dcu.go 8.65 KB
Newer Older
liming6's avatar
liming6 committed
1
2
3
4
5
6
7
8
9
10
package backend

import (
	"context"
	"fmt"
	"get-container/cmd/hytop/lib"
	"get-container/gpu"
	"get-container/utils"
	"math"
	"os"
liming6's avatar
liming6 committed
11
	"sort"
liming6's avatar
liming6 committed
12
13
14
15
16
17
	"strconv"
	"strings"
	"sync"
	"sync/atomic"
	"time"

18
	"github.com/shirou/gopsutil/v3/process"
liming6's avatar
liming6 committed
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
)

var (
	DCUSInfoMap   *DCUInfoMap       = nil // 记录dcu信息
	DockerPidInfo *DockerProcessMap = nil
	User                            = ""
	HostName                        = ""
	Rocmlib                         = lib.Rocmlib_instance
	DriverVersion                   = ""
	stopCtx       context.Context
	cancelFunc    context.CancelFunc
)

const (
	ENVUser = "USER"
)

// Init 初始化本包的数据
func Init() error {
	HostName, _ = os.Hostname()
	uid := os.Getuid()
	username, seted := os.LookupEnv(ENVUser)
	if !seted {
		User = strconv.Itoa(uid)
	} else {
		User = username
	}
	DCUSInfoMap = &DCUInfoMap{}
	DCUSInfoMap.qinfo = make(map[int]*DCUQuickInfo)
	DCUSInfoMap.sinfo = make(map[int]*DCUSlowInfo)
	Rocmlib = lib.GetRocmlib()
	_, err := Rocmlib.Init()
	if err != nil {
		return err
	}
	DriverVersion, err = Rocmlib.GetSystemDriverVersion()
	if err != nil {
		return err
	}
	_, err = Rocmlib.GetDevNumber()
	DockerPidInfo = &DockerProcessMap{}
	DockerPidInfo.Update()

	if err == nil {
		stopCtx, cancelFunc = context.WithCancel(context.Background())
		go func() {
			ticker := time.NewTicker(time.Second * 20)
			for {
				select {
				case <-ticker.C:
					DockerPidInfo.Update()
					DCUSInfoMap.UpdateSlowInfo()
				case <-stopCtx.Done():
					return
				}
			}
		}()
	}
	return err
}

func Shutdown() {
	Rocmlib.Shutdown()
	cancelFunc()
}

type DCUQuickInfo struct {
	lock             sync.RWMutex
	Id               int     //
	Name             string  //
	PerformanceLevel string  //
	Fan              string  //
	Temp             float32 //
	PwrAvg           float32 //
	PwrCap           float32 //
	BusId            string  //
	MemTotal         uint64  //
	MemUsed          uint64  //
	MemUsedPerent    float32 //
	DCUUTil          float32 //
}

type DCUSlowInfo struct {
	Id      atomic.Int32
	Ecc     atomic.Bool
	PwrMode atomic.Value // string
	Mig     atomic.Bool
}

type DCUInfo struct {
	SlowInfo  DCUSlowInfo
	QuickInfo DCUQuickInfo
}

type DCUInfoMap struct {
	qinfolock sync.RWMutex
	qinfo     map[int]*DCUQuickInfo
	sinfolock sync.RWMutex
	sinfo     map[int]*DCUSlowInfo
}

func (m *DCUInfoMap) UpdateQuickInfo() error {
	num, err := Rocmlib.GetDevNumber()
	if err != nil {
		return err
	}
	names, err := Rocmlib.GetDevName()
	if err != nil {
		return err
	}
	plevel, err := Rocmlib.GetPerfLevel()
	if err != nil {
		return err
	}
	fan, err := Rocmlib.GetFanSpeed()
	if err != nil {
		return err
	}
	temp, err := Rocmlib.GetTemp()
	if err != nil {
		return err
	}
	pwrAvg, err := Rocmlib.GetPowerAvg()
	if err != nil {
		return err
	}
	pwrCap, err := Rocmlib.GetPowerCap()
	if err != nil {
		return err
	}
	busid, err := Rocmlib.GetPCIBusId()
	if err != nil {
		return err
	}
	memTotal, err := Rocmlib.GetMemTotal()
	if err != nil {
		return err
	}
	memUsed, err := Rocmlib.GetMemUsed()
	if err != nil {
		return err
	}
	dcu, err := Rocmlib.GetBusyPercent()
	if err != nil {
		return err
	}

	set := make(map[int]bool)

	m.qinfolock.Lock()
	defer m.qinfolock.Unlock()

	for i := range num {
		qinfo, have := m.qinfo[i]
		if !have {
			qinfo = &DCUQuickInfo{}
			m.qinfo[i] = qinfo
		}
		qinfo.lock.Lock()
		qinfo.Id = i
		qinfo.Name = names[i]
		qinfo.PerformanceLevel = plevel[i]
		if rpm, have := fan[i]; !have || rpm == 0 {
			qinfo.Fan = "N/A"
		} else {
			qinfo.Fan = strconv.Itoa(int(rpm))
		}
		qinfo.Temp = float32(temp[i]) / 1000
		qinfo.PwrAvg = float32(pwrAvg[i]) / 1000000
		qinfo.PwrCap = float32(pwrCap[i]) / 1000000
		qinfo.BusId = busid[i]
		qinfo.MemTotal = memTotal[i]
		qinfo.MemUsed = memUsed[i]
		if qinfo.MemTotal == 0 {
			qinfo.MemUsedPerent = 0
		} else {
			qinfo.MemUsedPerent = float32(qinfo.MemUsed) / float32(qinfo.MemTotal) * 100
		}
		qinfo.DCUUTil = float32(dcu[i])
		qinfo.lock.Unlock()
		set[i] = true
	}
	for k := range m.qinfo {
		a, b := set[k]
		if a && b {
			continue
		}
		delete(m.qinfo, k)
	}
	return nil
}

func (m *DCUInfoMap) UpdateSlowInfo() error {
	num, err := Rocmlib.GetDevNumber()
	if err != nil {
		return err
	}
	ecc, err := gpu.GetEccInfo()
	if err != nil {
		return err
	}
	rinfo, err := gpu.GetRunningInfo()
	if err != nil {
		return err
	}
	set := make(map[int]bool)
	m.sinfolock.Lock()
	defer m.sinfolock.Unlock()
	for i := range num {
		sinfo, have := m.sinfo[i]
		if !have {
			sinfo = &DCUSlowInfo{}
			m.sinfo[i] = sinfo
		}
		sinfo.Id.Store(int32(i))
		sinfo.Mig.Store(false)
		if r, have := rinfo[i]; have {
			sinfo.PwrMode.Store(r.Mode)
		} else {
			sinfo.PwrMode.Store("Normal")
		}
		e, have := ecc[i]
		if have {
			sinfo.Ecc.Store(e)
		} else {
			sinfo.Ecc.Store(false)
		}
		set[i] = true
	}
	for k := range m.sinfo {
		a, b := set[k]
		if a && b {
			continue
		}
		delete(m.sinfo, k)
	}
	return nil
}

// GetSlowInfo 获取慢更新信息,读取完一定要释放锁
func (m *DCUInfoMap) GetSlowInfo() (map[int]*DCUSlowInfo, sync.Locker) {
	rl := m.sinfolock.RLocker()
	rl.Lock()
	return m.sinfo, rl
}

// GetQuitInfo 获取快更新信息,读取完一定要释放锁
func (m *DCUInfoMap) GetQuitInfo() (map[int]*DCUQuickInfo, sync.Locker) {
	rl := m.qinfolock.RLocker()
	rl.Lock()
	return m.qinfo, rl
}

type DCUProcessInfo struct {
liming6's avatar
liming6 committed
273
274
	DCU    int              // 使用的dcu号
	DCUMem utils.MemorySize // 使用的dcu内存容量
liming6's avatar
liming6 committed
275
276
277
278
279
	SDMA   int
	Info   ProcessInfo // 通用进程信息
}

type ProcessInfo struct {
liming6's avatar
liming6 committed
280
	Ppid     int32   // 父进程id
liming6's avatar
liming6 committed
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
	Pid      int32   // 进程号
	User     string  // 用户名或uid
	CPU      float64 // CPU使用率
	Mem      float32 // 内存使用率
	Time     string  // 占用的CPU时间
	Cmd      string  // 命令
	ContInfo *ContainerInfo
}

func getProcessInfo(pids []int32) map[int32]ProcessInfo {
	result := make(map[int32]ProcessInfo)
	if len(pids) == 0 {
		return result
	}
	dockerInfo, lock := DockerPidInfo.Get()
	defer lock.Unlock()
	for _, pid := range pids {
		p, err := process.NewProcess(int32(pid))
		if err != nil {
			continue
		}
		item := ProcessInfo{Pid: p.Pid}
		item.User, _ = p.Username()
		item.CPU, _ = p.CPUPercent()
		item.Mem, _ = p.MemoryPercent()
		t, err := p.Times()
		if err == nil {
liming6's avatar
liming6 committed
308
			item.Time = DurationStr(time.Duration(t.System+t.User) * time.Second)
liming6's avatar
liming6 committed
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
		}
		item.Cmd, _ = p.Cmdline()
		d, have := dockerInfo[item.Pid]
		if have {
			item.ContInfo = d
		}
		result[p.Pid] = item
	}
	return result
}

// GetDCUProcessInfo 返回值的key为dcu index
func (m *DCUInfoMap) GetDCUProcessInfo() map[int][]DCUProcessInfo {
	result := make(map[int][]DCUProcessInfo)
	mem := utils.MemorySize{}
	mem.Unit = utils.Byte
	info, err := Rocmlib.GetProcessInfo()
	if err != nil {
		return result
	}
	pids := make([]int32, 0)
	for _, v := range info {
		pids = append(pids, int32(v.Pid))
	}
	pinfo := getProcessInfo(pids)
	for _, v := range info {
		index := make([]int, 0)
		for _, i := range v.UsedGPUIndex {
			index = append(index, int(i))
		}
		for _, i := range index {
			l, have := result[i]
			if !have {
				result[i] = make([]DCUProcessInfo, 0)
				l = result[i]
			}
			item := DCUProcessInfo{DCU: i}
			item.Info = pinfo[int32(v.Pid)]
			mem.Num = v.VarmUsage
liming6's avatar
liming6 committed
348
			item.DCUMem = mem
liming6's avatar
liming6 committed
349
350
351
352
353
354
355
356
			item.SDMA = int(v.SdmaUsage)
			l = append(l, item)
			result[i] = l
		}
	}
	return result
}

357
358
359
360
361
362
363
364
365
366
367
368
// GetDCUProcessInfo2 返回值的key为dcu index
func (m *DCUInfoMap) GetDCUProcessInfo2() map[int][]DCUProcessInfo {
	result := make(map[int][]DCUProcessInfo)
	info, err := gpu.GetDCUPidInfo()
	if err != nil {
		return result
	}
	pids := make([]int32, 0)
	for _, v := range info {
		pids = append(pids, int32(v.Pid))
	}
	pinfo := getProcessInfo(pids)
liming6's avatar
liming6 committed
369
370
	for _, v := range info { // 按照进程遍历
		// 获取进程使用的所有dcu的索引
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
		index := make([]int, 0)
		for _, i := range v.HCUIndex {
			ii, err := strconv.Atoi(i)
			if err != nil {
				continue
			}
			index = append(index, ii)
		}
		for _, i := range index {
			l, have := result[i]
			if !have {
				result[i] = make([]DCUProcessInfo, 0)
				l = result[i]
			}
			item := DCUProcessInfo{DCU: i}
			item.Info = pinfo[int32(v.Pid)]
liming6's avatar
liming6 committed
387
			item.DCUMem = v.VRamUsed
388
389
			item.SDMA = v.SDMAUsed
			l = append(l, item)
liming6's avatar
liming6 committed
390

391
392
393
			result[i] = l
		}
	}
liming6's avatar
liming6 committed
394
395
396
397
398
399
400
	// 每个切片对pid排序
	for k, v := range result {
		sort.SliceStable(v, func(i, j int) bool {
			return v[i].Info.Pid < v[j].Info.Pid
		})
		result[k] = v
	}
401
402
403
	return result
}

liming6's avatar
liming6 committed
404
405
// DurationStr 将时间段格式化为 小时:分钟:秒 的格式
func DurationStr(d time.Duration) string {
liming6's avatar
liming6 committed
406
407
408
	h := int(math.Floor(d.Hours()))
	m := int(d.Minutes()) % 60
	s := int(math.Floor(d.Seconds())) % 60
liming6's avatar
liming6 committed
409
410
411
412
413
	if h <= 96 {
		return strings.Replace(fmt.Sprintf("%d:%2d:%2d", h, m, s), " ", "0", -1)
	} else {
		return fmt.Sprintf("%.1f days", d.Hours()/24)
	}
liming6's avatar
liming6 committed
414
}