Debian Codegraph @

drivers/accel/habanalabs/Kconfig

raw
1 # SPDX-License-Identifier: GPL-2.0-only
2 #
3 # HabanaLabs AI accelerators driver
4 #
5
6 config DRM_ACCEL_HABANALABS
7 tristate "HabanaLabs AI accelerators"
8 depends on DRM_ACCEL
9 depends on X86 && X86_64
10 depends on PCI && HAS_IOMEM
11 select GENERIC_ALLOCATOR
12 select HWMON
13 select DMA_SHARED_BUFFER
14 select CRC32
15 select FW_LOADER
16 help
17 Enables PCIe card driver for Habana's AI Processors (AIP) that are
18 designed to accelerate Deep Learning inference and training workloads.
19
20 The driver manages the PCIe devices and provides IOCTL interface for
21 the user to submit workloads to the devices.
22
23 The user-space interface is described in
24 include/uapi/drm/habanalabs_accel.h
25
26 If unsure, say N.
27
28 To compile this driver as a module, choose M here: the
29 module will be called habanalabs.
30
31 if DRM_ACCEL_HABANALABS
32
33 config HL_HLDIO
34 bool "Habanalabs NVMe Direct I/O (HLDIO)"
35 depends on PCI_P2PDMA
36 depends on BLOCK
37 help
38 Enable NVMe peer-to-peer direct I/O support for Habanalabs AI
39 accelerators.
40
41 This allows direct data transfers between NVMe storage devices
42 and Habanalabs accelerators without involving system memory,
43 using PCI peer-to-peer DMA capabilities.
44
45 Requirements:
46 - CONFIG_PCI_P2PDMA=y
47 - NVMe device and Habanalabs accelerator under same PCI root complex
48 - IOMMU disabled or in passthrough mode
49 - Hardware supporting PCI P2P DMA
50
51 If unsure, say N
52 endif # DRM_ACCEL_HABANALABS