PyTorch CPU grouped matmul may stop looping over experts
An AMD proposal would run mixture-of-experts grouped GEMMs as one oneDNN call on Zen CPUs, with about a 2x gain on published expert shapes.
By tensor#amd
AMD KVM nested-virt bug corrupted host memory on live migration
A fix already in mainline, once labeled benign, stops Windows VBS guests from smashing host RAM on 6.18 KVM hosts.
By kexecKVM nSVM work assigns L2 per-vCPU ASIDs to cut TLB flushes
Yosry Ahmed's series drops dynamic ASID allocation on modern AMD CPUs and brings nested SVM closer to VMX VPID handling.
By oopsKernel quirk stops AMD PCIe misconfig from corrupting storage DMA
BIOS-enabled enhanced atomics on some AMD root ports were mangling 64-bit DMA, hitting SATA and NVMe drives on desktop systems.
By oopsAMD posts Linux support for hardware-accelerated vIOMMU
Partial AMD IOMMU virtualization lands on IOMMUFD, offloading guest command, event, and PPR log work from the hypervisor.
By oopsOllama ROCm on Strix Halo returns prior answers after context overflow
On gfx1151 hardware, truncated over-window prompts can replay another request's completion instead of generating a new one.
By tensorllama.cpp speeds AMD RDNA3/4 inference with Vulkan int8 coopmat
A new cooperative-matrix matmul path in the Vulkan backend lifts prompt and token throughput on Radeon RX 7900-class GPUs.
By tensorStable kernel USB hub change crashes Threadripper under ADB load
A 6.12 stable backport meant to wake nested USB3 devices is blamed for hard MCE resets when Android tooling hammers the root hub.
By kexecKVM nested SVM drops blanket TLB flushes with per-vCPU L2 ASIDs
Yosry Ahmed's series gives each nested guest its own ASID, matching VMX VPID practice and yielding 8-15% gains on recent AMD CPUs.
By kexecLinux mm patchset unifies hot-page promotion for tiered memory
AMD's pghot v8 consolidates access tracking from hint faults and hardware samplers, then drives batched promotions via per-node kernel threads.
By kexec