Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
Mesa Clover OpenCL
View unanswered posts
View posts from last 24 hours

 
Reply to topic    Gentoo Forums Forum Index Other Things Gentoo
View previous topic :: View next topic  
Author Message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Tue Sep 13, 2016 10:12 pm    Post subject: Mesa Clover OpenCL Reply with quote

I'm trying to set up a OpenCL capable system. Well at least OpenCL 1.1 capable. ;)
This is just my own interest on testing this on open source AMD drivers.

I'm testing the OpenCL using bfgminer. bfgminer only acts as a test program to show me if I really have gotten things right. Unfortunately not.
bfgminer does not find my GPU, which is R9 Nano.

I've applied opencl USE-flags globally. Then rebuild @world with --newuse. I then used eselect opencl to set mesa as my OpenCL implementation.

I don't know what else to tell... I have no error logs or messages to paste. There just isn't any OpenCL platforms/devices on my system according to bfgminer and even to blender.
Any help?
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Mon Oct 10, 2016 11:07 am    Post subject: Reply with quote

Just a little update here.

I managed to get bfgminer to use OpenCL platform. I'm not sure how, but I at least got virtual/opencl from gentoo-gpu overlay.
However the mining speed is slow. The cause of that might reside in my gpu driver settings or in gpu bios.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Sun Oct 30, 2016 8:40 pm    Post subject: Reply with quote

It seems that enabling opencl and using Mesa Clover may cause Blender to Freeze. Even if Mesa Clover isn't selectable in Blender it's still somehow messing it up.
Also while bfgminer can utilize my GPU, Boinc is still unable to do so. I get tasks assigned to my GPU. The progress shows "100%" all the time and the GPU utilization is practically zero (looking at radeontop). This, however, should be possible with Mesa Clover. At least people have managed to get it working propely. I tried Einstein@home project which is the same as with one individual who got it working.

So. More investigation is needed.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Sun May 14, 2017 9:05 am    Post subject: Reply with quote

If someone is following this topic still... (based on views, it seems that people are actually reading this)

So. Clover seems like a dead project. But there's still ways to get OpenCL working on AMDGPU. One way is using OpenCL libraries from AMDGPU-PRO.

I've started to collect information and will eventually write an unoffical topic if I get things working. Spoiler: ROCm has been open-sourced. ;)
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
Ant P.
Watchman
Watchman


Joined: 18 Apr 2009
Posts: 6873

PostPosted: Sun May 14, 2017 4:45 pm    Post subject: Reply with quote

AFAIK clover was merged into mesa and lives there now. OpenCL "works", but is nearly useless because it still only supports the CL1.1 spec while most things are written for 1.2 or higher (or vulkan).
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Sun May 14, 2017 6:11 pm    Post subject: Reply with quote

And even CL1.1 is marked as WIP.

I'm going to try out pulling the libraries required for OpenCL from AMDGPU-PRO. Some have reported that to be a working solution. It is in no way offical way from any perspective, but worth trying out.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
szatox
Veteran
Veteran


Joined: 27 Aug 2013
Posts: 1988

PostPosted: Mon May 15, 2017 8:25 pm    Post subject: Reply with quote

I once had a working bitcoin miner running on radeon.
AFAIR all it took was enabling opencl flag and then pointing the miner to the correct cl kernel. It came with like 10 kernels and was unable to autodetect the correct one, so I had to specify it manually, and this part came unexpected which also made it confusing.

So... I am totally sure opencl used to work, though I gave up on mining after like a week. GPU can't compete with ASICs. My electricity bill would top any possible gain. (It was still 10 times faster than all CPU cores working together)
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Thu Jul 06, 2017 11:55 am    Post subject: Reply with quote

Looks like dev-libs/amdgpu-pro-opencl has appeared on portage (default gentoo overlay).
Darktable seems to utilize the OpenCL cababilities, although I needed to tell where to find the library.
Looks very promising, as I can still keep open source AMDGPU drivers, but use only OpenCL libraries from amdgpu-pro. ;)

EDIT: There's no download link on AMD's website for the specific driver package which amdgpu-pro-opencl (ebuild) needs (although I guess it only need to be one for Ubuntu 17.10). However I forced the exact download using this command:
Code:
curl -A '' -e 'https://support.amd.com/en-us/kb-articles/Pages/AMDGPU-PRO-Driver-for-Linux-Release-Notes.aspx' https://www2.ati.com/drivers/linux/ubuntu/amdgpu-pro-17.10-429170.tar.xz > amdgpu-pro-17.10-429170.tar.xz
... be sure to be in your DISTDIR, or move the archive there after the download has finished.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...


Last edited by Zucca on Sat Jul 08, 2017 4:20 pm; edited 2 times in total
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Thu Jul 06, 2017 6:23 pm    Post subject: Reply with quote

I tried to benchmark performance using ethminer. All I got was frozen X. amdgpu apparently died (horribly).
grep 'kernel: amdgpu':
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: VRAM: 4096M 0x0000000000000000 - 0x00000000FFFFFFFF (4096M used)
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: GTT: 4096M 0x0000000100000000 - 0x00000001FFFFFFFF
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: amdgpu: using MSI.
Jul 06 20:38:27 wren kernel: amdgpu: [powerplay] amdgpu: powerplay sw initialized
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000100000008, cpu addr 0xffff88080f23b008
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 1 use gpu addr 0x0000000100000018, cpu addr 0xffff88080f23b018
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 2 use gpu addr 0x0000000100000028, cpu addr 0xffff88080f23b028
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 3 use gpu addr 0x0000000100000038, cpu addr 0xffff88080f23b038
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 4 use gpu addr 0x0000000100000048, cpu addr 0xffff88080f23b048
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 5 use gpu addr 0x0000000100000058, cpu addr 0xffff88080f23b058
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 6 use gpu addr 0x0000000100000068, cpu addr 0xffff88080f23b068
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 7 use gpu addr 0x0000000100000078, cpu addr 0xffff88080f23b078
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 8 use gpu addr 0x0000000100000088, cpu addr 0xffff88080f23b088
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 9 use gpu addr 0x0000000100000098, cpu addr 0xffff88080f23b098
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 10 use gpu addr 0x00000001000000a8, cpu addr 0xffff88080f23b0a8
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 11 use gpu addr 0x00000001000000b8, cpu addr 0xffff88080f23b0b8
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 12 use gpu addr 0x000000000094ae90, cpu addr 0xffffc9000443fe90
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 13 use gpu addr 0x00000001000000d8, cpu addr 0xffff88080f23b0d8
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 14 use gpu addr 0x00000001000000e8, cpu addr 0xffff88080f23b0e8
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fence driver on ring 15 use gpu addr 0x00000001000000f8, cpu addr 0xffff88080f23b0f8
Jul 06 20:38:27 wren kernel: amdgpu: [powerplay] Can't find requested voltage id in vdd_dep_on_sclk table!
Jul 06 20:38:27 wren kernel: amdgpu: [powerplay] Failed to setup PCC HW register! Wrong GPIO assigned for VDDC_PCC_GPIO_PINID!
Jul 06 20:38:27 wren kernel: amdgpu 0000:01:00.0: fb0: amdgpudrmfb frame buffer device
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x08410402
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00E85508
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08104002
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x02, vmid 4) at page 15226120, read from '' (0x00000000) (260)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x0e810401
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x0526840D
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08188001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 86410253, read from 'TC4' (0x54433400) (392)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x0ce84401
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x03F0FFE1
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08084001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 66125793, read from '' (0x00000000) (132)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x08888802
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x06A4011A
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x080C8001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 111411482, read from '' (0x00000000) (200)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x0cd88801
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x06EFCCBF
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08084001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 116378815, read from '' (0x00000000) (132)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x0e818401
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x029343D0
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08184001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 43205584, read from 'TC5' (0x54433500) (388)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x03a9c802
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x02EB53E1
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08044001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 48976865, read from 'TC1' (0x54433100) (68)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x00308401
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x023268CB
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08084001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 36858059, read from '' (0x00000000) (132)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x0e308802
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x06F16ED8
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08148001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 116485848, read from '' (0x00000000) (328)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: GPU fault detected: 147 0x04984801
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x020B6C5B
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08004001
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: VM fault (0x01, vmid 4) at page 34303067, read from 'TC3' (0x54433300) (4)
Jul 06 20:54:34 wren kernel: amdgpu 0000:01:00.0: IH ring buffer overflow (0x00080630, 0x00000B50, 0x00000640)
Well. This combination of AMDGPU and propietary OpenCL libs isn't supported in any way. Backups are recommended. :)
I needed to reboot my PC with help of Magic SysRq. Managed to avoid data corruption.

At least darktable works with this setup, which is nice.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
skunk
l33t
l33t


Joined: 28 May 2003
Posts: 646
Location: granada, spain

PostPosted: Mon Feb 12, 2018 6:30 pm    Post subject: Reply with quote

hey!

did you manage to run any miner at a reasonable speed?
i've tried both dev-libs/amdgpu-pro-opencl and x11-drivers/amdgpu-pro (following this howto), however with the former no opencl devices are found, while with the latter the miner complains it cannot initialize...


thank you
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Thu Mar 01, 2018 7:50 pm    Post subject: Reply with quote

skunk wrote:
did you manage to run any miner at a reasonable speed?
I think you already got the answer on a another topic...
So if you still have problems use the amdgpu-pro-opencl.
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
skunk
l33t
l33t


Joined: 28 May 2003
Posts: 646
Location: granada, spain

PostPosted: Thu Mar 01, 2018 8:23 pm    Post subject: Reply with quote

Zucca wrote:
skunk wrote:
did you manage to run any miner at a reasonable speed?
I think you already got the answer on a another topic...
So if you still have problems use the amdgpu-pro-opencl.

yes, that's what i'm currently using, however the cards (rx480x2+rx580x2) hangs on an average of 40 times a day, so i had to implement a workaround that automatically reboots the rig and restarts the miner when it happens...
that's my only amd rig, other 6 nvidia based rigs runs fine without issues if there are no blackouts.
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Thu Mar 01, 2018 9:15 pm    Post subject: Reply with quote

skunk wrote:
however the cards (rx480x2+rx580x2) hangs on an average of 40 times a day
Sounds like a hardware problem. Maybe too much power draw, heat or interference...
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
skunk
l33t
l33t


Joined: 28 May 2003
Posts: 646
Location: granada, spain

PostPosted: Thu Mar 01, 2018 9:26 pm    Post subject: Reply with quote

Zucca wrote:
skunk wrote:
however the cards (rx480x2+rx580x2) hangs on an average of 40 times a day
Sounds like a hardware problem. Maybe too much power draw, heat or interference...

i don't think so as it happens on any card and uptime goes from few minutes to over 4 hours without a clear pattern...
I've even tried with other miners and lowering mining intensity to the minimum without success.
with Ubuntu it was even worst because the rig was hanging completely and needed to be rebooted manually...
Back to top
View user's profile Send private message
s_j_newbury
n00b
n00b


Joined: 05 Apr 2007
Posts: 55
Location: UK

PostPosted: Fri Oct 19, 2018 9:10 pm    Post subject: Reply with quote

I'm the maintainer of the gentoo-gpu overlay, I've been neglecting it quite badly but I have quite a few local changes to push, which I will do shortly if testing is successful.

I've recently acquired an RX470 so I can finally test Clover on actual hardware and fix a few issues. I've also taken to maintaining the old nvidia-drivers until such a time that Nouveau/Clover supports Fermi hardware (as on my laptop).

Right now I can confirm as working:

Mesa-git/LLVM-git/AMD: Works at least on POLARIS10, actually faster than amdgpu-pro for pyrit!

Beignet is working for me with IvyBridge again after a little hacking....

NVidia CUDA/OpenCL works at least for me.

I'm considering completely dropping CUDA support since it is dependent on really old gcc releases and maintaining just the GL and CL components.


Edited to add: All the above is working with ocl-icd so can operate simultaneously where supported.
Back to top
View user's profile Send private message
Zucca
Veteran
Veteran


Joined: 14 Jun 2007
Posts: 1866
Location: KUUSANKOSKI, Finland

PostPosted: Mon Oct 22, 2018 3:53 pm    Post subject: Reply with quote

I wonder if mesa clover will user ROCm (or Vulkan*) under the hood eventually...

* Some time ago I read that Vulkan could/would evetually replace OpenCL. So combining OpenGL and OpenCL basically...
Is there any truth to this?
_________________
..: Zucca :..

Code:
ERROR: '--failure' is not an option. Aborting...
Back to top
View user's profile Send private message
Marlo
Veteran
Veteran


Joined: 26 Jul 2003
Posts: 1492

PostPosted: Mon Oct 22, 2018 5:32 pm    Post subject: Reply with quote

s_j_newbury wrote:
I ... so I can finally test Clover on actual hardware...

Did you mean something like that?
Code:

...
  Platform Name                                   Clover
Number of devices                                 1
  Device Name                                     Radeon RX 560 Series (POLARIS11, DRM 3.26.0, 4.18.16-gentoo, LLVM 7.0.0)
  Device Vendor                                   AMD
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.1 Mesa 18.2.3
  Driver Version                                  18.2.3
  Device OpenCL C Version                         OpenCL C 1.1
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE
...

_________________
Thank you for your attention, interest and support.
------------------------------------------------------------------
http://radio.garden/
Back to top
View user's profile Send private message
s_j_newbury
n00b
n00b


Joined: 05 Apr 2007
Posts: 55
Location: UK

PostPosted: Tue Oct 23, 2018 5:50 pm    Post subject: Reply with quote

Marlo wrote:
s_j_newbury wrote:
I ... so I can finally test Clover on actual hardware...

Did you mean something like that?
Code:

...
  Platform Name                                   Clover
Number of devices                                 1
  Device Name                                     Radeon RX 560 Series (POLARIS11, DRM 3.26.0, 4.18.16-gentoo, LLVM 7.0.0)
  Device Vendor                                   AMD
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.1 Mesa 18.2.3
  Driver Version                                  18.2.3
  Device OpenCL C Version                         OpenCL C 1.1
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE
...


Yes, exactly. But it emerges correctly both with and without ocl-icd and along side both nvidia and Intel (Beignet) both also +-ocl-icd.

Code:
  Platform Name                                   Clover
Number of devices                                 1
  Device Name                                     AMD Radeon (TM) RX 470 Graphics (POLARIS10, DRM 3.26.0, 4.18.13-gentoo, LLVM 8.0.0)
  Device Vendor                                   AMD
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.1 Mesa 18.3.0-devel (git-bc021be78d)
  Driver Version                                  18.3.0-devel
  Device OpenCL C Version                         OpenCL C 1.1
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE

Back to top
View user's profile Send private message
s_j_newbury
n00b
n00b


Joined: 05 Apr 2007
Posts: 55
Location: UK

PostPosted: Tue Oct 23, 2018 9:16 pm    Post subject: Reply with quote

I was a little surprised Clover from git Mesa/LLVM is faster than amdgpu-pro at the "pyrit benchmark". It's a shame it's only OpenCL 1.1.
Back to top
View user's profile Send private message
Marlo
Veteran
Veteran


Joined: 26 Jul 2003
Posts: 1492

PostPosted: Tue Oct 23, 2018 9:52 pm    Post subject: Reply with quote

I had seen your gentoo-gpu overlay. Unfortunately it is not for AMD at the moment.
So I installed "ROCm on Gentoo" and "radeon-profile".
How did you do the "pyrite benchmark"?
Code:

 Platform Name                                   AMD Accelerated Parallel Processing
 Platform Vendor                                 Advanced Micro Devices, Inc.
 Platform Version                                OpenCL 2.1 AMD-APP (2639.3)


Code:

 # eselect opencl list
Available OpenCL implementations:
  [1]   ocl-icd
  [2]   roc *

_________________
Thank you for your attention, interest and support.
------------------------------------------------------------------
http://radio.garden/
Back to top
View user's profile Send private message
Marlo
Veteran
Veteran


Joined: 26 Jul 2003
Posts: 1492

PostPosted: Tue Oct 23, 2018 10:42 pm    Post subject: Reply with quote

Code:

# cat .pyrit/config
default_storage = file://
limit_ncpus = 0
rpc_announce = true
rpc_announce_broadcast = false
rpc_knownclients =
rpc_server = false
use_CUDA = false
use_OpenCL = true
workunit_size = 75000


use_OpenCL = true. Unfortunately does not work.
_________________
Thank you for your attention, interest and support.
------------------------------------------------------------------
http://radio.garden/
Back to top
View user's profile Send private message
s_j_newbury
n00b
n00b


Joined: 05 Apr 2007
Posts: 55
Location: UK

PostPosted: Sun Oct 28, 2018 10:45 pm    Post subject: Reply with quote

Marlo wrote:
Code:

# cat .pyrit/config
default_storage = file://
limit_ncpus = 0
rpc_announce = true
rpc_announce_broadcast = false
rpc_knownclients =
rpc_server = false
use_CUDA = false
use_OpenCL = true
workunit_size = 75000


use_OpenCL = true. Unfortunately does not work.


You need to emerge net-wireless/cpyrit-opencl for OpenCL support.

You will also need a patch to make it work with Cover:
Code:
--- a/_cpyrit_opencl.c
+++ b/_cpyrit_opencl.c
@@ -219,7 +219,7 @@ ocldevice_init(OpenCLDevice *self, PyObject *args, PyObject *kwds)
     cl_uint num_entries;
     cl_device_id *devices;
     cl_device_type device_type;
-    char dev_name[64];
+    char dev_name[128];
     size_t name_size, i;
     size_t *max_work_sizes;
     cl_int err;


I then get:

RX470/AMD FX-9370


Clover

Code:
Pyrit 0.5.0 (C) 2008-2011 Lukas Lueg - 2015 John Mora
https://github.com/JPaulMora/Pyrit
This code is distributed under the GNU General Public License v3+

Running benchmark (122897.1 PMKs/s)... |

Computed 122897.11 PMKs/s total.
#1: 'CPU-Core (SSE2/AES)': 1120.8 PMKs/s (RTT 2.9)
#2: 'CPU-Core (SSE2/AES)': 1120.5 PMKs/s (RTT 2.9)
OpenCL:
#1: 'OpenCL-Device 'AMD Radeon (TM) RX 470 Graphics (POLARIS10, DRM 3.26.0, 4.18.13-gentoo, LLVM 8.0.0)'': 134440.8 PMKs/s (RTT 1.3)


AMDGPU-PRO

Code:
Pyrit 0.5.0 (C) 2008-2011 Lukas Lueg - 2015 John Mora
https://github.com/JPaulMora/Pyrit
This code is distributed under the GNU General Public License v3+

Running benchmark (122289.2 PMKs/s)... -

Computed 122289.23 PMKs/s total.
#1: 'CPU-Core (SSE2/AES)': 1127.8 PMKs/s (RTT 2.9)
#2: 'CPU-Core (SSE2/AES)': 1128.5 PMKs/s (RTT 2.9)
OpenCL:
#1: 'OpenCL-Device 'Ellesmere'': 130793.7 PMKs/s (RTT 1.3)


I can't test ROCm since I don't have a supported CPU. :-(
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Other Things Gentoo All times are GMT
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum