Generating Native Code for the Processor Graphics

This topic only applies to Intel® 64 and IA-32 architectures targeting Intel® Graphics Technology.

You can build a kernel to run natively on the processor graphics. Native code is kernel code that is in the native Instruction Set Architecture (ISA) instead of a virtual Instruction Set Architecture (vISA), and can only run on platforms that support the native ISA. Native code generation is performed at link time, instead of invoking the JIT compiler at runtime.

If you are certain that the target platform will always be the same, you can consider using this option to avoid the runtime overhead of invoking the JIT compiler. If you are not certain that you can depend on the target platform always being the same, you shouldn't use this option. When you generate code to run natively, the offload kernel cannot run on other platforms.

To see the platform information for your machine, you can use the gfx_sys_check tool.

To generate native code for Intel® Graphics Technology kernels, include the mgpu-arch (Linux*) or Qgpu-arch (Windows*) compiler option when you invoke the compiler.