[rocprofiler-compute] Add database output format to analyze mode (#748)
Analysis data dump * Add `--output-format` and `--output-name` option to analyze mode * Remove `--output` and `-save-dfs` option to analyze mode * Add documentation on `rocpd` output format and analysis database file * Create sqlite3 database using object relation mapping (ORM) provided by sqlalchemy library * Fix metrics config to remove metrics marked as `null`, fix `Unit` header, add missing `title` * Add test cases to ensure analysis data dump work
이 커밋은 다음에 포함됨:
vedithal-amd
GitHub
@@ -213,4 +213,4 @@ extlinks = {
|
||||
}
|
||||
|
||||
# Uncomment if facing rate limit exceed issue with local build
|
||||
external_projects_remote_repository = ""
|
||||
# external_projects_remote_repository = ""
|
||||
|
||||
바이너리 파일은 표시되지 않음.
|
이후 너비: | 높이: | 크기: 66 KiB |
바이너리 파일은 표시되지 않음.
|
이후 너비: | 높이: | 크기: 17 KiB |
@@ -318,10 +318,12 @@ Per-kernel roofline analysis
|
||||
When analyzing specific kernels, the roofline analysis provides detailed metrics for each filtered kernel:
|
||||
|
||||
.. code-block:: shell-session
|
||||
|
||||
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -k 0 -b 4
|
||||
This generates enhanced roofline output showing per-kernel performance rates and arithmetic intensity calculations:
|
||||
|
||||
.. code-block:: text
|
||||
|
||||
================================================================================
|
||||
4. Roofline
|
||||
================================================================================
|
||||
@@ -372,6 +374,7 @@ Per-kernel roofline analysis
|
||||
Analyze multiple kernels for comparison:
|
||||
|
||||
.. code-block:: shell-session
|
||||
|
||||
$ rocprof-compute analyze -p workloads/vcopy/MI200/ -k 0 1 2 -b 4
|
||||
|
||||
Baseline comparison
|
||||
@@ -384,3 +387,102 @@ Baseline comparison
|
||||
.. code-block:: shell
|
||||
|
||||
rocprof-compute analyze -p workload1/path/ -k 0 -p workload2/path/ -k 1
|
||||
|
||||
Analysis output format
|
||||
======================
|
||||
|
||||
Use the ``--output-format <format>`` analyze mode option to specify the output format of the
|
||||
analysis report. Supported formats are ``stdout``, ``txt``, ``csv``, and ``db``. The default output
|
||||
format is ``stdout``.
|
||||
|
||||
* ``stdout`` format:
|
||||
* Print analysis report to the terminal.
|
||||
* NOTE: This option will not generate any file or folder.
|
||||
|
||||
* ``txt`` format:
|
||||
* Generate a file named ``rocprof_compute_<uuid>.txt`` in the current working directory.
|
||||
* This file contains the entire analysis report as printed on the terminal.
|
||||
* This is useful in case of searching across long analysis reports.
|
||||
* NOTE: This option will disable output of analysis report to terminal.
|
||||
|
||||
* ``csv`` format:
|
||||
* Generate a folder named ``rocprof_compute_<uuid>`` in the current working directory.
|
||||
* This folder contains multiple csv files representing the data in each metric table in the analysis report.
|
||||
* This is useful for further programmatic analysis of analysis reports.
|
||||
* NOTE: This will print the analysis report to the terminal as well.
|
||||
|
||||
* ``db`` format:
|
||||
* NOTE: This only works when provided workload paths are created using ``--format-rocprof-output rocpd`` profile mode option.
|
||||
* Generate a file named ``rocprof_compute_<uuid>.db`` in the current working directory.
|
||||
* This is a SQLite database file containing all the data in the analysis report structured according to :ref:`analysis database schema <analysis-database>`.
|
||||
* This is useful for further programmatic analysis of analysis reports.
|
||||
* NOTE: This option will disable output of analysis report to terminal.
|
||||
|
||||
Default file/folder name ``rocprofiler_compute_<uuid>`` can be overriden using ``--output-name <name>`` analyze mode option.
|
||||
|
||||
.. _analysis-database:
|
||||
|
||||
Analysis database schema
|
||||
========================
|
||||
|
||||
Analysis database tables
|
||||
|
||||
.. image:: ../../data/analyze/analysis_data_dump_schema.png
|
||||
:align: center
|
||||
:alt: Analysis database tables
|
||||
|
||||
Analysis database views
|
||||
|
||||
.. image:: ../../data/analyze/analysis_data_dump_views.png
|
||||
:align: center
|
||||
:alt: Analysis database views
|
||||
|
||||
Analysis database example
|
||||
|
||||
.. note::
|
||||
|
||||
Some metrics cannot be calculated when corresponding counters are missing as shown in the warnings below
|
||||
|
||||
.. note::
|
||||
|
||||
It is possible to merge the analysis data dump for multiple workload folders (resulting from multiple profiles) by repeating ``-p`` option for each workload
|
||||
|
||||
.. code-block:: shell-session
|
||||
|
||||
$ rocprof-compute analyze --verbose --db test -p workloads/vmem/MI300X_A1 -p workloads/vmem1/MI300X_A1
|
||||
DEBUG Execution mode = analyze
|
||||
|
||||
__ _
|
||||
_ __ ___ ___ _ __ _ __ ___ / _| ___ ___ _ __ ___ _ __ _ _| |_ ___
|
||||
| '__/ _ \ / __| '_ \| '__/ _ \| |_ _____ / __/ _ \| '_ ` _ \| '_ \| | | | __/ _ \
|
||||
| | | (_) | (__| |_) | | | (_) | _|_____| (_| (_) | | | | | | |_) | |_| | || __/
|
||||
|_| \___/ \___| .__/|_| \___/|_| \___\___/|_| |_| |_| .__/ \__,_|\__\___|
|
||||
|_| |_|
|
||||
|
||||
INFO Analysis mode = db
|
||||
DEBUG [omnisoc init]
|
||||
DEBUG [omnisoc init]
|
||||
DEBUG [analysis] prepping to do some analysis
|
||||
INFO [analysis] deriving rocprofiler-compute metrics...
|
||||
WARNING Roofline ceilings not found for /app/projects/rocprofiler-compute/workloads/vmem/MI300X_A1.
|
||||
WARNING Roofline ceilings not found for /app/projects/rocprofiler-compute/workloads/vmem1/MI300X_A1.
|
||||
WARNING PC sampling data not found for /app/projects/rocprofiler-compute/workloads/vmem/MI300X_A1.
|
||||
WARNING PC sampling data not found for /app/projects/rocprofiler-compute/workloads/vmem1/MI300X_A1.
|
||||
DEBUG Collected dispatch data
|
||||
DEBUG Applied analysis mode filters
|
||||
DEBUG Calculated dispatch data
|
||||
DEBUG Collected metrics data
|
||||
WARNING Failed to evaluate expression for 3.1.25 - Value: to_round(to_avg(
|
||||
(pmc_df.get("TCP_TCP_LATENCY_sum") / pmc_df.get("TCP_TA_TCP_STATE_READ_sum")).where((pmc_df.get("TCP_TA_TCP_STATE_READ_sum") != 0), None)), 0) - unsupported operand type(s) for /: 'NoneType' and 'float'
|
||||
WARNING Failed to evaluate expression for 3.1.39 - Value: to_round((to_avg(
|
||||
(pmc_df.get("pmc_perf_ACCUM") / pmc_df.get("SQC_ICACHE_REQ")).where((pmc_df.get("SQC_ICACHE_REQ") != 0), None)) * 100), 0) - unsupported operand type(s) for /: 'NoneType' and 'float'
|
||||
WARNING Failed to evaluate expression for 3.1.25 - Value: to_round(to_avg(
|
||||
(pmc_df.get("TCP_TCP_LATENCY_sum") / pmc_df.get("TCP_TA_TCP_STATE_READ_sum")).where((pmc_df.get("TCP_TA_TCP_STATE_READ_sum") != 0), None)), 0) - unsupported operand type(s) for /: 'NoneType' and 'float'
|
||||
WARNING Failed to evaluate expression for 3.1.39 - Value: to_round((to_avg(
|
||||
(pmc_df.get("pmc_perf_ACCUM") / pmc_df.get("SQC_ICACHE_REQ")).where((pmc_df.get("SQC_ICACHE_REQ") != 0), None)) * 100), 0) - unsupported operand type(s) for /: 'NoneType' and 'float'
|
||||
DEBUG Calculated metric values
|
||||
DEBUG Calculated roofline data points
|
||||
DEBUG [analysis] generating analysis
|
||||
DEBUG SQLite database initialized with name: test.db
|
||||
DEBUG Initialized database: test.db
|
||||
DEBUG Completed writing database
|
||||
@@ -22,7 +22,7 @@ options.
|
||||
* :doc:`cli`
|
||||
* :doc:`grafana-gui`
|
||||
* :doc:`standalone-gui`
|
||||
* :doc:`text-based user interface (TUI)`
|
||||
* :doc:`tui`
|
||||
|
||||
.. note::
|
||||
|
||||
|
||||
@@ -24,6 +24,15 @@ Profiling with ROCm Compute Profiler yields the following benefits.
|
||||
* :ref:`Automate counter collection <profiling-routine>`: ROCm Compute Profiler handles all
|
||||
of your profiling via pre-configured input files.
|
||||
|
||||
* :ref:`Profiling output format <profiling-output-format>`: ROCm Compute Profile can adjust the
|
||||
output format of underlying rocprof tool which changes the output format of raw performance
|
||||
counter data in the workload folder created during profiling. Supported output formats are
|
||||
``json``, ``csv``, and ``rocpd``. The default output format is ``csv``.
|
||||
|
||||
.. note::
|
||||
|
||||
The default output format will be changed to ``rocpd`` in a future release of ROCm Compute Profiler.
|
||||
|
||||
* :ref:`Filtering <filtering>`: Apply runtime filters to speed up the profiling
|
||||
process.
|
||||
|
||||
@@ -217,6 +226,32 @@ an Instinct MI210 vs an Instinct MI250.
|
||||
-rw-r--r-- 1 auser agroup 650 Mar 1 15:15 sysinfo.csv
|
||||
-rw-r--r-- 1 auser agroup 399 Mar 1 15:15 timestamps.csv
|
||||
|
||||
.. _profiling-output-format:
|
||||
|
||||
Profiling output format
|
||||
-----------------------
|
||||
|
||||
Use the ``--format-rocprof-output <format>`` profile mode option to specify the output format
|
||||
of the underlying ``rocprof`` tool. The following formats are supported:
|
||||
|
||||
* ``csv`` format:
|
||||
* Ask underlying rocprof tool to dump raw performance counter data in csv format.
|
||||
* The generated csv files across multiple runs of rocprof are processed and dumped into the workload directory as csv files.
|
||||
* Multiple csv files are merged into single pmc_perf.csv file in workload directory.
|
||||
|
||||
* ``json`` format:
|
||||
* Ask underlying rocprof tool to dump raw performance counter data in json format.
|
||||
* The generated json files across multiple runs of rocprof are processed and dumped into the workload directory as csv files.
|
||||
* Multiple csv files are merged into single pmc_perf.csv file in workload directory.
|
||||
|
||||
* ``rocpd`` format:
|
||||
* Ask underlying rocprof tool to dump raw performance counter data in rocpd format.
|
||||
* Multiple ``rocpd`` database files containding counter collection data are merged into a single csv under the workload folder.
|
||||
The database files are then removed.
|
||||
* Use ``--retain-rocpd-output`` profile mode option to preserve the ``rocpd`` database(s) in the workload folder.
|
||||
This is useful for custom analysis of profiling data.
|
||||
|
||||
|
||||
.. _filtering:
|
||||
|
||||
Filtering
|
||||
|
||||
새 이슈에서 참조
사용자 차단