To correctly map to all GPUs after an import, use the new extended
registration call that can import a virtual address without having to
specify a target node.
Change-Id: Ifca8f6f6ee24fa99b2af357dcc3ea1de3ab234f7
Currently registering graphics memory without specifying a target
node will return a memory handle that's not a virtual address.
As a result, ROCr is forced to register with a target node for
IPC usage.
Mapping memory without specifying a target node afterwards will
result in mapping to the target node that was imported because the
previous import call flags this node targeting action to future mapping.
For ROCr IPC usage, ROCr wants to map to all GPU nodes if the target node
is not specified.
Allow the caller to register graphics handles that returns a virtual
address without having to specify the target node so that the caller
can make a subsequent map call to all GPUs.
Change-Id: I5a935092b885cc3568e4f3a5dd951c7ec6c84fca
* Fixing some accumulate metrics
* Fixing some more accumulate metrics
---------
Co-authored-by: Benjamin Welton <bewelton@amd.com>
[ROCm/rocprofiler-sdk commit: fa7fc7ec5d]
PAL supports allocating from system memory once device memory is used up
or allocation is larger than the device memory.
Change-Id: Iccd3377e95a6cc6d23e45d4738a17af8b9ee32d7
A fixed sized std::array is used to store counter records in rocprofiler SDK. This limit was breached in SWDEV-484742. Upping the limit to 512 to be less likely to reach this limit again.
[ROCm/rocprofiler-sdk commit: a159fb47c7]
A fixed sized std::array is used to store counter records in rocprofiler SDK. This limit was breached in SWDEV-484742. Upping the limit to 512 to be less likely to reach this limit again.
CSV output truncates doubles to ints when it shouldn't. Derived metrics
are (mostly) doubles and lose precision (or become worthless) if treated
as an int. Converted these to double to match the format we return from
rocprof-sdk.
Co-authored-by: Benjamin Welton <ben@amd.com>
[ROCm/rocprofiler-sdk commit: 4f05850efd]
CSV output truncates doubles to ints when it shouldn't. Derived metrics
are (mostly) doubles and lose precision (or become worthless) if treated
as an int. Converted these to double to match the format we return from
rocprof-sdk.
Co-authored-by: Benjamin Welton <ben@amd.com>
This reverts commit 38eedf86cf.
Reason for revert: Observing issue SWDEV-488249 due to catch test permission settings.
Change-Id: Ifdc55e4eb3938c9bc0b9af0a4e9dc9ba366614b3
In static build, the dev and binary components are grouped to generate static package
Removed the line that was ignoring the component grouping
Change-Id: Ie0ca9db109f2002891260985634f2e6b1ea7f236
Changes to the README.md:
-Fixed the dead links in the examples.
-Updated the example to point to the ones in roctracer_spec.md instead of tests.
Changes to the roctracer_spec.md:
-Changed #include from #include <*.h> to #include <roctracer/*.h> to
reflect the latest code structure.
Change-Id: Iab44fc327821c1a01f0b949d3c9a3c20ca587599
[ROCm/roctracer commit: 725d289a2b]
Changes to the README.md:
-Fixed the dead links in the examples.
-Updated the example to point to the ones in roctracer_spec.md instead of tests.
Changes to the roctracer_spec.md:
-Changed #include from #include <*.h> to #include <roctracer/*.h> to
reflect the latest code structure.
Change-Id: Iab44fc327821c1a01f0b949d3c9a3c20ca587599
When hsa_amd_vmem_set_access is called, do not remove permissions for
unspecified agents. Also updating documentation in header to clarify
this.
Change-Id: I3bb4cf08ba399f85cc67b17fd13a4a40d862415f
Currently, KFDPerformanceTest.P2PBandWidthTest cannot work if there are
more than 16 KFD nodes in the system. This limit was put in to match the
number of SDMA queues supported on a single node.
This patch updates the test to make it run on systems with more than
16 KFD nodes.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Change-Id: I561d0cdef664cae84fb9c13a801052e2001256e5
* In a scenario where kernel is launched with hipExtLaunchKernelGGL and stop event is used, hipGraphInstantiate leaks. Since stop event is used, profiling is enabled and Timestamp (ReferencedCountedObject) is created, but it doesn't get released.
* The idea behind this solution is that profiling should be disabled when command is captured, hence the timestamp should not be created. Because information about capturing isn't available when kernel command is created, packet capturing state is used to determine whether to create a timestamp or not.
Change-Id: Ia23adac4592ded4fb5e236acf99e12e729f63692
Skip CPU agents when fetching metrics. Only fetch metrics for GPU
agents.
Co-authored-by: Benjamin Welton <ben@amd.com>
[ROCm/rocprofiler-sdk commit: 0c4a583e23]
* Updated validation of aborted-app test to use hipExtHostAlloc now that tracing for this command is supported
* Updated CHANGELOG
[ROCm/rocprofiler-sdk commit: 5dd2f94f74]
Peter Park
Branislav Brzak
Saleel Kudchadker
Jaydeep Patel
Jonathan Kim
Giovanni Lenzi Baraldi
Ben Richard
pghafari
AravindanC
Benjamin Welton
Ioannis Assiouras
Julia Jiang
Ranjith Ramakrishnan
Cole Ramos
Tim
Satyanvesh Dittakavi
Shweta Khatri
David Yat Sin
Rahul Manocha
Mukul Joshi
Charis Poag
Anusha GodavarthySurya
Vladana Stojiljkovic
Jeffrey Novotny
itrowbri