hy-smi.go 19.9 KB
Newer Older
liming6's avatar
liming6 committed
1
2
3
4
5
6
7
package gpu

import (
	"encoding/json"
	"get-container/utils"
	"os/exec"
	"regexp"
8
	"sort"
liming6's avatar
liming6 committed
9
10
11
12
	"strconv"
	"strings"
)

13
14
15
16
17
18
19
20
21
22
23
24
const (
	DCUBinaryFile = "hy-smi"

	PIDHeader             = "PID"
	PASIDHeader           = "PASID"
	HCUNodeHeader         = "HCU Node(Include CPU sort)"
	HCUIndexHeader        = "HCU Index"
	GPUIDHeader           = "GPUID"
	PCIBusHeader          = "PCI BUS"
	VRamUsedHeader        = "VRAM USED(MiB)"
	VRamUsedPercentHeader = "VRAM USED(%)"
	SDMAUsedHeader        = "SDMA USED"
liming6's avatar
liming6 committed
25
26

	NA = "N/A"
27
28
29
30
31
32
)

var (
	ReEmptyLine   = regexp.MustCompile(`^\s*$`)
	ReUselessLine = regexp.MustCompile(`^=[ =a-zA-Z0-9]*=$`)

liming6's avatar
liming6 committed
33
	ReInfoHeader     = regexp.MustCompile(`^HCU\[(\d+)]\s*:\s*(.*)$`)
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
	ReDriId          = regexp.MustCompile(`(?mi)^Device\s*ID\s*:\s*([x0-9]*)$`)
	ReVBIOSVersion   = regexp.MustCompile(`(?mi)^VBIOS\s*version\s*:\s*([0-9a-zA-Z.]*)$`)
	ReTempEdge       = regexp.MustCompile(`(?mi)^Temperature\s*\(Sensor\s*edge\)\s*\(C\)\s*:\s*([0-9.]*)$`)
	ReTempJunction   = regexp.MustCompile(`(?mi)^Temperature\s*\(Sensor\s*junction\)\s*\(C\)\s*:\s*([0-9.]*)$`)
	ReTempMem        = regexp.MustCompile(`(?mi)^Temperature\s*\(Sensor\s*mem\)\s*\(C\)\s*:\s*([0-9.]*)$`)
	ReTempCore       = regexp.MustCompile(`(?mi)^Temperature\s*\(Sensor\s*core\)\s*\(C\)\s*:\s*([0-9.]*)$`)
	ReFClk           = regexp.MustCompile(`(?mi)^fclk\s*clock\s*level\s*:\s*([0-9]*)\s*\(([0-9a-zA-Z]*)\)$`)
	ReMClk           = regexp.MustCompile(`(?mi)^mclk\s*clock\s*level\s*:\s*([0-9]*)\s*\(([0-9a-zA-Z]*)\)$`)
	ReSClk           = regexp.MustCompile(`(?mi)^sclk\s*clock\s*level\s*:\s*([0-9]*)\s*\(([0-9a-zA-Z]*)\)$`)
	ReSOCClk         = regexp.MustCompile(`(?mi)^socclk\s*clock\s*level\s*:\s*([0-9]*)\s*\(([0-9a-zA-Z]*)\)$`)
	RePCIClk         = regexp.MustCompile(`(?mi)^pcie\s*clock\s*level\s*([0-9]*)\s*\(([0-9.a-zA-Z/]*)\s*,\s*([x0-9]*)\s([0-9a-zA-Z]*)\)$`)
	RePreLevel       = regexp.MustCompile(`(?mi)^Performance\s*Level\s*:\s*([0-9a-zA-Z]*)$`)
	ReMaxPwr         = regexp.MustCompile(`(?mi)^Max\s*Graphics\s*Package\s*Power\s*\((.*)\)\s*:\s*([0-9.]*)$`)
	ReAvgPwr         = regexp.MustCompile(`(?mi)^Average\s*Graphics\s*Package\s*Power\s*\((.*)\)\s*:\s*([0-9.]*)$`)
	ReHCUUsage       = regexp.MustCompile(`(?mi)^HCU\s*use\s*\(.*\)\s*:\s*([0-9.]*)$`)
	ReHCUMemUsage    = regexp.MustCompile(`(?mi)^HCU\s*memory\s*use\s*\(.*\)\s*:\s*([0-9.]*)$`)
	ReHCUMemVendor   = regexp.MustCompile(`(?mi)^HCU\s*Memory\s*Vendor\s*:\s*(.*)$`)
	RePCIeRelay      = regexp.MustCompile(`(?mi)^PCIe\s*Replay\s*Count\s*:\s*([0-9]*)$`)
	ReSerialNum      = regexp.MustCompile(`(?mi)^Serial\s*Number\s*:\s*([0-9a-zA-Z]*)$`)
	ReVoltage        = regexp.MustCompile(`(?mi)^Voltage\s*\((.*)\)\s*:\s*([0-9.]*)$`)
liming6's avatar
liming6 committed
54
	RePCIBus         = regexp.MustCompile(`(?mi)^PCI\s*Bus\s*:\s*([0-9a-zA-Z.:]*).*$`)
55
56
57
58
59
60
61
62
	ReMECFWVersion   = regexp.MustCompile(`(?mi)^MEC\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReMEC2FWVersion  = regexp.MustCompile(`(?mi)^MEC2\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReRLCFWVersion   = regexp.MustCompile(`(?mi)^RLC\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReSDMAFWVersion  = regexp.MustCompile(`(?mi)^SDMA\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReSDMA2FWVersion = regexp.MustCompile(`(?mi)^SDMA2\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReSMCFWVersion   = regexp.MustCompile(`(?mi)^SMC\s*Firmware\s*Version\s*:\s*([0-9.]*)$`)
	ReCardSerial     = regexp.MustCompile(`(?mi)^Card\s*Series\s*:\s*(.*)$`)
	ReCardVendor     = regexp.MustCompile(`(?mi)^Card\s*Vendor\s*:\s*(.*)$`)
liming6's avatar
liming6 committed
63
64
65

	ReVersion    = regexp.MustCompile(`(?i)^version\s*([0-9a-zA-Z.]*)\s*\(.*\)$`)
	ReDriVersion = regexp.MustCompile(`(?mi)^driver\s*version\s*:\s*(.*)$`)
liming6's avatar
liming6 committed
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83

	ReEccOn  = regexp.MustCompile(`(?i)ecc\s*on`)
	ReEccOff = regexp.MustCompile(`(?i)ecc\s*off`)

	ReLowPower = regexp.MustCompile(`(?i)Low\s*Power`)

	HeaderMap = map[string]string{
		"Id":               "HCU",
		"Temp":             "Temp",
		"AvgPower":         "AvgPwr",
		"PerformanceLevel": "Perf",
		"MemPerc":          "VRAM%",
		"HCUPerc":          "HCU%",
		"Mode":             "Mode",
	}

	ReVramTotal = regexp.MustCompile(`(?i)total\s*memory\s*\((.*)\)\s*:\s*([0-9]*)`)
	ReVramUsed  = regexp.MustCompile(`(?i)used\s*memory\s*\((.*)\)\s*:\s*([0-9]*)`)
84
85
)

liming6's avatar
liming6 committed
86
type HYVersionInfo struct {
87
88
89
90
91
92
93
94
95
96
97
98
99
	SMIVersion    string // --version
	DriverVersion string // --showdriverversion
}

func GetHYVersionInfo() (*HYVersionInfo, error) {
	versionBytes, err := exec.Command(DCUBinaryFile, "--version").Output()
	if err != nil {
		return nil, err
	}
	driBytes, err := exec.Command(DCUBinaryFile, "--showdriverversion").Output()
	if err != nil {
		return nil, err
	}
liming6's avatar
liming6 committed
100
101
102
103
104
105
106
107
108
109
	result := HYVersionInfo{}
	mv := ReVersion.FindStringSubmatch(strings.TrimSpace(strings.Trim(string(versionBytes), "\n")))
	if len(mv) >= 2 {
		result.SMIVersion = mv[1]
	}
	mrv := ReDriVersion.FindStringSubmatch(strings.TrimSpace(strings.Trim(string(driBytes), "\n")))
	if len(mrv) >= 2 {
		result.DriverVersion = mrv[1]
	}
	return &result, nil
liming6's avatar
liming6 committed
110
111
112
}

type DCUInfo struct {
113
114
115
116
	Id               int     // id
	Name             string  // DCU名称
	PerformanceLevel string  // 性能等级
	FanSpeed         float32 // 风扇转速
liming6's avatar
liming6 committed
117
	Temperature      float32 // 平均温度
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
	PwrUsage         int16
	PwrCapacity      int16
	BusId            string
	MemTotal         int32
	MemUsed          int32
}

type ClockInfo struct {
	Level int    // 时钟等级
	Freq  string // 频率
}

type PcieClockInfo struct {
	Level     int    // 时钟等级
	Freq      string // 频率
	BandWidth string // 带宽
	Times     string // 倍率
}

// SMIAllOutput hy-smi -a输出的信息,列出了DCU全面的信息
type SMIAllOutput struct {
	Id             int
	DeviceId       string
	VBIOSVersion   string
	TempEdge       float32
	TempJunction   float32
	TempMem        float32
	TempCores      float32
	FClock         *ClockInfo
	MClock         *ClockInfo
	SClock         *ClockInfo
	SOCClock       *ClockInfo
	PCIEClock      *PcieClockInfo
	PerLevel       string
	MaxPwr         float32
	AvgPwr         float32
	HCUUsage       float32
	HCUMemUsage    float32
	HCUMemVendor   string
	PCIERelayCount int
	SerialNumber   string
	Voltage        float32
	PCIBus         string
	MECFWVersion   string
	MEC2FWVersion  string
	RLCFWVersion   string
	SDMAVersion    string
	SDMA2Version   string
	SMCVersion     string
	CardSeries     string
	CardVendor     string
}

liming6's avatar
liming6 committed
171
func GetSMIAllOutput() (map[int]*SMIAllOutput, error) {
liming6's avatar
liming6 committed
172
	b, err := exec.Command(DCUBinaryFile, "-a").CombinedOutput()
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
	if err != nil {
		return nil, err
	}
	lines := strings.Split(strings.Trim(strings.TrimSpace(string(b)), "\n"), "\n")
	info := make(map[int][]string)
	for _, line := range lines {
		if ReUselessLine.MatchString(line) || ReEmptyLine.MatchString(line) {
			continue
		}
		if ReInfoHeader.MatchString(line) {
			fields := ReInfoHeader.FindStringSubmatch(strings.TrimSpace(strings.ReplaceAll(line, "\t", " ")))
			if len(fields) < 2 {
				continue
			}
			id, innerErr := strconv.Atoi(fields[1])
			if innerErr != nil {
				return nil, innerErr
			}
			if v, ok := info[id]; !ok {
				info[id] = make([]string, 0)
				info[id] = append(v, fields[2])
			} else {
				info[id] = append(v, fields[2])
			}
		}
	}
liming6's avatar
liming6 committed
199
	result := make(map[int]*SMIAllOutput, 0)
200
201
202
203
204
205
	for k, v := range info {
		item, innerErr := parseSMIAllOutput(k, strings.Join(v, "\n"))
		if innerErr != nil {
			return nil, innerErr
		}
		if item != nil {
liming6's avatar
liming6 committed
206
			result[item.Id] = item
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
273
274
275
276
277
278
279
280
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
308
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
		}
	}
	return result, nil
}

func parseSMIAllOutput(id int, str string) (*SMIAllOutput, error) {
	if len(strings.TrimSpace(str)) == 0 {
		return nil, nil
	}
	result := SMIAllOutput{}
	result.Id = id
	if s := regMatch(ReDriId, str, 1); s != nil {
		result.DeviceId = s[0]
	}
	if s := regMatch(ReVBIOSVersion, str, 1); s != nil {
		result.VBIOSVersion = s[0]
	}
	if s := regMatch(ReTempEdge, str, 1); s != nil {
		if t, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32); err == nil {
			result.TempEdge = float32(t)
		}
	}
	if s := regMatch(ReTempJunction, str, 1); s != nil {
		if t, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32); err == nil {
			result.TempJunction = float32(t)
		}
	}
	if s := regMatch(ReTempMem, str, 1); s != nil {
		if t, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32); err == nil {
			result.TempMem = float32(t)
		}
	}
	if s := regMatch(ReTempCore, str, 1); s != nil {
		if t, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32); err == nil {
			result.TempCores = float32(t)
		}
	}
	if s := regMatch(ReFClk, str, 1, 2); s != nil {
		c := ClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.Freq = strings.TrimSpace(s[1])
		result.FClock = &c
	}
	if s := regMatch(ReMClk, str, 1, 2); s != nil {
		c := ClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.Freq = strings.TrimSpace(s[1])
		result.MClock = &c
	}
	if s := regMatch(ReSClk, str, 1, 2); s != nil {
		c := ClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.Freq = strings.TrimSpace(s[1])
		result.SClock = &c
	}
	if s := regMatch(ReSOCClk, str, 1, 2); s != nil {
		c := ClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.Freq = strings.TrimSpace(s[1])
		result.SOCClock = &c
	}
	if s := regMatch(RePCIClk, str, 1, 2); s != nil {
		c := ClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.Freq = strings.TrimSpace(s[1])
		result.SOCClock = &c
	}
	if s := regMatch(RePCIClk, str, 1, 2, 3, 4); s != nil {
		c := PcieClockInfo{}
		level, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			c.Level = level
		}
		c.BandWidth = strings.TrimSpace(s[1])
		c.Times = strings.TrimSpace(s[2])
		c.Freq = strings.TrimSpace(s[3])
		result.PCIEClock = &c
	}
	if s := regMatch(RePreLevel, str, 1); s != nil {
		result.PerLevel = s[0]
	}
	if s := regMatch(ReMaxPwr, str, 2); s != nil {
		p, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32)
		if err == nil {
			result.MaxPwr = float32(p)
		}
	}
	if s := regMatch(ReAvgPwr, str, 2); s != nil {
		p, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32)
		if err == nil {
			result.AvgPwr = float32(p)
		}
	}
	if s := regMatch(ReHCUUsage, str, 1); s != nil {
		p, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32)
		if err == nil {
			result.HCUUsage = float32(p)
		}
	}
	if s := regMatch(ReHCUMemUsage, str, 1); s != nil {
		p, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32)
		if err == nil {
			result.HCUMemUsage = float32(p)
		}
	}
	if s := regMatch(ReHCUMemVendor, str, 1); s != nil {
		result.HCUMemVendor = s[0]
	}
	if s := regMatch(RePCIeRelay, str, 1); s != nil {
		i, err := strconv.Atoi(strings.TrimSpace(s[0]))
		if err == nil {
			result.PCIERelayCount = i
		}
	}
	if s := regMatch(ReSerialNum, str, 1); s != nil {
		result.SerialNumber = s[0]
	}
	if s := regMatch(ReVoltage, str, 2); s != nil {
		p, err := strconv.ParseFloat(strings.TrimSpace(s[0]), 32)
		if err == nil {
			result.Voltage = float32(p)
		}
	}
liming6's avatar
liming6 committed
345
	if s := regMatch(RePCIBus, str, 1); s != nil {
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
		result.PCIBus = s[0]
	}
	if s := regMatch(ReMECFWVersion, str, 1); s != nil {
		result.MECFWVersion = s[0]
	}
	if s := regMatch(ReMEC2FWVersion, str, 1); s != nil {
		result.MEC2FWVersion = s[0]
	}
	if s := regMatch(ReRLCFWVersion, str, 1); s != nil {
		result.RLCFWVersion = s[0]
	}
	if s := regMatch(ReSDMAFWVersion, str, 1); s != nil {
		result.SDMAVersion = s[0]
	}
	if s := regMatch(ReSDMA2FWVersion, str, 1); s != nil {
		result.SDMA2Version = s[0]
	}
	if s := regMatch(ReSMCFWVersion, str, 1); s != nil {
		result.SMCVersion = s[0]
	}
	if s := regMatch(ReCardSerial, str, 1); s != nil {
		result.CardSeries = s[0]
	}
	if s := regMatch(ReCardVendor, str, 1); s != nil {
		result.CardVendor = s[0]
	}
	return &result, nil
}

// DCURunningInfo DCU运行状态信息
type DCURunningInfo struct {
	Id               int
	Temp             float32
	AvgPower         float32
	PerformanceLevel string
	MemPerc          float32
	HCUPerc          float32
liming6's avatar
liming6 committed
383
	Mode             string // Normal / Low Power
384
385
386
}

// GetRunningInfo 获取DCU运行相关信息
liming6's avatar
liming6 committed
387
func GetRunningInfo() (map[int]DCURunningInfo, error) {
liming6's avatar
liming6 committed
388
	output, err := exec.Command(DCUBinaryFile).CombinedOutput()
389
390
391
392
393
394
	if err != nil {
		return nil, err
	}
	return parseRunningInfo(string(output))
}

liming6's avatar
liming6 committed
395
func parseRunningInfo(info string) (map[int]DCURunningInfo, error) {
396
	lines := strings.Split(strings.Trim(strings.TrimSpace(info), "\n"), "\n")
liming6's avatar
liming6 committed
397
398
	result := make(map[int]DCURunningInfo)
	usefulLines := make([]string, 0)
399
400
401
402
	for _, line := range lines {
		if ReUselessLine.MatchString(line) || ReEmptyLine.MatchString(line) {
			continue
		}
liming6's avatar
liming6 committed
403
404
405
406
407
408
409
410
411
412
413
		usefulLines = append(usefulLines, line)
	}
	fields := strings.Fields(usefulLines[0])
	fieldMap := make(map[string]int)
	fieldNum := 0
	for i, k := range fields {
		fieldMap[k] = i
		fieldNum++
	}
	for _, line := range usefulLines {
		line = ReLowPower.ReplaceAllString(line, "LowPower")
414
		fields := strings.Fields(strings.TrimSpace(line))
liming6's avatar
liming6 committed
415
		if len(fields) < fieldNum {
416
417
418
419
420
421
422
423
			continue
		}
		item := DCURunningInfo{}
		id, err := strconv.Atoi(fields[0])
		if err != nil {
			continue
		}
		item.Id = id
liming6's avatar
liming6 committed
424
		temp, err := strconv.ParseFloat(strings.TrimSuffix(strings.ToLower(fields[fieldMap[HeaderMap["Temp"]]]), "c"), 32)
425
426
427
428
429
		if err != nil {
			return nil, err
		}
		item.Temp = float32(temp)

liming6's avatar
liming6 committed
430
		avgPwr, err := strconv.ParseFloat(strings.TrimSuffix(strings.ToLower(fields[fieldMap[HeaderMap["AvgPower"]]]), "w"), 32)
431
432
433
434
435
		if err != nil {
			return nil, err
		}
		item.AvgPower = float32(avgPwr)
		item.PerformanceLevel = fields[3]
liming6's avatar
liming6 committed
436
		vram, err := strconv.ParseFloat(strings.TrimSuffix(fields[fieldMap[HeaderMap["MemPerc"]]], "%"), 32)
437
438
439
440
		if err != nil {
			return nil, err
		}
		item.MemPerc = float32(vram)
liming6's avatar
liming6 committed
441
		utl, err := strconv.ParseFloat(strings.TrimSuffix(fields[fieldMap[HeaderMap["HCUPerc"]]], "%"), 32)
442
443
444
445
		if err != nil {
			return nil, err
		}
		item.HCUPerc = float32(utl)
liming6's avatar
liming6 committed
446
447
		item.Mode = fields[fieldMap[HeaderMap["Mode"]]]
		result[item.Id] = item
448
449
	}
	return result, nil
liming6's avatar
liming6 committed
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
}

type DCUPidInfo struct {
	Pid             uint64
	PASId           uint64
	HCUNode         []string
	HCUIndex        []string
	GPUID           []string
	PCIBus          []string
	VRamUsed        utils.MemorySize
	VRamUsedPercent int
	SDMAUsed        int
}

// GetDCUPidInfo 获取Pid相关信息
liming6's avatar
liming6 committed
465
// 需要剔除PCIBus为空的项
liming6's avatar
liming6 committed
466
func GetDCUPidInfo() ([]DCUPidInfo, error) {
467
	output, err := exec.Command(DCUBinaryFile, "--showpids").Output()
liming6's avatar
liming6 committed
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
	if err != nil {
		return nil, err
	}
	return parseDCUPidInfo(string(output))
}

func parseDCUPidInfo(s string) ([]DCUPidInfo, error) {
	lines := strings.Split(strings.Trim(string(s), "\n"), "\n")
	linesArray := make([][]string, 0)
	for _, line := range lines {
		if ReEmptyLine.MatchString(line) || ReUselessLine.MatchString(line) {
			continue
		}
		linesArray = append(linesArray, strings.SplitN(strings.TrimSpace(line), ":", 2))
	}
	infosArray := make([]map[string]string, 0)
	index := -1
	for _, line := range linesArray {
		if line == nil || len(line) != 2 {
			continue
		}
		line[1] = strings.TrimSpace(line[1])
		switch line[0] {
		case PIDHeader:
			index += 1
			infosArray = append(infosArray, make(map[string]string))
			infosArray[index][PIDHeader] = line[1]
		case PASIDHeader:
			infosArray[index][PASIDHeader] = line[1]
		case HCUNodeHeader:
			infosArray[index][HCUNodeHeader] = line[1]
		case HCUIndexHeader:
			infosArray[index][HCUIndexHeader] = line[1]
		case GPUIDHeader:
			infosArray[index][GPUIDHeader] = line[1]
		case PCIBusHeader:
			infosArray[index][PCIBusHeader] = line[1]
		case VRamUsedHeader:
			infosArray[index][VRamUsedHeader] = line[1]
		case VRamUsedPercentHeader:
			infosArray[index][VRamUsedPercentHeader] = line[1]
		case SDMAUsedHeader:
			infosArray[index][SDMAUsedHeader] = line[1]
		}
	}
	result := make([]DCUPidInfo, 0)
	if len(infosArray) == 0 {
		return result, nil
	}
	for _, info := range infosArray {
		i := DCUPidInfo{}
liming6's avatar
liming6 committed
519
520
521
522
523
524
525
		_ = json.Unmarshal([]byte(strings.ReplaceAll(info[HCUNodeHeader], "'", `"`)), &i.HCUNode)
		_ = json.Unmarshal([]byte(strings.ReplaceAll(info[HCUIndexHeader], "'", `"`)), &i.HCUIndex)
		_ = json.Unmarshal([]byte(strings.ReplaceAll(info[GPUIDHeader], "'", `"`)), &i.GPUID)
		_ = json.Unmarshal([]byte(strings.ReplaceAll(info[PCIBusHeader], "'", `"`)), &i.PCIBus)
		if len(i.PCIBus) == 0 {
			continue
		}
liming6's avatar
liming6 committed
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
		pid, innerErr := strconv.ParseUint(info[PIDHeader], 10, 64)
		if innerErr != nil {
			return nil, innerErr
		}
		i.Pid = pid
		i.PASId, innerErr = strconv.ParseUint(info[PASIDHeader], 10, 64)
		if innerErr != nil {
			return nil, innerErr
		}
		s, innerErr := strconv.ParseUint(info[VRamUsedHeader], 10, 64)
		if innerErr != nil {
			return nil, innerErr
		}
		i.VRamUsed = utils.MemorySize{Num: s, Unit: utils.MiB}
		if info[VRamUsedPercentHeader] == "" || info[VRamUsedPercentHeader] == "inf" {
			i.VRamUsedPercent = -1
		} else {
			i.VRamUsedPercent, innerErr = strconv.Atoi(info[VRamUsedPercentHeader])
			if innerErr != nil {
				return nil, innerErr
			}
		}
		i.SDMAUsed, innerErr = strconv.Atoi(info[SDMAUsedHeader])
		if innerErr != nil {
			return nil, innerErr
		}
		result = append(result, i)
	}
liming6's avatar
liming6 committed
554
555
556
557
558
559
560
561
562
563
564
565
	// 按照DCU index排序
	sort.Slice(result, func(i, j int) bool {
		ii, err := strconv.Atoi(result[i].HCUIndex[0])
		if err != nil {
			return false
		}
		jj, err := strconv.Atoi(result[j].HCUIndex[0])
		if err != nil {
			return false
		}
		return ii < jj
	})
liming6's avatar
liming6 committed
566
567
	return result, nil
}
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584

func regMatch(reg *regexp.Regexp, s string, index ...int) []string {
	i := reg.FindStringSubmatch(s)
	if i == nil {
		return nil
	}
	result := make([]string, 0)
	l := len(i)
	for _, ind := range index {
		if ind >= 0 && ind < l {
			result = append(result, i[ind])
		} else {
			return nil
		}
	}
	return result
}
liming6's avatar
liming6 committed
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696

// GetMigInfo 获取是否开启了mig模式
// 如果返回nil,表示都没有开启
func GetMigInfo() map[int]bool {
	return nil
}

// GetEccInfo 获取是否启用ecc信息
func GetEccInfo() (map[int]bool, error) {
	output, err := exec.Command(DCUBinaryFile, "--showecc").Output()
	if err != nil {
		return nil, err
	}
	str := strings.TrimSpace(strings.Trim(string(output), "\n"))
	return parseEccInfo(str)
}

func parseEccInfo(str string) (map[int]bool, error) {
	lines := strings.Split(str, "\n")
	result := make(map[int]bool)
	for _, l := range lines {
		if ReUselessLine.MatchString(l) || ReEmptyLine.MatchString(l) {
			continue
		}
		find := ReInfoHeader.FindStringSubmatch(l)
		if len(find) < 3 {
			continue
		}
		index, err := strconv.Atoi(find[1])
		if err != nil {
			return nil, err
		}
		if ReEccOn.MatchString(find[2]) {
			result[index] = true
		} else {
			result[index] = false
		}
	}
	return result, nil
}

type DCUMemInfo struct {
	Id    int
	Total utils.MemorySize
	Used  utils.MemorySize
}

func GetDCUMemInfo() (map[int]DCUMemInfo, error) {
	output, err := exec.Command(DCUBinaryFile, "--showmeminfo", "vram").Output()
	if err != nil {
		return nil, err
	}
	str := strings.TrimSpace(strings.Trim(string(output), "\n"))
	return parseDCUMemInfo(str)
}

func parseDCUMemInfo(str string) (map[int]DCUMemInfo, error) {
	lines := strings.Split(str, "\n")
	result := make(map[int]DCUMemInfo)

	for _, line := range lines {
		if ReUselessLine.MatchString(line) || ReEmptyLine.MatchString(line) {
			continue
		}
		matchs := ReInfoHeader.FindStringSubmatch(line)
		if len(matchs) < 3 {
			continue
		}
		id, err := strconv.Atoi(matchs[1])
		if err != nil {
			return nil, err
		}
		info, have := result[id]
		if !have {
			result[id] = DCUMemInfo{Id: id}
			info = result[id]
		}
		if ReVramTotal.MatchString(matchs[2]) {
			field := ReVramTotal.FindStringSubmatch(matchs[2])
			if len(field) < 3 {
				continue
			}
			unit, err := utils.ParseUnit(field[1])
			if err != nil {
				return nil, err
			}
			size, err := strconv.Atoi(field[2])
			if err != nil {
				return nil, err
			}
			info.Total = utils.MemorySize{Unit: unit, Num: uint64(size)}
			result[id] = info
		}
		if ReVramUsed.MatchString(matchs[2]) {
			field := ReVramUsed.FindStringSubmatch(matchs[2])
			if len(field) < 3 {
				continue
			}
			unit, err := utils.ParseUnit(field[1])
			if err != nil {
				return nil, err
			}
			size, err := strconv.Atoi(field[2])
			if err != nil {
				return nil, err
			}
			info.Used = utils.MemorySize{Unit: unit, Num: uint64(size)}
			result[id] = info
		}
	}
	return result, nil
}