Test GPU (AMD Radeon R9 380X)

[420noscope-exe]

I just recently got my new 16GB cm5 lite and upgraded my ARM server. I had an R9 380 sitting in the server, disconnected for quite some time, waiting for the day I could use it. I would like a server for testing custom game engine builds and compiling ARM binaries for them. I want vulkan support so I decided to use the amdgpu driver.

I didn't get it to work, but I think I got pretty far along:

sudo apt install git

git clone --depth=1 --branch rpi-6.6.y-gpu https://github.com/Coreforge/linux.git

cd linux

wget https://gist.githubusercontent.com/Coreforge/91da3d410ec7eb0ef5bc8dee24b91359/raw/b4848d1da9fff0cfcf7b601713efac1909e408e8/memcpy_unaligned.c

gcc -shared -fPIC -o memcpy.so memcpy_unaligned.c
sudo mv memcpy.so /usr/local/lib/memcpy.so
sudo nano /etc/ld.so.preload

Then I put the following line inside ld.so.preload:
/usr/local/lib/memcpy.so

sudo apt install bc bison flex libssl-dev make raspberrypi-kernel-headers

KERNEL=kernel_2712
make bcm2712_defconfig

sudo apt-get install libncurses-dev
make menuconfig

Kernel Features > Kernel support for 32-bit EL0 > Fix up misaligned multi-word loads and stores in user space
Kernel Features > Fix up misaligned loads and stores from userspace for 64bit code
Device Drivers > Graphics support > AMD GPU , as well as SI/CIK support
Device Drivers > Graphics support > AMD GPU>Display Engine Configuration>AMD DC Support for Southern Islands ASICs
Device Drivers > Graphics support > Direct Rendering Manager (XFree86 4.1.0 and higher DRI support) > Force Architecture can write-combine memory

saved as .config

make -j6 Image.gz modules dtbs
sudo make -j6 modules_install

sudo cp /boot/firmware/$KERNEL.img /boot/firmware/$KERNEL-backup.img
sudo cp arch/arm64/boot/Image.gz /boot/firmware/$KERNEL.img
sudo cp arch/arm64/boot/dts/broadcom/*.dtb /boot/firmware/
sudo cp arch/arm64/boot/dts/overlays/*.dtb* /boot/firmware/overlays/
sudo cp arch/arm64/boot/dts/overlays/README /boot/firmware/overlays/

sudo apt install firmware-amd-graphics

sudo reboot

I didn't get a display output, but I did see some errors with dmesg | grep amd

[    3.441096] [drm] amdgpu kernel modesetting enabled.
[    3.441160] amdgpu 0000:01:00.0: of_irq_parse_pci: failed with rc=134
[    3.441301] amdgpu 0000:01:00.0: Unable to change power state from D3cold to D0, device inaccessible
[   15.446930] amdgpu: Doesn't get ack from pf.
[   15.446935] amdgpu: Doesn't get ack from pf, give up
[   15.446940] amdgpu 0000:01:00.0: amdgpu: Fatal error during GPU init
[   15.446943] amdgpu 0000:01:00.0: amdgpu: amdgpu: finishing device.
[   27.447168] amdgpu: Doesn't get ack from pf.
[   39.447397] amdgpu: Doesn't get ack from pf.
[   39.447403] [drm:amdgpu_device_fini_hw [amdgpu]] *ERROR* failed to release exclusive mode on fini
[   39.447675] amdgpu: probe of 0000:01:00.0 failed with error -62

pasquale@PASARM01:~/linux $ sudo lspci -vvv
0000:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries BCM2712 PCIe Bridge (rev 30) (prog-if 00 [Normal decode])
	Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
	Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
	Latency: 0
	Interrupt: pin A routed to IRQ 38
	Bus: primary=00, secondary=01, subordinate=01, sec-latency=0
	Memory behind bridge: 80000000-800fffff [size=1M] [32-bit]
	Prefetchable memory behind bridge: 1800000000-1817ffffff [size=384M] [32-bit]
	Secondary status: 66MHz- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- <SERR- <PERR-
	BridgeCtl: Parity- SERR- NoISA- VGA- VGA16- MAbort- >Reset- FastB2B-
		PriDiscTmr- SecDiscTmr- DiscTmrStat- DiscTmrSERREn-
	Capabilities: [48] Power Management version 3
		Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold-)
		Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=1 PME-
	Capabilities: [ac] Express (v2) Root Port (Slot-), MSI 00
		DevCap:	MaxPayload 512 bytes, PhantFunc 0
			ExtTag- RBE+
		DevCtl:	CorrErr- NonFatalErr- FatalErr- UnsupReq-
			RlxdOrd+ ExtTag- PhantFunc- AuxPwr+ NoSnoop+
			MaxPayload 256 bytes, MaxReadReq 512 bytes
		DevSta:	CorrErr+ NonFatalErr- FatalErr- UnsupReq- AuxPwr- TransPend-
		LnkCap:	Port #0, Speed 5GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <2us
			ClockPM+ Surprise- LLActRep- BwNot+ ASPMOptComp+
		LnkCtl:	ASPM L0s L1 Enabled; RCB 64 bytes, Disabled- CommClk+
			ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
		LnkSta:	Speed 2.5GT/s, Width x1
			TrErr- Train- SlotClk+ DLActive- BWMgmt+ ABWMgmt+
		RootCap: CRSVisible+
		RootCtl: ErrCorrectable- ErrNon-Fatal- ErrFatal- PMEIntEna+ CRSVisible+
		RootSta: PME ReqID 0000, PMEStatus- PMEPending-
		DevCap2: Completion Timeout: Range ABCD, TimeoutDis+ NROPrPrP- LTR+
			 10BitTagComp- 10BitTagReq- OBFF Via WAKE#, ExtFmt- EETLPPrefix-
			 EmergencyPowerReduction Not Supported, EmergencyPowerReductionInit-
			 FRS- LN System CLS Not Supported, TPHComp- ExtTPHComp- ARIFwd+
			 AtomicOpsCap: Routing- 32bit- 64bit- 128bitCAS-
		DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis- LTR- 10BitTagReq- OBFF Disabled, ARIFwd+
			 AtomicOpsCtl: ReqEn- EgressBlck-
		LnkCap2: Supported Link Speeds: 2.5-5GT/s, Crosslink- Retimer- 2Retimers- DRS+
		LnkCtl2: Target Link Speed: 8GT/s, EnterCompliance- SpeedDis-
			 Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-
			 Compliance Preset/De-emphasis: -6dB de-emphasis, 0dB preshoot
		LnkSta2: Current De-emphasis Level: -3.5dB, EqualizationComplete- EqualizationPhase1-
			 EqualizationPhase2- EqualizationPhase3- LinkEqualizationRequest-
			 Retimer- 2Retimers- CrosslinkRes: unsupported, DRS-
			 DownstreamComp: Link Down - Not Determined
	Capabilities: [100 v1] Advanced Error Reporting
		UESta:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UEMsk:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UESvrt:	DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt- UnxCmplt- RxOF+ MalfTLP+ ECRC- UnsupReq- ACSViol-
		CESta:	RxErr- BadTLP- BadDLLP- Rollover- Timeout+ AdvNonFatalErr-
		CEMsk:	RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr+
		AERCap:	First Error Pointer: 00, ECRCGenCap+ ECRCGenEn- ECRCChkCap+ ECRCChkEn-
			MultHdrRecCap- MultHdrRecEn- TLPPfxPres- HdrLogCap-
		HeaderLog: 00000000 00000000 00000000 00000000
		RootCmd: CERptEn+ NFERptEn+ FERptEn+
		RootSta: CERcvd- MultCERcvd- UERcvd- MultUERcvd-
			 FirstFatal- NonFatalMsg- FatalMsg- IntMsg 0
		ErrorSrc: ERR_COR: 0000 ERR_FATAL/NONFATAL: 0000
	Capabilities: [160 v1] Virtual Channel
		Caps:	LPEVC=0 RefClk=100ns PATEntryBits=1
		Arb:	Fixed- WRR32- WRR64- WRR128-
		Ctrl:	ArbSelect=Fixed
		Status:	InProgress-
		VC0:	Caps:	PATOffset=00 MaxTimeSlots=1 RejSnoopTrans-
			Arb:	Fixed- WRR32- WRR64- WRR128- TWRR128- WRR256-
			Ctrl:	Enable+ ID=0 ArbSelect=Fixed TC/VC=ff
			Status:	NegoPending- InProgress-
	Capabilities: [180 v1] Vendor Specific Information: ID=0000 Rev=0 Len=028 <?>
	Capabilities: [240 v1] L1 PM Substates
		L1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2+ ASPM_L1.1+ L1_PM_Substates+
			  PortCommonModeRestoreTime=8us PortTPowerOnTime=10us
		L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1-
			   T_CommonMode=1us LTR1.2_Threshold=0ns
		L1SubCtl2: T_PwrOn=10us
	Capabilities: [300 v1] Secondary PCI Express
		LnkCtl3: LnkEquIntrruptEn- PerformEqu-
		LaneErrStat: 0
	Kernel driver in use: pcieport

0000:01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Tonga PRO [Radeon R9 285/380] (rev f1) (prog-if 00 [VGA controller])
	Subsystem: Tul Corporation / PowerColor Tonga PRO [Radeon R9 285/380]
	!!! Unknown header type 7f
	Region 0: Memory at 1800000000 (64-bit, prefetchable) [size=256M]
	Region 2: Memory at 1810000000 (64-bit, prefetchable) [size=2M]
	Region 5: Memory at 1b80000000 (32-bit, non-prefetchable) [size=256K]
	Expansion ROM at 1b80040000 [disabled] [size=128K]
	Kernel modules: amdgpu

0000:01:00.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Tonga HDMI Audio [Radeon R9 285/380]
	Subsystem: Tul Corporation / PowerColor Tonga HDMI Audio [Radeon R9 285/380]
	!!! Unknown header type 7f
	Interrupt: pin ? routed to IRQ 255
	Region 0: Memory at 1b80060000 (64-bit, non-prefetchable) [size=16K]

0001:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries BCM2712 PCIe Bridge (rev 30) (prog-if 00 [Normal decode])
	Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
	Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
	Latency: 0
	Interrupt: pin A routed to IRQ 39
	Bus: primary=00, secondary=01, subordinate=01, sec-latency=0
	Memory behind bridge: 00000000-005fffff [size=6M] [32-bit]
	Prefetchable memory behind bridge: 00000000fff00000-00000000000fffff [disabled] [64-bit]
	Secondary status: 66MHz- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- <SERR- <PERR-
	BridgeCtl: Parity- SERR- NoISA- VGA- VGA16- MAbort- >Reset- FastB2B-
		PriDiscTmr- SecDiscTmr- DiscTmrStat- DiscTmrSERREn-
	Capabilities: [48] Power Management version 3
		Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold-)
		Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=1 PME-
	Capabilities: [ac] Express (v2) Root Port (Slot-), MSI 00
		DevCap:	MaxPayload 512 bytes, PhantFunc 0
			ExtTag- RBE+
		DevCtl:	CorrErr- NonFatalErr- FatalErr- UnsupReq-
			RlxdOrd+ ExtTag- PhantFunc- AuxPwr+ NoSnoop+
			MaxPayload 256 bytes, MaxReadReq 512 bytes
		DevSta:	CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr- TransPend-
		LnkCap:	Port #0, Speed 5GT/s, Width x4, ASPM L1, Exit Latency L1 <2us
			ClockPM+ Surprise- LLActRep- BwNot+ ASPMOptComp+
		LnkCtl:	ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
			ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
		LnkSta:	Speed 5GT/s, Width x4
			TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt+
		RootCap: CRSVisible+
		RootCtl: ErrCorrectable- ErrNon-Fatal- ErrFatal- PMEIntEna+ CRSVisible+
		RootSta: PME ReqID 0000, PMEStatus- PMEPending-
		DevCap2: Completion Timeout: Range ABCD, TimeoutDis+ NROPrPrP- LTR+
			 10BitTagComp- 10BitTagReq- OBFF Via WAKE#, ExtFmt- EETLPPrefix-
			 EmergencyPowerReduction Not Supported, EmergencyPowerReductionInit-
			 FRS- LN System CLS Not Supported, TPHComp- ExtTPHComp- ARIFwd+
			 AtomicOpsCap: Routing- 32bit- 64bit- 128bitCAS-
		DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis- LTR- 10BitTagReq- OBFF Disabled, ARIFwd-
			 AtomicOpsCtl: ReqEn- EgressBlck-
		LnkCap2: Supported Link Speeds: 2.5-5GT/s, Crosslink- Retimer- 2Retimers- DRS+
		LnkCtl2: Target Link Speed: 5GT/s, EnterCompliance- SpeedDis-
			 Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-
			 Compliance Preset/De-emphasis: -6dB de-emphasis, 0dB preshoot
		LnkSta2: Current De-emphasis Level: -6dB, EqualizationComplete- EqualizationPhase1-
			 EqualizationPhase2- EqualizationPhase3- LinkEqualizationRequest-
			 Retimer- 2Retimers- CrosslinkRes: unsupported, DRS-
			 DownstreamComp: Link Up - Present
	Capabilities: [100 v1] Advanced Error Reporting
		UESta:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UEMsk:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UESvrt:	DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt- UnxCmplt- RxOF+ MalfTLP+ ECRC- UnsupReq- ACSViol-
		CESta:	RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr-
		CEMsk:	RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr+
		AERCap:	First Error Pointer: 00, ECRCGenCap+ ECRCGenEn- ECRCChkCap+ ECRCChkEn-
			MultHdrRecCap- MultHdrRecEn- TLPPfxPres- HdrLogCap-
		HeaderLog: 00000000 00000000 00000000 00000000
		RootCmd: CERptEn+ NFERptEn+ FERptEn+
		RootSta: CERcvd- MultCERcvd- UERcvd- MultUERcvd-
			 FirstFatal- NonFatalMsg- FatalMsg- IntMsg 0
		ErrorSrc: ERR_COR: 0000 ERR_FATAL/NONFATAL: 0000
	Capabilities: [160 v1] Virtual Channel
		Caps:	LPEVC=0 RefClk=100ns PATEntryBits=1
		Arb:	Fixed- WRR32- WRR64- WRR128-
		Ctrl:	ArbSelect=Fixed
		Status:	InProgress-
		VC0:	Caps:	PATOffset=00 MaxTimeSlots=1 RejSnoopTrans-
			Arb:	Fixed- WRR32- WRR64- WRR128- TWRR128- WRR256-
			Ctrl:	Enable+ ID=0 ArbSelect=Fixed TC/VC=ff
			Status:	NegoPending- InProgress-
	Capabilities: [180 v1] Vendor Specific Information: ID=0000 Rev=0 Len=028 <?>
	Capabilities: [240 v1] L1 PM Substates
		L1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2+ ASPM_L1.1+ L1_PM_Substates+
			  PortCommonModeRestoreTime=8us PortTPowerOnTime=10us
		L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1-
			   T_CommonMode=1us LTR1.2_Threshold=0ns
		L1SubCtl2: T_PwrOn=10us
	Capabilities: [300 v1] Secondary PCI Express
		LnkCtl3: LnkEquIntrruptEn- PerformEqu-
		LaneErrStat: 0
	Kernel driver in use: pcieport

0001:01:00.0 Ethernet controller: Raspberry Pi Ltd RP1 PCIe 2.0 South Bridge
	Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+
	Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
	Latency: 0
	Interrupt: pin A routed to IRQ 39
	Region 0: Memory at 1f00410000 (32-bit, non-prefetchable) [size=16K]
	Region 1: Memory at 1f00000000 (32-bit, non-prefetchable) [virtual] [size=4M]
	Region 2: Memory at 1f00400000 (32-bit, non-prefetchable) [size=64K]
	Capabilities: [40] Power Management version 3
		Flags: PMEClk- DSI- D1+ D2- AuxCurrent=375mA PME(D0+,D1+,D2-,D3hot+,D3cold+)
		Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-
	Capabilities: [70] Express (v2) Endpoint, MSI 00
		DevCap:	MaxPayload 256 bytes, PhantFunc 0, Latency L0s unlimited, L1 unlimited
			ExtTag- AttnBtn- AttnInd- PwrInd- RBE+ FLReset- SlotPowerLimit 0W
		DevCtl:	CorrErr+ NonFatalErr+ FatalErr+ UnsupReq+
			RlxdOrd+ ExtTag- PhantFunc- AuxPwr- NoSnoop-
			MaxPayload 256 bytes, MaxReadReq 512 bytes
		DevSta:	CorrErr+ NonFatalErr- FatalErr- UnsupReq- AuxPwr+ TransPend-
		LnkCap:	Port #0, Speed 5GT/s, Width x4, ASPM L0s L1, Exit Latency L0s <1us, L1 <2us
			ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp+
		LnkCtl:	ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
			ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
		LnkSta:	Speed 5GT/s, Width x4
			TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
		DevCap2: Completion Timeout: Not Supported, TimeoutDis+ NROPrPrP- LTR-
			 10BitTagComp- 10BitTagReq- OBFF Not Supported, ExtFmt- EETLPPrefix-
			 EmergencyPowerReduction Not Supported, EmergencyPowerReductionInit-
			 FRS- TPHComp- ExtTPHComp-
			 AtomicOpsCap: 32bit- 64bit- 128bitCAS-
		DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis- LTR- 10BitTagReq- OBFF Disabled,
			 AtomicOpsCtl: ReqEn-
		LnkCap2: Supported Link Speeds: 2.5-5GT/s, Crosslink- Retimer- 2Retimers- DRS-
		LnkCtl2: Target Link Speed: 5GT/s, EnterCompliance- SpeedDis-
			 Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-
			 Compliance Preset/De-emphasis: -6dB de-emphasis, 0dB preshoot
		LnkSta2: Current De-emphasis Level: -6dB, EqualizationComplete- EqualizationPhase1-
			 EqualizationPhase2- EqualizationPhase3- LinkEqualizationRequest-
			 Retimer- 2Retimers- CrosslinkRes: unsupported
	Capabilities: [b0] MSI-X: Enable+ Count=61 Masked-
		Vector table: BAR=0 offset=00000000
		PBA: BAR=0 offset=00002000
	Capabilities: [100 v2] Advanced Error Reporting
		UESta:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UEMsk:	DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
		UESvrt:	DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt- UnxCmplt- RxOF+ MalfTLP+ ECRC- UnsupReq- ACSViol-
		CESta:	RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr-
		CEMsk:	RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr+
		AERCap:	First Error Pointer: 00, ECRCGenCap+ ECRCGenEn- ECRCChkCap+ ECRCChkEn-
			MultHdrRecCap- MultHdrRecEn- TLPPfxPres- HdrLogCap-
		HeaderLog: 00000000 00000000 00000000 00000000
	Kernel driver in use: rp1

Using git clone --depth=1 clone --single-branch --branch rpi-6.6.y https://github.com/raspberrypi/linux gave no display at all, no dmesg errors, Fix up misaligned loads and stores from userspace for 64bit code and Force Architecture can write-combine memory were missing as options in menuconfig. Using the official raspberry pi kernel, I also manually applied CoreForge's patch.

wget -O amdgpu-pi5.patch https://github.com/geerlingguy/linux/pull/8.patch
git apply -v amdgpu-pi5.patch 

I also tried with the most recent commit as per the raspberry pi documentation. This acted the same as the 6.6 kernel, but the patch also failed to apply.

[bagong]

As someone who went through basically the same steps 2 days ago (for rpi5 & RX6800XT) with partial success (no audio), and failure (RX6700XT) may I generallize by asking two questions:

• is it necessary to apply any patches (memcopy and/or gerlingguy) if one straightforwardly used coreforges gpu branches?
• should one keep using the 6.6.y-gpu branch for certain models, and the 6.12.y-gpu branch for others?

This is a really cool project, however the info bits are a bit scattered over various texts and videos, and one isn't always sure they represent the current state of development. Thanks for your considerations!

[420noscope-exe]

I agree, the info is a bit scattered across the net.

1. As far as I know the memcopy patch is still required, even if you use the coreforge fork.

2. I did not test 6.12.y-gpu yet, but I did get the 6.6.y-gpu working with another gpu, with the same steps as above.

Side note: I swapped out the 380x for a RX 6400. It worked immediately, display and full acceleration in chromium.

Looks like some work needs to be done for 300 series cards.

[geerlingguy]

Yeah, I'm not sure anyones tried any 300 series AMD cards yet, you may be the first.

[Coreforge]

6.12.y-gpu is a little bit more up to date than 6.6.y-gpu. Unless there are some issues with 6.12, I don't really see a reason to keep 6.6.

The memcpy library isn't needed anymore on the pi 5, it should now all be handled in the kernel instead. There is a chance it gives a little bit better performance, but I don't know if it really makes a difference in practice. If there are any things that only run with the memcpy library but not without, please tell me (and check dmesg), as that might mean some op codes aren't implemented yet, or aren't implemented correctly.
Note that this doesn't apply to the pi 4, but that doesn't work with as many GPUs anyways.

It's interesting that the driver errors out and it's not an alignment fault like usually. That probably will make finding the issue a bit harder.

[420noscope-exe]

Thank you @geerlingguy and @Coreforge for the info! I ran a couple more tests and got some nice results!

I actually wasn't aware that there was a 6.12.y-gpu branch! I did find it however, and then I cloned it using git clone --depth=1 --branch rpi-6.12.y-gpu https://github.com/Coreforge/linux.git. I used the above intructions, skipping the memcpy steps. Once I rebooted, the R9 380X displayed an image!

It was a little unstable. The system locked up a few times, but I haven't ruled out connection issues with the risers that I am using. The system would be usable over SSH, but the image on the screen was frozen, and would not refresh with sudo reboot. I would have to disconnect and reconnect power via the ATX watchdog HAT that I have. One thing that I noticed was that 3D performance would drop dramatically whenever I started a vnc session. I would use this WebGL game I made as a test. The FPS would drop from a stable, locked 60, to like 12-18. Nvtop showed that the gpu was being used, however.

I decided to re-image the SD card and try again:

sudo apt install git bc bison flex libssl-dev make raspberrypi-kernel-headers libncurses-dev

git clone --depth=1 --branch rpi-6.12.y-gpu https://github.com/Coreforge/linux.git

cd linux

KERNEL=kernel_2712

make bcm2712_defconfig

make menuconfig

Kernel Features > Kernel support for 32-bit EL0 > Fix up misaligned multi-word loads and stores in user space
Kernel Features > Fix up misaligned loads and stores from userspace for 64bit code
Device Drivers > Graphics support > AMD GPU , as well as SI/CIK support
Device Drivers > Graphics support > AMD GPU>Display Engine Configuration>AMD DC Support for Southern Islands ASICs
Device Drivers > Graphics support > Direct Rendering Manager (XFree86 4.1.0 and higher DRI support) > Force Architecture can write-combine memory

saved as .config

make -j6 Image.gz modules dtbs
sudo make -j6 modules_install

sudo cp /boot/firmware/$KERNEL.img /boot/firmware/$KERNEL-backup.img
sudo cp arch/arm64/boot/Image.gz /boot/firmware/$KERNEL.img
sudo cp arch/arm64/boot/dts/broadcom/*.dtb /boot/firmware/
sudo cp arch/arm64/boot/dts/overlays/*.dtb* /boot/firmware/overlays/
sudo cp arch/arm64/boot/dts/overlays/README /boot/firmware/overlays/

sudo apt install firmware-amd-graphics

sudo reboot

I haven't had it running like this for very long, but let me tell you that it runs so much better! So far, the FPS in that specific WebGL game only drops to around 52FPS now! vkcube runs a lot smoother as well!

demsg | grep amdgpu now gives me:

[    2.538974] [drm] amdgpu kernel modesetting enabled.
[    2.539223] amdgpu 0001:01:00.0: enabling device (0000 -> 0002)
[    2.788671] amdgpu 0001:01:00.0: amdgpu: Fetched VBIOS from ROM BAR
[    2.788679] amdgpu: ATOM BIOS: 113-C7660300-005
[    2.790669] amdgpu 0001:01:00.0: amdgpu: Trusted Memory Zone (TMZ) feature not supported
[    2.790673] amdgpu 0001:01:00.0: amdgpu: PCIE atomic ops is not supported
[    2.797327] amdgpu 0001:01:00.0: BAR 2 [mem 0x1810000000-0x18101fffff 64bit pref]: releasing
[    2.797334] amdgpu 0001:01:00.0: BAR 0 [mem 0x1800000000-0x180fffffff 64bit pref]: releasing
[    2.797371] amdgpu 0001:01:00.0: BAR 0 [mem 0x1800000000-0x18ffffffff 64bit pref]: assigned
[    2.797379] amdgpu 0001:01:00.0: BAR 2 [mem 0x1900000000-0x19001fffff 64bit pref]: assigned
[    2.797396] amdgpu 0001:01:00.0: amdgpu: VRAM: 4096M 0x000000F400000000 - 0x000000F4FFFFFFFF (4096M used)
[    2.797399] amdgpu 0001:01:00.0: amdgpu: GART: 1024M 0x000000FF00000000 - 0x000000FF3FFFFFFF
[    2.797568] [drm] amdgpu: 4096M of VRAM memory ready
[    2.797572] [drm] amdgpu: 8109M of GTT memory ready.
[    2.814087] amdgpu: hwmgr_sw_init smu backed is tonga_smu
[    2.893676] amdgpu 0001:01:00.0: [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!
[    2.893711] amdgpu 0001:01:00.0: [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!
[    2.893733] amdgpu 0001:01:00.0: [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!
[    3.183258] amdgpu 0001:01:00.0: amdgpu: SE 4, SH per SE 1, CU per SH 8, active_cu_number 28
[    3.714121] amdgpu 0001:01:00.0: amdgpu: Using BACO for runtime pm
[    3.714430] [drm] Initialized amdgpu 3.61.0 for 0001:01:00.0 on minor 1
[    3.768557] amdgpu 0001:01:00.0: [drm] fb0: amdgpudrmfb frame buffer device

I still need to go back and test my CM4.

[Coreforge]

Nice to see that it works!
amdgpu_top should show you more detailed usage information of different parts of the GPU. My guess is that the framerate drops so much with VNC because encoding is probably done on the CPU, so the framebuffer needs to be transferred to the CPU. I've seen the SDMA block being the one with the highest usage a decent bit (and depending on how the copy is done, it might also run into the alignment trap a lot, which would be a big performance hit), and the load on the PCIe link from frequent framebuffer copies isn't too low either.
If it's still unstable, you can also try disabling Force architecture can write-combine memory again. I haven't run into issues with it on the pi5, but that doesn't mean there can't be any.

It'd be interesting to see what happens on the CM4, I never got amdgpu to work there, only radeon. Since this card is supported by both, it might be different, but it also might not be. Debugging on the cm4 is much more annoying though, as it tends to just lock up and not even have the CPU properly reachable over JTAG anymore.