ROCm 6.1.2 includes enhancements to SMI tools and improvements to some libraries.
ROCm 6.1.2 has been tested against a pre-release version of Ubuntu 22.04.5 (kernel: 5.15 [GA], 6.8 [HWE]).
AMD SMI for ROCm 6.1.2
- Added process isolation and clean shader APIs and CLI commands.
amdsmi_get_gpu_process_isolation()
amdsmi_set_gpu_process_isolation()
amdsmi_set_gpu_clear_sram_data()
- Added the
MIN_POWER
metric to output provided byamd-smi static --limit
.
- Updated the
amd-smi monitor --pcie
output to prevent delays with themonitor
command.
- Updated
amismi_get_power_cap_info
to return values in uW instead of W. - Updated Python library return types for
amdsmi_get_gpu_memory_reserved_pages
andamdsmi_get_gpu_bad_page_info
. - Updated the output of
amd-smi metric --ecc-blocks
to show counters available from blocks.
amdsmi_get_gpu_board_info()
no longer returns junk character strings.amd-smi metric --power
now correctly details power output for RDNA3, RDNA2, and MI1x devices.- Fixed the
amdsmitstReadWrite.TestPowerCapReadWrite
test for RDNA3, RDNA2, and MI100 devices. - Fixed an issue with the
amdsmi_get_gpu_memory_reserved_pages
andamdsmi_get_gpu_bad_page_info
Python interface calls.
- Removed the
amdsmi_get_gpu_process_info
API from the Python library. It was removed from the C library in an earlier release.
See the AMD SMI [detailed changelog](https://github.com/ROCm/amdsmi/blob/rocm-6.1.x/CHANGELOG.md) with code samples for more information.
ROCm SMI for ROCm 6.1.2
- Added the ring hang event to the
amdsmi_evt_notification_type_t
enum.
- Fixed an issue causing ROCm SMI to incorrectly report GPU utilization for RDNA3 GPUs. See the issue on GitHub.
- Fixed the parsing of
pp_od_clk_voltage
inget_od_clk_volt_info
to work better with MI-series hardware.
Library | Version |
---|---|
AMDMIGraphX | 2.9 |
composable_kernel | 0.2.0 |
hipBLAS | 2.1.0 |
hipBLASLt | 0.7.0 |
hipCUB | 3.1.0 |
hipFFT | 1.0.14 |
hipRAND | 2.10.17 |
hipSOLVER | 2.1.1 |
hipSPARSE | 3.0.1 |
hipSPARSELt | 0.2.0 |
hipTensor | 1.2.0 |
MIOpen | 3.1.0 |
MIVisionX | 2.5.0 |
rccl | 2.18.6 |
rocALUTION | 3.1.1 |
rocBLAS | 4.1.0 ⇒ 4.1.2 |
rocDecode | 0.5.0 ⇒ 0.6.0 |
rocFFT | 1.0.27 |
rocm-cmake | 0.12.0 |
rocPRIM | 3.1.0 |
rocRAND | 3.0.1 |
rocSOLVER | 3.25.0 |
rocSPARSE | 3.1.2 |
rocThrust | 3.0.1 |
rocWMMA | 1.4.0 |
rpp | 1.5.0 |
Tensile | 4.40.0 |
RCCL 2.18.6 for ROCm 6.1.2
- Reduced
NCCL_TOPO_MAX_NODES
to limit stack usage and avoid stack overflow.
rocBLAS 4.1.2 for ROCm 6.1.2
- Tuned BBS TN and TT operations on the CDNA3 architecture.
- Fixed an issue related to obtaining solutions for BF16 TT operations.
rocDecode 0.6.0 for ROCm 6.1.2
- Added support for FFmpeg v5.x.
- Updated error checking in the
rocDecode-setup.py
script.
- Updated core dependencies.
- Updated to support the use of public LibVA headers.
- Fixed some package dependencies.
- A future release will enable the use of HIPCC compiled binaries
hipcc.bin
andhipconfig.bin
by default. No action is needed by users; you may continue calling high-level Perl scriptshipcc
andhipconfig
.hipcc.bin
andhipconfig.bin
will be invoked by the high-level Perl scripts. To revert to the previous behavior and invokehipcc.pl
andhipconfig.pl
, set theHIP_USE_PERL_SCRIPTS
environment variable to1
. - A subsequent release will remove high-level HIPCC Perl scripts from
hipcc
andhipconfig
. This release will remove theHIP_USE_PERL_SCRIPTS
environment variable. It will renamehipcc.bin
andhipconfig.bin
tohipcc
andhipconfig
respectively. No action is needed by the users. To revert to the previous behavior, invokehipcc.pl
andhipconfig.pl
explicitly. - A subsequent release will remove
hipcc.pl
andhipconfig.pl
for HIPCC.