[Issue]: Video decoding acceleration via libva breaks after installing ROCm(Cannot find target for triple amdgcn-- Unable to find target for this triple (no targets are registered))
Closed this issue · 11 comments
Problem Description
NOTE: The GPU field of this form was originally wrongly selected as no applicable model is enumerated.
After installing rocm 6.2.2 I can no longer play local video files using VA-API decoding acceleration.
The application crashed shortly after launching:
$ gst-play-1.0 --videosink vaapidecodebin /path/to/video.mp4
Press 'k' to see a list of keyboard shortcuts.
Now playing /home/brlin/??/Animation_ New recruit learns military customs from his sergeant ? Saka(23_sakapicdora_ENG).mp4
Redistribute latency...
Redistribute latency...
ac: Unknown GPU, using 0 for raster_config
Cannot find target for triple amdgcn-- Unable to find target for this triple (no targets are registered)
Segmentation fault (core dumped)
Operating System
Ubuntu 24.04
CPU
AMD Ryzen 7 7840U w/ Radeon 780M Graphics
GPU
Radeon 780M Graphics
c1:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Phoenix1 (rev c4)
ROCm Version
ROCm 6.2.0
ROCm Component
ROCm
Steps to Reproduce
-
Install ROCm via the Quick start installation guide — ROCm installation (Linux).
-
Play any video which decoding is hardware-accelerable using libva:
totem /path/to/video.mp4
(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support
ROCk module version 6.8.5 is loaded
=====================
HSA System Attributes
=====================
Runtime Version: 1.14
Runtime Ext Version: 1.6
System Timestamp Freq.: 1000.000000MHz
Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
Machine Model: LARGE
System Endianness: LITTLE
Mwaitx: DISABLED
DMAbuf Support: YES
==========
HSA Agents
==========
*******
Agent 1
*******
Name: AMD Ryzen 7 7840U w/ Radeon 780M Graphics
Uuid: CPU-XX
Marketing Name: AMD Ryzen 7 7840U w/ Radeon 780M Graphics
Vendor Name: CPU
Feature: None specified
Profile: FULL_PROFILE
Float Round Mode: NEAR
Max Queue Number: 0(0x0)
Queue Min Size: 0(0x0)
Queue Max Size: 0(0x0)
Queue Type: MULTI
Node: 0
Device Type: CPU
Cache Info:
L1: 32768(0x8000) KB
Chip ID: 0(0x0)
ASIC Revision: 0(0x0)
Cacheline Size: 64(0x40)
Max Clock Freq. (MHz): 5132
BDFID: 0
Internal Node ID: 0
Compute Unit: 16
SIMDs per CU: 0
Shader Engines: 0
Shader Arrs. per Eng.: 0
WatchPts on Addr. Ranges:1
Memory Properties:
Features: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: FINE GRAINED
Size: 61515984(0x3aaa8d0) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 2
Segment: GLOBAL; FLAGS: KERNARG, FINE GRAINED
Size: 61515984(0x3aaa8d0) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 3
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 61515984(0x3aaa8d0) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
ISA Info:
*******
Agent 2
*******
Name: gfx1103
Uuid: GPU-XX
Marketing Name: AMD Radeon Graphics
Vendor Name: AMD
Feature: KERNEL_DISPATCH
Profile: BASE_PROFILE
Float Round Mode: NEAR
Max Queue Number: 128(0x80)
Queue Min Size: 64(0x40)
Queue Max Size: 131072(0x20000)
Queue Type: MULTI
Node: 1
Device Type: GPU
Cache Info:
L1: 32(0x20) KB
L2: 2048(0x800) KB
Chip ID: 5567(0x15bf)
ASIC Revision: 7(0x7)
Cacheline Size: 64(0x40)
Max Clock Freq. (MHz): 2700
BDFID: 49408
Internal Node ID: 1
Compute Unit: 12
SIMDs per CU: 2
Shader Engines: 1
Shader Arrs. per Eng.: 2
WatchPts on Addr. Ranges:4
Coherent Host Access: FALSE
Memory Properties: APU
Features: KERNEL_DISPATCH
Fast F16 Operation: TRUE
Wavefront Size: 32(0x20)
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Max Waves Per CU: 32(0x20)
Max Work-item Per CU: 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 4294967295(0xffffffff)
y 4294967295(0xffffffff)
z 4294967295(0xffffffff)
Max fbarriers/Workgrp: 32
Packet Processor uCode:: 39
SDMA engine uCode:: 18
IOMMU Support:: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 30757992(0x1d55468) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 30757992(0x1d55468) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 3
Segment: GROUP
Size: 64(0x40) KB
Allocatable: FALSE
Alloc Granule: 0KB
Alloc Recommended Granule:0KB
Alloc Alignment: 0KB
Accessible by all: FALSE
ISA Info:
ISA 1
Name: amdgcn-amd-amdhsa--gfx1103
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 4294967295(0xffffffff)
y 4294967295(0xffffffff)
z 4294967295(0xffffffff)
FBarrier Max Size: 32
*** Done ***
Additional Information
System: Framework Laptop 13(AMD 7040 series)
Please help remove the invalid AMD Instinct MI300X tag, thanks!
Similar situation here
Similar situation here, Framework laptop, gfx1103, similar software setup.
Applications that need VA-API or VDPAU support abort during start with the mentioned error message.
Work-around for me
$ LIBVA_DRIVER_NAME="vdpau" <application>
or
$ LIBVA_DRIVER_NAME="vaapi" <application>
Looks like a detection error of existing video hardware acceleration.
I guess a system-wide export of this variable could work as well, but unclear if there are some side effects.
After re-installing ROCm I can no longer reproduce this issue, I would say that the workaround you enumerated doesn't work IIRC, though.
Was there any difference in your installation? We were not able to reproduce this issue either after adding the graphics use case to our installation i.e
sudo amdgpu-install --usecase=graphics,rocm
Adding this use case is recommended for graphical workloads such as hardware video acceleration. @frzb and @brlin-tw, could you please give this a try and let me know if it resolves your issue?
Was there any difference in your installation? We were not able to reproduce this issue either after adding the graphics use case to our installation i.e
sudo amdgpu-install --usecase=graphics,rocmAdding this use case is recommended for graphical workloads such as hardware video acceleration. @frzb and @brlin-tw, could you please give this a try and let me know if it resolves your issue?
I believe I was using the package installation(e.g. by following the quickstart guide) when I reproduced the issue.
The second installation attempt I'm using the exact command you mentioned and it worked.
I'm glad to hear that it's now working on your side!
@harkgill-amd Thanks a lot for your hints!
Using sudo amdgpu-install --usecase=graphics which reflects to sudo apt install amdgpu-lib amdgpu-lib32 made hardware video acceleration for the application work again without setting LIBVA_DRIVER_NAME=... manually.
Case closed for me. 😃
Only left-over remark from my side is that the official Quick start installation guide for ROCm should not break libva hardware acceleration in the first place, or I am wrong? 😉
Technically these releases are meant for headless configurations, which is why the quick start guide can cause some issues with graphical applications when physical displays are involved. We also have official ROCm on Radeon releases (such as 6.1.3) which are specifically intended for and tested on such configurations, as pointed out above the quick start instructions:
If you’re using ROCm with AMD Radeon or Radeon Pro GPUs for graphics workloads, see the Use ROCm on Radeon GPU documentation for installation instructions .
That being said, the regular releases with graphics usecase tend to work on these systems as well, so you shouldn't need to downgrade to the Radeon-specific version unless you run into further issues. Thanks for your input!
Technically these releases are meant for headless configurations, which is why the quick start guide can cause some issues with graphical applications when physical displays are involved. We also have official ROCm on Radeon releases (such as 6.1.3) which are specifically intended for and tested on such configurations, as pointed out above the quick start instructions:
@schung-amd Oh yes, I see! RTFM 👍
Would a simple heuristic guess checking in amdgpu-install if $XDG_SESSION_TYPE contains a value and if so picking or recommending --usecase=graphics make sense to improve the UX?
Reasoning behind this
I thought about why I missed this documentation from a UX perspective.
And I think it is caused by the misalignment between AMD product brand naming and the technical naming. Radeon™ is used as brand name but abandoned as software name on a technical level in Linux since some time. In my perception radeon was percepted as this legacy software. This is confusing and misleading.
Would a simple heuristic guess checking in amdgpu-install if $XDG_SESSION_TYPE contains a value and if so picking or recommending --usecase=graphics make sense to improve the UX?
Thanks for your feedback! Ultimately, I think the issue lies in a lack of visibility and clarity about when the graphics usecase (or equivalently the ROCm on Radeon release) is required. While the graphics usecase is required for some graphical workloads with a physical display attached, ROCm still functions in most other cases with a physical display even without the graphics usecase, so it would be helpful if we could be more specific about when --usecase=graphics is warranted (versus, say, the multimedia usecase, or no additional usecases).
And I think it is caused by the misalignment between AMD product brand naming and the technical naming. Radeon™ is used as brand name but abandoned as software name on a technical level in Linux since some time. In my perception radeon was percepted as this legacy software. This is confusing and misleading.
This is interesting, and I'll bring this up to the internal team. We're working on improving the installation methods and documentation, so your feedback is greatly appreciated.
IMHO the quick start guide should simply instruct the user to use the amdgpu-install script instead of:
sudo apt install amdgpu-dkms rocm
As the script is essentially a wrapper of package installation there's little reason to instruct the users to run the package installation command.