Hello,

referring to this list project Collatz Conjecture, there is Linux Intel GPU OpenCL support now.

On which processors can I use it? Is there finally linux gpu driver support?

Greetings
MPW

referring to this list project Collatz Conjecture, there is Linux Intel GPU OpenCL support now.

How do you make that connection? All I see is that there is Windows, Mac OS X, Linux, Android and Rasp Pi support, and separately Nvidia, AMD and Intel GPU. That doesn't mean that there's CUDA and OpenCL support for all operating systems.

But when Intel releases any Linux drivers that have OpenCL support, they'll be available through Intel's driver download site.

---

It's hidden in the details.

mpw, you could try Beignet.

I see what you mean, but cannot find any mention of it in the past 6 months on the Collatz forums. I would think something as such is a big deal. I also wonder if it isn't a hiccup of sorts. Collatz is researching OpenCL on the CPU. Could be that they mean that?

---

Wow that Beignet Implementation seems to be interessting.

Actually, I don't even get cpu tasks here from that Collatz project, seems to be weird. Before I put time into this, I'd like to see if this project does even work...

Does it work for you? Did someone try it up or does someone support it?

Did anyone check through the rest of the Collatz pages?

I see:

On the front page, "For GPU processing: Intel HD 2500 or later embedded GPU with Intel OpenCL driver"

On the applications page, plan class "(opencl_intel_gpu)" listed for three Linux platforms - 32-bit for Collatz Mini and Solo, and 64-bit for Mini only.

Unfortunately, their apps page doesn't have the enhancement (like SETI Beta) which enables us to see which application versions are completely unproductive - zero GigaFLOPS.

Finally, the server status page lets us see that tasks are available for all four task types, so - providing you have the right sub-projects enabled in your preferences - it should be possible to fetch work and find out whether the listed application types are in fact viable.

That project has some serious issues or is just out of work units. I detached and reattached it and my client got one single work unit. Now it's out of work again. (All tried with standard cpu tasks.)

But in general, I'm really looking forward to that beignet implementation. Maybe more projects will support openCL with Linux on GPUs.

Is there any difference in coding OpenCL for ATI, Nvidia or Intel?

There have been some recent (but unsuccessful) attempts to run beigenet at Einstein@Home, where the project administrator is game to help with the attempt.

BRP4 Intel GPU app feedback thread

Read from that post to the end of the thread before you start.

There is a year old thread about Linux+Intel GPU. opencl app on intel_gpu . Not too positive results but that's old version of Beignet as is in the Einstein thread.

mpw, if you decide to give Beignet a try I suggest you try the latest version, compiling it yourself if necessary. I can't try it myself because I don't have Intel hardware.

Or if you have too much money burning in your pockets you could buy the OpenCL drivers from Intel. As a bonus the drivers comes with some developer software.

I think Collatz just deployed their standard Linux OpenCL app for Linux for when Intel OpenCL GPU support becomes available.

Urs at Lunatics had another go at getting beignet running:

What a lucky morning ! Finally got BOINC to see my Intel GPU on openSUSE 13.1 .

Mi 01 Apr 2015 06:30:06 CEST | | Starting BOINC client version 7.0.65 for x86_64-pc-linux-gnu
Mi 01 Apr 2015 06:30:06 CEST | | log flags: file_xfer, sched_ops, task, sched_op_debug, work_fetch_debug
Mi 01 Apr 2015 06:30:06 CEST | | Libraries: libcurl/7.40.0 OpenSSL/1.0.1k zlib/1.2.8 libidn/1.25 libssh2/1.5.0
Mi 01 Apr 2015 06:30:06 CEST | | Data directory: /home/uje/BOINC
Mi 01 Apr 2015 06:30:06 CEST | | Processor: 4 GenuineIntel Intel(R) Core(TM) i7-3770T CPU @ 2.50GHz [Family 6 Model 58 Stepping 9]
Mi 01 Apr 2015 06:30:06 CEST | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm ida arat epb xsaveopt pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase smep erms
Mi 01 Apr 2015 06:30:06 CEST | | OS: Linux: 3.11.10-25-desktop
Mi 01 Apr 2015 06:30:06 CEST | | Memory: 15.58 GB physical, 16.00 GB virtual
Mi 01 Apr 2015 06:30:06 CEST | | Disk: 125.86 GB total, 18.04 GB free
Mi 01 Apr 2015 06:30:06 CEST | | Local time is UTC +2 hours
Mi 01 Apr 2015 06:30:06 CEST | | CAL: ATI GPU 0: AMD Radeon HD 7870/7950/7970 series (Tahiti) (CAL version 1.4.1848, 2048MB, 1974MB available, 7680 GFLOPS peak)
Mi 01 Apr 2015 06:30:06 CEST | | OpenCL: AMD/ATI GPU 0: AMD Radeon HD 7870/7950/7970 series (Tahiti) (driver version 1642.5 (VM), device version OpenCL 1.2 AMD-APP (1642.5), 2048MB, 1974MB available, 7680 GFLOPS peak)
Mi 01 Apr 2015 06:30:06 CEST | | OpenCL: Intel GPU 0: Intel(R) HD Graphics IvyBridge GT2 (driver version 1.1, device version OpenCL 1.2 beignet 1.1, 2048MB, 2048MB available, 128 GFLOPS peak)

But it wasn't all good news:

>:( No luck with Luxmark and intel_beignet, some basics like clGetDeviceIDs(-1) or clBuildProgramm(-3) fail :

[size=8pt]2015-04-02 18:02:38 - LuxMark v2.0
2015-04-02 18:02:38 - Based on SmallLuxGPU v2.0devel4 (LuxRays demo: http://www.luxrender.net)
2015-04-02 18:02:38 - [RenderConfig] Reading configuration file: scenes/sala/render.cfg
2015-04-02 18:02:38 - [RenderConfig] Configuration:
2015-04-02 18:02:38 - [RenderConfig] accelerator.type = -1
2015-04-02 18:02:38 - [RenderConfig] batch.haltspp = 0
2015-04-02 18:02:38 - [RenderConfig] batch.halttime = 0
2015-04-02 18:02:38 - [RenderConfig] batch.periodicsave = 0
2015-04-02 18:02:38 - [RenderConfig] film.filter.type = 0
2015-04-02 18:02:38 - [RenderConfig] film.gamma = 2.2
2015-04-02 18:02:38 - [RenderConfig] film.tonemap.linear.scale = 1.6
2015-04-02 18:02:38 - [RenderConfig] film.tonemap.type = 0
2015-04-02 18:02:38 - [RenderConfig] image.filename = sala.png
2015-04-02 18:02:38 - [RenderConfig] image.height = 600
2015-04-02 18:02:38 - [RenderConfig] image.width = 600
2015-04-02 18:02:38 - [RenderConfig] opencl.cpu.use = 0
2015-04-02 18:02:38 - [RenderConfig] opencl.gpu.use = 1
2015-04-02 18:02:38 - [RenderConfig] opencl.gpu.workgroup.size = 64
2015-04-02 18:02:38 - [RenderConfig] path.filter.B = 0.333333
2015-04-02 18:02:38 - [RenderConfig] path.filter.C = 0.333333
2015-04-02 18:02:38 - [RenderConfig] path.filter.alpha = 2.
2015-04-02 18:02:38 - [RenderConfig] path.filter.type = GAUSSIAN
2015-04-02 18:02:38 - [RenderConfig] path.filter.width.x = 1.5
2015-04-02 18:02:38 - [RenderConfig] path.filter.width.y = 1.5
2015-04-02 18:02:38 - [RenderConfig] path.maxdepth = 10
2015-04-02 18:02:38 - [RenderConfig] path.pixelatomics.enable = 0
2015-04-02 18:02:38 - [RenderConfig] path.russianroulette.cap = 0.25
2015-04-02 18:02:38 - [RenderConfig] path.russianroulette.depth = 10
2015-04-02 18:02:38 - [RenderConfig] path.russianroulette.strategy = 1
2015-04-02 18:02:38 - [RenderConfig] path.sampler.largesteprate = 0.4
2015-04-02 18:02:38 - [RenderConfig] path.sampler.maxconsecutivereject = 512.
2015-04-02 18:02:38 - [RenderConfig] path.sampler.type = METROPOLIS
2015-04-02 18:02:38 - [RenderConfig] renderengine.type = 4
2015-04-02 18:02:38 - [RenderConfig] scene.epsilon = 0.0001
2015-04-02 18:02:38 - [RenderConfig] scene.file = scenes/sala/sala.scn
2015-04-02 18:02:39 - [LuxRays] OpenCL Platform 0: Intel
2015-04-02 18:02:39 - [LuxRays] OpenCL Platform 1: Advanced Micro Devices, Inc.
2015-04-02 18:02:39 - [LuxRays] Device 0 NativeThread name: NativeThread-000
2015-04-02 18:02:39 - [LuxRays] Device 1 NativeThread name: NativeThread-001
2015-04-02 18:02:39 - [LuxRays] Device 2 NativeThread name: NativeThread-002
2015-04-02 18:02:39 - [LuxRays] Device 3 NativeThread name: NativeThread-003
2015-04-02 18:02:39 - [LuxRays] Device 4 NativeThread name: NativeThread-004
2015-04-02 18:02:39 - [LuxRays] Device 5 NativeThread name: NativeThread-005
2015-04-02 18:02:39 - [LuxRays] Device 6 NativeThread name: NativeThread-006
2015-04-02 18:02:39 - [LuxRays] Device 7 NativeThread name: NativeThread-007
2015-04-02 18:02:39 - [LuxRays] Device 8 OpenCL name: Intel(R) HD Graphics IvyBridge GT2
2015-04-02 18:02:39 - [LuxRays] Device 8 OpenCL type: GPU
2015-04-02 18:02:39 - [LuxRays] Device 8 OpenCL units: 16
2015-04-02 18:02:39 - [LuxRays] Device 8 OpenCL max allocable memory: 2048MBytes
2015-04-02 18:02:39 - [LuxRays] Device 9 OpenCL name: Tahiti
2015-04-02 18:02:39 - [LuxRays] Device 9 OpenCL type: GPU
2015-04-02 18:02:39 - [LuxRays] Device 9 OpenCL units: 24
2015-04-02 18:02:39 - [LuxRays] Device 9 OpenCL max allocable memory: 1877MBytes
2015-04-02 18:02:39 - [LuxRays] Device 10 OpenCL name: Intel(R) Core(TM) i7-3770T CPU @ 2.50GHz
2015-04-02 18:02:39 - [LuxRays] Device 10 OpenCL type: CPU
2015-04-02 18:02:39 - [LuxRays] Device 10 OpenCL units: 8
2015-04-02 18:02:39 - [LuxRays] Device 10 OpenCL max allocable memory: 15956MBytes
2015-04-02 18:02:39 - [LuxRays] Creating 2 intersection device(s)
2015-04-02 18:02:39 - [LuxRays] Allocating intersection device 0: Intel(R) HD Graphics IvyBridge GT2 (Type = OPENCL)
2015-04-02 18:02:39 - [LuxRays] Allocating intersection device 1: Tahiti (Type = OPENCL)
2015-04-02 18:02:39 - [RenderEngine] OpenCL Devices used:
2015-04-02 18:02:39 - [RenderEngine] [Intel(R) HD Graphics IvyBridge GT2Intersect]
2015-04-02 18:02:39 - [RenderEngine] [TahitiIntersect]
2015-04-02 18:02:39 - [LuxRays] Preprocessing DataSet
2015-04-02 18:02:39 - [LuxRays] Total vertex count: 616582
2015-04-02 18:02:39 - [LuxRays] Total triangle count: 488451
2015-04-02 18:02:39 - [LuxRays] Total vertices memory usage: 7225Kbytes
2015-04-02 18:02:39 - [LuxRays] Total triangles memory usage: 5724Kbytes
2015-04-02 18:02:39 - [LuxRays] Building QBVH, primitives: 488451, initial nodes: 40710
2015-04-02 18:02:39 - [LuxRays] QBVH completed with 80908/81420 nodes
2015-04-02 18:02:39 - [LuxRays] Total QBVH memory usage: 8849Kbytes
2015-04-02 18:02:39 - [LuxRays] Total QBVH QuadTriangle count: 161119
2015-04-02 18:02:39 - [LuxRays] Max. QBVH Depth: 29
2015-04-02 18:02:39 - [LuxRays] [OpenCL device::Intel(R) HD Graphics IvyBridge GT2Intersect] Ray buffer size: 2048Kbytes
2015-04-02 18:02:39 - [LuxRays] [OpenCL device::Intel(R) HD Graphics IvyBridge GT2Intersect] Ray hits buffer size: 1024Kbytes
2015-04-02 18:02:39 - [LuxRays] [OpenCL device::Intel(R) HD Graphics IvyBridge GT2Intersect] QBVH max. stack size: 24
2015-04-02 18:02:39 - [LuxRays] [OpenCL device::Intel(R) HD Graphics IvyBridge GT2Intersect] QBVH compilation error:
2015-04-02 18:02:39 - OpenCL ERROR: clBuildProgram(-3)
2015-04-02 18:03:12 - Set Benchmark Custom mode
2015-04-02 18:03:12 - [RenderConfig] Reading configuration file: scenes/sala/render.cfg
2015-04-02 18:03:12 - [RenderConfig] Configuration:
2015-04-02 18:03:12 - [RenderConfig] accelerator.type = -1
2015-04-02 18:03:12 - [RenderConfig] batch.haltspp = 0
2015-04-02 18:03:12 - [RenderConfig] batch.halttime = 0
2015-04-02 18:03:12 - [RenderConfig] batch.periodicsave = 0
2015-04-02 18:03:12 - [RenderConfig] film.filter.type = 0
2015-04-02 18:03:12 - [RenderConfig] film.gamma = 2.2
2015-04-02 18:03:12 - [RenderConfig] film.tonemap.linear.scale = 1.6
2015-04-02 18:03:12 - [RenderConfig] film.tonemap.type = 0
2015-04-02 18:03:12 - [RenderConfig] image.filename = sala.png
2015-04-02 18:03:12 - [RenderConfig] image.height = 600
2015-04-02 18:03:12 - [RenderConfig] image.width = 600
2015-04-02 18:03:12 - [RenderConfig] opencl.cpu.use = 0
2015-04-02 18:03:12 - [RenderConfig] opencl.gpu.use = 1
2015-04-02 18:03:12 - [RenderConfig] opencl.gpu.workgroup.size = 64
2015-04-02 18:03:12 - [RenderConfig] path.filter.B = 0.333333
2015-04-02 18:03:12 - [RenderConfig] path.filter.C = 0.333333
2015-04-02 18:03:12 - [RenderConfig] path.filter.alpha = 2.
2015-04-02 18:03:12 - [RenderConfig] path.filter.type = GAUSSIAN
2015-04-02 18:03:12 - [RenderConfig] path.filter.width.x = 1.5
2015-04-02 18:03:12 - [RenderConfig] path.filter.width.y = 1.5
2015-04-02 18:03:12 - [RenderConfig] path.maxdepth = 10
2015-04-02 18:03:12 - [RenderConfig] path.pixelatomics.enable = 0
2015-04-02 18:03:12 - [RenderConfig] path.russianroulette.cap = 0.25
2015-04-02 18:03:12 - [RenderConfig] path.russianroulette.depth = 10
2015-04-02 18:03:12 - [RenderConfig] path.russianroulette.strategy = 1
2015-04-02 18:03:12 - [RenderConfig] path.sampler.largesteprate = 0.4
2015-04-02 18:03:12 - [RenderConfig] path.sampler.maxconsecutivereject = 512.
2015-04-02 18:03:12 - [RenderConfig] path.sampler.type = METROPOLIS
2015-04-02 18:03:12 - [RenderConfig] renderengine.type = 4
2015-04-02 18:03:12 - [RenderConfig] scene.epsilon = 0.0001
2015-04-02 18:03:12 - [RenderConfig] scene.file = scenes/sala/sala.scn
2015-04-02 18:03:12 - [LuxRays] OpenCL Platform 0: Intel
2015-04-02 18:03:12 - [LuxRays] OpenCL Platform 1: Advanced Micro Devices, Inc.
2015-04-02 18:03:12 - OpenCL ERROR: clGetDeviceIDs(-1)
2015-04-02 18:03:56 - Set LuxBall scene
2015-04-02 18:04:01 - Set Interactive mode
2015-04-02 18:04:01 - [RenderConfig] Reading configuration file: scenes/luxball/render.cfg
2015-04-02 18:04:01 - [RenderConfig] Configuration:
2015-04-02 18:04:01 - [RenderConfig] film.filter.type = 0
2015-04-02 18:04:01 - [RenderConfig] image.filename = luxball.png
2015-04-02 18:04:01 - [RenderConfig] image.height = 480
2015-04-02 18:04:01 - [RenderConfig] image.width = 640
2015-04-02 18:04:01 - [RenderConfig] opencl.cpu.use = 0
2015-04-02 18:04:01 - [RenderConfig] opencl.gpu.use = 1
2015-04-02 18:04:01 - [RenderConfig] opencl.gpu.workgroup.size = 64
2015-04-02 18:04:01 - [RenderConfig] path.filter.type = GAUSSIAN
2015-04-02 18:04:01 - [RenderConfig] path.maxdepth = 16
2015-04-02 18:04:01 - [RenderConfig] path.sampler.type = METROPOLIS
2015-04-02 18:04:01 - [RenderConfig] scene.fieldofview = 45
2015-04-02 18:04:01 - [RenderConfig] scene.file = scenes/luxball/luxball.scn
2015-04-02 18:04:01 - [LuxRays] OpenCL Platform 0: Intel
2015-04-02 18:04:01 - [LuxRays] OpenCL Platform 1: Advanced Micro Devices, Inc.
2015-04-02 18:04:01 - OpenCL ERROR: clGetDeviceIDs(-1)
2015-04-02 18:04:04 - Set Benchmark Custom mode
2015-04-02 18:04:04 - [RenderConfig] Reading configuration file: scenes/luxball/render.cfg
2015-04-02 18:04:04 - [RenderConfig] Configuration:
2015-04-02 18:04:04 - [RenderConfig] film.filter.type = 0
2015-04-02 18:04:04 - [RenderConfig] image.filename = luxball.png
2015-04-02 18:04:04 - [RenderConfig] image.height = 480
2015-04-02 18:04:04 - [RenderConfig] image.width = 640
2015-04-02 18:04:04 - [RenderConfig] opencl.cpu.use = 0
2015-04-02 18:04:04 - [RenderConfig] opencl.gpu.use = 1
2015-04-02 18:04:04 - [RenderConfig] opencl.gpu.workgroup.size = 64
2015-04-02 18:04:04 - [RenderConfig] path.filter.type = GAUSSIAN
2015-04-02 18:04:04 - [RenderConfig] path.maxdepth = 16
2015-04-02 18:04:04 - [RenderConfig] path.sampler.type = METROPOLIS
2015-04-02 18:04:04 - [RenderConfig] scene.fieldofview = 45
2015-04-02 18:04:04 - [RenderConfig] scene.file = scenes/luxball/luxball.scn
2015-04-02 18:04:05 - [LuxRays] OpenCL Platform 0: Intel
2015-04-02 18:04:05 - [LuxRays] OpenCL Platform 1: Advanced Micro Devices, Inc.
2015-04-02 18:04:05 - OpenCL ERROR: clGetDeviceIDs(-1)
2015-04-02 18:04:15 - Set Benchmark CPUs mode
2015-04-02 18:04:15 - [RenderConfig] Reading configuration file: scenes/luxball/render.cfg
2015-04-02 18:04:15 - [RenderConfig] Configuration:
2015-04-02 18:04:15 - [RenderConfig] film.filter.type = 0
2015-04-02 18:04:15 - [RenderConfig] image.filename = luxball.png
2015-04-02 18:04:15 - [RenderConfig] image.height = 480
2015-04-02 18:04:15 - [RenderConfig] image.width = 640
2015-04-02 18:04:15 - [RenderConfig] opencl.cpu.use = 0
2015-04-02 18:04:15 - [RenderConfig] opencl.gpu.use = 1
2015-04-02 18:04:15 - [RenderConfig] opencl.gpu.workgroup.size = 64
2015-04-02 18:04:15 - [RenderConfig] path.filter.type = GAUSSIAN
2015-04-02 18:04:15 - [RenderConfig] path.maxdepth = 16
2015-04-02 18:04:15 - [RenderConfig] path.sampler.type = METROPOLIS
2015-04-02 18:04:15 - [RenderConfig] scene.fieldofview = 45
2015-04-02 18:04:15 - [RenderConfig] scene.file = scenes/luxball/luxball.scn
2015-04-02 18:04:15 - [LuxRays] OpenCL Platform 0: Intel
2015-04-02 18:04:15 - [LuxRays] OpenCL Platform 1: Advanced Micro Devices, Inc.
2015-04-02 18:04:15 - OpenCL ERROR: clGetDeviceIDs(-1)

Claggy

Hello,

thanks for your replys. I read the whole thread. Even Windows with the official drivers from Intel has some serious issues with Intel GPU crunching. So it's not Linux fault ;)

I will not buy that driver, hell where would we come to. But I'm willing to try any open source attempts.

Unfortunately I can't post there yet, as I don't run Einstein and so I don't have credits and the bug and support area is restricted to users with credits. So I attached Einstein to my boinc client and we'll see how long it takes till I can post there.

I googled around to find a beignet + boinc howto, but couldn't find any. Maybe I should just give it a shot.

I already read, that beignet needs an X server. My computer is a linux server without X and without attached monitor. I read that on windows you need at least a vga dummy to make it work. I hope the beignet developer's worked around that silly restriction.

Greetings
MPW

@Claggy: Do you have a link where you found that?

I think you shouldn't forget, that beignet is under heavy development. There are commits each week. So maybe we can work it out with a little help of the developers.

/edit: And another point, beignet only supports OpenCL 1.1 right now, most boinc applications need 1.2 I guess. So we'll have to wait, till they get there.

@Claggy: Do you have a link where you found that?

I think you shouldn't forget, that beignet is under heavy development. There are commits each week. So maybe we can work it out with a little help of the developers.

From a developer thread (that's only viewable to devs and testers):

Beignet 0.8, 0.9.3, 1.01, 1.02

Claggy

OpenCL ERROR: clGetDeviceIDs(-1)

Urs has two platforms, Intel and AMD. Maybe that confuses one of them. BOINC probably gets the same error and then just skips the platform.

I already read, that beignet needs an X server. My computer is a linux server without X and without attached monitor. I read that on windows you need at least a vga dummy to make it work. I hope the beignet developer's worked around that silly restriction.

Beignet wrote:

Normally, beignet needs to run under X server environment as normal user. If there isn't X server, beignet provides two alternative to run: * Run as root without X. * Enable the drm render nodes by passing drm.rnodes=1 to the kernel boot args, then you can run beignet with non-root and without X.

OpenCL ERROR: clGetDeviceIDs(-1)

Urs has two platforms, Intel and AMD. Maybe that confuses one of them. BOINC probably gets the same error and then just skips the platform.

That wasn't with Boinc, but with Luxmark:

>:( No luck with Luxmark and intel_beignet, some basics like clGetDeviceIDs(-1) or clBuildProgramm(-3) fail :

2015-04-02 18:02:38 - LuxMark v2.0
2015-04-02 18:02:38 - Based on SmallLuxGPU v2.0devel4 (LuxRays demo: http://www.luxrender.net)

The Einstein app also fails to get the OpenCL device id:

GPU not found: type=intel_gpu, opencl_device_index=0, device_num=0
[23:05:30][2388][ERROR] Failed to get OpenCL platform/device info from BOINC (error: -1)!

Claggy

OpenCL ERROR: clGetDeviceIDs(-1)

Urs has two platforms, Intel and AMD. Maybe that confuses one of them. BOINC probably gets the same error and then just skips the platform.

That wasn't with Boinc, but with Luxmark:

Yes, I did see that. What BOINC does is it gets all the OpenCL platforms and then loops over the platforms and calls clGetDeviceIDs() for each platform. If clGetDeviceIDs() returns CL_DEVICE_NOT_FOUND=-1 BOINC skips over that platform without printing an error message.

The Luxmark run shows two platforms, Intel and AMD. Since Luxmark and BOINC both did detect AMD and Intel GPUs one of the platforms works and the other doesn't.

You may have heard that you can't really have AMD and NVIDIA GPUs on the same computer, use proprietary drivers and run X at the same time. The reason is AMD drivers install an accelerator module that doesn't know how to talk to NVIDIA GPU and vice versa. Since Beignet prefers to use X to talk to hardware I wouldn't be surprised if it's not compatible with AMD and NVIDIA drivers.

Also, Beignet people write that Beignet passes almost all Piglit test cases. Piglit is an OpenGL and OpenCL test suite. You'd think that being able to compile OpenCL kernels would be a rather important part of the tests. Since Luxmark can't compile kernels I'd say that hints strongly that the source of the problems is something other than Beignet.

If this were my machine and I were interested in using Intel GPU with BOINC, I'd start by removing AMD/NVIDIA drivers and the cards too for good measure, connect monitor to Intel GPU, not run CPUs at full tilt and do everything else necessary to make the setup as basic as possible. Only if Beignet works in that basic setup then I'd start thinking about running multiple GPUs and other things.

(Beignet page gives instructions on how to run without X. The same instructions might apply to running Beignet alongside AMD/NVIDIA drivers.)