umr won't work. (Can't find pci.did file in database)

[StDenis, Tom]

[AMD Official Use Only]

Hi Mikhail,

I've added the amd-gfx list since a wider audience might shed light on this.  I successfully run umr on my navi10 (with gfxoff disabled) fairly frequently.  I don't have other gfx10 ASICs in my office though.

So maybe others could shed light on this.  For completeness I've attached your dmesg and shell script to this email.

Tom

________________________________________
From: Mikhail Gavrilov <mikhail.v.gavrilov at gmail.com>
Sent: Monday, June 7, 2021 10:57
To: StDenis, Tom
Subject: Re: umr won't work. (Can't find pci.did file in database)

On Mon, 7 Jun 2021 at 18:41, StDenis, Tom <Tom.StDenis at amd.com> wrote:

> As for the hang that's probably gfxoff getting in the way.  Try adding "--gfxoff 0" before your -wa command (and you can put a --gfxoff 1 after to re-enable it).

Unfortunately it couldn't help and I can easily cause GPU hangs by
several umr launches.

I tried two variants:
1)
./umr -dbp ../../../database/ --gfxoff 0 -O halt_waves -wa gfx_0.0.0 --gfxoff 1
2)
./umr -dbp ../../../database/ --gfxoff 0
./umr -dbp ../../../database/ -O halt_waves -wa gfx_0.0.0 --gfxoff 1
./umr -dbp ../../../database/ --gfxoff 1

The second variant is part of the script which dumps gfx,
mmGRBM_STATUS, mmCP_EOP_,  mmCP_PFP_HEADER_DUMP, mmCP_ME_HEADER_DUMP
and kernel logs.
I attach my collect log script.
For reproduction it is enough to run this script several times.

--
Best Regards,
Mike Gavrilov.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: dmesg.txt
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20210607/38076eb6/attachment-0001.txt>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: collect-umr-logs.sh
Type: application/x-shellscript
Size: 1274 bytes
Desc: collect-umr-logs.sh
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20210607/38076eb6/attachment-0001.bin>