When running a sample on Windows, SPIRV internal error appears.

The SPIRV error could be occurring if the default processor is a GPU but you are running a sample for a CPU. To force the sample to compile on the CPU:

1. In the src file for the sample, open the .cpp file. For example, C:\samples\vector-add\src\vector-add-buffers.cpp.
2. Find the #else statement that starts with //The default device selector will select the most performant device.
3. Replace the entire #else #endif statement with:

   #else
     // The default device selector will select the most performant device.
     // default_selector d_selector;
     cpu_selector d_selector;
   #endif
   

4. Compile the sample again, then run the sample.

In Visual Studio, the Intel menu does not appear and/or the samples from the installed toolkit are missing.

If Visual Studio is installed after a oneAPI toolkit, the plug-ins that come as part of the toolkit are not installed.

To fix this problem, uninstall and reinstall your oneAPI toolkits:

1. In Windows settings, open Add or Remove Programs.
2. Select on Intel® oneAPI toolkits.
3. Click Uninstall.
4. After the Uninstall process is complete, verify that Visual Studio 2017 and/or Visual Studio 2019 is installed.
5. Install your oneAPI toolkits following the same procedure you used the first time. The software is available for download at the Intel® oneAPI Toolkits page.

If you are unable to access the samples from the oneAPI CLI Samples Browser, set the value of https_proxy and http_proxy, then run oneapi-cli.exe in the same command line window. To set the proxy enter this command, where your.proxy is the name of your proxy server and 8080 is your default port.

set http_proxy=http://your.proxy:8080
set https_proxy=https://your.proxy:8080

After you set the proxy, return to the instructions for Running a Sample Using the Command Line and try again.

Long-running applications are terminated.

If your application has long-running GPU compute workloads in native environments, those workloads may be terminated by the hardware. This functionality is intended to prevent process hangs to run indefinitely; however, this behavior is not recommended for virtualizations or other standard usages of GPU, such as gaming.

A workload that takes more than four seconds for GPU hardware to execute is considered a long-running workload. By default, individual threads that qualify as long-running workloads are considered hung and are terminated.

To modify the termination process, see Timeout Detection and Recovery (TDR) Registry Keys from Microsoft.