Sample Gaussian-Blur.cpp fails in CUDA Device

SYCL development
1

Hello,
I installed llvm project for sycl in WSL for windows, I am using windows 11Pro and Ubuntu 20.04, I have a NVIDIA GTX1650 card with CUDA 11.5 software installed.

Then I started to compile the examples in ComputeCpp-sdk.
The example images.cpp compiles and works.
The example Gaussian-Blur.cpp compiles but when I executed I get error
pi_die: PI CUDA kernels only support images with channel types int32, uint32, float, and half.

Looking at the code I see that images.cpp has the lines for image as
image<2> srcImage(src, image_channel_order::rgba, image_channel_type::fp32, range<2>(16, 16));
image<2> destImage(dest, image_channel_order::rgba, image_channel_type::fp32, range<2>(16, 16));

Looking at the code in Gaussian-Blur.cpp the lines for image are
image<2> image_in(inputData, co::rgba, ct::unorm_int8, imgRange);
image<2> image_out(outputData, co::rgba, ct::unorm_int8, imgRange);

It seems image_channel_type unorm_int8 is not supported.
I wonder if there is a solution for this problem.
Is it possible to change the pixels in the image file to int32 or float, do you know any example?

Thank you for your attention.

2

This seems to be an expected error, i.e. when I look at the code in DPC++ it throws that error so the behaviour you see is expected. We are investigating a similar issue and I am trying to confirm if this is the same and if we can fix this or provide a workaround. Thanks for your patience.