Intel Graphics Compiler Merges New Vector Compute Backend

While Intel on the hardware manufacturing side continues facing stiff challenges, on the open-source software side the company continues making legendary progress. Out in today's Intel Graphics Compiler and in turn Intel Compute Runtime releases as part of their GPGPU toolchain is the recent open-sourcing and integration of their Vector Compute back-end.

One of the big tasks recently was on developing this new "Vector Compute" back-end for the Intel Graphics Compiler. It was just earlier this month that they open-sourced this Vector Compute back-end. In turn it's part of today's IGC 1.0.4361 compiler update.

Public documentation is light concerning this new Vector Compute back-end but it appears to be part of their GPGPU toolchain effort for supporting new vector instructions on their GPUs and in turn allowing inline Assembly in SPIR-V and a new vector compute extension for SPIR-V (SPV_INTEL_vector_compute). That new extension opens up new Intel vector compute features for SPIR-V shaders.

Going along with that work is the recently open-sourced Intel VC Intrinsics library as new intrinsics built on top of LLVM IR for representing SIMD semantics targeting Intel GPUs. VC Intrinsics in turn is used by not only the Intel Graphics Compiler but also their C for Metal Compiler, Data Parallel C++, and ISPC compiler.

This new Intel Graphics Compiler release with the VC back-end in turn was picked up by today's Intel Compute Runtime 20.29.17408 releasethat also makes note of the new VC support in IGC while also enabling cl_khr_il_program support for all platforms and other code updates for this open-source project enabling OpenCL and oneAPI Level Zero support on Intel graphics processors.