Looking at the coproc_info.xml file and i spied a warning

failed to get CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD

Is this a bug to report?

Full contents below.

    <coprocs>
   <ati_opencl>
      <name>Ellesmere</name>
      <vendor>Advanced Micro Devices, Inc.</vendor>
      <vendor_id>4098</vendor_id>
      <available>1</available>
      <half_fp_config>0</half_fp_config>
      <single_fp_config>190</single_fp_config>
      <double_fp_config>63</double_fp_config>
      <endian_little>1</endian_little>
      <execution_capabilities>1</execution_capabilities>
      <extensions>cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_ex
tended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_fp16 cl_khr_gl_sharing cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_image2d_fro
m_buffer cl_khr_spir cl_khr_gl_event </extensions>
      <global_mem_size>8491732992</global_mem_size>
      <local_mem_size>32768</local_mem_size>
      <max_clock_frequency>555</max_clock_frequency>
      <max_compute_units>14</max_compute_units>
      <nv_compute_capability_major>0</nv_compute_capability_major>
      <nv_compute_capability_minor>0</nv_compute_capability_minor>
      <amd_simd_per_compute_unit>4</amd_simd_per_compute_unit>
      <amd_simd_width>16</amd_simd_width>
      <amd_simd_instruction_width>1</amd_simd_instruction_width>
      <opencl_platform_version>OpenCL 2.0 AMD-APP (2117.7)</opencl_platform_version>
      <opencl_device_version>OpenCL 1.2 AMD-APP (2117.7)</opencl_device_version>
      <opencl_driver_version>2117.7 (VM)</opencl_driver_version>
      <device_num>0</device_num>
      <peak_flops>994560000000.000000</peak_flops>
      <opencl_available_ram>8491732992.000000</opencl_available_ram>
      <opencl_device_index>0</opencl_device_index>
      <warn_bad_cuda>0</warn_bad_cuda>
   </ati_opencl>
<warning>NVIDIA: libcuda.so: cannot open shared object file: No such file or directory</warning>
<warning>ATI: libaticalrt.so: cannot open shared object file: No such file or directory</warning>
<warning>clGetDeviceInfo failed to get CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD for device 0</warning>
    </coprocs>

Looking at the coproc_info.xml file and i spied a warning

You should get the same warning when you have the coproc_debug flag enabled. If I read what it does right, your GPU shouldn't even be able to do OpenCL at this moment, because the SIMD on the compute units takes care of the parallellisation of the work.

Driver problem?
Hardware error?

To be complete, what OS and driver combo is this with?
Can you run work with it? Do you have more errors than feels normal?

---

Ageless wrote:

You should get the same warning when you have the coproc_debug flag enabled. If I read what it does right, your GPU shouldn't even be able to do OpenCL at this moment, because the SIMD on the compute units takes care of the parallellisation of the work.

Driver problem?
Hardware error?

No errors.

To be complete, what OS and driver combo is this with?

Ubuntu 16.04LTS and amdgpu-pro-driver-16.30.3

Can you run work with it? Do you have more errors than feels normal?

Working absolutely perfectly, with the debug flag set yes the warning shows.

Fri 16 Sep 2016 00:32:50 BST |  | [coproc] launching child process at /usr/bin/boinc
Fri 16 Sep 2016 00:32:50 BST |  | [coproc] relative to directory /
Fri 16 Sep 2016 00:32:50 BST |  | [coproc] with data directory /var/lib/boinc-client
Fri 16 Sep 2016 00:32:50 BST |  | OpenCL: AMD/ATI GPU 0: Ellesmere (driver version 2117.7 (VM), device version OpenCL 1.2 AMD-APP (2117.7), 6717MB, 6717MB available, 995 GFLOPS peak)
Fri 16 Sep 2016 00:32:50 BST |  | [coproc] NVIDIA: libcuda.so: cannot open shared object file: No such file or directory
Fri 16 Sep 2016 00:32:50 BST |  | [coproc] ATI: libaticalrt.so: cannot open shared object file: No such file or directory
Fri 16 Sep 2016 00:32:50 BST |  | [coproc] clGetDeviceInfo failed to get CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD for device 0

Well, there is something wrong somewhere.

CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD, CL_DEVICE_SIMD_WIDTH_AMD and CL_DEVICE_SIMD_INSTRUCTION_WIDTH_AMD are AMD extensions. If they are available BOINC uses them to compute peak flops. The driver says it supports them:

<extensions>... cl_amd_device_attribute_query ...</extensions>

And BOINC has managed to extract that info:

      <amd_simd_per_compute_unit>4</amd_simd_per_compute_unit>
      <amd_simd_width>16</amd_simd_width>
      <amd_simd_instruction_width>1</amd_simd_instruction_width>

The odd thing is, if BOINC has problem getting any OpenCL property it skips the device in question. And somehow BOINC still reports the AMD GPU.

What clinfo has to say?

Well, there is something wrong somewhere.

I thought that was odd also.

What clinfo has to say?

clinfo is silent on simd

Edit: ... but GPU speed thread had this error elsewhere

$ clinfo 
Number of platforms:				 1
  Platform Profile:				 FULL_PROFILE
  Platform Version:				 OpenCL 2.0 AMD-APP (2117.7)
  Platform Name:				 AMD Accelerated Parallel Processing
  Platform Vendor:				 Advanced Micro Devices, Inc.
  Platform Extensions:				 cl_khr_icd cl_amd_event_callback cl_amd_offline_devices 


  Platform Name:				 AMD Accelerated Parallel Processing
Number of devices:				 2
  Device Type:					 CL_DEVICE_TYPE_GPU
  Vendor ID:					 1002h
  Board name:					 
  Device Topology:				 PCI[ B#1, D#0, F#0 ]
  Max compute units:				 14
  Max work items dimensions:			 3
    Max work items[0]:				 256
    Max work items[1]:				 256
    Max work items[2]:				 256
  Max work group size:				 256
  Preferred vector width char:			 4
  Preferred vector width short:			 2
  Preferred vector width int:			 1
  Preferred vector width long:			 1
  Preferred vector width float:			 1
  Preferred vector width double:		 1
  Native vector width char:			 4
  Native vector width short:			 2
  Native vector width int:			 1
  Native vector width long:			 1
  Native vector width float:			 1
  Native vector width double:			 1
  Max clock frequency:				 555Mhz
  Address bits:					 64
  Max memory allocation:			 4244635648
  Image support:				 Yes
  Max number of images read arguments:		 128
  Max number of images write arguments:		 8
  Max image 2D width:				 16384
  Max image 2D height:				 16384
  Max image 3D width:				 2048
  Max image 3D height:				 2048
  Max image 3D depth:				 2048
  Max samplers within kernel:			 16
  Max size of kernel argument:			 1024
  Alignment (bits) of base address:		 2048
  Minimum alignment (bytes) for any datatype:	 128
  Single precision floating point capability
    Denorms:					 No
    Quiet NaNs:					 Yes
    Round to nearest even:			 Yes
    Round to zero:				 Yes
    Round to +ve and infinity:			 Yes
    IEEE754-2008 fused multiply-add:		 Yes
  Cache type:					 Read/Write
  Cache line size:				 64
  Cache size:					 16384
  Global memory size:				 6963351552
  Constant buffer size:				 65536
  Max number of constant args:			 8
  Local memory type:				 Scratchpad
  Local memory size:				 32768
  Max pipe arguments:				 0
  Max pipe active reservations:			 0
  Max pipe packet size:				 0
  Max global variable size:			 0
  Max global variable preferred total size:	 0
  Max read/write image args:			 0
  Max on device events:				 0
  Queue on device max size:			 0
  Max on device queues:				 0
  Queue on device preferred size:		 0
  SVM capabilities:				 
    Coarse grain buffer:			 No
    Fine grain buffer:				 No
    Fine grain system:				 No
    Atomics:					 No
  Preferred platform atomic alignment:		 0
  Preferred global atomic alignment:		 0
  Preferred local atomic alignment:		 0
  Kernel Preferred work group size multiple:	 64
  Error correction support:			 0
  Unified memory for Host and Device:		 0
  Profiling timer resolution:			 1
  Device endianess:				 Little
  Available:					 Yes
  Compiler available:				 Yes
  Execution capabilities:				 
    Execute OpenCL kernels:			 Yes
    Execute native function:			 No
  Queue on Host properties:				 
    Out-of-Order:				 No
    Profiling :					 Yes
  Queue on Device properties:				 
    Out-of-Order:				 No
    Profiling :					 No
  Platform ID:					 0x7fe5c9cb48f8
  Name:						 Ellesmere
  Vendor:					 Advanced Micro Devices, Inc.
  Device OpenCL C version:			 OpenCL C 1.2 
  Driver version:				 2117.7 (VM)
  Profile:					 FULL_PROFILE
  Version:					 OpenCL 1.2 AMD-APP (2117.7)
  Extensions:					 cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_fp16 cl_khr_gl_sharing cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_image2d_from_buffer cl_khr_spir cl_khr_gl_event 

Device Type:					 CL_DEVICE_TYPE_CPU
<snip CPU clinfo>

Ubuntu 16.04LTS and amdgpu-pro-driver-16.30.3

I wonder if that combination there isn't just the problem. Or is all this no longer true all of a sudden?

---

Or is all this no longer true all of a sudden?

Correct that is old news! Stay tuned to ubuntu-1604-lts-deprecating-amds-fglrx-catalyst

In fact if you dig around you'll see this SIMD error has been around a while. See my Edit link earlier.

Edit: ... but GPU speed thread had this error elsewhere

A-ha. BOINC isn't reporting OpenCL CPU, is it?

Stay tuned to ubuntu-1604-lts-deprecating-amds-fglrx-catalyst

I'd rather not go to those forums ever again in my life if I can help it. No matter what the information is that can only be found there.

In fact if you dig around you'll see this SIMD error has been around a while. See my Edit link earlier.

I had seen that error and thread before I posted in this thread. Before you find it again as well, I have also seen that the actual error is done by the client, it comes from gpu_opencl.cpp line 907 and further.

The code is reading this directly from the drivers.

My earlier comment on no parallellisation being capable was sourced directly from the GCN Architecture Whitepapers:

In GCN, each CU includes 4 separate SIMD units for vector processing. Each of these SIMD units simultaneously executes a single operation across 16 work items, but each can be working on a separate wavefront. This places emphasis on finding many wavefronts to be processed in parallel, rather than relying on the compiler to find independent operations within a single wavefront.

(Page 3, above the Figure 2 image)

Could this be a BOINC bug? Possibly, but then again it does catch it in its code if it cannot read from the drivers.
So, wouldn't this then be a bug in the drivers? In my opinion, more likely.
You don't completely remove support for hardware from an operating system to then without any problems whatsoever fix that again by adding drivers back.

But you ought to be able to 'simply' test it: install a previous Ubuntu, with its drivers still intact... does that also throw that warning? If that one does as well, we have some basis to work with.

And then we can go figure out if it's hardware or BOINC.

Now on the other hand, if you feel there are no further problems, it's not throwing errors, it's just showing a low max clock frequency (also something that comes from the drivers) and this warning about the SIMD, is there something that needs fixing?

---

Edit: ... but GPU speed thread had this error elsewhere

A-ha. BOINC isn't reporting OpenCL CPU, is it?

No it is not, and i guess it should (as clinfo reports it)

But you ought to be able to 'simply' test it: install a previous Ubuntu, with its drivers still intact... does that also throw that warning? If that one does as well, we have some basis to work with.

RX-480 does not work with the older AMD driver fgrlx and not on older versions of ubuntu so sadly that is not possible.

Now on the other hand, if you feel there are no further problems, it's not throwing errors, it's just showing a low max clock frequency (also something that comes from the drivers) and this warning about the SIMD, is there something that needs fixing?

The clock frequency reported i think uses the simd value to calculate it, i believe based on the documentation you quoted it should equal 4, assuming 4th gen GCN works as described in the documentation.

The warning message should be fixed if there is nothing to warn about, especially as this is the current version of hardware, driver and OS. (my assumption here that nothing is wrong - perhaps it could be running N times faster but it's running as i expected)

If the fixing is for AMD to do (also no simd in clinfo) then we need to raise that with AMD. If boinc is making a (no longer) or unsupported call then then that needs fixing.

Is it a priority? i very much doubt it.

Ok, the error likely comes from OpenCL CPU detection. The driver says it supports AMD device attribute query but then doesn't support it for CPUs. Or something like that.

I don't know if it's BOINC or driver bug. Would have to read the OpenCL extensions specification carefully first. Either way, BOINC could handle that better.

BOINC could also have better logging. The error message wasn't exactly clear on what was failing.

If boinc is making a (no longer) or unsupported call then then that needs fixing.

It's a normal function of OpenCL, backward compatible all the way to the earliest versions. It's also something that's required for the correct function of OpenCL, so it should just be there.

I'm wondering if the card isn't just warped.
Global memory size: 6963351552
6963351552 / 1024 = 6800148 /1024 = 6640.77 / 1024 = 6.49 GB memory.

By comparison, the value for my HD7870 - 2GB is 2147483648 which when calculated comes out at exactly 2GB, so do you have an RX-480 - 6.49GB?

---

What clinfo says for the CPU?

btw. The max clock frequency comes straight from the driver. You might want to report that to AMD. Is the driver still in beta?

edit: In the first post the global memory size is reasonable 7,9 GB but less in later posts. Is the driver reporting available memory there?

Juha wrote:

btw. The max clock frequency comes straight from the driver. You might want to report that to AMD. Is the driver still in beta?

AMD Linux-beta-Driver suggests beta (by virtue of the URL name), although not mentioned beta in a quick scan on web site and release notes.

Edit: this thread at AMD support shows number of compute units and other things also wrong.

edit: In the first post the global memory size is reasonable 7,9 GB but less in later posts. Is the driver reporting available memory there?

Yes it appears so, it does change downward, and https://community.amd.com/message/1305298 suggests this is not always the physical maximum.

What clinfo says for the CPU?

  Device Type:					 CL_DEVICE_TYPE_CPU
  Vendor ID:					 1002h
  Board name:					 
  Max compute units:				 4
  Max work items dimensions:			 3
    Max work items[0]:				 1024
    Max work items[1]:				 1024
    Max work items[2]:				 1024
  Max work group size:				 1024
  Preferred vector width char:			 16
  Preferred vector width short:			 8
  Preferred vector width int:			 4
  Preferred vector width long:			 2
  Preferred vector width float:			 8
  Preferred vector width double:		 4
  Native vector width char:			 16
  Native vector width short:			 8
  Native vector width int:			 4
  Native vector width long:			 2
  Native vector width float:			 8
  Native vector width double:			 4
  Max clock frequency:				 4499Mhz
  Address bits:					 64
  Max memory allocation:			 4131714048
  Image support:				 Yes
  Max number of images read arguments:		 128
  Max number of images write arguments:		 64
  Max image 2D width:				 8192
  Max image 2D height:				 8192
  Max image 3D width:				 2048
  Max image 3D height:				 2048
  Max image 3D depth:				 2048
  Max samplers within kernel:			 16
  Max size of kernel argument:			 4096
  Alignment (bits) of base address:		 1024
  Minimum alignment (bytes) for any datatype:	 128
  Single precision floating point capability
    Denorms:					 Yes
    Quiet NaNs:					 Yes
    Round to nearest even:			 Yes
    Round to zero:				 Yes
    Round to +ve and infinity:			 Yes
    IEEE754-2008 fused multiply-add:		 Yes
  Cache type:					 Read/Write
  Cache line size:				 64
  Cache size:					 32768
  Global memory size:				 16526856192
  Constant buffer size:				 65536
  Max number of constant args:			 8
  Local memory type:				 Global
  Local memory size:				 32768
  Max pipe arguments:				 16
  Max pipe active reservations:			 16
  Max pipe packet size:				 4131714048
  Max global variable size:			 1879048192
  Max global variable preferred total size:	 1879048192
  Max read/write image args:			 64
  Max on device events:				 0
  Queue on device max size:			 0
  Max on device queues:				 0
  Queue on device preferred size:		 0
  SVM capabilities:				 
    Coarse grain buffer:			 No
    Fine grain buffer:				 No
    Fine grain system:				 No
    Atomics:					 No
  Preferred platform atomic alignment:		 0
  Preferred global atomic alignment:		 0
  Preferred local atomic alignment:		 0
  Kernel Preferred work group size multiple:	 1
  Error correction support:			 0
  Unified memory for Host and Device:		 1
  Profiling timer resolution:			 1
  Device endianess:				 Little
  Available:					 Yes
  Compiler available:				 Yes
  Execution capabilities:				 
    Execute OpenCL kernels:			 Yes
    Execute native function:			 Yes
  Queue on Host properties:				 
    Out-of-Order:				 No
    Profiling :					 Yes
  Queue on Device properties:				 
    Out-of-Order:				 No
    Profiling :					 No
  Platform ID:					 0x7f744609b8f8
  Name:						 Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz
  Vendor:					 GenuineIntel
  Device OpenCL C version:			 OpenCL C 1.2 
  Driver version:				 2117.7 (sse2,avx)
  Profile:					 FULL_PROFILE
  Version:					 OpenCL 1.2 AMD-APP (2117.7)
  Extensions:					 cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_device_fission cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_spir cl_khr_gl_event 

  Device Type:	CL_DEVICE_TYPE_CPU
  Extensions:	... cl_amd_device_attribute_query ...

Ok, I'd call it driver bug then.

While researching an answer I gave someone in PM, I came across this post (this one) on the Ethereum Community forums where someone shows what I suspect is his clinfo output of his RX-480 and it's identical to that of Mike's here. Well, bar the memory readings. Same driver, though.

---

Hmmm... I just turned coproc_debug on and find the following in my BOINC log:

26/09/2016 14:19:42 |  | Starting BOINC client version 7.6.33 for windows_x86_64
26/09/2016 14:19:42 |  | Data directory: M:\ProgramData\BOINC
26/09/2016 14:19:42 |  | [coproc] launching child process at P:\BOINCProgram\BOINC\boinc.exe
26/09/2016 14:19:42 |  | [coproc] relative to directory M:\ProgramData\BOINC
26/09/2016 14:19:42 |  | [coproc] with data directory "M:\ProgramData\BOINC"
26/09/2016 14:19:42 |  | CAL: ATI GPU 0: AMD Radeon HD 7850/7870 series (Pitcairn) (CAL version 1.4.1848, 2048MB, 2008MB available, 6400 GFLOPS peak)
26/09/2016 14:19:42 |  | OpenCL: AMD/ATI GPU 0: AMD Radeon HD 7850/7870 series (Pitcairn) (driver version 2117.9 (VM), device version OpenCL 1.2 AMD-APP (2117.9), 2048MB, 2008MB available, 6400 GFLOPS peak)
26/09/2016 14:19:42 |  | [coproc] No NVIDIA library found
26/09/2016 14:19:42 |  | [coproc] clGetDeviceInfo failed to get CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD for device 0

That's with the Crimson 16.8.3 drivers on Windows 7.

Suspect that's something the drivers don't do then.

---