Intel® oneAPI DPC++/C++ Compiler
Talk to fellow users of Intel® oneAPI DPC++/C++ Compiler and companion tools like Intel® oneAPI DPC++ Library, Intel® DPC++ Compatibility Tool, and Intel® Distribution for GDB*
683 Discussions

Target Intel HD Graphics with OneAPI 2024

Sal1
Beginner
576 Views

I see that OneAPI 2024 dropped support for 10th gen integrated HD Graphics and below. My dev machine has an HD 630 gpu. Now when I try to target it in a SYCL program, I get null pointer dereferencing exceptions at runtime when opencl builds the program. Is there a way to still use older integrated gpus? Or is there a way to not accidentally select a device that will make my program crash? I know I can manually select a device, but I'd rather not rely on comparing device names.


This is reproduceable with oneapi 2024.1 and the vector add sample here https://github.com/oneapi-src/oneAPI-samples/tree/master/DirectProgramming/C%2B%2BSYCL/DenseLinearAlgebra/vector-add.

Labels (1)
0 Kudos
4 Replies
Alex_Y_Intel
Moderator
470 Views

Can you please attach a screen shot of the problem? Please also try with the newest compiler 2024.2

0 Kudos
Sal1
Beginner
426 Views

This is using the 2024.2 compiler with an i7-7700k and and Intel(R) HD Graphics 630 device. 

hd630.png

0 Kudos
Joel_L_Intel
Moderator
303 Views

is there a way to not accidentally select a device that will make my program crash? --> you can try using cpu_selector_v instead of default_selector_v

0 Kudos
Sal1
Beginner
285 Views

Yes I can and that will work. The trouble comes when you want to deploy a sycl application and you don't know what hardware it's running on. Right now I have to use cpu_selector_v then parse the processor name to see if it's an Intel iX processor that's 11th gen or later. Only then can I replace the device with gpu_selector_v or default_selector_v. I would think that default_selector_v returning the gpu device on an unsupported gpu is a bug because any attempted use of that device causes segfaults.

0 Kudos
Reply