Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Dmitry_Savin
New Contributor I
669 Views

Silent exit with execution policy on NEO.

Jump to solution

I am trying to use PSTL with std::vector.
A simple program fails to compile for FPGA Emulation Device and OpenCL CPU device with cl::sycl::compile_program_error, which provides some information, at least.
But on the OpenCL HD Graphics it silently exits.
Why does it happen and how to debug such fails?

The command to compile is

 

 

dpcpp -tbb -lpthread main.C -O0 -g -std=c++20

 

 

The output, gdb output and the source are attached to this post.

On the SYCL host device a program that skips NEO gives

 

 

terminate called after throwing an instance of 'cl::sycl::runtime_error'
  what():  Sub-group feature is not supported on HOST device. -33 (CL_INVALID_DEVICE)

 

 

So it compiles, at least.

Suggestions how to make it work on any of devices are appreciated.

PS It removes main.cpp attachment because "The attachment's main.cpp content type (text/x-c++src) does not match its file extension", so changed it to main.C.

Tags (1)
0 Kudos
1 Solution
Dmitry_Savin
New Contributor I
490 Views

The non-free drivers are required for an integrated GPU, too.

After the reinstallation the programs work on the Intel(R) OpenCL HD Graphics
 platform.

I think that this topic should be renamed to "Silent exit on NEO" because the issue was not with the PSTL, and can be closed,

unless someone can explain why there is no Level-Zero platform for the GPU, while it is present in a system with the same driver installation and a similar GPU? This one is reported as Intel Corporation UHD Graphics 630 (Mobile) in i7-8750H CPU, the other - Intel UHD Graphics 620 (rev 07) in i5-8250U CPU. Both systems are up-to-date Ubuntu 20.04 LTS.

View solution in original post

10 Replies
Dmitry_Savin
New Contributor I
600 Views

A simplified version that calls fill like the example fails the same way.

Dmitry_Savin
New Contributor I
594 Views

An asynchronous exception handler does not give any additional information.

GouthamK_Intel
Moderator
515 Views

Hi Dmitry,

Thanks for providing the reproducer and steps to reproduce.

>>"A simplified version that calls fill like the example fails the same way."

We tried to compile to run the code on our end, and we didn't face any issues, the code was executed successfully.

Please refer to the below screenshot.

GouthamK_Intel_3-1616157158859.png

 

Could you share more details about your system environment?

OSVersion:

DPCPP Compiler Version:

GPU Device:

Driver Version:

 

Thanks & Regards

Goutham

 

Dmitry_Savin
New Contributor I
500 Views

OSVersion: Ubuntu 20.04.2 LTS

DPCPP Compiler Version: 2021.1.2 (2020.10.0.1214)

GPU Device: Intel Corporation UHD Graphics 630 (Mobile)

Driver Version: 5.4.0-67-generic

UPD Below the tail of the strace output is attached.

UPD The guide says that additional GPU drivers are to be installed if either it is te first installation of the OneAPI, ore the Intel GPU is discrete.
Anyway, I noticed the Level Zero device in your output, and I am updating by the guide.
Please notify the repository maintainers that it is impossible to update with i386 packages installed due to unsatisfied dependencies.

Dmitry_Savin
New Contributor I
491 Views

The non-free drivers are required for an integrated GPU, too.

After the reinstallation the programs work on the Intel(R) OpenCL HD Graphics
 platform.

I think that this topic should be renamed to "Silent exit on NEO" because the issue was not with the PSTL, and can be closed,

unless someone can explain why there is no Level-Zero platform for the GPU, while it is present in a system with the same driver installation and a similar GPU? This one is reported as Intel Corporation UHD Graphics 630 (Mobile) in i7-8750H CPU, the other - Intel UHD Graphics 620 (rev 07) in i5-8250U CPU. Both systems are up-to-date Ubuntu 20.04 LTS.

View solution in original post

GouthamK_Intel
Moderator
463 Views

Hi Dmitry,

>>"why there is no Level-Zero platform for the GPU, while it is present in a system with the same driver installation and a similar GPU?"

Could you please share the complete output of the below two commands for both the systems you are referring to?

  1. clinfo
  2. sycl-ls --verbose


Regards

Goutham


Dmitry_Savin
New Contributor I
456 Views

UPD found that intel-level-zero-gpu wasn't installed on one of the systems, installing fixed the issue.
The attached files contain the original output.

Hi Goutham,

Find them attached below.

This post was edited to include the output from the other device.

Regards
Dmitry

 

GouthamK_Intel
Moderator
432 Views

Hi Dmitry,

>>"found that intel-level-zero-gpu wasn't installed on one of the systems, installing fixed the issue."

Glad to know that your issue is resolved, could you please let us know if we can close this thread from our end?


Thanks

Goutham


Dmitry_Savin
New Contributor I
423 Views

Hi Goutham,

You can close this thread.

Thanks
Dmitry

GouthamK_Intel
Moderator
412 Views

Hi Dmitry, 

Thanks for the confirmation!

As this issue has been resolved, we will no longer respond to this thread. 

If you require any additional assistance from Intel, please start a new thread. 

Any further interaction in this thread will be considered community only. 

Have a Good day.


Thanks & Regards

Goutham


Reply