197bb86f7SLois Curfman McInnes /* 27588ac45SBarry Smith Defines profile/logging in PETSc. 397bb86f7SLois Curfman McInnes */ 46524c165SJacob Faibussowitsch #ifndef PETSCLOG_H 526bd1501SBarry Smith #define PETSCLOG_H 6ac09b921SBarry Smith 7aaa7dc30SBarry Smith #include <petscsys.h> 8958c4211Shannah_mairs #include <petsctime.h> 9c8d78d4dSSatish Balay 10ac09b921SBarry Smith /* SUBMANSEC = Sys */ 11ac09b921SBarry Smith 12fa2bb9feSLisandro Dalcin /* General logging of information; different from event logging */ 133ca90d2dSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfo_Private(const char[], PetscObject, const char[], ...) PETSC_ATTRIBUTE_FORMAT(3, 4); 14fa2bb9feSLisandro Dalcin #if defined(PETSC_USE_INFO) 157d3de750SJacob Faibussowitsch #define PetscInfo(A, ...) PetscInfo_Private(PETSC_FUNCTION_NAME, ((PetscObject)A), __VA_ARGS__) 16fa2bb9feSLisandro Dalcin #else 173ba16761SJacob Faibussowitsch #define PetscInfo(A, ...) PETSC_SUCCESS 18fa2bb9feSLisandro Dalcin #endif 19e94e781bSJacob Faibussowitsch 207d3de750SJacob Faibussowitsch #define PetscInfo1(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 217d3de750SJacob Faibussowitsch #define PetscInfo2(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 225899db6dSLisandro Dalcin #define PetscInfo3(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 237d3de750SJacob Faibussowitsch #define PetscInfo4(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 247d3de750SJacob Faibussowitsch #define PetscInfo5(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 257d3de750SJacob Faibussowitsch #define PetscInfo6(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 267d3de750SJacob Faibussowitsch #define PetscInfo7(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 277d3de750SJacob Faibussowitsch #define PetscInfo8(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 287d3de750SJacob Faibussowitsch #define PetscInfo9(...) PETSC_DEPRECATED_MACRO("GCC warning \"Use PetscInfo() (since version 3.17)\"") PetscInfo(__VA_ARGS__) 297d3de750SJacob Faibussowitsch 30e94e781bSJacob Faibussowitsch /*E 3187497f52SBarry Smith PetscInfoCommFlag - Describes the method by which to filter `PetscInfo()` by communicator size 32e94e781bSJacob Faibussowitsch 33*667f096bSBarry Smith Values: 34d1f92df0SBarry Smith + `PETSC_INFO_COMM_ALL` - Default uninitialized value. `PetscInfo()` will not filter based on 358b86dd2eSJacob Faibussowitsch communicator size (i.e. will print for all communicators) 36d1f92df0SBarry Smith . `PETSC_INFO_COMM_NO_SELF` - `PetscInfo()` will NOT print for communicators with size = 1 (i.e. *_COMM_SELF) 37d1f92df0SBarry Smith - `PETSC_INFO_COMM_ONLY_SELF` - `PetscInfo()` will ONLY print for communicators with size = 1 38e94e781bSJacob Faibussowitsch 39e94e781bSJacob Faibussowitsch Level: intermediate 40e94e781bSJacob Faibussowitsch 41*667f096bSBarry Smith Note: 42*667f096bSBarry Smith Used as an input for `PetscInfoSetFilterCommSelf()` 43*667f096bSBarry Smith 44db781477SPatrick Sanan .seealso: `PetscInfo()`, `PetscInfoSetFromOptions()`, `PetscInfoSetFilterCommSelf()` 45e94e781bSJacob Faibussowitsch E*/ 46e94e781bSJacob Faibussowitsch typedef enum { 47e94e781bSJacob Faibussowitsch PETSC_INFO_COMM_ALL = -1, 48e94e781bSJacob Faibussowitsch PETSC_INFO_COMM_NO_SELF = 0, 49e94e781bSJacob Faibussowitsch PETSC_INFO_COMM_ONLY_SELF = 1 50e94e781bSJacob Faibussowitsch } PetscInfoCommFlag; 51e94e781bSJacob Faibussowitsch 52e94e781bSJacob Faibussowitsch PETSC_EXTERN const char *const PetscInfoCommFlags[]; 53fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscInfoDeactivateClass(PetscClassId); 54fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscInfoActivateClass(PetscClassId); 55e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoEnabled(PetscClassId, PetscBool *); 56e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoAllow(PetscBool); 57e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoSetFile(const char[], const char[]); 58e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoGetFile(char **, FILE **); 59e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoSetClasses(PetscBool, PetscInt, const char *const *); 60e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoGetClass(const char *, PetscBool *); 61e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoGetInfo(PetscBool *, PetscBool *, PetscBool *, PetscBool *, PetscInfoCommFlag *); 628b86dd2eSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoProcessClass(const char[], PetscInt, const PetscClassId[]); 63e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoSetFilterCommSelf(PetscInfoCommFlag); 64e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoSetFromOptions(PetscOptions); 65e94e781bSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscInfoDestroy(void); 66fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscBool PetscLogPrintInfo; /* if true, indicates PetscInfo() is turned on */ 67fa2bb9feSLisandro Dalcin 68b859824cSBarry Smith /*MC 69b859824cSBarry Smith PetscLogEvent - id used to identify PETSc or user events which timed portions (blocks of executable) 70b859824cSBarry Smith code. 71b859824cSBarry Smith 72b859824cSBarry Smith Level: intermediate 73b859824cSBarry Smith 74d1f92df0SBarry Smith .seealso: [](ch_profiling), `PetscLogEventRegister()`, `PetscLogEventBegin()`, `PetscLogEventEnd()`, `PetscLogStage` 75b859824cSBarry Smith M*/ 76b859824cSBarry Smith typedef int PetscLogEvent; 77b859824cSBarry Smith 78b859824cSBarry Smith /*MC 79b859824cSBarry Smith PetscLogStage - id used to identify user stages (phases, sections) of runs - for logging 80b859824cSBarry Smith 81b859824cSBarry Smith Level: intermediate 82b859824cSBarry Smith 83d1f92df0SBarry Smith .seealso: [](ch_profiling), `PetscLogStageRegister()`, `PetscLogStagePush()`, `PetscLogStagePop()`, `PetscLogEvent` 84b859824cSBarry Smith M*/ 85b859824cSBarry Smith typedef int PetscLogStage; 86b859824cSBarry Smith 878ba1e511SMatthew Knepley #define PETSC_EVENT 1311311 88014dd563SJed Brown PETSC_EXTERN PetscLogEvent PETSC_LARGEST_EVENT; 8997bb86f7SLois Curfman McInnes 90c708d6e3SStefano Zampini /* Handle multithreading */ 91c708d6e3SStefano Zampini #if defined(PETSC_HAVE_THREADSAFETY) 92c708d6e3SStefano Zampini #if defined(__cplusplus) 93c708d6e3SStefano Zampini #define PETSC_TLS thread_local 94c708d6e3SStefano Zampini #else 95c708d6e3SStefano Zampini #define PETSC_TLS _Thread_local 96c708d6e3SStefano Zampini #endif 97c708d6e3SStefano Zampini #define PETSC_EXTERN_TLS extern PETSC_TLS PETSC_VISIBILITY_PUBLIC 98c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscAddLogDouble(PetscLogDouble *, PetscLogDouble *, PetscLogDouble); 99c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscAddLogDoubleCnt(PetscLogDouble *, PetscLogDouble *, PetscLogDouble *, PetscLogDouble *, PetscLogDouble); 100c708d6e3SStefano Zampini #else 101c708d6e3SStefano Zampini #define PETSC_EXTERN_TLS PETSC_EXTERN 102c708d6e3SStefano Zampini #define PETSC_TLS 1033ba16761SJacob Faibussowitsch #define PetscAddLogDouble(a, b, c) ((PetscErrorCode)((*(a) += (c), PETSC_SUCCESS) || ((*(b) += (c)), PETSC_SUCCESS))) 1043ba16761SJacob Faibussowitsch #define PetscAddLogDoubleCnt(a, b, c, d, e) ((PetscErrorCode)(PetscAddLogDouble(a, c, 1) || PetscAddLogDouble(b, d, e))) 105c708d6e3SStefano Zampini #endif 10619b02663SBarry Smith 10731d06abdSBarry Smith /* We must make the following structures available to access the event 10831d06abdSBarry Smith activation flags in the PetscLogEventBegin/End() macros. These are not part of the PETSc public 10931d06abdSBarry Smith API and are not intended to be used by other parts of PETSc or by users. 110d49d4b11SBarry Smith 111fa2bb9feSLisandro Dalcin The code that manipulates these structures is in src/sys/logging/utils. 11231d06abdSBarry Smith */ 113aa213bdcSJed Brown typedef struct _n_PetscIntStack *PetscIntStack; 1146a6a9b46SSatish Balay 115217044c2SLisandro Dalcin /* -----------------------------------------------------------------------------------------------------*/ 116c60ec953SBarry Smith /* 117aa213bdcSJed Brown PetscClassRegInfo, PetscClassPerfInfo - Each class has two data structures associated with it. The first has 118c60ec953SBarry Smith static information about it, the second collects statistics on how many objects of the class are created, 119c60ec953SBarry Smith how much memory they use, etc. 120c60ec953SBarry Smith 121aa213bdcSJed Brown PetscClassRegLog, PetscClassPerfLog - arrays of the PetscClassRegInfo and PetscClassPerfInfo for all classes. 122c60ec953SBarry Smith */ 1236a6a9b46SSatish Balay typedef struct { 124c60ec953SBarry Smith char *name; /* The class name */ 125c60ec953SBarry Smith PetscClassId classid; /* The integer identifying this class */ 126aa213bdcSJed Brown } PetscClassRegInfo; 1276a6a9b46SSatish Balay 1286a6a9b46SSatish Balay typedef struct { 1290700a824SBarry Smith PetscClassId id; /* The integer identifying this class */ 1306a6a9b46SSatish Balay int creations; /* The number of objects of this class created */ 1316a6a9b46SSatish Balay int destructions; /* The number of objects of this class destroyed */ 1324851f57dSBarry Smith PetscLogDouble mem; /* The total memory allocated by objects of this class; this is completely wrong and should possibly be removed */ 1334851f57dSBarry Smith PetscLogDouble descMem; /* The total memory allocated by descendents of these objects; this is completely wrong and should possibly be removed */ 134aa213bdcSJed Brown } PetscClassPerfInfo; 1356a6a9b46SSatish Balay 136aa213bdcSJed Brown typedef struct _n_PetscClassRegLog *PetscClassRegLog; 137aa213bdcSJed Brown struct _n_PetscClassRegLog { 138c60ec953SBarry Smith int numClasses; /* The number of classes registered */ 139c60ec953SBarry Smith int maxClasses; /* The maximum number of classes */ 140aa213bdcSJed Brown PetscClassRegInfo *classInfo; /* The structure for class information (classids are monotonicly increasing) */ 141c60ec953SBarry Smith }; 1426a6a9b46SSatish Balay 143aa213bdcSJed Brown typedef struct _n_PetscClassPerfLog *PetscClassPerfLog; 144aa213bdcSJed Brown struct _n_PetscClassPerfLog { 145c60ec953SBarry Smith int numClasses; /* The number of logging classes */ 146c60ec953SBarry Smith int maxClasses; /* The maximum number of classes */ 147aa213bdcSJed Brown PetscClassPerfInfo *classInfo; /* The structure for class information (classids are monotonicly increasing) */ 148c60ec953SBarry Smith }; 149c60ec953SBarry Smith /* -----------------------------------------------------------------------------------------------------*/ 150c60ec953SBarry Smith /* 151aa213bdcSJed Brown PetscEventRegInfo, PetscEventPerfInfo - Each event has two data structures associated with it. The first has 152c60ec953SBarry Smith static information about it, the second collects statistics on how many times the event is used, how 153c60ec953SBarry Smith much time it takes, etc. 154c60ec953SBarry Smith 155aa213bdcSJed Brown PetscEventRegLog, PetscEventPerfLog - an array of all PetscEventRegInfo and PetscEventPerfInfo for all events. There is one 156c60ec953SBarry Smith of these for each stage. 157c60ec953SBarry Smith 158c60ec953SBarry Smith */ 1596a6a9b46SSatish Balay typedef struct { 1606a6a9b46SSatish Balay char *name; /* The name of this event */ 161c60ec953SBarry Smith PetscClassId classid; /* The class the event is associated with */ 162217044c2SLisandro Dalcin PetscBool collective; /* Flag this event as collective */ 163a0c7f9aaSSamuel Khuvis #if defined(PETSC_HAVE_TAU_PERFSTUBS) 164a0c7f9aaSSamuel Khuvis void *timer; /* Associated external tool timer for this event */ 165a0c7f9aaSSamuel Khuvis #endif 1666a6a9b46SSatish Balay #if defined(PETSC_HAVE_MPE) 1676a6a9b46SSatish Balay int mpe_id_begin; /* MPE IDs that define the event */ 1686a6a9b46SSatish Balay int mpe_id_end; 1696a6a9b46SSatish Balay #endif 170aa213bdcSJed Brown } PetscEventRegInfo; 1716a6a9b46SSatish Balay 172c60ec953SBarry Smith typedef struct { 173c60ec953SBarry Smith int id; /* The integer identifying this event */ 174ace3abfcSBarry Smith PetscBool active; /* The flag to activate logging */ 175ace3abfcSBarry Smith PetscBool visible; /* The flag to print info in summary */ 176c60ec953SBarry Smith int depth; /* The nesting depth of the event call */ 177c60ec953SBarry Smith int count; /* The number of times this event was executed */ 178c708d6e3SStefano Zampini PetscLogDouble flops; /* The flops used in this event */ 179c708d6e3SStefano Zampini PetscLogDouble flops2; /* The square of flops used in this event */ 180c708d6e3SStefano Zampini PetscLogDouble flopsTmp; /* The accumulator for flops used in this event */ 181c708d6e3SStefano Zampini PetscLogDouble time; /* The time taken for this event */ 182c708d6e3SStefano Zampini PetscLogDouble time2; /* The square of time taken for this event */ 183c708d6e3SStefano Zampini PetscLogDouble timeTmp; /* The accumulator for time taken for this event */ 184217044c2SLisandro Dalcin PetscLogDouble syncTime; /* The synchronization barrier time */ 185891e75beSMatthew G. Knepley PetscLogDouble dof[8]; /* The number of degrees of freedom associated with this event */ 1865d68e14cSMatthew G. Knepley PetscLogDouble errors[8]; /* The errors (user-defined) associated with this event */ 187c60ec953SBarry Smith PetscLogDouble numMessages; /* The number of messages in this event */ 188c60ec953SBarry Smith PetscLogDouble messageLength; /* The total message lengths in this event */ 189c60ec953SBarry Smith PetscLogDouble numReductions; /* The number of reductions in this event */ 190e3ed9ee7SBarry Smith PetscLogDouble memIncrease; /* How much the resident memory has increased in this event */ 191e3ed9ee7SBarry Smith PetscLogDouble mallocIncrease; /* How much the maximum malloced space has increased in this event */ 192e3ed9ee7SBarry Smith PetscLogDouble mallocSpace; /* How much the space was malloced and kept during this event */ 193e3ed9ee7SBarry Smith PetscLogDouble mallocIncreaseEvent; /* Maximum of the high water mark with in event minus memory available at the end of the event */ 194091bc5acSJunchao Zhang #if defined(PETSC_HAVE_DEVICE) 195bec0b493Shannah_mairs PetscLogDouble CpuToGpuCount; /* The total number of CPU to GPU copies */ 196bec0b493Shannah_mairs PetscLogDouble GpuToCpuCount; /* The total number of GPU to CPU copies */ 197bec0b493Shannah_mairs PetscLogDouble CpuToGpuSize; /* The total size of CPU to GPU copies */ 198bec0b493Shannah_mairs PetscLogDouble GpuToCpuSize; /* The total size of GPU to CPU copies */ 199958c4211Shannah_mairs PetscLogDouble GpuFlops; /* The flops done on a GPU in this event */ 200958c4211Shannah_mairs PetscLogDouble GpuTime; /* The time spent on a GPU in this event */ 201bec0b493Shannah_mairs #endif 202aa213bdcSJed Brown } PetscEventPerfInfo; 203c60ec953SBarry Smith 204aa213bdcSJed Brown typedef struct _n_PetscEventRegLog *PetscEventRegLog; 205aa213bdcSJed Brown struct _n_PetscEventRegLog { 2066a6a9b46SSatish Balay int numEvents; /* The number of registered events */ 2076a6a9b46SSatish Balay int maxEvents; /* The maximum number of events */ 208aa213bdcSJed Brown PetscEventRegInfo *eventInfo; /* The registration information for each event */ 2096a6a9b46SSatish Balay }; 2106a6a9b46SSatish Balay 211aa213bdcSJed Brown typedef struct _n_PetscEventPerfLog *PetscEventPerfLog; 212aa213bdcSJed Brown struct _n_PetscEventPerfLog { 2136a6a9b46SSatish Balay int numEvents; /* The number of logging events */ 2146a6a9b46SSatish Balay int maxEvents; /* The maximum number of events */ 215aa213bdcSJed Brown PetscEventPerfInfo *eventInfo; /* The performance information for each event */ 2166a6a9b46SSatish Balay }; 217c60ec953SBarry Smith /* ------------------------------------------------------------------------------------------------------------*/ 218c60ec953SBarry Smith /* 219aa213bdcSJed Brown PetscStageInfo - Contains all the information about a particular stage. 2206a6a9b46SSatish Balay 221aa213bdcSJed Brown PetscStageLog - An array of PetscStageInfo for each registered stage. There is a single one of these in the code. 222c60ec953SBarry Smith */ 223aa213bdcSJed Brown typedef struct _PetscStageInfo { 2246a6a9b46SSatish Balay char *name; /* The stage name */ 225ace3abfcSBarry Smith PetscBool used; /* The stage was pushed on this processor */ 226aa213bdcSJed Brown PetscEventPerfInfo perfInfo; /* The stage performance information */ 227aa213bdcSJed Brown PetscEventPerfLog eventLog; /* The event information for this stage */ 228aa213bdcSJed Brown PetscClassPerfLog classLog; /* The class information for this stage */ 229a0c7f9aaSSamuel Khuvis #if defined(PETSC_HAVE_TAU_PERFSTUBS) 230a0c7f9aaSSamuel Khuvis void *timer; /* Associated external tool timer for this stage */ 231a0c7f9aaSSamuel Khuvis #endif 232aa213bdcSJed Brown } PetscStageInfo; 2336a6a9b46SSatish Balay 234aa213bdcSJed Brown typedef struct _n_PetscStageLog *PetscStageLog; 235aa213bdcSJed Brown struct _n_PetscStageLog { 2366a6a9b46SSatish Balay int numStages; /* The number of registered stages */ 2376a6a9b46SSatish Balay int maxStages; /* The maximum number of stages */ 238aa213bdcSJed Brown PetscIntStack stack; /* The stack for active stages */ 239aa213bdcSJed Brown int curStage; /* The current stage (only used in macros so we don't call PetscIntStackTop) */ 240aa213bdcSJed Brown PetscStageInfo *stageInfo; /* The information for each stage */ 241aa213bdcSJed Brown PetscEventRegLog eventLog; /* The registered events */ 242aa213bdcSJed Brown PetscClassRegLog classLog; /* The registered classes */ 2436a6a9b46SSatish Balay }; 244217044c2SLisandro Dalcin /* -----------------------------------------------------------------------------------------------------*/ 245fee518ebSJed Brown 246d71ae5a4SJacob Faibussowitsch PETSC_DEPRECATED_FUNCTION("PetscLogObjectParent() is deprecated (since version 3.18)") static inline PetscErrorCode PetscLogObjectParent(PetscObject o, PetscObject p) 247d71ae5a4SJacob Faibussowitsch { 2484f839811SJacob Faibussowitsch (void)o; 2494f839811SJacob Faibussowitsch (void)p; 2503ba16761SJacob Faibussowitsch return PETSC_SUCCESS; 2514dfa11a4SJacob Faibussowitsch } 2524dfa11a4SJacob Faibussowitsch 253d71ae5a4SJacob Faibussowitsch PETSC_DEPRECATED_FUNCTION("PetscLogObjectMemory() is deprecated (since version 3.18)") static inline PetscErrorCode PetscLogObjectMemory(PetscObject o, PetscLogDouble m) 254d71ae5a4SJacob Faibussowitsch { 2554f839811SJacob Faibussowitsch (void)o; 2564f839811SJacob Faibussowitsch (void)m; 2573ba16761SJacob Faibussowitsch return PETSC_SUCCESS; 2584dfa11a4SJacob Faibussowitsch } 2593bb1ff40SBarry Smith 260aa482453SBarry Smith #if defined(PETSC_USE_LOG) /* --- Logging is turned on --------------------------------*/ 261aeeda9f6SJohn Fettig PETSC_EXTERN PetscStageLog petsc_stageLog; 262f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogGetStageLog(PetscStageLog *); 263f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetCurrent(PetscStageLog, int *); 264f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetEventPerfLog(PetscStageLog, int, PetscEventPerfLog *); 265614700edSBarry Smith 266fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscGetFlops(PetscLogDouble *); 26777c4ece6SBarry Smith 268aa482453SBarry Smith #if defined(PETSC_HAVE_MPE) 269014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEBegin(void); 270014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEDump(const char[]); 27177c4ece6SBarry Smith #endif 27277c4ece6SBarry Smith 273014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLB)(PetscLogEvent, int, PetscObject, PetscObject, PetscObject, PetscObject); 274014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLE)(PetscLogEvent, int, PetscObject, PetscObject, PetscObject, PetscObject); 275014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHC)(PetscObject); 276014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHD)(PetscObject); 27777c4ece6SBarry Smith 2789566063dSJacob Faibussowitsch #define PetscLogObjectParents(p, n, d) PetscMacroReturnStandard(for (int _i = 0; _i < (n); ++_i) PetscCall(PetscLogObjectParent((PetscObject)(p), (PetscObject)(d)[_i]));) 2793ba16761SJacob Faibussowitsch #define PetscLogObjectCreate(h) ((PetscLogPHC) ? (*PetscLogPHC)((PetscObject)(h)) : PETSC_SUCCESS) 2803ba16761SJacob Faibussowitsch #define PetscLogObjectDestroy(h) ((PetscLogPHD) ? (*PetscLogPHD)((PetscObject)(h)) : PETSC_SUCCESS) 2813ca90d2dSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject, const char[], ...) PETSC_ATTRIBUTE_FORMAT(2, 3); 282fa2bb9feSLisandro Dalcin 2838ba1e511SMatthew Knepley /* Initialization functions */ 284bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogDefaultBegin(void); 285014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogAllBegin(void); 286bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogNestedBegin(void); 287014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogTraceBegin(FILE *); 288014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogActions(PetscBool); 289014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogObjects(PetscBool); 290fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogSetThreshold(PetscLogDouble, PetscLogDouble *); 2919371c9d4SSatish Balay PETSC_EXTERN PetscErrorCode PetscLogSet(PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject), PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject)); 292fa2bb9feSLisandro Dalcin 2938ba1e511SMatthew Knepley /* Output functions */ 294014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogView(PetscViewer); 295f14045dbSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogViewFromOptions(void); 296014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogDump(const char[]); 29731d06abdSBarry Smith 2984dd65854SConnor Ward /* Status checking functions */ 2994dd65854SConnor Ward PETSC_EXTERN PetscErrorCode PetscLogIsActive(PetscBool *); 3004dd65854SConnor Ward 301fa2bb9feSLisandro Dalcin /* Stage functions */ 302014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageRegister(const char[], PetscLogStage *); 303014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePush(PetscLogStage); 304014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePop(void); 305014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetActive(PetscLogStage, PetscBool); 306014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetActive(PetscLogStage, PetscBool *); 307014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetVisible(PetscLogStage, PetscBool); 308014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetVisible(PetscLogStage, PetscBool *); 309014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetId(const char[], PetscLogStage *); 310f5d6ab90SLisandro Dalcin 3118ba1e511SMatthew Knepley /* Event functions */ 312014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventRegister(const char[], PetscClassId, PetscLogEvent *); 313217044c2SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSetCollective(PetscLogEvent, PetscBool); 314fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventIncludeClass(PetscClassId); 315fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventExcludeClass(PetscClassId); 316014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivate(PetscLogEvent); 317014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivate(PetscLogEvent); 318c00cb57fSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventDeactivatePush(PetscLogEvent); 319c00cb57fSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventDeactivatePop(PetscLogEvent); 320014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventSetActiveAll(PetscLogEvent, PetscBool); 321014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivateClass(PetscClassId); 322014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivateClass(PetscClassId); 3238c49a8d9SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventGetId(const char[], PetscLogEvent *); 32413230571SMichael Lange PETSC_EXTERN PetscErrorCode PetscLogEventGetPerfInfo(int, PetscLogEvent, PetscEventPerfInfo *); 325891e75beSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetDof(PetscLogEvent, PetscInt, PetscLogDouble); 3265d68e14cSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetError(PetscLogEvent, PetscInt, PetscLogDouble); 3277a101e5eSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogPushCurrentEvent_Internal(PetscLogEvent); 3287a101e5eSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogPopCurrentEvent_Internal(void); 3298ba1e511SMatthew Knepley 330e3ed9ee7SBarry Smith PETSC_EXTERN PetscBool PetscLogMemory; 331e3ed9ee7SBarry Smith 332fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscBool PetscLogSyncOn; /* true if logging synchronization is enabled */ 33362872c28SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSynchronize(PetscLogEvent, MPI_Comm); 33462872c28SLisandro Dalcin 33562872c28SLisandro Dalcin #define PetscLogEventSync(e, comm) \ 3363ba16761SJacob Faibussowitsch ((PetscErrorCode)(((PetscLogPLB && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? PetscLogEventSynchronize((e), (comm)) : PETSC_SUCCESS))) 33777c4ece6SBarry Smith 33852e6d16bSBarry Smith #define PetscLogEventBegin(e, o1, o2, o3, o4) \ 3393ba16761SJacob Faibussowitsch ((PetscErrorCode)((PetscLogPLB && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? (PetscErrorCode)(((*PetscLogPLB)((e), 0, (PetscObject)(o1), (PetscObject)(o2), (PetscObject)(o3), (PetscObject)(o4))) || PetscLogPushCurrentEvent_Internal(e)) : PETSC_SUCCESS)) 340043328b6SSatish Balay 34152e6d16bSBarry Smith #define PetscLogEventEnd(e, o1, o2, o3, o4) \ 3423ba16761SJacob Faibussowitsch ((PetscErrorCode)((PetscLogPLE && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? (PetscErrorCode)(((*PetscLogPLE)((e), 0, (PetscObject)(o1), (PetscObject)(o2), (PetscObject)(o3), (PetscObject)(o4))) || PetscLogPopCurrentEvent_Internal()) : PETSC_SUCCESS)) 34377c4ece6SBarry Smith 344014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventGetFlops(PetscLogEvent, PetscLogDouble *); 345014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventZeroFlops(PetscLogEvent); 346003131ecSBarry Smith 347c708d6e3SStefano Zampini /* Global flop counter */ 348c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_TotalFlops; 349c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_irecv_ct; 350c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_isend_ct; 351c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_recv_ct; 352c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_send_ct; 353c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_irecv_len; 354c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_isend_len; 355c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_recv_len; 356c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_send_len; 357c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_allreduce_ct; 358c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gather_ct; 359c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_scatter_ct; 360c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_ct; 361c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_any_ct; 362c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_all_ct; 363c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_sum_of_waits_ct; 364c708d6e3SStefano Zampini 365c708d6e3SStefano Zampini /* Thread local storage */ 366c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_TotalFlops_th; 367c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_irecv_ct_th; 368c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_isend_ct_th; 369c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_recv_ct_th; 370c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_send_ct_th; 371c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_irecv_len_th; 372c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_isend_len_th; 373c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_recv_len_th; 374c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_send_len_th; 375c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_allreduce_ct_th; 376c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gather_ct_th; 377c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_scatter_ct_th; 378c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_ct_th; 379c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_any_ct_th; 380c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_all_ct_th; 381c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_sum_of_waits_ct_th; 382c708d6e3SStefano Zampini 383c708d6e3SStefano Zampini /* 384c708d6e3SStefano Zampini Flop counting: We count each arithmetic operation (e.g., addition, multiplication) separately. 385c708d6e3SStefano Zampini 386c708d6e3SStefano Zampini For the complex numbers version, note that 387c708d6e3SStefano Zampini 1 complex addition = 2 flops 388c708d6e3SStefano Zampini 1 complex multiplication = 6 flops, 389c708d6e3SStefano Zampini where we define 1 flop as that for a double precision scalar. We roughly approximate 390c708d6e3SStefano Zampini flop counting for complex numbers by multiplying the total flops by 4; this corresponds 391c708d6e3SStefano Zampini to the assumption that we're counting mostly additions and multiplications -- and 392c708d6e3SStefano Zampini roughly the same number of each. More accurate counting could be done by distinguishing 393c708d6e3SStefano Zampini among the various arithmetic operations. 394c708d6e3SStefano Zampini */ 395c708d6e3SStefano Zampini 396c708d6e3SStefano Zampini #if defined(PETSC_USE_COMPLEX) 397c708d6e3SStefano Zampini #define PETSC_FLOPS_PER_OP 4.0 398c708d6e3SStefano Zampini #else 399c708d6e3SStefano Zampini #define PETSC_FLOPS_PER_OP 1.0 400c708d6e3SStefano Zampini #endif 401c708d6e3SStefano Zampini 402c708d6e3SStefano Zampini /*@C 403c708d6e3SStefano Zampini PetscLogFlops - Log how many flops are performed in a calculation 404c708d6e3SStefano Zampini 405c708d6e3SStefano Zampini Input Parameter: 406c708d6e3SStefano Zampini . flops - the number of flops 407c708d6e3SStefano Zampini 408c708d6e3SStefano Zampini Level: intermediate 409c708d6e3SStefano Zampini 410c200f75eSSatish Balay Note: 411c200f75eSSatish Balay To limit the chance of integer overflow when multiplying by a constant, represent the constant as a double, 412c200f75eSSatish Balay not an integer. Use `PetscLogFlops`(4.0*n) not `PetscLogFlops`(4*n) 413c200f75eSSatish Balay 414c200f75eSSatish Balay .seealso: [](ch_profiling), `PetscLogView()`, `PetscLogGpuFlops()` 415c708d6e3SStefano Zampini @*/ 416c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogFlops(PetscLogDouble n) 417c708d6e3SStefano Zampini { 418c708d6e3SStefano Zampini PetscAssert(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Cannot log negative flops"); 419c708d6e3SStefano Zampini return PetscAddLogDouble(&petsc_TotalFlops, &petsc_TotalFlops_th, PETSC_FLOPS_PER_OP * n); 420c708d6e3SStefano Zampini } 421c708d6e3SStefano Zampini 422ce85283eSBarry Smith /* 423f621e05eSBarry Smith These are used internally in the PETSc routines to keep a count of MPI messages and 424f621e05eSBarry Smith their sizes. 425f621e05eSBarry Smith 426af0996ceSBarry Smith This does not work for MPI-Uni because our include/petsc/mpiuni/mpi.h file 427ce85283eSBarry Smith uses macros to defined the MPI operations. 42815308354SBarry Smith 42915308354SBarry Smith It does not work correctly from HP-UX because it processes the 430bb4af37aSBarry Smith macros in a way that sometimes it double counts, hence 431b6410449SSatish Balay PETSC_HAVE_BROKEN_RECURSIVE_MACRO 4327c1e34a4SSatish Balay 433f569fd43SBarry Smith It does not work with Windows because winmpich lacks MPI_Type_size() 434ce85283eSBarry Smith */ 435d5bc873cSPierre Jolivet #if !defined(MPIUNI_H) && !defined(PETSC_HAVE_BROKEN_RECURSIVE_MACRO) 43677a39924SBarry Smith /* 43777a39924SBarry Smith Logging of MPI activities 43877a39924SBarry Smith */ 439c708d6e3SStefano Zampini static inline PetscErrorCode PetscMPITypeSize(PetscInt count, MPI_Datatype type, PetscLogDouble *length, PetscLogDouble *length_th) 440d71ae5a4SJacob Faibussowitsch { 441f12d10ccSJunchao Zhang PetscMPIInt typesize; 4425f80ce2aSJacob Faibussowitsch 4433ba16761SJacob Faibussowitsch if (type == MPI_DATATYPE_NULL) return PETSC_SUCCESS; 4449566063dSJacob Faibussowitsch PetscCallMPI(MPI_Type_size(type, &typesize)); 445c708d6e3SStefano Zampini return PetscAddLogDouble(length, length_th, (PetscLogDouble)(count * typesize)); 446f95db71bSBarry Smith } 44777a39924SBarry Smith 448c708d6e3SStefano Zampini static inline PetscErrorCode PetscMPITypeSizeComm(MPI_Comm comm, const PetscMPIInt *counts, MPI_Datatype type, PetscLogDouble *length, PetscLogDouble *length_th) 449d71ae5a4SJacob Faibussowitsch { 450f12d10ccSJunchao Zhang PetscMPIInt typesize, size, p; 451c708d6e3SStefano Zampini PetscLogDouble l; 4523b9284c0SMatthew G Knepley 4533ba16761SJacob Faibussowitsch if (type == MPI_DATATYPE_NULL) return PETSC_SUCCESS; 4549566063dSJacob Faibussowitsch PetscCallMPI(MPI_Comm_size(comm, &size)); 4559566063dSJacob Faibussowitsch PetscCallMPI(MPI_Type_size(type, &typesize)); 456c708d6e3SStefano Zampini for (p = 0, l = 0.0; p < size; ++p) l += (PetscLogDouble)(counts[p] * typesize); 457c708d6e3SStefano Zampini return PetscAddLogDouble(length, length_th, l); 4583b9284c0SMatthew G Knepley } 4593b9284c0SMatthew G Knepley 46072b11a25SBarry Smith /* 46172b11a25SBarry Smith Returns 1 if the communicator is parallel else zero 46272b11a25SBarry Smith */ 463d71ae5a4SJacob Faibussowitsch static inline int PetscMPIParallelComm(MPI_Comm comm) 464d71ae5a4SJacob Faibussowitsch { 4659371c9d4SSatish Balay PetscMPIInt size; 4669371c9d4SSatish Balay MPI_Comm_size(comm, &size); 4679371c9d4SSatish Balay return size > 1; 46872b11a25SBarry Smith } 46972b11a25SBarry Smith 470c708d6e3SStefano Zampini #define MPI_Irecv(buf, count, datatype, source, tag, comm, request) \ 471c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_irecv_ct, &petsc_irecv_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_irecv_len), &(petsc_irecv_len_th)) || MPI_Irecv((buf), (count), (datatype), (source), (tag), (comm), (request))) 47215308354SBarry Smith 473c708d6e3SStefano Zampini #define MPI_Irecv_c(buf, count, datatype, source, tag, comm, request) \ 474c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_irecv_ct, &petsc_irecv_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_irecv_len), &(petsc_irecv_len_th)) || MPI_Irecv_c((buf), (count), (datatype), (source), (tag), (comm), (request))) 475c87b50c4SJunchao Zhang 476c708d6e3SStefano Zampini #define MPI_Isend(buf, count, datatype, dest, tag, comm, request) \ 477c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_isend_ct, &petsc_isend_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_isend_len), &(petsc_isend_len_th)) || MPI_Isend((buf), (count), (datatype), (dest), (tag), (comm), (request))) 47815308354SBarry Smith 479c708d6e3SStefano Zampini #define MPI_Isend_c(buf, count, datatype, dest, tag, comm, request) \ 480c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_isend_ct, &petsc_isend_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_isend_len), &(petsc_isend_len_th)) || MPI_Isend_c((buf), (count), (datatype), (dest), (tag), (comm), (request))) 481c87b50c4SJunchao Zhang 482c708d6e3SStefano Zampini #define MPI_Startall_irecv(count, datatype, number, requests) \ 483c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_irecv_ct, &petsc_irecv_ct_th, number) || PetscMPITypeSize((count), (datatype), &(petsc_irecv_len), &(petsc_irecv_len_th)) || ((number) && MPI_Startall((number), (requests)))) 4840d4b0b6cSBarry Smith 485c708d6e3SStefano Zampini #define MPI_Startall_isend(count, datatype, number, requests) \ 486c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_isend_ct, &petsc_isend_ct_th, number) || PetscMPITypeSize((count), (datatype), &(petsc_isend_len), &(petsc_isend_len_th)) || ((number) && MPI_Startall((number), (requests)))) 4870d4b0b6cSBarry Smith 488c708d6e3SStefano Zampini #define MPI_Start_isend(count, datatype, requests) (PetscAddLogDouble(&petsc_isend_ct, &petsc_isend_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_isend_len), (&petsc_isend_len_th)) || MPI_Start((requests))) 4890d4b0b6cSBarry Smith 490c708d6e3SStefano Zampini #define MPI_Recv(buf, count, datatype, source, tag, comm, status) \ 491c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_recv_ct, &petsc_recv_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_recv_len), (&petsc_recv_len_th)) || MPI_Recv((buf), (count), (datatype), (source), (tag), (comm), (status))) 49215308354SBarry Smith 493c708d6e3SStefano Zampini #define MPI_Recv_c(buf, count, datatype, source, tag, comm, status) \ 494c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_recv_ct, &petsc_recv_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_recv_len), &(petsc_recv_len_th)) || MPI_Recv_c((buf), (count), (datatype), (source), (tag), (comm), (status))) 495c87b50c4SJunchao Zhang 496c708d6e3SStefano Zampini #define MPI_Send(buf, count, datatype, dest, tag, comm) \ 497c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_send_ct, &petsc_send_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Send((buf), (count), (datatype), (dest), (tag), (comm))) 49877a39924SBarry Smith 499c708d6e3SStefano Zampini #define MPI_Send_c(buf, count, datatype, dest, tag, comm) \ 500c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_send_ct, &petsc_send_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Send_c((buf), (count), (datatype), (dest), (tag), (comm))) 501c87b50c4SJunchao Zhang 502c708d6e3SStefano Zampini #define MPI_Wait(request, status) (PetscAddLogDouble(&petsc_wait_ct, &petsc_wait_ct_th, 1) || PetscAddLogDouble(&petsc_sum_of_waits_ct, &petsc_sum_of_waits_ct_th, 1) || MPI_Wait((request), (status))) 50377a39924SBarry Smith 504c708d6e3SStefano Zampini #define MPI_Waitany(a, b, c, d) (PetscAddLogDouble(&petsc_wait_any_ct, &petsc_wait_any_ct_th, 1) || PetscAddLogDouble(&petsc_sum_of_waits_ct, &petsc_sum_of_waits_ct_th, 1) || MPI_Waitany((a), (b), (c), (d))) 50577a39924SBarry Smith 506c708d6e3SStefano Zampini #define MPI_Waitall(count, array_of_requests, array_of_statuses) \ 507c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_wait_all_ct, &petsc_wait_all_ct_th, 1) || PetscAddLogDouble(&petsc_sum_of_waits_ct, &petsc_sum_of_waits_ct_th, count) || MPI_Waitall((count), (array_of_requests), (array_of_statuses))) 50877a39924SBarry Smith 509c708d6e3SStefano Zampini #define MPI_Allreduce(sendbuf, recvbuf, count, datatype, op, comm) (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || MPI_Allreduce((sendbuf), (recvbuf), (count), (datatype), (op), (comm))) 5103914022bSBarry Smith 511c708d6e3SStefano Zampini #define MPI_Bcast(buffer, count, datatype, root, comm) (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || MPI_Bcast((buffer), (count), (datatype), (root), (comm))) 51254fe5c21SBarry Smith 513c708d6e3SStefano Zampini #define MPI_Reduce_scatter_block(sendbuf, recvbuf, recvcount, datatype, op, comm) \ 514c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || MPI_Reduce_scatter_block((sendbuf), (recvbuf), (recvcount), (datatype), (op), (comm))) 515d6e4c47cSJed Brown 5163b9284c0SMatthew G Knepley #define MPI_Alltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm) \ 517c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Alltoall((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm))) 5183b9284c0SMatthew G Knepley 5193b9284c0SMatthew G Knepley #define MPI_Alltoallv(sendbuf, sendcnts, sdispls, sendtype, recvbuf, recvcnts, rdispls, recvtype, comm) \ 520c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || PetscMPITypeSizeComm((comm), (sendcnts), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Alltoallv((sendbuf), (sendcnts), (sdispls), (sendtype), (recvbuf), (recvcnts), (rdispls), (recvtype), (comm))) 5213b9284c0SMatthew G Knepley 522c708d6e3SStefano Zampini #define MPI_Allgather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm) \ 523c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Allgather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm))) 52401faf4e4SMatthew Knepley 52501faf4e4SMatthew Knepley #define MPI_Allgatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, comm) \ 526c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Allgatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (comm))) 52701faf4e4SMatthew Knepley 52801faf4e4SMatthew Knepley #define MPI_Gather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm) \ 529c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, 1) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Gather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm))) 53001faf4e4SMatthew Knepley 53101faf4e4SMatthew Knepley #define MPI_Gatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, root, comm) \ 532c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, 1) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Gatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (root), (comm))) 53301faf4e4SMatthew Knepley 53401faf4e4SMatthew Knepley #define MPI_Scatter(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm) \ 535c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_scatter_ct, &petsc_scatter_ct_th, 1) || PetscMPITypeSize((recvcount), (recvtype), (&petsc_recv_len), &(petsc_recv_len_th)) || MPI_Scatter((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm))) 53601faf4e4SMatthew Knepley 53701faf4e4SMatthew Knepley #define MPI_Scatterv(sendbuf, sendcount, displs, sendtype, recvbuf, recvcount, recvtype, root, comm) \ 538c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_scatter_ct, &petsc_scatter_ct_th, 1) || PetscMPITypeSize((recvcount), (recvtype), (&petsc_recv_len), &(petsc_recv_len_th)) || MPI_Scatterv((sendbuf), (sendcount), (displs), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm))) 539f12d10ccSJunchao Zhang 540f12d10ccSJunchao Zhang #define MPI_Ialltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm, request) \ 541c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Ialltoall((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm), (request))) 542f12d10ccSJunchao Zhang 543f12d10ccSJunchao Zhang #define MPI_Ialltoallv(sendbuf, sendcnts, sdispls, sendtype, recvbuf, recvcnts, rdispls, recvtype, comm, request) \ 544c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || PetscMPITypeSizeComm((comm), (sendcnts), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Ialltoallv((sendbuf), (sendcnts), (sdispls), (sendtype), (recvbuf), (recvcnts), (rdispls), (recvtype), (comm), (request))) 545f12d10ccSJunchao Zhang 546f12d10ccSJunchao Zhang #define MPI_Iallgather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm, request) \ 547c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Iallgather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm), (request))) 548f12d10ccSJunchao Zhang 549f12d10ccSJunchao Zhang #define MPI_Iallgatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, comm, request) \ 550c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Iallgatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (comm), (request))) 551f12d10ccSJunchao Zhang 552f12d10ccSJunchao Zhang #define MPI_Igather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \ 553c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, 1) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Igather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm), (request))) 554f12d10ccSJunchao Zhang 555f12d10ccSJunchao Zhang #define MPI_Igatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, root, comm, request) \ 556c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, 1) || PetscMPITypeSize((sendcount), (sendtype), (&petsc_send_len), (&petsc_send_len_th)) || MPI_Igatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (root), (comm), (request))) 557f12d10ccSJunchao Zhang 558f12d10ccSJunchao Zhang #define MPI_Iscatter(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \ 559c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_scatter_ct, &petsc_scatter_ct_th, 1) || PetscMPITypeSize((recvcount), (recvtype), (&petsc_recv_len), (&petsc_recv_len_th)) || MPI_Iscatter((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm), (request))) 560f12d10ccSJunchao Zhang 561f12d10ccSJunchao Zhang #define MPI_Iscatterv(sendbuf, sendcount, displs, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \ 562c708d6e3SStefano Zampini (PetscAddLogDouble(&petsc_scatter_ct, &petsc_scatter_ct_th, 1) || PetscMPITypeSize((recvcount), (recvtype), (&petsc_recv_len), (&petsc_recv_len_th)) || MPI_Iscatterv((sendbuf), (sendcount), (displs), (sendtype), (recvbuf), (recvcount), (recvtype), (root), (comm), (request))) 563f12d10ccSJunchao Zhang 5640d4b0b6cSBarry Smith #else 5650d4b0b6cSBarry Smith 5669371c9d4SSatish Balay #define MPI_Startall_irecv(count, datatype, number, requests) ((number) && MPI_Startall((number), (requests))) 5670d4b0b6cSBarry Smith 5689371c9d4SSatish Balay #define MPI_Startall_isend(count, datatype, number, requests) ((number) && MPI_Startall((number), (requests))) 5690d4b0b6cSBarry Smith 5709371c9d4SSatish Balay #define MPI_Start_isend(count, datatype, requests) (MPI_Start((requests))) 5710d4b0b6cSBarry Smith 572c9b973beSBarry Smith #endif /* !MPIUNI_H && ! PETSC_HAVE_BROKEN_RECURSIVE_MACRO */ 573614700edSBarry Smith 574df8cf0b5SBarry Smith #else /* ---Logging is turned off --------------------------------------------*/ 575614700edSBarry Smith 576e3ed9ee7SBarry Smith #define PetscLogMemory PETSC_FALSE 577e3ed9ee7SBarry Smith 5783ba16761SJacob Faibussowitsch #define PetscLogFlops(n) ((void)(n), PETSC_SUCCESS) 5793ba16761SJacob Faibussowitsch #define PetscGetFlops(a) (*(a) = 0.0, PETSC_SUCCESS) 580614700edSBarry Smith 5813ba16761SJacob Faibussowitsch #define PetscLogStageRegister(a, b) PETSC_SUCCESS 5823ba16761SJacob Faibussowitsch #define PetscLogStagePush(a) PETSC_SUCCESS 5833ba16761SJacob Faibussowitsch #define PetscLogStagePop() PETSC_SUCCESS 5843ba16761SJacob Faibussowitsch #define PetscLogStageSetActive(a, b) PETSC_SUCCESS 5853ba16761SJacob Faibussowitsch #define PetscLogStageGetActive(a, b) PETSC_SUCCESS 5863ba16761SJacob Faibussowitsch #define PetscLogStageGetVisible(a, b) PETSC_SUCCESS 5873ba16761SJacob Faibussowitsch #define PetscLogStageSetVisible(a, b) PETSC_SUCCESS 5883ba16761SJacob Faibussowitsch #define PetscLogStageGetId(a, b) (*(b) = 0, PETSC_SUCCESS) 589f5d6ab90SLisandro Dalcin 5903ba16761SJacob Faibussowitsch #define PetscLogEventRegister(a, b, c) PETSC_SUCCESS 5913ba16761SJacob Faibussowitsch #define PetscLogEventSetCollective(a, b) PETSC_SUCCESS 5923ba16761SJacob Faibussowitsch #define PetscLogEventIncludeClass(a) PETSC_SUCCESS 5933ba16761SJacob Faibussowitsch #define PetscLogEventExcludeClass(a) PETSC_SUCCESS 5943ba16761SJacob Faibussowitsch #define PetscLogEventActivate(a) PETSC_SUCCESS 5953ba16761SJacob Faibussowitsch #define PetscLogEventDeactivate(a) PETSC_SUCCESS 5963ba16761SJacob Faibussowitsch #define PetscLogEventDeactivatePush(a) PETSC_SUCCESS 5973ba16761SJacob Faibussowitsch #define PetscLogEventDeactivatePop(a) PETSC_SUCCESS 5983ba16761SJacob Faibussowitsch #define PetscLogEventActivateClass(a) PETSC_SUCCESS 5993ba16761SJacob Faibussowitsch #define PetscLogEventDeactivateClass(a) PETSC_SUCCESS 6003ba16761SJacob Faibussowitsch #define PetscLogEventSetActiveAll(a, b) PETSC_SUCCESS 6013ba16761SJacob Faibussowitsch #define PetscLogEventGetId(a, b) (*(b) = 0, PETSC_SUCCESS) 6023ba16761SJacob Faibussowitsch #define PetscLogEventGetPerfInfo(a, b, c) PETSC_SUCCESS 6033ba16761SJacob Faibussowitsch #define PetscLogEventSetDof(a, b, c) PETSC_SUCCESS 6043ba16761SJacob Faibussowitsch #define PetscLogEventSetError(a, b, c) PETSC_SUCCESS 60577c4ece6SBarry Smith 6063ba16761SJacob Faibussowitsch #define PetscLogPLB PETSC_SUCCESS 6073ba16761SJacob Faibussowitsch #define PetscLogPLE PETSC_SUCCESS 6083ba16761SJacob Faibussowitsch #define PetscLogPHC PETSC_SUCCESS 6093ba16761SJacob Faibussowitsch #define PetscLogPHD PETSC_SUCCESS 610f5d6ab90SLisandro Dalcin 6113ba16761SJacob Faibussowitsch #define PetscLogObjectParents(p, n, c) PETSC_SUCCESS 6123ba16761SJacob Faibussowitsch #define PetscLogObjectCreate(h) PETSC_SUCCESS 6133ba16761SJacob Faibussowitsch #define PetscLogObjectDestroy(h) PETSC_SUCCESS 6143ca90d2dSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject, const char[], ...) PETSC_ATTRIBUTE_FORMAT(2, 3); 615fa2bb9feSLisandro Dalcin 6163ba16761SJacob Faibussowitsch #define PetscLogDefaultBegin() PETSC_SUCCESS 6173ba16761SJacob Faibussowitsch #define PetscLogAllBegin() PETSC_SUCCESS 6183ba16761SJacob Faibussowitsch #define PetscLogNestedBegin() PETSC_SUCCESS 6193ba16761SJacob Faibussowitsch #define PetscLogTraceBegin(file) PETSC_SUCCESS 6203ba16761SJacob Faibussowitsch #define PetscLogActions(a) PETSC_SUCCESS 6213ba16761SJacob Faibussowitsch #define PetscLogObjects(a) PETSC_SUCCESS 6223ba16761SJacob Faibussowitsch #define PetscLogSetThreshold(a, b) PETSC_SUCCESS 6233ba16761SJacob Faibussowitsch #define PetscLogSet(lb, le) PETSC_SUCCESS 6243ba16761SJacob Faibussowitsch #define PetscLogIsActive(flag) (*(flag) = PETSC_FALSE, PETSC_SUCCESS) 625fa2bb9feSLisandro Dalcin 6263ba16761SJacob Faibussowitsch #define PetscLogView(viewer) PETSC_SUCCESS 6273ba16761SJacob Faibussowitsch #define PetscLogViewFromOptions() PETSC_SUCCESS 6283ba16761SJacob Faibussowitsch #define PetscLogDump(c) PETSC_SUCCESS 629fa2bb9feSLisandro Dalcin 6303ba16761SJacob Faibussowitsch #define PetscLogEventSync(e, comm) PETSC_SUCCESS 6313ba16761SJacob Faibussowitsch #define PetscLogEventBegin(e, o1, o2, o3, o4) PETSC_SUCCESS 6323ba16761SJacob Faibussowitsch #define PetscLogEventEnd(e, o1, o2, o3, o4) PETSC_SUCCESS 633ce6058e1SBarry Smith 63423c86b1bSSatish Balay /* If PETSC_USE_LOG is NOT defined, these still need to be! */ 63523c86b1bSSatish Balay #define MPI_Startall_irecv(count, datatype, number, requests) ((number) && MPI_Startall(number, requests)) 63623c86b1bSSatish Balay #define MPI_Startall_isend(count, datatype, number, requests) ((number) && MPI_Startall(number, requests)) 63723c86b1bSSatish Balay #define MPI_Start_isend(count, datatype, requests) MPI_Start(requests) 63823c86b1bSSatish Balay 63923c86b1bSSatish Balay #endif /* PETSC_USE_LOG */ 64023c86b1bSSatish Balay 641204a6943SJed Brown #define PetscPreLoadBegin(flag, name) \ 642204a6943SJed Brown do { \ 643204a6943SJed Brown PetscBool PetscPreLoading = flag; \ 644204a6943SJed Brown int PetscPreLoadMax, PetscPreLoadIt; \ 645166c7f25SBarry Smith PetscLogStage _stageNum; \ 6469566063dSJacob Faibussowitsch PetscCall(PetscOptionsGetBool(NULL, NULL, "-preload", &PetscPreLoading, NULL)); \ 647204a6943SJed Brown PetscPreLoadMax = (int)(PetscPreLoading); \ 648204a6943SJed Brown PetscPreLoadingUsed = PetscPreLoading ? PETSC_TRUE : PetscPreLoadingUsed; \ 649204a6943SJed Brown for (PetscPreLoadIt = 0; PetscPreLoadIt <= PetscPreLoadMax; PetscPreLoadIt++) { \ 650204a6943SJed Brown PetscPreLoadingOn = PetscPreLoading; \ 6519566063dSJacob Faibussowitsch PetscCall(PetscBarrier(NULL)); \ 6529566063dSJacob Faibussowitsch if (PetscPreLoadIt > 0) PetscCall(PetscLogStageGetId(name, &_stageNum)); \ 6539566063dSJacob Faibussowitsch else PetscCall(PetscLogStageRegister(name, &_stageNum)); \ 6549566063dSJacob Faibussowitsch PetscCall(PetscLogStageSetActive(_stageNum, (PetscBool)(!PetscPreLoadMax || PetscPreLoadIt))); \ 6559566063dSJacob Faibussowitsch PetscCall(PetscLogStagePush(_stageNum)); 6568e58c17dSMatthew Knepley 657204a6943SJed Brown #define PetscPreLoadEnd() \ 6589566063dSJacob Faibussowitsch PetscCall(PetscLogStagePop()); \ 659204a6943SJed Brown PetscPreLoading = PETSC_FALSE; \ 660043328b6SSatish Balay } \ 6619371c9d4SSatish Balay } \ 6629371c9d4SSatish Balay while (0) 6638e58c17dSMatthew Knepley 6649371c9d4SSatish Balay #define PetscPreLoadStage(name) \ 6659371c9d4SSatish Balay do { \ 6669566063dSJacob Faibussowitsch PetscCall(PetscLogStagePop()); \ 6679566063dSJacob Faibussowitsch if (PetscPreLoadIt > 0) PetscCall(PetscLogStageGetId(name, &_stageNum)); \ 6689566063dSJacob Faibussowitsch else PetscCall(PetscLogStageRegister(name, &_stageNum)); \ 6699566063dSJacob Faibussowitsch PetscCall(PetscLogStageSetActive(_stageNum, (PetscBool)(!PetscPreLoadMax || PetscPreLoadIt))); \ 6709566063dSJacob Faibussowitsch PetscCall(PetscLogStagePush(_stageNum)); \ 671204a6943SJed Brown } while (0) 672e9fa29b7SSatish Balay 673ad14c47eSJed Brown /* some vars for logging */ 674ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingUsed; /* true if we are or have done preloading */ 675ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingOn; /* true if we are currently in a preloading calculation */ 676ad14c47eSJed Brown 677c708d6e3SStefano Zampini #if defined(PETSC_USE_LOG) && defined(PETSC_HAVE_DEVICE) 678c708d6e3SStefano Zampini 679c708d6e3SStefano Zampini /* Global GPU counters */ 680c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_ct; 681c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_ct; 682c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_sz; 683c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_sz; 684c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_ct_scalar; 685c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_ct_scalar; 686c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_sz_scalar; 687c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_sz_scalar; 688c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gflops; 689c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtime; 690c708d6e3SStefano Zampini 691c708d6e3SStefano Zampini /* Thread local storage */ 692c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_ct_th; 693c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_ct_th; 694c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_sz_th; 695c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_sz_th; 696c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_ct_scalar_th; 697c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_ct_scalar_th; 698c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_sz_scalar_th; 699c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_sz_scalar_th; 700c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gflops_th; 701c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtime_th; 702c708d6e3SStefano Zampini 703c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTime(void); 704c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTimeBegin(void); 705c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTimeEnd(void); 706c708d6e3SStefano Zampini 707c708d6e3SStefano Zampini /*@C 708c708d6e3SStefano Zampini PetscLogGpuFlops - Log how many flops are performed in a calculation on the device 709c708d6e3SStefano Zampini 710c708d6e3SStefano Zampini Input Parameter: 711c708d6e3SStefano Zampini . flops - the number of flops 712c708d6e3SStefano Zampini 713c200f75eSSatish Balay Level: intermediate 714c200f75eSSatish Balay 715c708d6e3SStefano Zampini Notes: 716c708d6e3SStefano Zampini To limit the chance of integer overflow when multiplying by a constant, represent the constant as a double, 717c200f75eSSatish Balay not an integer. Use `PetscLogFlops`(4.0*n) not `PetscLogFlops`(4*n) 718c708d6e3SStefano Zampini 719c708d6e3SStefano Zampini The values are also added to the total flop count for the MPI rank that is set with `PetscLogFlops()`; hence the number of flops 720c708d6e3SStefano Zampini just on the CPU would be the value from set from `PetscLogFlops()` minus the value set from `PetscLogGpuFlops()` 721c708d6e3SStefano Zampini 722c200f75eSSatish Balay .seealso: [](ch_profiling), `PetscLogView()`, `PetscLogFlops()`, `PetscLogGpuTimeBegin()`, `PetscLogGpuTimeEnd()` 723c708d6e3SStefano Zampini @*/ 724c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuFlops(PetscLogDouble n) 725c708d6e3SStefano Zampini { 726c708d6e3SStefano Zampini PetscAssert(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Cannot log negative flops"); 7273ba16761SJacob Faibussowitsch PetscCall(PetscAddLogDouble(&petsc_TotalFlops, &petsc_TotalFlops_th, PETSC_FLOPS_PER_OP * n)); 7283ba16761SJacob Faibussowitsch PetscCall(PetscAddLogDouble(&petsc_gflops, &petsc_gflops_th, PETSC_FLOPS_PER_OP * n)); 7293ba16761SJacob Faibussowitsch return PETSC_SUCCESS; 730c708d6e3SStefano Zampini } 731c708d6e3SStefano Zampini 732c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuTimeAdd(PetscLogDouble t) 733c708d6e3SStefano Zampini { 7343ba16761SJacob Faibussowitsch return PetscAddLogDouble(&petsc_gtime, &petsc_gtime_th, t); 735c708d6e3SStefano Zampini } 736c708d6e3SStefano Zampini 737c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogCpuToGpu(PetscLogDouble size) 738c708d6e3SStefano Zampini { 7393ba16761SJacob Faibussowitsch return PetscAddLogDoubleCnt(&petsc_ctog_ct, &petsc_ctog_sz, &petsc_ctog_ct_th, &petsc_ctog_sz_th, size); 740c708d6e3SStefano Zampini } 741c708d6e3SStefano Zampini 742c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuToCpu(PetscLogDouble size) 743c708d6e3SStefano Zampini { 7443ba16761SJacob Faibussowitsch return PetscAddLogDoubleCnt(&petsc_gtoc_ct, &petsc_gtoc_sz, &petsc_gtoc_ct_th, &petsc_gtoc_sz_th, size); 745c708d6e3SStefano Zampini } 746c708d6e3SStefano Zampini 747c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogCpuToGpuScalar(PetscLogDouble size) 748c708d6e3SStefano Zampini { 7493ba16761SJacob Faibussowitsch return PetscAddLogDoubleCnt(&petsc_ctog_ct_scalar, &petsc_ctog_sz_scalar, &petsc_ctog_ct_scalar_th, &petsc_ctog_sz_scalar_th, size); 750c708d6e3SStefano Zampini } 751c708d6e3SStefano Zampini 752c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuToCpuScalar(PetscLogDouble size) 753c708d6e3SStefano Zampini { 7543ba16761SJacob Faibussowitsch return PetscAddLogDoubleCnt(&petsc_gtoc_ct_scalar, &petsc_gtoc_sz_scalar, &petsc_gtoc_ct_scalar_th, &petsc_gtoc_sz_scalar_th, size); 755c708d6e3SStefano Zampini } 756c708d6e3SStefano Zampini #else 757c708d6e3SStefano Zampini 7583ba16761SJacob Faibussowitsch #define PetscLogCpuToGpu(a) PETSC_SUCCESS 7593ba16761SJacob Faibussowitsch #define PetscLogGpuToCpu(a) PETSC_SUCCESS 7603ba16761SJacob Faibussowitsch #define PetscLogCpuToGpuScalar(a) PETSC_SUCCESS 7613ba16761SJacob Faibussowitsch #define PetscLogGpuToCpuScalar(a) PETSC_SUCCESS 7623ba16761SJacob Faibussowitsch #define PetscLogGpuFlops(a) PETSC_SUCCESS 7633ba16761SJacob Faibussowitsch #define PetscLogGpuTime() PETSC_SUCCESS 7643ba16761SJacob Faibussowitsch #define PetscLogGpuTimeAdd(a) PETSC_SUCCESS 7653ba16761SJacob Faibussowitsch #define PetscLogGpuTimeBegin() PETSC_SUCCESS 7663ba16761SJacob Faibussowitsch #define PetscLogGpuTimeEnd() PETSC_SUCCESS 767c708d6e3SStefano Zampini 768c708d6e3SStefano Zampini #endif /* PETSC_USE_LOG && PETSC_HAVE_DEVICE */ 769c708d6e3SStefano Zampini 770c708d6e3SStefano Zampini /* remove TLS defines */ 771c708d6e3SStefano Zampini #undef PETSC_EXTERN_TLS 772c708d6e3SStefano Zampini #undef PETSC_TLS 773c708d6e3SStefano Zampini 77497bb86f7SLois Curfman McInnes #endif 775