Run pre-commit's whitespace related hooks on projects/rocprofiler-systems (#2123)
In order for pre-commit to be useful, everything needs to meet a common baseline. Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
This commit is contained in:
@@ -244,8 +244,8 @@ Sampling types:
|
||||
To enable sampling:
|
||||
|
||||
1. Use ``rocprof-sys-sample`` (auto-enables sampling).
|
||||
2. Set ``ROCPROFSYS_USE_SAMPLING=ON`` and ``ROCPROFSYS_MODE=sampling``.
|
||||
3. Use ``-S`` or ``--sample`` with ``rocprof-sys-run``.
|
||||
2. Set ``ROCPROFSYS_USE_SAMPLING=ON`` and ``ROCPROFSYS_MODE=sampling``.
|
||||
3. Use ``-S`` or ``--sample`` with ``rocprof-sys-run``.
|
||||
4. Use ``-M sampling`` or ``--mode sampling`` with ``rocprof-sys-instrument``. Use of ``rocprof-sys-sample`` is **recommended** over ``rocprof-sys-instrument -M sampling`` when binary instrumentation is not necessary. For more details, see :doc:`Sampling the call stack <../how-to/sampling-call-stack>`.
|
||||
|
||||
Causal mode
|
||||
|
||||
@@ -64,9 +64,9 @@ GPU metrics
|
||||
* JPEG activity
|
||||
* XGMI interconnect metrics (link width, link speed, read/write data)
|
||||
* PCIe metrics (link width, link speed, bandwidth)
|
||||
|
||||
|
||||
.. note::
|
||||
|
||||
|
||||
The availability of VCN, JPEG, XGMI, and PCIe metrics depends on device support and system topology. If unsupported, values will be reported as ``N/A`` in the output of ``amd-smi metric --usage``.
|
||||
|
||||
CPU metrics
|
||||
|
||||
@@ -187,8 +187,8 @@ There are two distinct approaches for collecting PAPI-based hardware counters, e
|
||||
|
||||
**Example 1: Using ``papi_array`` for a fixed list of events**
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
# Enable profiling mode (required)
|
||||
export ROCPROFSYS_PROFILE=ON
|
||||
|
||||
@@ -197,30 +197,30 @@ There are two distinct approaches for collecting PAPI-based hardware counters, e
|
||||
|
||||
# Specify which PAPI events to collect
|
||||
export ROCPROFSYS_PAPI_EVENTS="PAPI_TOT_CYC,PAPI_TOT_INS"
|
||||
|
||||
|
||||
|
||||
**Example 2: Using ``papi_vector`` for dynamically allocated array of events**
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
# Include papi_vector for dynamic event lists
|
||||
export ROCPROFSYS_TIMEMORY_COMPONENTS="wall_clock,papi_vector"
|
||||
|
||||
# Alternative: Use perf event names
|
||||
export ROCPROFSYS_PAPI_EVENTS="perf::INSTRUCTIONS,perf::CACHE-REFERENCES,perf::CACHE-MISSES"
|
||||
|
||||
|
||||
|
||||
2. **Sampling-based collection:** Periodically interrupts program execution to capture hardware counters along with call stack information. This works with the sampling mode.
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
# Enable sampling mode (required)
|
||||
export ROCPROFSYS_USE_SAMPLING=ON
|
||||
|
||||
# Specify PAPI events for sampling
|
||||
export ROCPROFSYS_PAPI_EVENTS="PAPI_TOT_CYC,PAPI_TOT_INS"
|
||||
|
||||
You can also enable overflow sampling for PAPI events with ``ROCPROFSYS_SAMPLING_OVERFLOW_EVENT``:
|
||||
You can also enable overflow sampling for PAPI events with ``ROCPROFSYS_SAMPLING_OVERFLOW_EVENT``:
|
||||
|
||||
.. code-block:: shell
|
||||
|
||||
@@ -309,8 +309,8 @@ Use the following command to view the available domains:
|
||||
|
||||
.. note::
|
||||
|
||||
Some settings can enable tracing for multiple domains, such as:
|
||||
|
||||
Some settings can enable tracing for multiple domains, such as:
|
||||
|
||||
* ``hip_api`` which will enable both ``hip_runtime_api`` and ``hip_compiler_api``.
|
||||
* ``hsa_api`` which will enable all hsa domains, ``hsa_core_api``, ``hsa_amd_ext_api``, ``hsa_image_exit_api``, and ``hsa_finalize_ext_api``.
|
||||
* ``marker_api`` or ``roctx`` can be used to enable the roctx marker API tracing.
|
||||
|
||||
@@ -56,7 +56,7 @@ environment variable.
|
||||
.. code-block:: shell
|
||||
|
||||
export LD_LIBRARY_PATH=:/opt/venv/lib/python3.10/site-packages/torch/lib:$LD_LIBRARY_PATH
|
||||
|
||||
|
||||
Running ROCm Systems Profiler on a Python script
|
||||
================================================
|
||||
|
||||
|
||||
@@ -78,7 +78,7 @@ Metadata JSON Sample
|
||||
|
||||
.. code-block:: json
|
||||
:linenos:
|
||||
|
||||
|
||||
{
|
||||
"rocprofiler-systems": {
|
||||
"metadata": {
|
||||
|
||||
@@ -55,8 +55,8 @@ These metrics help identify GPU-to-GPU communication patterns and bandwidth util
|
||||
|
||||
.. note::
|
||||
|
||||
XGMI metrics are only available on systems with multiple GPUs connected via XGMI links.
|
||||
The availability depends on the system topology and GPU architecture. If unsupported or not
|
||||
XGMI metrics are only available on systems with multiple GPUs connected via XGMI links.
|
||||
The availability depends on the system topology and GPU architecture. If unsupported or not
|
||||
available, the values will be reported as N/A in the output.
|
||||
|
||||
PCIe metrics
|
||||
|
||||
مرجع در شماره جدید
Block a user