ROCm in Raspberry PI5

[tanreinama]

In my environment, RadeonRX480 is running on RaspberryPI5.
https://github.com/geerlingguy/raspberry-pi-pcie-devices
I used this repository as a reference.

And I found that the ROCm source code is compatible with AARCH64.
#231

Therefore, I think I can run ROCm on RaspberryPI and use it for edge AI, but it doesn't work.

I downloaded the source code using repo, compiled and installed rocm-core, but I'm having trouble installing the kernel module.

I feel like I'm simply not doing it enough.
Any ideas?

[fluidnumerics-joe]

What does uname -r show for your Linux kernel version?

Which raspberry pi has a discrete AMD GPU? Integrated AMD graphics won't work with ROCm

Edit: checked the link. I didn't realize Raspberry Pi now has PCIe! Quite neat. Disregard my last question about the discrete GPU

What are the specific errors you're seeing when building the AMDGPU drivers?

[tanreinama]

$ uname -r
6.1.61-v8_16k-amd+

The "-amd+" part is the SUSFIX I added when I compiled the kernel.
Not integrated graphics.
I installed an external RadeonRX480 on the PCIe port of RaspberryPI5 via a PCIe->M.2 adapter and M.2->PCIe x16 riser card.
The power supply is also wired and the graphics board is working.

Simply, I don't know how to compile and install ROCm from source code.
Once please forgetting about RaspberryPI, could you tell me how to install ROCm from the source code (not apt-get) on PC?
When you sync using repo, many directories are created. I was able to build some of them using cmake, but I don't know how to build the whole thing.

extra hw imformation.

$ lspci
0000:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries Device 2712 (rev 21)
0000:01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] (rev c7)
0000:01:00.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere HDMI Audio [Radeon RX 470/480 / 570/580/590]
0001:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries Device 2712 (rev 21)
0001:01:00.0 Ethernet controller: Device 1de4:0001

 $ dmesg|grep amd
[    0.000000] Linux version 6.1.61-v8_16k-amd+ (root@raspberrypi) (gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40) #1 SMP PREEMPT Sun Mar 17 14:29:09 JST 2024
[    0.729754] usb usb1: Manufacturer: Linux 6.1.61-v8_16k-amd+ xhci-hcd
[    0.730135] usb usb2: Manufacturer: Linux 6.1.61-v8_16k-amd+ xhci-hcd
[    0.789729] usb usb3: Manufacturer: Linux 6.1.61-v8_16k-amd+ xhci-hcd
[    0.790067] usb usb4: Manufacturer: Linux 6.1.61-v8_16k-amd+ xhci-hcd
[   15.025562] [drm] amdgpu kernel modesetting enabled.
[   15.025664] amdgpu: CRAT table disabled by module option
[   15.025667] amdgpu: DSDT table not found for OEM information
[   15.025669] amdgpu: IO link not available for non x86 platforms
[   15.025671] amdgpu: Virtual CRAT table created for CPU
[   15.025686] amdgpu: Topology: Add CPU node
[   15.025860] amdgpu 0000:01:00.0: enabling device (0000 -> 0002)
[   15.237875] amdgpu 0000:01:00.0: amdgpu: Fetched VBIOS from ROM BAR
[   15.237883] amdgpu: ATOM BIOS: 113-V34111-F1
[   15.237923] amdgpu 0000:01:00.0: amdgpu: Trusted Memory Zone (TMZ) feature not supported
[   15.237927] amdgpu 0000:01:00.0: amdgpu: PCIE atomic ops is not supported
[   15.373214] amdgpu 0000:01:00.0: BAR 2: releasing [mem 0x1810000000-0x18101fffff 64bit pref]
[   15.373220] amdgpu 0000:01:00.0: BAR 0: releasing [mem 0x1800000000-0x180fffffff 64bit pref]
[   15.373255] amdgpu 0000:01:00.0: BAR 0: assigned [mem 0x1800000000-0x19ffffffff 64bit pref]
[   15.373265] amdgpu 0000:01:00.0: BAR 2: assigned [mem 0x1a00000000-0x1a001fffff 64bit pref]
[   15.373288] amdgpu 0000:01:00.0: amdgpu: VRAM: 8192M 0x000000F400000000 - 0x000000F5FFFFFFFF (8192M used)
[   15.373291] amdgpu 0000:01:00.0: amdgpu: GART: 256M 0x000000FF00000000 - 0x000000FF0FFFFFFF
[   15.373342] [drm] amdgpu: 8192M of VRAM memory ready
[   15.373345] [drm] amdgpu: 4026M of GTT memory ready.
[   15.393599] amdgpu: hwmgr_sw_init smu backed is polaris10_smu
[   15.871888] kfd kfd: amdgpu: skipped device 1002:67df, PCI rejects atomics 730<0
[   15.871907] amdgpu 0000:01:00.0: amdgpu: SE 4, SH per SE 1, CU per SH 9, active_cu_number 36
[   15.875210] amdgpu 0000:01:00.0: amdgpu: Using BACO for runtime pm
[   15.880076] [drm] Initialized amdgpu 3.49.0 20150101 for 0000:01:00.0 on minor 2
[   15.885274] amdgpu 0000:01:00.0: [drm] Cannot find any crtc or sizes

[fortunerains]

How's it going now