0001 # SPDX-License-Identifier: MIT
0002 #
0003 # Heterogeneous system architecture configuration
0004 #
0005
0006 config HSA_AMD
0007 bool "HSA kernel driver for AMD GPU devices"
0008 depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64)
0009 imply AMD_IOMMU_V2 if X86_64
0010 select HMM_MIRROR
0011 select MMU_NOTIFIER
0012 select DRM_AMDGPU_USERPTR
0013 help
0014 Enable this if you want to use HSA features on AMD GPU devices.
0015
0016 config HSA_AMD_SVM
0017 bool "Enable HMM-based shared virtual memory manager"
0018 depends on HSA_AMD && DEVICE_PRIVATE
0019 default y
0020 select HMM_MIRROR
0021 select MMU_NOTIFIER
0022 help
0023 Enable this to use unified memory and managed memory in HIP. This
0024 memory manager supports two modes of operation. One based on
0025 preemptions and one based on page faults. To enable page fault
0026 based memory management on most GFXv9 GPUs, set the module
0027 parameter amdgpu.noretry=0.
0028
0029 config HSA_AMD_P2P
0030 bool "HSA kernel driver support for peer-to-peer for AMD GPU devices"
0031 depends on HSA_AMD && PCI_P2PDMA && DMABUF_MOVE_NOTIFY
0032 help
0033 Enable peer-to-peer (P2P) communication between AMD GPUs over
0034 the PCIe bus. This can improve performance of multi-GPU compute
0035 applications and libraries by enabling GPUs to access data directly
0036 in peer GPUs' memory without intermediate copies in system memory.
0037
0038 This P2P feature is only enabled on compatible chipsets, and between
0039 GPUs with large memory BARs that expose the entire VRAM in PCIe bus
0040 address space within the physical address limits of the GPUs.
0041
