* SWDEV-518214: GPU Metrics 1.8 (#31)
- Updates:
- Adding the following metrics to allow new calculations for violation status:
- Per XCP metrics gfx_below_host_limit_ppt_acc
- Per XCP metrics gfx_below_host_limit_thm_acc
- Per XCP metrics gfx_low_utilization_acc
- Per XCP metrics gfx_below_host_limit_total_acc
- Increasing available JPEG engines to 40. Current ASICs may not support all 40. These will be indicated as UINT16_MAX or N/A in CLI.
Signed-off-by: Juan Castillo <juan.castillo@amd.com>
Co-authored-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: f69e65f7bd]
Changes: - Fixed Device Name (market name)
- Added new API rsmi_dev_market_name_get()
- Updated tests
- Updated amdgpu_drm.h to match latest mainline kernel
- Fixed subsystem ID to only show hex value (not subsystem name)
- rocm_smi_lib now has a recommended requirement for libdrm
Change-Id: Ic438529e16c8c3dbbdd620da664918148c40c997
[ROCm/rocm_smi_lib commit: 6a5e94c451]
Changes:
- Added new GPU metrics:
1) XGMI link status - Up/Down; 1 = up; 0 = down
2) Graphics clocks below host limit (per XCP)
accumulators -> used to help calculate a violation status
3) VRAM max bandwidth at max memory clock
- Updated rocm-smi --showmetrics to include new metrics.
Units/values reflect as indicated by driver, may differ
from AMD SMI or other ROCm SMI interfaces which
use these fields.
- N/A fields means the device does not support providing this
data.
Change-Id: I17b313345f15070a76b3a30dd8d5645d212d601b
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 4de2168866]
The reset gpu partition support for both compute and memory were removed
Code changes related to the following:
* rsmi_dev_compute_partition_reset()
* rsmi_dev_memory_partition_reset()
* CLI
* Unit tests
* Documentation
Change-Id: I3fb8570dbf9e755ae70369587ef44bbf64e17fe8
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: a1295714f2]
Changes:
- Added new GPU metrics:
1) Violation status' (ex. PVIOL/TVIOL) accumulators
2) XCP (Graphics Compute Partitions) statistics
3) pcie other end recovery counter
- Added rocm-smi --showmetrics
Units/values reflect as indicated by driver, may differ
from AMD SMI or other ROCm SMI interfaces which
use these fields.
- N/A fields means the device does not support providing this
data.
Change-Id: Ia2cd3bb65c4f474ebdb39db8062ea716f2b4d8ee
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 0609cbf1d0]
Checks and forces rereading gpu metrics unconditionally
Code changes related to the following:
* Device::dev_log_gpu_metrics()
* Examples
* Unit tests
Change-Id: Ic1c4f34a39f2bf197263f80ddbb84da26345807d
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: b4d37caa70]
Cleans up individual gpu metric APIs which will be implemented according to 'unified-headers' standards
Code changes related to the following:
* 'rsmi_dev_metrics_' APIs
* Functional tests
* Examples
Change-Id: I7d562a95889361ee6f8f7588f8a790f42c8eb262
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: ce36198cb1]
Updated:
* [CLI] Fixed vram % - printf style formatting causes many data errors
This fix updates to the recommended way of outputting formatted data.
https://docs.python.org/3/library/stdtypes.html#printf-style-string-formatting
* [API/CLI] Added gpu_id / GUID from kfd (rsmi_dev_guid_get)
-> CLI name: "GUID"
-> ROCm SMI calls: no arg, -i, --showhw, --showproduct
* [API/CLI] Added node_id from kfd (rsmi_dev_node_get)
-> CLI name: "Node"
-> ROCm SMI calls: no arg, --showhw, --showproduct
* [CLI] Added target gfx version from kfd
-> CLI name: "GFX Version" or "GFX VER"
-> ROCm SMI calls: --showhw, --showproduct
* [CLI] Base ROCm CLI
-> Removed - stacked id formatting:
This is to simplify identifiers helpful to users.
More identifiers can be found on -i --showhw, --showproduct
* [CLI] Update -i, --showhw, --showproduct, w/out arg
-> Card ID/DID/Model/SKU/VBIOS:
All unsupported values now display "N/A" instead
of "unknown" or "unsupported"
* [CLI] Showhw now expands data based on content
Change-Id: Ifb8586f9f545892b8a5aa7903608273cdd77e075
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 4b5ccb57f0]
Updates:
- [API] rsmi_dev_target_graphics_version_get, takes
reported value from KFD -> parses into human-readable
values. If device does not support, returns MAX UINT64
value and RSMI_STATUS_NOT_SUPPORTED.
Otherwise, puts into base10 format removing
extra 0's + putting in correct format. If user
provides nullptr, returning RSMI_STATUS_INVALID_ARGS.
- [Test/Example] sys_info_read updated to include
new rsmi_dev_target_graphics_version_get tests
Change-Id: I50f94e06b8733a5dec2eb08f284b44927f36abcd
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 5d2cd0c271]
Adds support and implement APIs for 'gpu_metrics_v1_5'
Code changes related to the following:
* gpu metrics 1.5 support
* Unit tests
* Examples
Build changes related to the following: None
Change-Id: Ie8917dd63c1dd1a94467b100fa44b634cebe62b6
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: 373621aed3]
Code changes related to the following:
* Check smallest copy size for multi-valued metrics
* Unit tests: gpu_metric_read
* ROCMSMI examples
Build changes related to the following:
* CMakeLists.txt
Change-Id: Ieb2363020fa21c93fbacd0edcc1d394eed183051
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: 8e0d3d5a39]
Uses new support for 'gpu_metrics_v1_4'
Code changes related to the following:
* rsmi gpu_metrics APIs
* rsmi gpu_metrics Logs
* new data structure fields added in 1.4
* added APIs for all other existing metrics before 1.4
* added support to older metrics; 1.1, and 1.2
* public APIs renamed to start with prefix 'rsmi_dev_metrics_'
* Unit tests updated
* Examples updated
Build changes related to the following: None
Change-Id: Ibdaf031be9d916020b4049544dbd725858c7711d
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: 2c8ba4cae9]
* Updates:
- [API/CLI] rsmi_dev_*_partition_set &
rsmi_dev_*_partition_reset - exposed RSMI_STATUS_BUSY for
EBUSY writes + cleaned up accidental map insertions
(maplookup[] can insert values that are not in the map,
map.at(key) fixes this potential issue)
- [API] rsmi_dev_gpu_metrics_info_get() - returns
RSMI_STATUS_NOT_SUPPORTED for unsupported metric tables
outside of 1v1/1v2/1v3
- [API] writeDevInfoStr() - exposes RSMI_STATUS_BUSY for
EBUSY write errors; kept backward compatibility
for other writes which do not care about these states
- [API] rsmi_dev_od_volt_info_get()
& rsmi_dev_od_volt_curve_regions_get() have better logging
+ Expose more details on why they are erroring
- [Utils/logs/example] Expose AMD GPU gfx target version to aid in
system troubleshooting
- [Utils] Added test methods that look at od volt
freq & regions into here - for easier access across
several tests
- [Utils] Updated getRSMIStatusString(new argument - fullstatus;
default to true for backwards compatibility)
-> true shows shortened RSMI STATUS response
- [Utils] Added splitString to cut out noisy return responses
(used in getRSMIStatusString(), when fullstatus = true)
- [Utils] Added getFileCreationDate() to expose build date
of the library - helpful for local builds or experimental builds
- [Utils] Macro cleanup
- [Example] Added a few gpu_metric checks - helpful for upcoming
updates
- [Device] SYSFS/DebugFS - now have better r/w displayed in logs
- [LOGS] Expose library build date - see above for details
- [Tests] Add more warnings/errors to test builds
- [Tests] Moved up Partition tests for ordered test runs - helped
identify issues with GPU BUSY writes
- [Tests] compute_partition_read_write - handles RSMI_STATUS_BUSY
with waits for busy status found & cleaned up how we checked
for partition changes - with RSMI responses exposed more clearly
- [Tests] perf_determinism - multi gpu now properly runs through
with full resets as needed
- [Tests] volt_freq_curv_read - better error handling with more
verbose output
Change-Id: Ie94c6abb6a9aab95c345996d3ad3843cf6734977
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 57b6135e54]
Upstream soversion is at 5 for a while, but Debian's soversion has been set to
1 in the beginning of the rocm-smi-lib package. This is probably erroneous,
and the library should probably be better off being synchronized with upstream
so there is some kind of ABI compatibility between the two distributions.
.
FIXME: please use upstream soversion next time an ABI breakage justifies an
SOVERSION bump, instead of just incrementing the present version by one.
Author: Étienne Mollier <emollier@debian.org>
Forwarded: not-needed
Last-Update: 2023-09-17
Change-Id: I6c4d28bd26889359c0b83c474d5ae58a81741cf4
Co-authored-by: Étienne Mollier <emollier@debian.org>
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
[ROCm/rocm_smi_lib commit: 1775ae4b8d]
* Updates:
- [API] After discovering all amd gpus, we now properly
map correct bdf (xgmi nodes). Especially important for
partition changes - aka secondary nodes.
- [API] While adding new secondary nodes we now have
better grouping -> due to resorting based on
kfd properties list & matching to primary uniqueid
- [API] All secondary nodes are now AddToDeviceList
with correct bdf (location id), provided by kfd
- [API] Modified AddToDeviceList(..., uint64_t bdfid):
providing an optional field - bdfid. This allows working
around primary pcie cards with xgmi nodes
- [API] Utils - cpplint minor fixes
- [Example] Removed all endl references w/ newline, fixed
spacing, and some incorrect values displaying as hex
(needed dec representation)
- [API] kfd node functions - now print full path of file
for trace logs
- [Tests] power_read.cc: Added in generic power test to
confirm guaranteeing specific return values
Change-Id: I143474e8d64c4915a966e789be6bcea4fa7f4472
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 6f1afd2678]
* Updates:
- [API] Added rsmi_dev_power_get(uint32_t dv_ind,
uint64_t *power,
RSMI_POWER_TYPE
*type)
provides generic get to average or
current power & provides backwards
compatibility
- Added a utility function to get MonitorTypes
(monitor_type_string(type)) &
RSMI_POWER_TYPE (power_type_string(type))
strings
- [Tests] Added rsmi_dev_power_get tests and
provided better verification of return values for
all power APIs
- [Tests] Updated power outputs to show correct
units
- [example] Now uses avg, current, and generic
power functions with type output response
Change-Id: I5ca06ca37fd5f61e100f2835b664d6cdd1ca42e6
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 31a1fcce7d]
* Updates:
- rocm_smi_lib + CLI:
Rename all "NPS mode" -> "memory partition"
related files/functions/API/CLI to align with correct
technical naming
- rocm_smi_main: fixed identifying primary card's unique id
utilize rsmi_dev_unique_id_get to map which
KFD nodes belong to it
- rsmi_dev_*_partition*: now have better logging output
- compute partition tests:
Added 20 sec delay for workaround until GPU
busy is confirmed as the issue
- CPPLint fixes/formatting
- [Example] Moved all endl to "\n" for efficiency
- [Example] Added Edge & Junction temperature examples
- [Example] Added rsmi_minmax_bandwidth_get() example - WIP
Change-Id: Ida6db6fda7e0ac9d696a34cb15b4746e69d58d51
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: b251bb0c9f]
* Updates:
- Fixed infinit loop on systems
which did not have VRAM files
- Fixed concise info from throwing exception
with no amdgpu driver loaded
- Fix for ability to see all nodes when
after switching partitions (mirrors
original card display/settings)
- Added to logs build type, lib path,
and set env. variables
Change-Id: Ic0333df355144ce2242cecea93fe4ce51caf311c
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: ed6777a8e7]
Sending RSMI_STATUS_UNEXPECTED_DATA for drivers
which do not set some clock freqs
Change-Id: I43a9515c2757dddd412bb25cfd54095e63367030
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: f191c2753c]
Code changes related to the following:
* Added 'rsmi_dev_revision_get()' related code
* Test code
* Functional tests
Change-Id: I8c2097c65384a028c8c8437b717d05d52fe45250
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
[ROCm/rocm_smi_lib commit: 573620f586]
The librocm_smi64.so is used for development, while
librocm_smi64.so.MAJOR is used for runtime, thus the python front end
should not be loading the .so binary, but rather the .so.MAJOR binary.
As well, it's good not to hardcode "lib" as some distros will change
this.
rsmiBindings.py is now generated with CMake
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
Change-Id: I7cb745f8936fdf10d3ebd6c1e606031f713184ca
[ROCm/rocm_smi_lib commit: 2d2c73a5e6]
When building from github, these tags don't exist, so the defaults
should try to match the internal tags
Change-Id: Id570341f27e21916b1a7f3605ee2b5b9716cad9b
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
[ROCm/rocm_smi_lib commit: 74dc98114f]
The keyword "PROGRAMS" should be used in place of "FILES" in order to
make sure executable scripts have the correct permissions.
Change-Id: I6c287dc1291774ad6d97a04d621957dea0a1b697
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
[ROCm/rocm_smi_lib commit: d00d885394]
Updates:
* VoltRead - needed to properly send out RSMI_STATUS_NOT_SUPPORTED
when device does not have voltage hwmon files
* ComputePart. - test failure was likely caused due to EvtNotif
causing conflicts (unknown exactly why). Test passes when
moving it ahead of the event notifier. Both API calls may have
a system resource issue, TBD.
* rocm_smi_example - now indicates when an API call
returns RSMI_STATUS_NOT_SUPPORTED or
RSMI_STATUS_NOT_YET_IMPLEMENTED. Allows example to fully complete
on systems which may not provide support for all API calls.
Change-Id: I520b8584e078d412414e8e5797c664220a7e823a
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 78a0812f7f]
Updates:
* Fixed rsmi_dev_compute_partition_get
& rsmi_dev_nps_mode_get to properly check
for invalid arguments
* Updated compute partition & NPS mode tests
- Now properly confirms the invalid
argument is seen
- Spacing for multiple devices is added
to better see distinction between
separate device's tests (for verbose output)
- Changed expect to assert calls, so errors
are observed faster for test failures
- Fixed multiple device testing where a
variable should have been unset, but
having multiple devices caused it to
set
- Updated multiple device testing to iterate
accross all devices (previously returned,
instead of continuing checking support
after RSMI_STATUS_NOT_SUPPORTED detected)
- Fixed a few spelling errors & verbose output
Change-Id: Ieba9e5b46763c6cd880fbf27fcdf58be8ecbc683
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: c252ecccd1]
Updates:
* Added RSMI_STATUS_SETTING_UNAVAILABLE for
rsmi_dev_compute_partition_set - gives users
better error output when attempting to set
compute partition to values not listed in
available_compute_partition SYSFS
* Updated python --setcomputepartition to
provide better output when receiving
RSMI_STATUS_SETTING_UNAVAILABLE
* Updated all test & example files to check for
RSMI_STATUS_SETTING_UNAVAILABLE when doing
rsmi_dev_compute_partition_set
Change-Id: Ida5d54880d9b9b6e4a0468cdb962fdc0c18d6257
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 0d3558945b]
Updates:
* Added rsmi_dev_compute_partition_reset & rsmi_dev_nps_mode_reset
* Added --resetcomputepartition and --resetnpsmode python smi calls
* Added temp data files rocmsmi_boot_compute_partition_<device num>
& rocmsmi_boot_nps_mode_partition_<device num>, writes UNKNOWN
if data cannot be read or device does not support
* Cleaned up NPS & compute API documentation
* Added creation and reading of API temp files (used in reset
functionality)
* Cleaned up output of rocm_smi_example
* Updated rocm_smi_example to check if running with sudo permission
before executing write API calls (cleans up erroneous output)
* Added template specialization for storing temp data, requires
specific rsmi_type_t enums (restrics what data can be stored)
* Added storage of temp data, if temp files do not exist
* Updated google tests for NPS & compute to include reset API calls
Change-Id: I69895a466b97107617e6dbb355737b84499a76c9
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 77c950a4bf]
Updates:
* Added rsmi_dev_nps_mode_set and rsmi_dev_nps_mode_get
* Added ability to set multiple SYSFS files in debug build
* Added ability to see user's env variables set for debug build
* Added tests for rsmi_dev_nps_mode_set and rsmi_dev_nps_mode_get
* Added ability to restart AMD GPU driver, used in nps_mode_set
* Updated ROCm_SMI_Manual.pdf to include new APIs
* Added progress bar for long running python_smi_tools, used
in setting nps_mode if runs longer than .1 seconds
Change-Id: I6d61bedd28d7cba6aff432ad2d127ba741b7d15a
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/rocm_smi_lib commit: 9ef376cd61]
Original updates:
* Added .gitignore to help with future commits
* Updated/added copyrights on modified or added files
* Updated rocm_smi.h/.cc
- Added 3 new SMI API functions:
rsmi_dev_compute_partition_set &
rsmi_dev_compute_partition_get
- Added helpful maps/enums used in
new get/set compute_partition API calls
* Updated rocm_smi.py
- Added --showcomputepartition
- Added --setcomputepartition
- Fixed a few mistypes
* Updated rsmiBindings.py - added helpful class/dict/list
* Updated rocm_smi_example.cc
- Added helpful MACRO to detect if api is not supported.
- Added current_compute_partition set/get rocm lib calls
- Added helpful macro to discover future RSMI errors
- Commented out test_set_freq, was having permission issues
on a Navi21
* Updated rocm_smi_main.cc
- Added helpful map to debug API calls, left in for future use
- Added comment to better understand a non-class function returns
* Added computepartition_read_write.cc/.h
- Added get/set compute partition API test calls
- Confirmed on devices that do not support the API calls, tests pass
* Updated rocm_smi_test/main.cc
- Calls new compute partition gtests
Added following updates from review feedback:
* Updated rocm_smi.h/cc
- Removed C++ API calls, adding support for both C/C++
API calls could cause confusion and adds extra work for us
- rsmi_dev_compute_partition_get -> Fixed an edge case where
user gives a small buffer length size (smaller than data
received), but does not receive the partial buffer back.
google Tests are updated to reflect this find.
* Updated rocm_smi_example.cc
- Fixed test_set_freq, issue was that file was not writable.
We now indicate this warning, so prior errors make sense.
- General test code cleanup. Removed extra code,
by creating loops for tests.
* Updated rocm_smi_main.cc
- Moved and got rid of an external reference to a map used
for debugging RSMI enums, now is a const public reference.
* Updated rocm_smi.py
- Updated python code to identify NOT_SUPPORTED due to
(currently) only a few GPU support the feature
Change-Id: I4a567acbb59d6771fb64df08d19175fe3604fd1b
[ROCm/rocm_smi_lib commit: 4d7f3f2bc7]
Install interface should provide /opt/rocm-ver/include as the include path
Path /opt/rocm-ver/rocm_smi/include should be used only as part of backward compatibility support
Change-Id: Idc1f663069356c6b1fbd492f45ef4637fc90e4eb
[ROCm/rocm_smi_lib commit: 9a650b1378]
Changes made on rsmi_perf_determinism_mode_set function documentation
as well for styling consistency.
Change-Id: I09ce8139eb9cbda94352ac7725c4c9b9bb06bd59
[ROCm/rocm_smi_lib commit: aa5cba122c]
Use GNUInstallDirs variables to determine the location of LIBDIR, BINDIR, INCLUDEDIR, DOCDIR
Note that CMAKE_INSTALL_LIBDIR is overriden, since the default for RHEL
is lib64, but ROCm packaging wants it to be lib always. Distros or users
can easily override this.
Change-Id: I616152ccd2bc1f5a60bffa940312b38ca6e88c04
[ROCm/rocm_smi_lib commit: b72c464ac0]
Also update copyright years
Signed-off-by: Kent Russell <kent.russell@amd.com>
Change-Id: Ic9ead543c4937680afc1957623c4d5fcbfbd58b0
[ROCm/rocm_smi_lib commit: 85571318e2]
Wrapper header files
Soft link to libraries and binaries
rocm_smi.py and rsmiBindings.py installed in libexec/rocm_smi
Binaries, libraries and header files installed as per File Reorg folder structure
Change-Id: I3166ab67f89c2ae4aafbc87bb00c9a5233221ade
[ROCm/rocm_smi_lib commit: f1da5591b5]
When build the release, it will strip the library file instead of link.
Change-Id: Ib2d4cea614e8938bdb2be0fd74f046680158d256
[ROCm/rocm_smi_lib commit: 77502bed2a]
systems-assistant[bot]
Arif, Maisam
Ranjith Ramakrishnan
Peter Park
Stella Laurenzo
Afzal Patel
Poag, Charis
Castillo, Juan
Oliveira, Daniel
Istvan Kiss
Galantsev, Dmitrii
Étienne Mollier
Jeremy Newton
Sam Wu
AravindanC
Sreekant Somasekharan
Kent Russell
Bill(Shuzhou) Liu