Intel IGC 2.38.2 Brings Latest Round Of Graphics Compiler Improvements

Ahead of the next Intel Compute Runtime release, IGC 2.38.2 was released today as the newest feature update to this open-source graphics compiler used by Intel iGPU/dGPU hardware on both Windows and Linux.

There are a wide assortment of fixes, new feature additions, and performance optimizations with the Intel Graphics Compiler 2.38.2 release.

On the performance optimization side, there are some optimizations benefiting Xe2+ hardware. There is also now IGC preferring SIMD16 over SIMD32 on Panther Lake when register pressure favors it.

Enabling Address Sanitizer "ASan" builds for IGC on Linux is one of the new improvements in this release. IGC 2.38.2 also brings adjustments for newer LLVM versions, namely LLVM 22.

Support for the OpenCL cl_khr_subgroup_rotate extension is also new to the Intel Graphics Compiler. The cl_khr_subgroup_rotate adds built-in OpenCL C functions for a new sub-group data exchange operation for rotating values through the work items in a sub-group.

IGC 2.38.2 is now available from GitHub. Typically when there is a new IGC compiler release, a new Intel Compute Runtime release quickly follows.