The Intel Graphics Compiler "IGC" 2.34.4 release is out today as this compiler used by the Intel Compute Runtime for Level Zero and OpenCL compute on Intel graphics hardware plus is also used as the graphics shader compiler under Windows.

The IGC 2.34.4 release has more enablement work around the Crescent Island accelerator as well as Xe3P graphics at large. IGC 2.34.4 also adds [1]support for WideMulMad , [2]code scheduling improvements , Panther Lake optimizations, enabling the 2D load splitting pass by default, and officially supporting SPV_INTEL_bfloat16_arithmetic SPIR-V on Xe3P. SPV_INTEL_bfloat16_arithmetic has been experimental on Xe HPG+ platforms.

[3]

Downloads and more details on the IGC 2.34.4 compiler release via [4]GitHub . Given the timing of this new IGC release, a new Intel Compute Runtime update will also likely be released very soon.

Meanwhile on the open-source graphics compiler side for Intel Linux users, [5]the new Jay compiler continues being developed within Mesa to promising results.



[1] https://github.com/intel/intel-graphics-compiler/commit/af75aea1df7f6171d1224240c221e7fec417937b

[2] https://github.com/intel/intel-graphics-compiler/commit/d415ebea14c9c8b9e59c13fad32379e7fdb4d515

[3] https://www.phoronix.com/image-viewer.php?id=2026&image=intel_arc_b70_4_lrg

[4] https://github.com/intel/intel-graphics-compiler/releases/tag/v2.34.4

[5] https://www.phoronix.com/news/Intel-Jay-Mesa-Shader-Compiler