GOSTSec/ccminer
1
0
Fork 0
mirror of https://github.com/GOSTSec/ccminer synced 2025-01-08 22:07:56 +00:00
Code Issues Releases Wiki Activity

nvml: api enhancements, add power usage

Browse Source 
fallbacks on v2 nvml exports and check api implementation on call
This commit is contained in:
Tanguy Pruvot 2015-06-24 05:02:46 +02:00
parent f73f5f6a2e
commit b9cfee0bb3
5 changed files with 134 additions and 62 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
api/index.php
View File

@ -62,6 +62,7 @@ function translateField($key)
$intl['TEMP'] = 'T°c'; $intl['TEMP'] = 'T°c';
$intl['FAN'] = 'Fan %'; $intl['FAN'] = 'Fan %';
$intl['FREQ'] = 'Freq.'; $intl['FREQ'] = 'Freq.';
$intl['POWER'] = 'Power';
$intl['PST'] = 'P-State'; $intl['PST'] = 'P-State';
// pool infos // pool infos
@ -96,6 +97,9 @@ function translateValue($key,$val,$data=array())
case 'FREQ': case 'FREQ':
$val = sprintf("%d MHz", round(floatval($val)/1000.0)); $val = sprintf("%d MHz", round(floatval($val)/1000.0));
break; break;
case 'POWER':
$val = sprintf("%d W", round(floatval($val)/1000.0));
break;
case 'TS': case 'TS':
$val = strftime("%H:%M:%S", (int) $val); $val = strftime("%H:%M:%S", (int) $val);
break; break;

2
configure.ac
View File

@ -1,4 +1,4 @@
AC_INIT([ccminer], [1.6.5]) AC_INIT([ccminer], [1.6.6-dev])
AC_PREREQ([2.59c]) AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM AC_CANONICAL_SYSTEM

6
cpuminer-config.h
View File

@ -162,7 +162,7 @@
#define PACKAGE_NAME "ccminer" #define PACKAGE_NAME "ccminer"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "ccminer 1.6.5" #define PACKAGE_STRING "ccminer 1.6.6-dev"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "ccminer" #define PACKAGE_TARNAME "ccminer"
@ -171,7 +171,7 @@
#define PACKAGE_URL "http://github.com/tpruvot/ccminer" #define PACKAGE_URL "http://github.com/tpruvot/ccminer"
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.6.5" #define PACKAGE_VERSION "1.6.6-dev"
/* If using the C implementation of alloca, define if you know the /* If using the C implementation of alloca, define if you know the
direction of stack growth for your system; otherwise it will be direction of stack growth for your system; otherwise it will be
@ -185,7 +185,7 @@
#define STDC_HEADERS 1 #define STDC_HEADERS 1
/* Version number of package */ /* Version number of package */
#define VERSION "1.6.5" #define VERSION "1.6.6-dev"
/* Define curl_free() as free() if our version of curl lacks curl_free. */ /* Define curl_free() as free() if our version of curl lacks curl_free. */
/* #undef curl_free */ /* #undef curl_free */

143
nvml.cpp
View File

@ -15,6 +15,7 @@
* *
*/ */
#include <errno.h>
#include <stdio.h> #include <stdio.h>
#include <stdlib.h> #include <stdlib.h>
#include <string.h> #include <string.h>
@ -113,14 +114,12 @@ nvml_handle * nvml_create()
nvmlh->nvml_dll = nvml_dll; nvmlh->nvml_dll = nvml_dll;
nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) wrap_dlsym(nvmlh->nvml_dll, "nvmlInit_v2");
wrap_dlsym(nvmlh->nvml_dll, "nvmlInit_v2"); if (!nvmlh->nvmlInit)
if (!nvmlh->nvmlInit) { nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) wrap_dlsym(nvmlh->nvml_dll, "nvmlInit");
nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) nvmlh->nvmlDeviceGetCount = (nvmlReturn_t (*)(int *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCount_v2");
wrap_dlsym(nvmlh->nvml_dll, "nvmlInit"); if (!nvmlh->nvmlDeviceGetCount)
} nvmlh->nvmlDeviceGetCount = (nvmlReturn_t (*)(int *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCount");
nvmlh->nvmlDeviceGetCount = (nvmlReturn_t (*)(int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCount_v2");
nvmlh->nvmlDeviceGetHandleByIndex = (nvmlReturn_t (*)(int, nvmlDevice_t *)) nvmlh->nvmlDeviceGetHandleByIndex = (nvmlReturn_t (*)(int, nvmlDevice_t *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetHandleByIndex_v2"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetHandleByIndex_v2");
nvmlh->nvmlDeviceGetAPIRestriction = (nvmlReturn_t (*)(nvmlDevice_t, nvmlRestrictedAPI_t, nvmlEnableState_t *)) nvmlh->nvmlDeviceGetAPIRestriction = (nvmlReturn_t (*)(nvmlDevice_t, nvmlRestrictedAPI_t, nvmlEnableState_t *))
@ -139,15 +138,31 @@ nvml_handle * nvml_create()
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedGraphicsClocks"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedGraphicsClocks");
nvmlh->nvmlDeviceGetSupportedMemoryClocks = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *count, unsigned int *clocksMHz)) nvmlh->nvmlDeviceGetSupportedMemoryClocks = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *count, unsigned int *clocksMHz))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedMemoryClocks"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedMemoryClocks");
/* NVML_ERROR_NOT_SUPPORTED
nvmlh->nvmlDeviceGetAutoBoostedClocksEnabled = (nvmlReturn_t (*)(nvmlDevice_t, nvmlEnableState_t *isEnabled, nvmlEnableState_t *defaultIsEnabled))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetAutoBoostedClocksEnabled");
nvmlh->nvmlDeviceSetAutoBoostedClocksEnabled = (nvmlReturn_t (*)(nvmlDevice_t, nvmlEnableState_t enabled))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetAutoBoostedClocksEnabled"); */
nvmlh->nvmlDeviceGetClockInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock)) nvmlh->nvmlDeviceGetClockInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetClockInfo"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetClockInfo");
nvmlh->nvmlDeviceGetPciInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPciInfo_t *)) nvmlh->nvmlDeviceGetMaxClockInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPciInfo_v2"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxClockInfo");
nvmlh->nvmlDeviceGetPciInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPciInfo_t *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPciInfo_v2");
if (!nvmlh->nvmlDeviceGetPciInfo)
nvmlh->nvmlDeviceGetPciInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPciInfo_t *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPciInfo");
nvmlh->nvmlDeviceGetCurrPcieLinkGeneration = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *gen))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCurrPcieLinkGeneration");
nvmlh->nvmlDeviceGetCurrPcieLinkWidth = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *width))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCurrPcieLinkWidth");
nvmlh->nvmlDeviceGetMaxPcieLinkGeneration = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *gen))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxPcieLinkGeneration");
nvmlh->nvmlDeviceGetMaxPcieLinkWidth = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *width))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxPcieLinkWidth");
nvmlh->nvmlDeviceGetPowerUsage = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerUsage");
nvmlh->nvmlDeviceGetPowerManagementDefaultLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementDefaultLimit");
nvmlh->nvmlDeviceGetPowerManagementLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementLimit");
nvmlh->nvmlDeviceGetPowerManagementLimitConstraints = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *min, unsigned int *max))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementLimitConstraints");
nvmlh->nvmlDeviceSetPowerManagementLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetPowerManagementLimit");
nvmlh->nvmlDeviceGetName = (nvmlReturn_t (*)(nvmlDevice_t, char *, int)) nvmlh->nvmlDeviceGetName = (nvmlReturn_t (*)(nvmlDevice_t, char *, int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetName"); wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetName");
nvmlh->nvmlDeviceGetTemperature = (nvmlReturn_t (*)(nvmlDevice_t, int, unsigned int *)) nvmlh->nvmlDeviceGetTemperature = (nvmlReturn_t (*)(nvmlDevice_t, int, unsigned int *))
@ -168,25 +183,26 @@ nvml_handle * nvml_create()
wrap_dlsym(nvmlh->nvml_dll, "nvmlErrorString"); wrap_dlsym(nvmlh->nvml_dll, "nvmlErrorString");
nvmlh->nvmlShutdown = (nvmlReturn_t (*)()) nvmlh->nvmlShutdown = (nvmlReturn_t (*)())
wrap_dlsym(nvmlh->nvml_dll, "nvmlShutdown"); wrap_dlsym(nvmlh->nvml_dll, "nvmlShutdown");
// v331
/* nvmlh->nvmlDeviceGetEnforcedPowerLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
nvmlDeviceGetBrand(nvmlDevice_t device, nvmlBrandType_t *type) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetEnforcedPowerLimit");
nvmlDeviceGetMaxPcieLinkWidth(nvmlDevice_t device, unsigned int *maxLinkWidth) // v340
nvmlDeviceGetMaxPcieLinkGeneration(nvmlDevice_t device, unsigned int *maxLinkGen) /* NVML_ERROR_NOT_SUPPORTED
nvmlDeviceGetCurrPcieLinkWidth(nvmlDevice_t device, unsigned int *currLinkWidth) nvmlh->nvmlDeviceGetAutoBoostedClocksEnabled = (nvmlReturn_t (*)(nvmlDevice_t, nvmlEnableState_t *isEnabled, nvmlEnableState_t *defaultIsEnabled))
nvmlDeviceGetCurrPcieLinkGeneration(nvmlDevice_t device, unsigned int *currLinkGen) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetAutoBoostedClocksEnabled");
*/ nvmlh->nvmlDeviceSetAutoBoostedClocksEnabled = (nvmlReturn_t (*)(nvmlDevice_t, nvmlEnableState_t enabled))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetAutoBoostedClocksEnabled"); */
// v346
nvmlh->nvmlDeviceGetPcieThroughput = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPcieUtilCounter_t, unsigned int *value))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPcieThroughput");
if (nvmlh->nvmlInit == NULL || if (nvmlh->nvmlInit == NULL ||
nvmlh->nvmlShutdown == NULL || nvmlh->nvmlShutdown == NULL ||
nvmlh->nvmlErrorString == NULL || nvmlh->nvmlErrorString == NULL ||
nvmlh->nvmlSystemGetDriverVersion == NULL ||
nvmlh->nvmlDeviceGetCount == NULL || nvmlh->nvmlDeviceGetCount == NULL ||
nvmlh->nvmlDeviceGetHandleByIndex == NULL || nvmlh->nvmlDeviceGetHandleByIndex == NULL ||
nvmlh->nvmlDeviceGetPciInfo == NULL || nvmlh->nvmlDeviceGetPciInfo == NULL ||
nvmlh->nvmlDeviceGetName == NULL || nvmlh->nvmlDeviceGetName == NULL)
nvmlh->nvmlDeviceGetTemperature == NULL ||
nvmlh->nvmlDeviceGetFanSpeed == NULL)
{ {
if (opt_debug) if (opt_debug)
applog(LOG_DEBUG, "Failed to obtain required NVML function pointers"); applog(LOG_DEBUG, "Failed to obtain required NVML function pointers");
@ -196,6 +212,7 @@ nvml_handle * nvml_create()
} }
nvmlh->nvmlInit(); nvmlh->nvmlInit();
if (nvmlh->nvmlSystemGetDriverVersion)
nvmlh->nvmlSystemGetDriverVersion(driver_version, sizeof(driver_version)); nvmlh->nvmlSystemGetDriverVersion(driver_version, sizeof(driver_version));
nvmlh->nvmlDeviceGetCount(&nvmlh->nvml_gpucount); nvmlh->nvmlDeviceGetCount(&nvmlh->nvml_gpucount);
@ -280,7 +297,7 @@ int nvml_set_clocks(nvml_handle *nvmlh, int dev_id)
uint32_t gpu_clk = 0, mem_clk = 0; uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id]; int n = nvmlh->cuda_nvml_device_id[dev_id];
if (n < 0 || n >= nvmlh->nvml_gpucount) if (n < 0 || n >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
// prevent double operations on the same gpu... to enhance // prevent double operations on the same gpu... to enhance
if (gpu_clocks_changed[dev_id]) if (gpu_clocks_changed[dev_id])
@ -291,18 +308,18 @@ int nvml_set_clocks(nvml_handle *nvmlh, int dev_id)
if (nvmlh->app_clocks[n] != NVML_FEATURE_ENABLED) { if (nvmlh->app_clocks[n] != NVML_FEATURE_ENABLED) {
applog(LOG_WARNING, "GPU #%d: NVML application clock feature is not allowed!", dev_id); applog(LOG_WARNING, "GPU #%d: NVML application clock feature is not allowed!", dev_id);
return -1; return -EPERM;
} }
nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_MEM, &mem_clk); nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_MEM, &mem_clk);
rc = nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_GRAPHICS, &gpu_clk); rc = nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_GRAPHICS, &gpu_clk);
if (rc != NVML_SUCCESS) { if (rc != NVML_SUCCESS) {
applog(LOG_WARNING, "GPU #%d: unable to query application clocks", dev_id); applog(LOG_WARNING, "GPU #%d: unable to query application clocks", dev_id);
return -1; return -EINVAL;
} }
if (opt_debug) if (opt_debug)
applog(LOG_DEBUG, "GPU #%d: default clocks are %u/%u", dev_id, mem_clk, gpu_clk); applog(LOG_DEBUG, "GPU #%d: default application clocks are %u/%u", dev_id, mem_clk, gpu_clk);
// get application config values // get application config values
if (device_mem_clocks[dev_id]) mem_clk = device_mem_clocks[dev_id]; if (device_mem_clocks[dev_id]) mem_clk = device_mem_clocks[dev_id];
@ -352,7 +369,7 @@ int nvml_reset_clocks(nvml_handle *nvmlh, int dev_id)
uint32_t gpu_clk = 0, mem_clk = 0; uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id]; int n = nvmlh->cuda_nvml_device_id[dev_id];
if (n < 0 || n >= nvmlh->nvml_gpucount) if (n < 0 || n >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!gpu_clocks_changed[dev_id]) if (!gpu_clocks_changed[dev_id])
return 0; // nothing to do return 0; // nothing to do
@ -383,7 +400,10 @@ int nvml_get_gpu_name(nvml_handle *nvmlh, int cudaindex, char *namebuf, int bufs
{ {
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetName)
return -ENOSYS;
if (nvmlh->nvmlDeviceGetName(nvmlh->devs[gpuindex], namebuf, bufsize) != NVML_SUCCESS) if (nvmlh->nvmlDeviceGetName(nvmlh->devs[gpuindex], namebuf, bufsize) != NVML_SUCCESS)
return -1; return -1;
@ -397,7 +417,10 @@ int nvml_get_tempC(nvml_handle *nvmlh, int cudaindex, unsigned int *tempC)
nvmlReturn_t rc; nvmlReturn_t rc;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetTemperature)
return -ENOSYS;
rc = nvmlh->nvmlDeviceGetTemperature(nvmlh->devs[gpuindex], 0u /* NVML_TEMPERATURE_GPU */, tempC); rc = nvmlh->nvmlDeviceGetTemperature(nvmlh->devs[gpuindex], 0u /* NVML_TEMPERATURE_GPU */, tempC);
if (rc != NVML_SUCCESS) { if (rc != NVML_SUCCESS) {
@ -413,7 +436,10 @@ int nvml_get_fanpcnt(nvml_handle *nvmlh, int cudaindex, unsigned int *fanpcnt)
nvmlReturn_t rc; nvmlReturn_t rc;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetFanSpeed)
return -ENOSYS;
rc = nvmlh->nvmlDeviceGetFanSpeed(nvmlh->devs[gpuindex], fanpcnt); rc = nvmlh->nvmlDeviceGetFanSpeed(nvmlh->devs[gpuindex], fanpcnt);
if (rc != NVML_SUCCESS) { if (rc != NVML_SUCCESS) {
@ -428,12 +454,15 @@ int nvml_get_power_usage(nvml_handle *nvmlh, int cudaindex, unsigned int *milliw
{ {
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetPowerUsage)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetPowerUsage(nvmlh->devs[gpuindex], milliwatts); nvmlReturn_t res = nvmlh->nvmlDeviceGetPowerUsage(nvmlh->devs[gpuindex], milliwatts);
if (res != NVML_SUCCESS) { if (res != NVML_SUCCESS) {
if (opt_debug) //if (opt_debug)
applog(LOG_DEBUG, "nvmlDeviceGetPowerUsage: %s", nvmlh->nvmlErrorString(res)); // applog(LOG_DEBUG, "nvmlDeviceGetPowerUsage: %s", nvmlh->nvmlErrorString(res));
return -1; return -1;
} }
@ -445,7 +474,10 @@ int nvml_get_pstate(nvml_handle *nvmlh, int cudaindex, int *pstate)
{ {
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetPerformanceState)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetPerformanceState(nvmlh->devs[gpuindex], pstate); nvmlReturn_t res = nvmlh->nvmlDeviceGetPerformanceState(nvmlh->devs[gpuindex], pstate);
if (res != NVML_SUCCESS) { if (res != NVML_SUCCESS) {
@ -461,7 +493,7 @@ int nvml_get_busid(nvml_handle *nvmlh, int cudaindex, int *busid)
{ {
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
(*busid) = nvmlh->nvml_pci_bus_id[gpuindex]; (*busid) = nvmlh->nvml_pci_bus_id[gpuindex];
return 0; return 0;
@ -474,13 +506,17 @@ int nvml_get_serial(nvml_handle *nvmlh, int cudaindex, char *sn, int maxlen)
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
nvmlReturn_t res; nvmlReturn_t res;
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (nvmlh->nvmlDeviceGetSerial) {
res = nvmlh->nvmlDeviceGetSerial(nvmlh->devs[gpuindex], sn, maxlen); res = nvmlh->nvmlDeviceGetSerial(nvmlh->devs[gpuindex], sn, maxlen);
if (res == NVML_SUCCESS) { if (res == NVML_SUCCESS)
return 0; return 0;
} }
if (!nvmlh->nvmlDeviceGetUUID)
return -ENOSYS;
// nvmlDeviceGetUUID: GPU-f2bd642c-369f-5a14-e0b4-0d22dfe9a1fc // nvmlDeviceGetUUID: GPU-f2bd642c-369f-5a14-e0b4-0d22dfe9a1fc
// use a part of uuid to generate an unique serial // use a part of uuid to generate an unique serial
// todo: check if there is vendor id is inside // todo: check if there is vendor id is inside
@ -501,7 +537,10 @@ int nvml_get_bios(nvml_handle *nvmlh, int cudaindex, char *desc, int maxlen)
uint32_t subids = 0; uint32_t subids = 0;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
if (!nvmlh->nvmlDeviceGetVbiosVersion)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetVbiosVersion(nvmlh->devs[gpuindex], desc, maxlen); nvmlReturn_t res = nvmlh->nvmlDeviceGetVbiosVersion(nvmlh->devs[gpuindex], desc, maxlen);
if (res != NVML_SUCCESS) { if (res != NVML_SUCCESS) {
@ -517,7 +556,7 @@ int nvml_get_info(nvml_handle *nvmlh, int cudaindex, uint16_t &vid, uint16_t &pi
uint32_t subids = 0; uint32_t subids = 0;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex]; int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount) if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -1; return -ENODEV;
subids = nvmlh->nvml_pci_subsys_id[gpuindex]; subids = nvmlh->nvml_pci_subsys_id[gpuindex];
pid = subids >> 16; pid = subids >> 16;
@ -561,7 +600,7 @@ int nvapi_temperature(unsigned int devNum, unsigned int *temperature)
NvAPI_Status ret; NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
NV_GPU_THERMAL_SETTINGS thermal; NV_GPU_THERMAL_SETTINGS thermal;
thermal.version = NV_GPU_THERMAL_SETTINGS_VER; thermal.version = NV_GPU_THERMAL_SETTINGS_VER;
@ -584,7 +623,7 @@ int nvapi_fanspeed(unsigned int devNum, unsigned int *speed)
NvAPI_Status ret; NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
NvU32 fanspeed = 0; NvU32 fanspeed = 0;
ret = NvAPI_GPU_GetTachReading(phys[devNum], &fanspeed); ret = NvAPI_GPU_GetTachReading(phys[devNum], &fanspeed);
@ -606,7 +645,7 @@ int nvapi_getpstate(unsigned int devNum, unsigned int *power)
NvAPI_Status ret; NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
NV_GPU_PERF_PSTATE_ID CurrentPstate = NVAPI_GPU_PERF_PSTATE_UNDEFINED; /* 16 */ NV_GPU_PERF_PSTATE_ID CurrentPstate = NVAPI_GPU_PERF_PSTATE_UNDEFINED; /* 16 */
ret = NvAPI_GPU_GetCurrentPstate(phys[devNum], &CurrentPstate); ret = NvAPI_GPU_GetCurrentPstate(phys[devNum], &CurrentPstate);
@ -631,7 +670,7 @@ int nvapi_getusage(unsigned int devNum, unsigned int *pct)
NvAPI_Status ret; NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
NV_GPU_DYNAMIC_PSTATES_INFO_EX info; NV_GPU_DYNAMIC_PSTATES_INFO_EX info;
info.version = NV_GPU_DYNAMIC_PSTATES_INFO_EX_VER; info.version = NV_GPU_DYNAMIC_PSTATES_INFO_EX_VER;
@ -657,7 +696,7 @@ int nvapi_getinfo(unsigned int devNum, uint16_t &vid, uint16_t &pid)
NvU32 pDeviceId, pSubSystemId, pRevisionId, pExtDeviceId; NvU32 pDeviceId, pSubSystemId, pRevisionId, pExtDeviceId;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
ret = NvAPI_GPU_GetPCIIdentifiers(phys[devNum], &pDeviceId, &pSubSystemId, &pRevisionId, &pExtDeviceId); ret = NvAPI_GPU_GetPCIIdentifiers(phys[devNum], &pDeviceId, &pSubSystemId, &pRevisionId, &pExtDeviceId);
if (ret != NVAPI_OK) { if (ret != NVAPI_OK) {
@ -678,7 +717,7 @@ int nvapi_getserial(unsigned int devNum, char *serial, unsigned int maxlen)
{ {
// NvAPI_Status ret; // NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
sprintf(serial, ""); sprintf(serial, "");
@ -702,7 +741,7 @@ int nvapi_getbios(unsigned int devNum, char *desc, unsigned int maxlen)
{ {
NvAPI_Status ret; NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt) if (devNum >= nvapi_dev_cnt)
return -1; return -ENODEV;
if (maxlen < 64) // Short String if (maxlen < 64) // Short String
return -1; return -1;
@ -899,6 +938,7 @@ static int translate_vendor_id(uint16_t vid, char *vendorname)
// { 0x10DE, "NVIDIA" }, // { 0x10DE, "NVIDIA" },
{ 0x1458, "Gigabyte" }, { 0x1458, "Gigabyte" },
{ 0x1462, "MSI" }, { 0x1462, "MSI" },
{ 0x19DA, "Zotac" },
{ 0x3842, "EVGA" }, { 0x3842, "EVGA" },
{ 0, "" } { 0, "" }
}; };
@ -919,7 +959,6 @@ static int translate_vendor_id(uint16_t vid, char *vendorname)
#ifdef HAVE_PCIDEV #ifdef HAVE_PCIDEV
extern "C" { extern "C" {
#include <errno.h>
#include <pci/pci.h> #include <pci/pci.h>
} }
static int linux_gpu_vendor(uint8_t pci_bus_id, char* vendorname, uint16_t &pid) static int linux_gpu_vendor(uint8_t pci_bus_id, char* vendorname, uint16_t &pid)

35
nvml.h
View File

@ -66,6 +66,20 @@ enum nvmlClockType_t {
NVML_CLOCK_MEM = 2 NVML_CLOCK_MEM = 2
}; };
enum nvmlPcieUtilCounter_t {
NVML_PCIE_UTIL_TX_BYTES = 0,
NVML_PCIE_UTIL_RX_BYTES = 1,
NVML_PCIE_UTIL_COUNT
};
enum nvmlValueType_t {
NVML_VALUE_TYPE_DOUBLE = 0,
NVML_VALUE_TYPE_UNSIGNED_INT = 1,
NVML_VALUE_TYPE_UNSIGNED_LONG = 2,
NVML_VALUE_TYPE_UNSIGNED_LONG_LONG = 3,
NVML_VALUE_TYPE_COUNT
};
#define NVML_DEVICE_SERIAL_BUFFER_SIZE 30 #define NVML_DEVICE_SERIAL_BUFFER_SIZE 30
#define NVML_DEVICE_UUID_BUFFER_SIZE 80 #define NVML_DEVICE_UUID_BUFFER_SIZE 80
#define NVML_DEVICE_VBIOS_VERSION_BUFFER_SIZE 32 #define NVML_DEVICE_VBIOS_VERSION_BUFFER_SIZE 32
@ -97,10 +111,17 @@ typedef struct {
nvmlReturn_t (*nvmlDeviceResetApplicationsClocks)(nvmlDevice_t); nvmlReturn_t (*nvmlDeviceResetApplicationsClocks)(nvmlDevice_t);
nvmlReturn_t (*nvmlDeviceGetSupportedGraphicsClocks)(nvmlDevice_t, uint32_t mem, uint32_t *num, uint32_t *arr); nvmlReturn_t (*nvmlDeviceGetSupportedGraphicsClocks)(nvmlDevice_t, uint32_t mem, uint32_t *num, uint32_t *arr);
nvmlReturn_t (*nvmlDeviceGetSupportedMemoryClocks)(nvmlDevice_t, unsigned int *count, unsigned int *clocksMHz); nvmlReturn_t (*nvmlDeviceGetSupportedMemoryClocks)(nvmlDevice_t, unsigned int *count, unsigned int *clocksMHz);
nvmlReturn_t (*nvmlDeviceGetAutoBoostedClocksEnabled)(nvmlDevice_t, nvmlEnableState_t *isEnabled, nvmlEnableState_t *defaultIsEnabled);
nvmlReturn_t (*nvmlDeviceSetAutoBoostedClocksEnabled)(nvmlDevice_t, nvmlEnableState_t enabled);
nvmlReturn_t (*nvmlDeviceGetClockInfo)(nvmlDevice_t, nvmlClockType_t, unsigned int *); nvmlReturn_t (*nvmlDeviceGetClockInfo)(nvmlDevice_t, nvmlClockType_t, unsigned int *);
nvmlReturn_t (*nvmlDeviceGetMaxClockInfo)(nvmlDevice_t, nvmlClockType_t, unsigned int *);
nvmlReturn_t (*nvmlDeviceGetPowerManagementDefaultLimit)(nvmlDevice_t, unsigned int *limit);
nvmlReturn_t (*nvmlDeviceGetPowerManagementLimit)(nvmlDevice_t, unsigned int *limit);
nvmlReturn_t (*nvmlDeviceGetPowerManagementLimitConstraints)(nvmlDevice_t, unsigned int *min, unsigned int *max);
nvmlReturn_t (*nvmlDeviceSetPowerManagementLimit)(nvmlDevice_t device, unsigned int limit);
nvmlReturn_t (*nvmlDeviceGetPciInfo)(nvmlDevice_t, nvmlPciInfo_t *); nvmlReturn_t (*nvmlDeviceGetPciInfo)(nvmlDevice_t, nvmlPciInfo_t *);
nvmlReturn_t (*nvmlDeviceGetCurrPcieLinkGeneration)(nvmlDevice_t device, unsigned int *gen);
nvmlReturn_t (*nvmlDeviceGetCurrPcieLinkWidth)(nvmlDevice_t device, unsigned int *width);
nvmlReturn_t (*nvmlDeviceGetMaxPcieLinkGeneration)(nvmlDevice_t device, unsigned int *gen);
nvmlReturn_t (*nvmlDeviceGetMaxPcieLinkWidth)(nvmlDevice_t device, unsigned int *width);
nvmlReturn_t (*nvmlDeviceGetName)(nvmlDevice_t, char *, int); nvmlReturn_t (*nvmlDeviceGetName)(nvmlDevice_t, char *, int);
nvmlReturn_t (*nvmlDeviceGetTemperature)(nvmlDevice_t, int, unsigned int *); nvmlReturn_t (*nvmlDeviceGetTemperature)(nvmlDevice_t, int, unsigned int *);
nvmlReturn_t (*nvmlDeviceGetFanSpeed)(nvmlDevice_t, unsigned int *); nvmlReturn_t (*nvmlDeviceGetFanSpeed)(nvmlDevice_t, unsigned int *);
@ -112,6 +133,15 @@ typedef struct {
nvmlReturn_t (*nvmlSystemGetDriverVersion)(char *version, unsigned int len); nvmlReturn_t (*nvmlSystemGetDriverVersion)(char *version, unsigned int len);
char* (*nvmlErrorString)(nvmlReturn_t); char* (*nvmlErrorString)(nvmlReturn_t);
nvmlReturn_t (*nvmlShutdown)(void); nvmlReturn_t (*nvmlShutdown)(void);
// v331
nvmlReturn_t (*nvmlDeviceGetEnforcedPowerLimit)(nvmlDevice_t, unsigned int *limit);
// v340
//nvmlReturn_t (*nvmlDeviceGetCpuAffinity)(nvmlDevice_t, unsigned int cpuSetSize, unsigned long* cpuSet);
//nvmlReturn_t (*nvmlDeviceSetCpuAffinity)(nvmlDevice_t);
//nvmlReturn_t (*nvmlDeviceGetAutoBoostedClocksEnabled)(nvmlDevice_t, nvmlEnableState_t *isEnabled, nvmlEnableState_t *defaultIsEnabled);
//nvmlReturn_t (*nvmlDeviceSetAutoBoostedClocksEnabled)(nvmlDevice_t, nvmlEnableState_t enabled);
// v346
nvmlReturn_t (*nvmlDeviceGetPcieThroughput)(nvmlDevice_t, nvmlPcieUtilCounter_t, unsigned int *value);
} nvml_handle; } nvml_handle;
@ -170,7 +200,6 @@ unsigned int gpu_fanpercent(struct cgpu_info *gpu);
unsigned int gpu_fanrpm(struct cgpu_info *gpu); unsigned int gpu_fanrpm(struct cgpu_info *gpu);
float gpu_temp(struct cgpu_info *gpu); float gpu_temp(struct cgpu_info *gpu);
unsigned int gpu_power(struct cgpu_info *gpu); unsigned int gpu_power(struct cgpu_info *gpu);
unsigned int gpu_usage(struct cgpu_info *gpu);
int gpu_pstate(struct cgpu_info *gpu); int gpu_pstate(struct cgpu_info *gpu);
int gpu_busid(struct cgpu_info *gpu); int gpu_busid(struct cgpu_info *gpu);