The wavefront size is currently only exposed as an agent level
attribute. This is not correctyl, because while the agent has a default
wave front size that is usually correct, it can easily be overridden via
options like -mwavefrontsize64 on various ISAs. The wavefrontsize
attribute is actually more of a calling convention that is consistent
within a callgraph. Because the root of each call graph is a kernel in
this architecture, we need to be able to query this on a per-kernel
basis. This information is already avialable in the kernel descriptor
packet, but it wasn't exported.
This patch adds HSA_CODE_SYMBOL_INFO_KERNEL_WAVEFRONT_SIZE as a new
option to query on the executable symbol.
Change-Id: I744815c89cc9d4c82f25479bdd48ae1f32e859ff
In addition to be able to set clock range, new setextremum option
is added to set only min/max clock as sometimes one of them may
not be supported.
Change-Id: I7c91ba308f3fc6c78efc88117509c515d403a6cb
* escape sbios, can have commas on some systems
Signed-off-by: Nicholas Curtis <nicurtis@amd.com>
* Prefer to cast to str in original function
Signed-off-by: colramos-amd <colramos@amd.com>
---------
Signed-off-by: Nicholas Curtis <nicurtis@amd.com>
Signed-off-by: colramos-amd <colramos@amd.com>
Co-authored-by: Nicholas Curtis <nicurtis@amd.com>
Co-authored-by: colramos-amd <colramos@amd.com>
* OMNITRACE_ROCM_SMI_METRICS
- configuration variable OMNITRACE_ROCM_SMI_METRICS for specifying which rocm-smi metrics to collect
- auto-disable metric collection when rsmi_dev_X_get returns RSMI_STATUS_NOT_SUPPORTED
* Bump version to 1.11.1
* Python formatting
* Update python/libpyomnitrace.cpp
- fix usage of substr (ignored return value)
* Update python/gui/source/gui.py
- Fix E721
- do not compare types, for exact checks use `is` / `is not`, for instance checks use `isinstance()`
* Add support for AQL dimension changes
Adds support for returning dimensions from AQLProfile through rocprofiler
to tools. Includes a much larger expanded test suite that covers nearly
all files in counter collection.
Specific changes below:
samples/counter_collection/print_functional_counters: Modified to check
the validity of dimensions returned in comparison to the actual underlying
data obtained from a kernel execution.
rocprofiler-sdk/aql/helpers: adds function calls to support fetching
dimension information from AQLProfile.
rocprofiler-sdk/aql/packet_construct: modified to allow for events
to be exported to aid evaluate_ast in decoding the output buffer.
lib/rocprofiler-sdk/counters: Instance count now derived from dimension
sizes. rocprofiler_query_counter_dimensions now moved to a callback format
to improve usability.
rocprofiler-sdk/counters/core: Code migrations and exports of functions
for testing.
rocprofiler-sdk/counters/dimensions: Generates a dimension cache to be
used when querying dimension information for a counter id.
rocprofiler-sdk/counters/evaluate_ast: Modified to pass back correct
dimension information and to check/determine output dimensions for derived
counters.
rocprofiler-sdk/counters/id_decode: Modified to have a map between
dimension name -> dimension along with a conversion from the aql profile
id for a dimension (string) -> integer based id (happens only once during
init).
rocprofiler-sdk/hsa/queue: Modified to allow for making testing easier.
Specifically to allow Queue to now be mocked in unit tests for counter
collection.
* Merge with changes for serialization
* Added suggestions
* source formatting (clang-format v11) (#457)
Co-authored-by: bwelton <bwelton@users.noreply.github.com>
* Minor fix
* Test change
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: bwelton <bwelton@users.noreply.github.com>
Instead of caching shared memory fds for export on the exporter side,
only export the FD in the async handler when requested.
The importer should request export fd closure once import is done.
Change-Id: I469e0cd1749beeb9c506c8a6461745fb039d9c3b
Add @ammarwa and @bgopesh as CODEOWNERS.
This is for GitHub upstream.
Signed-off-by: David Galiffi <David.Galiffi@amd.com>
Change-Id: I84ddc63e20c8b3ea4d4c8349a95d46b2925da2a2
Add @ammarwa and @bgopesh as CODEOWNERS.
This is for GitHub upstream.
Signed-off-by: David Galiffi <David.Galiffi@amd.com>
Change-Id: I788f5ad550f91e8d3ce45bbeb527283bf11c4fd7
Adjust code to allow the use of non-contiguous chunks of memory to be
mapped within a single VA range.
Change-Id: Ida21ba202927229347b3a32d9b7106df10819cf5
Add tests to catch whether ROCr breaks ABI compatibility with the
hsa_amd_pointer_info API in case the hsa_amd_pointer_info struct is
extended.
Change-Id: I4e69bf30db9791e59f895b2798b87985c41242e5
Currently, the definition of hsa_amd_memory_fault_reason_t tries to
set a constant of 0x8000_0000 by using the definition "1 << 31".
However, the 1 in this definition is a signed integer by C++ rules.
On our architectures, shifting a signed integer by 31 results in
signed integer overflow. Signed integer overflow results in
undefined behavior.
Forcing the 1 to be unsigned avoids this.
Change-Id: I860431eeede4eff29598f646abf3c1337b048d71
Added global flag g_IsEmuMode and set it when running under emulator.
Adjusted delays in KFDCWSRTest for emulator.
Change-Id: Ia5c0be40816ac2219add943e306ee16438f5b852
Signed-off-by: David Belanger <david.belanger@amd.com>
Users can import device memory without specifying the target node.
DMA buf imports return a Thunk handle that's not useful for
gpu mapping calls.
Fix this by using the import node information to re-import and
map with the correct target GPU.
Also fix IPC detach calls by deregistering the Thunk handle
import immediately during attach instead of failing to do it later
on detach since Thunk handles aren't placed into ROCr allocation
map.
Finally refactor the IPC attach function for cleaner logic flow.
Change-Id: Ib2bf178110b2be98bd6917c765f724e4e613f5f2
If the card does not have edge temperature, fallback to junction
temperature. If the card only have socket power, then use socket
power instead.
Change-Id: I053a67a89cf3b29a34e82123f522c08d7dd68916
Joseph Huber
Bill(Shuzhou) Liu
Galantsev, Dmitrii
colramos-amd
Nick Curtis
Jonathan R. Madsen
Manjunath-Jakaraddi
Benjamin Welton
Jonathan Kim
Mythreya
David Galiffi
Shweta Khatri
David Yat Sin
Joseph Greathouse
David Belanger
Gopesh Bhardwaj
Karl W Schulz
SrirakshaNag
JoseSantosAMD