1//===- AMDGPUArchByKFD.cpp - list AMDGPU installed ------*- C++ -*---------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file implements a tool for detecting name of AMD GPUs installed in
10// system using the Linux sysfs interface for the AMD KFD driver. This file does
11// not respect ROCR_VISIBLE_DEVICES like the ROCm environment would.
12//
13//===----------------------------------------------------------------------===//
14
15#include "llvm/Support/FileSystem.h"
16#include "llvm/Support/Format.h"
17#include "llvm/Support/LineIterator.h"
18#include "llvm/Support/MemoryBuffer.h"
19#include "llvm/Support/Path.h"
20#include "llvm/Support/raw_ostream.h"
21#include <memory>
22#include <tuple>
23
24using namespace llvm;
25
26constexpr static const char *KFD_SYSFS_NODE_PATH =
27 "/sys/devices/virtual/kfd/kfd/topology/nodes";
28constexpr static long GFX1250_VERSION = 120500;
29
30// See the ROCm implementation for how this is handled.
31// https://github.com/ROCm/ROCT-Thunk-Interface/blob/master/src/libhsakmt.h#L126
32constexpr static long getMajor(long Ver) { return (Ver / 10000) % 100; }
33constexpr static long getMinor(long Ver) { return (Ver / 100) % 100; }
34constexpr static long getStep(long Ver) { return Ver % 100; }
35
36// For A0, print gfx1250-strict to match rocminfo
37static StringRef getRevisionSuffix(long GFXVersion, long ASICRevision) {
38 return (GFXVersion == GFX1250_VERSION && ASICRevision == 0) ? "-strict" : "";
39}
40
41// Exposed for testing
42int printGPUsByKFD(StringRef NodePath) {
43 struct KFDNode {
44 long Node;
45 long GFXVersion;
46 long ASICRevision;
47 };
48
49 SmallVector<KFDNode> Devices;
50 std::error_code EC;
51 sys::fs::directory_iterator Begin(NodePath, EC), End;
52
53 // Fail if the sysfs topology does not exist (e.g., WSL)
54 if (EC)
55 return 1;
56
57 for (; Begin != End; Begin.increment(ec&: EC)) {
58 if (EC)
59 return 1;
60
61 long Node = 0;
62 if (sys::path::stem(path: Begin->path()).consumeInteger(Radix: 10, Result&: Node))
63 return 1;
64
65 SmallString<0> Path(Begin->path());
66 sys::path::append(path&: Path, a: "properties");
67
68 ErrorOr<std::unique_ptr<MemoryBuffer>> BufferOrErr =
69 MemoryBuffer::getFileOrSTDIN(Filename: Path);
70 if (std::error_code EC = BufferOrErr.getError())
71 return 1;
72
73 long GFXVersion = 0;
74 uint64_t Capability = 0;
75 for (line_iterator Lines(**BufferOrErr, false); !Lines.is_at_end();
76 ++Lines) {
77 StringRef Line(*Lines);
78 if (Line.consume_front(Prefix: "gfx_target_version")) {
79 if (Line.drop_while(F: [](char C) { return std::isspace(C); })
80 .consumeInteger(Radix: 10, Result&: GFXVersion))
81 return 1;
82 // Differentiate between capability and capability2
83 } else if (Line.consume_front(Prefix: "capability") && !Line.starts_with(Prefix: '2')) {
84 if (Line.drop_while(F: [](char C) { return std::isspace(C); })
85 .consumeInteger(Radix: 10, Result&: Capability))
86 return 1;
87 }
88 }
89
90 // If this is zero the node is a CPU.
91 if (GFXVersion == 0)
92 continue;
93 // ASIC revision is bits 25:22 in capability
94 long ASICRevision = (Capability >> 22) & 0xf;
95 Devices.push_back(Elt: {.Node: Node, .GFXVersion: GFXVersion, .ASICRevision: ASICRevision});
96 }
97
98 // Sort the devices by their node to make sure it prints in order.
99 llvm::sort(C&: Devices, Comp: [](auto &L, auto &R) { return L.Node < R.Node; });
100 for (const auto &[Node, GFXVersion, ASICRevision] : Devices) {
101 outs() << "gfx" << getMajor(Ver: GFXVersion) << getMinor(Ver: GFXVersion)
102 << format_hex_no_prefix(N: getStep(Ver: GFXVersion), Width: 1)
103 << getRevisionSuffix(GFXVersion, ASICRevision) << '\n';
104 }
105
106 return 0;
107}
108
109int printGPUsByKFD() { return printGPUsByKFD(NodePath: KFD_SYSFS_NODE_PATH); }
110