections become
available with detailed metrics specific to that area of AMD hardware. These
detailed sections mirror the data displayed in Omniperf's
detailed sections mirror the data displayed in ROCm Compute Profiler's
:doc:`Grafana interface <grafana-gui>`.
docs/how-to/profile/mode.rst
+80 -81
View File
@@ -1,27 +1,27 @@
.. meta::
:description: How to use Omniperf's profile mode
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, AMD,
:description: How to use ROCm Compute Profiler's profile mode
:keywords: ROCm Compute Profiler, ROCm, profiler, tool, Instinct, accelerator, AMD,
profiling, profile mode
************
Profile mode
************
The following chapter walks you through Omniperf's core profiling features by
The following chapter walks you through ROCm Compute Profiler's core profiling features by
example.
Learn about analysis with Omniperf in :doc:`../analyze/mode`. For an overview of
Omniperf's other modes, see :ref:`modes`.
Learn about analysis with ROCm Compute Profiler in :doc:`../analyze/mode`. For an overview of
ROCm Compute Profiler's other modes, see :ref:`modes`.
Profiling
=========
Use the ``omniperf`` executable to acquire all necessary performance monitoring
Use the ``rocprof-compute`` executable to acquire all necessary performance monitoring
data through analysis of compute workloads.
Profiling with Omniperf yields the following benefits.
Profiling with ROCm Compute Profiler yields the following benefits.
* :ref:`Automate counter collection <profiling-routine>`: Omniperf handles all
* :ref:`Automate counter collection <profiling-routine>`: ROCm Compute Profiler handles all
of your profiling via pre-configured input files.
* :ref:`Filtering <filtering>`: Apply runtime filters to speed up the profiling
@@ -30,7 +30,7 @@ Profiling with Omniperf yields the following benefits.
* :ref:`Standalone roofline <standalone-roofline>`: Isolate a subset of built-in
metrics or build your own profiling configuration.
Run ``omniperf profile -h`` for more details. See
Run ``rocprof-compute profile -h`` for more details. See
:ref:`Basic usage <modes-profile>`.
.. _profile-example:
@@ -38,14 +38,14 @@ Run ``omniperf profile -h`` for more details. See
Profiling example
-----------------
The `<https://github.com/ROCm/omniperf/blob/amd-mainline/sample/vcopy.cpp>`__ repository
The `<https://github.com/ROCm/rocprofiler-compute/blob/amd-mainline/sample/vcopy.cpp>`__ repository
includes source code for a sample GPU compute workload, ``vcopy.cpp``. A copy of
this file is available in the ``share/sample`` subdirectory after a normal
Omniperf installation, or via the ``$OMNIPERF_SHARE/sample`` directory when
ROCm Compute Profiler installation, or via the ``$ROCPROFCOMPUTE_SHARE/sample`` directory when
using the supplied modulefile.
The examples in this section use a compiled version of the ``vcopy`` workload to
demonstrate the use of Omniperf in MI accelerator performance analysis. Unless
demonstrate the use of ROCm Compute Profiler in MI accelerator performance analysis. Unless
otherwise noted, the performance analysis is done on the
:ref:`MI200 platform <def-soc>`.
@@ -54,7 +54,7 @@ Workload compilation
The following example demonstrates compilation of ``vcopy``.
.. code-block:: shell
.. code-block:: shell-session
$ hipcc vcopy.cpp -o vcopy
$ ls
@@ -74,20 +74,20 @@ The following example demonstrates compilation of ``vcopy``.
The following sample command profiles the ``vcopy`` workload.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf profile --name vcopy -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile --name vcopy -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Omniperf version: 2.0.0
rocprofiler-compute version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Path: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
@@ -98,10 +98,10 @@ The following sample command profiles the ``vcopy`` workload.
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[profiling] Current input file: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt
|-> [rocprof] RPL: on '240312_174329' from '/opt/rocm-5.2.1' in '/home/auser/repos/omniperf/src/omniperf'
[profiling] Current input file: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt
|-> [rocprof] RPL: on '240312_174329' from '/opt/rocm-5.2.1' in '/home/auser/repos/rocprofiler-compute/src/rocprof-compute'
|-> [rocprof] RPL: profiling '""./vcopy -n 1048576 -b 256""'
|-> [rocprof] RPL: input file '/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt'
|-> [rocprof] RPL: input file '/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt'
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240312_174329_692890'
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240312_174329_692890/input0_results_240312_174329'
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240312_174329_692890/input0.xml"
@@ -121,15 +121,15 @@ The following sample command profiles the ``vcopy`` workload.
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|-> [rocprof] Releasing GPU memory
|-> [rocprof] Releasing CPU memory
|-> [rocprof]
|-> [rocprof]
|-> [rocprof] ROCPRofiler: 1 contexts collected, output directory /tmp/rpl_data_240312_174329_692890/input0_results_240312_174329
|-> [rocprof] File '/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_IFETCH_LEVEL.csv' is generating
|-> [rocprof]
[profiling] Current input file: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_INST_LEVEL_LDS.txt
|-> [rocprof] File '/home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/SQ_IFETCH_LEVEL.csv' is generating
|-> [rocprof]
[profiling] Current input file: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200/perfmon/SQ_INST_LEVEL_LDS.txt
...
[roofline] Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
[roofline] Checking for roofline.csv in /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
[roofline] No roofline data found. Generating...
Empirical Roofline Calculation
Copyright © 2022 Advanced Micro Devices, Inc. All rights reserved.
@@ -171,9 +171,9 @@ The following sample command profiles the ``vcopy`` workload.
.. _profiling-routine:
Notice the two main stages in Omniperf's *default* profiling routine.
Notice the two main stages in ROCm Compute Profiler's *default* profiling routine.
1. The first stage collects all the counters needed for Omniperf analysis
1. The first stage collects all the counters needed for ROCm Compute Profiler analysis
(omitting any filters you have provided).
2. The second stage collects data for the roofline analysis (this stage can be
@@ -187,18 +187,18 @@ example:
* "MI200" for the AMD Instinct MI200 family of accelerators
* "MI100" for the AMD Instinct MI100 family of accelerators
The SoC names are generated as a part of Omniperf, and do not *always*
The SoC names are generated as a part of ROCm Compute Profiler, and do not *always*
distinguish between different accelerators in the same family; for instance,
an Instinct MI210 vs an Instinct MI250.
.. note::
Additionally, you will notice a few extra files. An SoC parameters file,
Additionally, you will notice a few extra files. An SoC parameters file,
``sysinfo.csv``, is created to reflect the target device settings. All
profiling output is stored in ``log.txt``. Roofline-specific benchmark
results are stored in ``roofline.csv``.
.. code-block:: shell
.. code-block:: shell-session
$ ls workloads/vcopy/MI200/
total 112
@@ -222,8 +222,8 @@ Filtering
To reduce profiling time and the counters collected, you should use profiling
filters. Profiling filters and their functionality depend on the underlying
profiler being used. While Omniperf is profiler-agnostic, this following is a
detailed description of profiling filters available when using Omniperf with
profiler being used. While ROCm Compute Profiler is profiler-agnostic, this following is a
detailed description of profiling filters available when using ROCm Compute Profiler with
:doc:`ROCProfiler <rocprofiler:index>`.
Filtering options
@@ -255,7 +255,7 @@ Hardware component filtering
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
You can profile specific hardware components to speed up the profiling process.
In Omniperf, the term hardware block to refers to a hardware component or a
In ROCm Compute Profiler, the term hardware block to refers to a hardware component or a
group of hardware components. All profiling results are accumulated in the same
target directory without overwriting those for other hardware components. This
enables incremental profiling and analysis.
@@ -263,16 +263,16 @@ enables incremental profiling and analysis.
The following example only gathers hardware counters for the shader sequencer
(SQ) and L2 cache (TCC) components, skipping all other hardware components.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf profile --name vcopy -b SQ TCC -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile --name vcopy -b SQ TCC -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
fname: pmc_cpc_perf: Skipped
fname: pmc_spi_perf: Skipped
@@ -289,9 +289,9 @@ The following example only gathers hardware counters for the shader sequencer
fname: pmc_sqc_perf1: Skipped
fname: pmc_sq_perf6: Added
fname: pmc_sq_perf2: Added
Omniperf version: 2.0.0
rocprofiler-compute version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Path: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
@@ -314,20 +314,20 @@ kernel name substring list to isolate desired kernels.
The following example demonstrates profiling isolating the kernel matching
substring ``vecCopy``.
.. code-block:: shell
.. code-block:: shell-session
$ omniperf profile --name vcopy -k vecCopy -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile --name vcopy -k vecCopy -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Omniperf version: 2.0.0
rocprofiler-compute version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Path: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: ['vecCopy']
@@ -344,25 +344,25 @@ substring ``vecCopy``.
Dispatch filtering
^^^^^^^^^^^^^^^^^^
Dispatch filtering is based on the *global* dispatch index of kernels in a run.
Dispatch filtering is based on the *global* dispatch index of kernels in a run.
The following example profiles only the first kernel dispatch in the execution
of the application (note zero-based indexing).
.. code-block:: shell
.. code-block:: shell-session
$ omniperf profile --name vcopy -d 0 -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile --name vcopy -d 0 -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
__ _
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|_| |_|
Omniperf version: 2.0.0
rocprofiler-compute version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Path: /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
@@ -381,7 +381,7 @@ Standalone roofline
If you are only interested in generating roofline analysis data try using
``--roof-only``. This will only collect counters relevant to roofline, as well
as generate a standalone ``.pdf`` output of your roofline plot.
as generate a standalone ``.pdf`` output of your roofline plot.
Roofline options
----------------
@@ -408,14 +408,14 @@ Roofline only
The following example demonstrates profiling roofline data only:
.. code-block:: shell
.. code-block:: shell-session
$ omniperf profile --name vcopy --roof-only -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile --name vcopy --roof-only -- ./vcopy -n 1048576 -b 256
...
[roofline] Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
[roofline] Checking for roofline.csv in /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
[roofline] No roofline data found. Generating...
Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Checking for roofline.csv in /home/auser/repos/rocprofiler-compute/sample/workloads/vcopy/MI200
Empirical Roofline Calculation
Copyright © 2022 Advanced Micro Devices, Inc. All rights reserved.
Total detected GPU devices: 4
@@ -427,7 +427,7 @@ The following example demonstrates profiling roofline data only:
An inspection of our workload output folder shows ``.pdf`` plots were generated
successfully.
.. code-block:: shell
.. code-block:: shell-session
$ ls workloads/vcopy/MI200/
total 48
@@ -441,15 +441,14 @@ successfully.
.. note::
Omniperf generates two roofline outputs to organize results and reduce
ROCm Compute Profiler generates two roofline outputs to organize results and reduce
clutter. One chart plots FP32/FP64 performance while the other plots I8/FP16
performance.
The following image is a sample ``empirRoof_gpu-ALL_fp32_fp64.pdf`` roofline
The following image is a sample ``empirRoof_gpu-0_int8_fp16.pdf`` roofline
plot.
.. image:: ../../data/profile/sample-roof-plot.png
.. image:: ../../data/profile/sample-roof-plot.jpg
:align: center
:alt: Sample Omniperf roofline output
:alt: Sample ROCm Compute Profiler roofline output
:width: 800
docs/how-to/use.rst
+26 -27
View File
@@ -1,13 +1,13 @@
.. meta::
:description: Omniperf basic usage
:keywords: Omniperf, ROCm, profiler, tool, Instinct, accelerator, AMD,
:description: ROCm Compute Profiler basic usage
:keywords: ROCm Compute Profiler, ROCm, profiler, tool, Instinct, accelerator, AMD,
basics, usage, operations
***********
Basic usage
***********
The following section outlines basic Omniperf workflows, modes, options, and
The following section outlines basic ROCm Compute Profiler workflows, modes, options, and
operations.
Command line profiler
@@ -18,7 +18,7 @@ Launch and profile the target application using the command line profiler.
The command line profiler launches the target application, calls the
ROCProfiler API via the ``rocprof`` binary, and collects profile results for
the specified kernels, dispatches, and hardware components. If not
specified, Omniperf defaults to collecting all available counters for all
specified, ROCm Compute Profiler defaults to collecting all available counters for all
kernels and dispatches launched by the your executable.
To collect the default set of data for all kernels in the target
@@ -26,7 +26,7 @@ application, launch, for example:
.. code-block:: shell
$ omniperf profile -n vcopy_data -- ./vcopy -n 1048576 -b 256
$ rocprof-compute profile -n vcopy_data -- ./vcopy -n 1048576 -b 256
This runs the app, launches each kernel, and generates profiling results. By
default, results are written to a subdirectory with your accelerator's name;
@@ -35,7 +35,7 @@ via the ``-n`` argument.
.. note::
To collect all requested profile information, Omniperf might replay kernels
To collect all requested profile information, ROCm Compute Profiler might replay kernels
multiple times.
.. _basic-filter-data-collection:
@@ -67,7 +67,7 @@ argument:
.. code-block:: shell
$ omniperf analyze --list-metrics <sys_arch>
$ rocprof-compute analyze --list-metrics <sys_arch>
.. _basic-analyze-cli:
@@ -87,7 +87,7 @@ To interact with profiling results from a different session, provide the
workload path.
``-p``, ``--path``
Enables you to analyze existing profiling data in the Omniperf CLI.
Enables you to analyze existing profiling data in the ROCm Compute Profiler CLI.
See :doc:`analyze/cli` for more detailed information.
@@ -97,16 +97,16 @@ Analyze in the Grafana GUI
--------------------------
To conduct a more in-depth analysis of profiling results, it's suggested to use
a Grafana GUI with Omniperf. To interact with profiling results, import your
data to the MongoDB instance included in the Omniperf Dockerfile. See
a Grafana GUI with ROCm Compute Profiler. To interact with profiling results, import your
data to the MongoDB instance included in the ROCm Compute Profiler Dockerfile. See
:doc:`/install/grafana-setup`.
To interact with Grafana data, stored in the Omniperf database, enter
To interact with Grafana data, stored in the ROCm Compute Profiler database, enter
``database`` :ref:`mode <modes-database>`; for example:
.. code-block:: shell
$ omniperf database --import [CONNECTION OPTIONS]
$ rocprof-compute database --import [CONNECTION OPTIONS]
See :doc:`/how-to/analyze/grafana-gui` for more detailed information.
@@ -115,7 +115,7 @@ See :doc:`/how-to/analyze/grafana-gui` for more detailed information.
Modes
=====
Modes change the fundamental behavior of the Omniperf command line tool.
Modes change the fundamental behavior of the ROCm Compute Profiler command line tool.
Depending on which mode you choose, different command line options become
available.
@@ -133,10 +133,10 @@ Profile mode
.. code-block:: shell
$ omniperf profile --help
$ rocprof-compute profile --help
See :doc:`profile/mode` to learn about this mode in depth and to get started
profiling with Omniperf.
profiling with ROCm Compute Profiler.
.. _modes-analyze:
@@ -144,24 +144,24 @@ Analyze mode
------------
``analyze``
Loads profiling data from the ``--path`` (``-p``) directory into the Omniperf
Loads profiling data from the ``--path`` (``-p``) directory into the ROCm Compute Profiler
CLI analyzer where you have immediate access to profiling results and
generated metrics. It generates metrics from the entirety of your profiled
application or a subset identified through the Omniperf CLI analysis filters.
application or a subset identified through the ROCm Compute Profiler CLI analysis filters.
To generate a lightweight GUI interface, you can add the ``--gui`` flag to your
analysis command.
This mode is a middle ground to the highly detailed Omniperf Grafana GUI and
This mode is a middle ground to the highly detailed ROCm Compute Profiler Grafana GUI and
is great if you want immediate access to a hardware component you’re already
familiar with.
.. code-block:: shell
$ omniperf analyze --help
$ rocprof-compute analyze --help
See :doc:`analyze/mode` to learn about this mode in depth and to get started
with analysis using Omniperf.
with analysis using ROCm Compute Profiler.
.. _modes-database:
@@ -178,7 +178,7 @@ Database mode
.. code-block:: shell
$ omniperf database --help
$ rocprof-compute database --help
See :doc:`/install/grafana-setup` to learn about setting up a Grafana server and
database instance to make your profiling data more digestible and shareable.
@@ -188,11 +188,11 @@ database instance to make your profiling data more digestible and shareable.
Global options
==============
The Omniperf command line tool has a set of *global* utility options that are
available across all modes.
The ROCm Compute Profiler command line tool has a set of *global* utility options that are
available across all modes.
``-v``, ``--version``
Prints the Omniperf version and exits.
Prints the ROCm Compute Profiler version and exits.
``-V``, ``--verbose``
Increases output verbosity. Use multiple times for higher levels of
@@ -206,7 +206,7 @@ available across all modes.
.. note::
Omniperf also recognizes the project variable, ``OMNIPERF_COLOR`` should you
ROCm Compute Profiler also recognizes the project variable, ``ROCPROFCOMPUTE_COLOR`` should you
choose to disable colorful output. To disable default colorful behavior, set
this variable to ``0``.
@@ -215,7 +215,7 @@ available across all modes.
Basic operations
================
The following table lists Omniperf's basic operations, their
The following table lists ROCm Compute Profiler's basic operations, their
:ref:`modes <modes>`, and required arguments.
.. list-table::
@@ -248,4 +248,3 @@ The following table lists Omniperf's basic operations, their
* - :doc:`Interact with profiling results from CLI </how-to/analyze/cli>`
- ``analyze``
- ``--path``