Linux 6.3 Preparing Support For Arm's Scalable Matrix Extension 2 (SME2)
While initial Scalable Matrix Extension (SME) support for the Linux kernel only was mainlined last year to the kernel tree, Arm already has SME 2 and SME 2.1 support on the way to mainline.
While it's still not too common finding SVE/SVE2 Arm SoCs or even more so for SME that was announced in 2021 for Armv9-A as a superset of SVE2, Arm is already pushing ahead with mainline kernel support for SME 2/2.1. Scalable Matrix Extension Version 2 and 2.1 add the ZT0 register and new architectural state over SME Version 1 that is already supported by the mainline kernel since Linux 5.19.
There have been the Linux kernel SME 2/2.1 patches floating around the mailing list the past few months for review while now they look set for introduction in Linux 6.3.
Via ARM64's Git tree is a for-next/sme2 branch created toward the end of January and carrying the various SME 2/2.1 patches. With it now going the way of "for-next", it should be ready for sending in when the Linux 6.3 merge window opens later this month -- in two weeks. So at least the kernel-side enablement around Scalable Matrix Extension 2 is ready well in advance of SoCs with this capability.
While it's still not too common finding SVE/SVE2 Arm SoCs or even more so for SME that was announced in 2021 for Armv9-A as a superset of SVE2, Arm is already pushing ahead with mainline kernel support for SME 2/2.1. Scalable Matrix Extension Version 2 and 2.1 add the ZT0 register and new architectural state over SME Version 1 that is already supported by the mainline kernel since Linux 5.19.
There have been the Linux kernel SME 2/2.1 patches floating around the mailing list the past few months for review while now they look set for introduction in Linux 6.3.
Via ARM64's Git tree is a for-next/sme2 branch created toward the end of January and carrying the various SME 2/2.1 patches. With it now going the way of "for-next", it should be ready for sending in when the Linux 6.3 merge window opens later this month -- in two weeks. So at least the kernel-side enablement around Scalable Matrix Extension 2 is ready well in advance of SoCs with this capability.
Add A Comment