cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

nickpenny
Adept I

How to enable HSA Beta SDK?

I have been working through the README.  All goes well until I get to the section:
'Force Enabling HSA with Environment Variables' and the line:
To verify usage of the HSA stack, run clinfo and look for the name of the GPU device. It should be “AMD HSA Device”.

My clinfo still reports 'Spectre'.

System is Windows 7 pro SP1 64 bit.
ASUS A88x-PRO (containing an AMD Radeon R7)
BIOS is 0802

I have set both AMDAPPSDKROOT and HSA_RUNTIME in
System-­->Advanced system settings-­->Environment variables
and rebooted.  Both environment variables show up in the command window.

Thanks for your attention,
Nick.

x86_64\clinfo output:
Number of platforms:     1
  Platform Profile:     FULL_PROFILE
  Platform Version:     OpenCL 1.2 AMD-APP (1411.4)
  Platform Name:     AMD Accelerated Parallel Processing
  Platform Vendor:     Advanced Micro Devices, Inc.
  Platform Extensions:     cl_khr_icd cl_khr_d3d10_sharing cl_khr_d3d11_sharing

cl_khr_dx9_media_sharing cl_amd_event_callback cl_amd_offline_devices cl_amd_hsa


  Platform Name:     AMD Accelerated Parallel Processing
Number of devices:     2
  Device Type:      CL_DEVICE_TYPE_GPU
  Device ID:      4098
  Board name:      AMD Radeon(TM) R7 Graphics
  Device Topology:     PCI[ B#0, D#1, F#0 ]
  Max compute units:     8
  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:     720Mhz
  Address bits:      32
  Max memory allocation:    603193344
  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:     2146959360
  Constant buffer size:     65536
  Max number of constant args:    8
  Local memory type:     Scratchpad
  Local memory size:     32768
  Kernel Preferred work group size multiple:  64
  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:    No
  Queue properties:    
    Out-of-Order:     No
    Profiling :      Yes
  Platform ID:      000007FEE3E10FB0
  Name:       Spectre
  Vendor:      Advanced Micro Devices, Inc.
  Device OpenCL C version:    OpenCL C 1.2
  Driver version:     1411.4 (VM)
  Profile:      FULL_PROFILE
  Version:      OpenCL 1.2 AMD-APP (1411.4)
  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_atomic_counters_32 cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops

cl_amd_media_ops2 cl_amd_popcnt cl_khr_d3d10_sharing cl_khr_dx9_media_sharing cl_khr_image2d_from_buffer cl_khr_spir


  Device Type:      CL_DEVICE_TYPE_CPU
  Device ID:      4098
  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:     3700Mhz
  Address bits:      64
  Max memory allocation:    2147483648
  Image support:     Yes
  Max number of images read arguments:   128
  Max number of images write arguments:   8
  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:      16384
  Global memory size:     7450910720
  Constant buffer size:     65536
  Max number of constant args:    8
  Local memory type:     Global
  Local memory size:     32768
  Kernel Preferred work group size multiple:  1
  Error correction support:    0
  Unified memory for Host and Device:   1
  Profiling timer resolution:    276
  Device endianess:     Little
  Available:      Yes
  Compiler available:     Yes
  Execution capabilities:    
    Execute OpenCL kernels:    Yes
    Execute native function:    Yes
  Queue properties:    
    Out-of-Order:     No
    Profiling :      Yes
  Platform ID:      000007FEE3E10FB0
  Name:       AMD A10-7850K APU with Radeon(TM) R7 Graphics 
  Vendor:      AuthenticAMD
  Device OpenCL C version:    OpenCL C 1.2
  Driver version:     1411.4 (sse2,avx,fma4)
  Profile:      FULL_PROFILE
  Version:      OpenCL 1.2 AMD-APP (1411.4)
  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_d3d10_sharing cl_khr_spir cl_amd_svm

0 Likes
12 Replies
dipak
Big Boss

Hi Nick,

I installed the HSA beta driver [http://developer.amd.com/tools-and-sdks/opencl-zone/opencl-tools-sdks/opencl-1-2-beta-driver/] on AMD A10-7850K APU with Radeon(TM) R7 Graphics and followed the steps as described in section 'Force Enabling HSA with Environment Variables' and it worked for me.

Here are my steps (as per "README" doc):

1. Installed the Catalyst driver and reboot. At this time, clinfo showed the GPU device name as Spectre.

2. Skip the registry setting steps as not needed for this test.

3. Set environmental variables AMDAPPSDKROOT and HSA_RUNTIME via control panel as mentioned in the doc.

4. Run clinfo and it showed device name as "AMD HSA Device" (please find the attached clinfo output)

Please make sure HSA_RUNTIME is set to 1 (no extra character say a space).

Regards,

0 Likes

Thnakx

0 Likes
nickpenny
Adept I

Hi Dipak,

  Thanks for the details you provided. Sorry for the late reply - I have been on holiday.

  As you suggested I have checked my environment variables. I have also attached the output from 'set' here.

  There seems to be some differences between our 'clinfo' outputs E.g. you have an earlier version of OpenCL to me:

Platform Version: OpenCL 1.2 AMD-APP (1526.3) VS OpenCL 1.2 AMD-APP (1411.4)

and you have the platform extension 'cl_amd_hsa' which is missing from mine.

Do I need to roll back to version 1411.4 for this to work?

Are there any diagnostic things I can do to see why the HSA is not starting?  Are there reports to a log file or the event log?

Can I manually start it to see if stderr receives any output?

Regards,

Nick.

ALLUSERSPROFILE=C:\ProgramData

AMDAPPSDKROOT=C:\Users\spa\Desktop\HSA-Devel-Beta\sdk\

APPDATA=C:\Users\spa\AppData\Roaming

CommonProgramFiles=C:\Program Files\Common Files

CommonProgramFiles(x86)=C:\Program Files (x86)\Common Files

CommonProgramW6432=C:\Program Files\Common Files

COMPUTERNAME=KAVERI2

ComSpec=C:\Windows\system32\cmd.exe

FP_NO_HOST_CHECK=NO

HOMEDRIVE=C:

HOMEPATH=\Users\spa

HSA_RUNTIME=1

LOCALAPPDATA=C:\Users\spa\AppData\Local

LOGONSERVER=\\KAVERI2

NUMBER_OF_PROCESSORS=4

OPENSSL_CONF=C:\OpenSSL-Win64\bin\openssl.cfg

OS=Windows_NT

Path=C:\Windows\system32;C:\Windows;C:\Windows\System32\Wbem;C:\Windows\System32\WindowsPowerShell\v1.0\;C:\Program Files (x86)\ATI Technologies\ATI.ACE\Core-Static;C:\Program Files\Microsoft SQL Server\110\Tools\Binn\

PATHEXT=.COM;.EXE;.BAT;.CMD;.VBS;.VBE;.JS;.JSE;.WSF;.WSH;.MSC

PROCESSOR_ARCHITECTURE=AMD64

PROCESSOR_IDENTIFIER=AMD64 Family 21 Model 48 Stepping 1, AuthenticAMD

PROCESSOR_LEVEL=21

PROCESSOR_REVISION=3001

ProgramData=C:\ProgramData

ProgramFiles=C:\Program Files

ProgramFiles(x86)=C:\Program Files (x86)

ProgramW6432=C:\Program Files

PROMPT=$P$G

PSModulePath=C:\Windows\system32\WindowsPowerShell\v1.0\Modules\

PUBLIC=C:\Users\Public

SESSIONNAME=Console

SystemDrive=C:

SystemRoot=C:\Windows

TEMP=C:\Users\spa\AppData\Local\Temp

TMP=C:\Users\spa\AppData\Local\Temp

USERDOMAIN=KAVERI2

USERNAME=spa

USERPROFILE=C:\Users\spa

VS100COMNTOOLS=c:\Program Files (x86)\Microsoft Visual Studio 10.0\Common7\Tools\

VS110COMNTOOLS=C:\Program Files (x86)\Microsoft Visual Studio 11.0\Common7\Tools\

windir=C:\Windows

windows_tracing_flags=3

windows_tracing_logfile=C:\BVTBin\Tests\installpackage\csilogfile.log

0 Likes

Hi Nick,

As you posted in your earlier mail:

@@@ Nick -


x86_64\clinfo output:
Number of platforms:     1
  Platform Profile:     FULL_PROFILE
  Platform Version:     OpenCL 1.2 AMD-APP (1411.4)
  Platform Name:     AMD Accelerated Parallel Processing
  Platform Vendor:     Advanced Micro Devices, Inc.
  Platform Extensions:     cl_khr_icd cl_khr_d3d10_sharing cl_khr_d3d11_sharing


cl_khr_dx9_media_sharing cl_amd_event_callback cl_amd_offline_devices cl_amd_hsa



@@@@@ Nick


There seems to be some differences between our 'clinfo' outputs E.g. you have an earlier version of OpenCL to me:









Platform Version: OpenCL 1.2 AMD-APP (1526.3) VS OpenCL 1.2 AMD-APP (1411.4)

I think your version (see bold under-line text in your clinfo) with  was same as mine [1411.4] [As I mentioned, I installed the latest HSA beta driver from here http://developer.amd.com/tools-and-sdks/opencl-zone/opencl-tools-sdks/opencl-1-2-beta-driver/].


@@@@@ Nick


you have the platform extension 'cl_amd_hsa' which is missing from mine.


You see, cl_amd_hsa extension (see bold under-line text) is also available in your clinfo.

FYI: There is a HSA forum (newly created) for HSA related issues, so now onwards you can post your queries over there.

Regards,

0 Likes

Hi Dipak,

  Thanks for the quick response.

  Yes - I have been investigating why my version number has changed and it is based on the directory I am in at the time I run clinfo.

  If I am in C:\Users\spa\HSA-Devel-Beta\sdk\bin\x86  (and also x86_64)

I get Platform Version:     OpenCL 1.2 AMD-APP (1411.4)


If I am elsewhere (the clinfo.exe from c:\windows\system32 is called)

then I get Platform Version: OpenCL 1.2 AMD-APP (1526.3)


The cl_amd_hsa extension is only missing from the system32 version.


All 3 clinfo's still report Name:       Spectre


As to my diagnostics questions:

Are there any diagnostic things I can do to see why the HSA is not starting?  Are there reports to a log file or the event log?

Can I manually start it to see if stderr receives any output?

I had a look at the new HSA forum and I see you have moved this question into it. So I hope replying here is what you intended.

Thanks again,

Nick.

0 Likes

Hi Nick,

As per the section "Driver installation instructions for AMD Kaveri" in README-Devel-Beta.pdf :


Pre-­‐install notes


1. Windows 8.1 is required for this Beta.


2. You will need to start with a fresh install of the OS with no previous version of Catalyst or AMD graphics drivers installed.


In your case, I think, other version of driver co-exists and may causing the problem. Please try to install the HSA driver on a fresh OS and check.

As you wrote:


As to my diagnostics questions:


Are there any diagnostic things I can do to see why the HSA is not starting?  Are there reports to a log file or the event log?



Can I manually start it to see if stderr receives any output?


I don't have much idea how to diagnostics manually. If you face the same problem after fresh installation of HSA driver, please let us know. In that case, it will be better to move this thread (haven't moved it yet) to HSA forum where you may get a solution.

Regards,

0 Likes

Hi Dipak,

  Thanks for your continued help.

Yesterday I did get a colleague to check my findings on duplicate hardware but a completely fresh version of Windows 7 pro SP1 64 bit.

He has the same issue - Name:       Spectre

I'm sure I could provide better information if there was a diagnostic process to follow?

I'm happy for you to move this thread to HSA.

Regards,

Nick.

0 Likes

Hi Nick,

Don't know why it is not working for you (also your colleague). As I told it worked for me fine.

I'll move this thread to HSA forum.

BTW, did you check the BIOS version? Because as per the section "HW Requirements" in README-Devel-Beta.pdf :


If the motherboard BIOS version is lower than 0802 it must be updated. All testing has been done with the 0802 revision and we recommend you use the 0802 BIOS for now even if new versions are released.


Regards,

0 Likes

Hi Dipak,

  Yes - my bios is at 0802.

  As to why it does not work here - that is why a diagnostic process is useful

  So lets try and build one.  First things first - what process takes notice of HSA_RUNTIME?

Regards,

Nick.

0 Likes

Hi Nick,

As I told, I'm moving this thread to HSA forum. Hope you'll get your solution soon or may get the diagnostic process you wanted.

Regards,

0 Likes

You say that you use Windows 7 x64 SP1. As far as I see on the download page of the driver, only Windows 8.1 x64 is supported. May be you should try with Windows 8.1.

(I'm not sure it would fix your issue, I have not tested HSA yet).

0 Likes

Hi Ziple,

  Thanks for your suggestion. I've been working through the Beta SDK (which contains the driver) so I haven't downloaded the driver directly.  In the ReadMe that comes with the Beta SDK it says:

HW Requirements

This release is intended for use with a specific hardware configuration – an AMD A10-7850K “Kaveri” APU in an ASUS A88X-PRO motherboard with 8GB of RAM, running a 64‐bit version of either Microsoft Windows 7 or Windows 8.1, without a discrete GPU.

The ReadMe may, of course, be wrong - it would be nice to know.

In the mean time I am trialing the SDK on win 8.1 - which is throwing up different issues but I would like to keep this thread to win 7.

Regards,

Nick.

0 Likes