Intel Compute Runtime 24.31.30508.7 was released this morning as the newest version of this OpenCL and oneAPI Level Zero compute stack used on both Windows and Linux platforms. Notable with the Intel Compute Runtime 24.31.30508.7 is having initial Xe2 support.

The Intel Compute Runtime 24.31.30508.7 release has initial Xe2 platform support that is used by both upcoming Lunar Lake integrated graphics as well as the Battlemage discrete GPUs. Intel open-source/Linux driver engineers have been quite busy in the past number of months working out the Xe2 support. With [1]Linux 6.12 there will be the support enabled out-of-the-box on both Lunar Lake and Battlemage graphics. It's good seeing the Compute Runtime support making its debut ahead of the products being announced. The Compute Runtime feature matrix has yet to be updated for Xe2 BMG / LNL so it's not clear at this time if the support is fully-baked or how Intel is positioning the support status in this inaugural release.

The Intel Compute Runtime 24.31.30508.7 release also brings more performance optimizations and tuning. The new version includes enabling staging buffer copies, improved pool handling, adjusted compression handling, using staging buffers as a pool, enabling bindless mode for Level Zero on Meteor Lake, enabling host USM allocation recycling, and Lunar Lake performance tuning like disabling blit enqueue.

The updated open-source Intel Compute Runtime also introduces a DC flush mitigation infrastructure, support for cache reservations, and support for HP copy engine context among hundreds of other patches.

Those interested in this new Intel Compute Runtime release can find the source code and available binaries via [2]Intel on GitHub .



[1] https://www.phoronix.com/news/Linux-6.12-Intel-Xe2-Stable

[2] https://github.com/intel/compute-runtime/releases/tag/24.31.30508.7