xref: /petsc/include/petsclog.h (revision edd03b47ed9fa9b7ef9dd67df42d5194593a501d)
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 
20*edd03b47SJacob Faibussowitsch #define PetscInfo1(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
21*edd03b47SJacob Faibussowitsch #define PetscInfo2(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
22*edd03b47SJacob Faibussowitsch #define PetscInfo3(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
23*edd03b47SJacob Faibussowitsch #define PetscInfo4(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
24*edd03b47SJacob Faibussowitsch #define PetscInfo5(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
25*edd03b47SJacob Faibussowitsch #define PetscInfo6(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
26*edd03b47SJacob Faibussowitsch #define PetscInfo7(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
27*edd03b47SJacob Faibussowitsch #define PetscInfo8(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
28*edd03b47SJacob Faibussowitsch #define PetscInfo9(...) PETSC_DEPRECATED_MACRO(3, 17, 0, "PetscInfo()", ) PetscInfo(__VA_ARGS__)
297d3de750SJacob Faibussowitsch 
30e94e781bSJacob Faibussowitsch /*E
3116a05f60SBarry Smith   PetscInfoCommFlag - Describes the method by which to filter information displayed by `PetscInfo()` by communicator size
32e94e781bSJacob Faibussowitsch 
33667f096bSBarry 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 
41667f096bSBarry Smith   Note:
42667f096bSBarry Smith   Used as an input for `PetscInfoSetFilterCommSelf()`
43667f096bSBarry 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 
115c60ec953SBarry Smith /*
116aa213bdcSJed Brown     PetscClassRegInfo, PetscClassPerfInfo - Each class has two data structures associated with it. The first has
117c60ec953SBarry Smith        static information about it, the second collects statistics on how many objects of the class are created,
118c60ec953SBarry Smith        how much memory they use, etc.
119c60ec953SBarry Smith 
120aa213bdcSJed Brown     PetscClassRegLog, PetscClassPerfLog - arrays of the PetscClassRegInfo and PetscClassPerfInfo for all classes.
121c60ec953SBarry Smith */
1226a6a9b46SSatish Balay typedef struct {
123c60ec953SBarry Smith   char        *name;    /* The class name */
124c60ec953SBarry Smith   PetscClassId classid; /* The integer identifying this class */
125aa213bdcSJed Brown } PetscClassRegInfo;
1266a6a9b46SSatish Balay 
1276a6a9b46SSatish Balay typedef struct {
1280700a824SBarry Smith   PetscClassId   id;           /* The integer identifying this class */
1296a6a9b46SSatish Balay   int            creations;    /* The number of objects of this class created */
1306a6a9b46SSatish Balay   int            destructions; /* The number of objects of this class destroyed */
1314851f57dSBarry Smith   PetscLogDouble mem;          /* The total memory allocated by objects of this class; this is completely wrong and should possibly be removed */
1324851f57dSBarry Smith   PetscLogDouble descMem;      /* The total memory allocated by descendents of these objects; this is completely wrong and should possibly be removed */
133aa213bdcSJed Brown } PetscClassPerfInfo;
1346a6a9b46SSatish Balay 
135aa213bdcSJed Brown typedef struct _n_PetscClassRegLog *PetscClassRegLog;
136aa213bdcSJed Brown struct _n_PetscClassRegLog {
137c60ec953SBarry Smith   int                numClasses; /* The number of classes registered */
138c60ec953SBarry Smith   int                maxClasses; /* The maximum number of classes */
139aa213bdcSJed Brown   PetscClassRegInfo *classInfo;  /* The structure for class information (classids are monotonicly increasing) */
140c60ec953SBarry Smith };
1416a6a9b46SSatish Balay 
142aa213bdcSJed Brown typedef struct _n_PetscClassPerfLog *PetscClassPerfLog;
143aa213bdcSJed Brown struct _n_PetscClassPerfLog {
144c60ec953SBarry Smith   int                 numClasses; /* The number of logging classes */
145c60ec953SBarry Smith   int                 maxClasses; /* The maximum number of classes */
146aa213bdcSJed Brown   PetscClassPerfInfo *classInfo;  /* The structure for class information (classids are monotonicly increasing) */
147c60ec953SBarry Smith };
14816a05f60SBarry Smith 
149c60ec953SBarry Smith /*
150aa213bdcSJed Brown     PetscEventRegInfo, PetscEventPerfInfo - Each event has two data structures associated with it. The first has
151c60ec953SBarry Smith        static information about it, the second collects statistics on how many times the event is used, how
152c60ec953SBarry Smith        much time it takes, etc.
153c60ec953SBarry Smith 
154aa213bdcSJed Brown     PetscEventRegLog, PetscEventPerfLog - an array of all PetscEventRegInfo and PetscEventPerfInfo for all events. There is one
155c60ec953SBarry Smith       of these for each stage.
156c60ec953SBarry Smith 
157c60ec953SBarry Smith */
1586a6a9b46SSatish Balay typedef struct {
1596a6a9b46SSatish Balay   char        *name;       /* The name of this event */
160c60ec953SBarry Smith   PetscClassId classid;    /* The class the event is associated with */
161217044c2SLisandro Dalcin   PetscBool    collective; /* Flag this event as collective */
162a0c7f9aaSSamuel Khuvis #if defined(PETSC_HAVE_TAU_PERFSTUBS)
163a0c7f9aaSSamuel Khuvis   void *timer; /* Associated external tool timer for this event */
164a0c7f9aaSSamuel Khuvis #endif
1656a6a9b46SSatish Balay #if defined(PETSC_HAVE_MPE)
1666a6a9b46SSatish Balay   int mpe_id_begin; /* MPE IDs that define the event */
1676a6a9b46SSatish Balay   int mpe_id_end;
1686a6a9b46SSatish Balay #endif
169aa213bdcSJed Brown } PetscEventRegInfo;
1706a6a9b46SSatish Balay 
171c60ec953SBarry Smith typedef struct {
172c60ec953SBarry Smith   int            id;                  /* The integer identifying this event */
173ace3abfcSBarry Smith   PetscBool      active;              /* The flag to activate logging */
174ace3abfcSBarry Smith   PetscBool      visible;             /* The flag to print info in summary */
175c60ec953SBarry Smith   int            depth;               /* The nesting depth of the event call */
176c60ec953SBarry Smith   int            count;               /* The number of times this event was executed */
177c708d6e3SStefano Zampini   PetscLogDouble flops;               /* The flops used in this event */
178c708d6e3SStefano Zampini   PetscLogDouble flops2;              /* The square of flops used in this event */
179c708d6e3SStefano Zampini   PetscLogDouble flopsTmp;            /* The accumulator for flops used in this event */
180c708d6e3SStefano Zampini   PetscLogDouble time;                /* The time taken for this event */
181c708d6e3SStefano Zampini   PetscLogDouble time2;               /* The square of time taken for this event */
182c708d6e3SStefano Zampini   PetscLogDouble timeTmp;             /* The accumulator for time taken for this event */
183217044c2SLisandro Dalcin   PetscLogDouble syncTime;            /* The synchronization barrier time */
184891e75beSMatthew G. Knepley   PetscLogDouble dof[8];              /* The number of degrees of freedom associated with this event */
1855d68e14cSMatthew G. Knepley   PetscLogDouble errors[8];           /* The errors (user-defined) associated with this event */
186c60ec953SBarry Smith   PetscLogDouble numMessages;         /* The number of messages in this event */
187c60ec953SBarry Smith   PetscLogDouble messageLength;       /* The total message lengths in this event */
188c60ec953SBarry Smith   PetscLogDouble numReductions;       /* The number of reductions in this event */
189e3ed9ee7SBarry Smith   PetscLogDouble memIncrease;         /* How much the resident memory has increased in this event */
190e3ed9ee7SBarry Smith   PetscLogDouble mallocIncrease;      /* How much the maximum malloced space has increased in this event */
191e3ed9ee7SBarry Smith   PetscLogDouble mallocSpace;         /* How much the space was malloced and kept during this event */
192e3ed9ee7SBarry Smith   PetscLogDouble mallocIncreaseEvent; /* Maximum of the high water mark with in event minus memory available at the end of the event */
193091bc5acSJunchao Zhang #if defined(PETSC_HAVE_DEVICE)
194bec0b493Shannah_mairs   PetscLogDouble CpuToGpuCount; /* The total number of CPU to GPU copies */
195bec0b493Shannah_mairs   PetscLogDouble GpuToCpuCount; /* The total number of GPU to CPU copies */
196bec0b493Shannah_mairs   PetscLogDouble CpuToGpuSize;  /* The total size of CPU to GPU copies */
197bec0b493Shannah_mairs   PetscLogDouble GpuToCpuSize;  /* The total size of GPU to CPU copies */
198958c4211Shannah_mairs   PetscLogDouble GpuFlops;      /* The flops done on a GPU in this event */
199958c4211Shannah_mairs   PetscLogDouble GpuTime;       /* The time spent on a GPU in this event */
200bec0b493Shannah_mairs #endif
201aa213bdcSJed Brown } PetscEventPerfInfo;
202c60ec953SBarry Smith 
203aa213bdcSJed Brown typedef struct _n_PetscEventRegLog *PetscEventRegLog;
204aa213bdcSJed Brown struct _n_PetscEventRegLog {
2056a6a9b46SSatish Balay   int                numEvents; /* The number of registered events */
2066a6a9b46SSatish Balay   int                maxEvents; /* The maximum number of events */
207aa213bdcSJed Brown   PetscEventRegInfo *eventInfo; /* The registration information for each event */
2086a6a9b46SSatish Balay };
2096a6a9b46SSatish Balay 
210aa213bdcSJed Brown typedef struct _n_PetscEventPerfLog *PetscEventPerfLog;
211aa213bdcSJed Brown struct _n_PetscEventPerfLog {
2126a6a9b46SSatish Balay   int                 numEvents; /* The number of logging events */
2136a6a9b46SSatish Balay   int                 maxEvents; /* The maximum number of events */
214aa213bdcSJed Brown   PetscEventPerfInfo *eventInfo; /* The performance information for each event */
2156a6a9b46SSatish Balay };
21616a05f60SBarry Smith 
217c60ec953SBarry Smith /*
218aa213bdcSJed Brown    PetscStageInfo - Contains all the information about a particular stage.
2196a6a9b46SSatish Balay 
220aa213bdcSJed Brown    PetscStageLog - An array of PetscStageInfo for each registered stage. There is a single one of these in the code.
221c60ec953SBarry Smith */
222aa213bdcSJed Brown typedef struct _PetscStageInfo {
2236a6a9b46SSatish Balay   char              *name;     /* The stage name */
224ace3abfcSBarry Smith   PetscBool          used;     /* The stage was pushed on this processor */
225aa213bdcSJed Brown   PetscEventPerfInfo perfInfo; /* The stage performance information */
226aa213bdcSJed Brown   PetscEventPerfLog  eventLog; /* The event information for this stage */
227aa213bdcSJed Brown   PetscClassPerfLog  classLog; /* The class information for this stage */
228a0c7f9aaSSamuel Khuvis #if defined(PETSC_HAVE_TAU_PERFSTUBS)
229a0c7f9aaSSamuel Khuvis   void *timer; /* Associated external tool timer for this stage */
230a0c7f9aaSSamuel Khuvis #endif
231aa213bdcSJed Brown } PetscStageInfo;
2326a6a9b46SSatish Balay 
233aa213bdcSJed Brown typedef struct _n_PetscStageLog *PetscStageLog;
234aa213bdcSJed Brown struct _n_PetscStageLog {
2356a6a9b46SSatish Balay   int              numStages; /* The number of registered stages */
2366a6a9b46SSatish Balay   int              maxStages; /* The maximum number of stages */
237aa213bdcSJed Brown   PetscIntStack    stack;     /* The stack for active stages */
238aa213bdcSJed Brown   int              curStage;  /* The current stage (only used in macros so we don't call PetscIntStackTop) */
239aa213bdcSJed Brown   PetscStageInfo  *stageInfo; /* The information for each stage */
240aa213bdcSJed Brown   PetscEventRegLog eventLog;  /* The registered events */
241aa213bdcSJed Brown   PetscClassRegLog classLog;  /* The registered classes */
2426a6a9b46SSatish Balay };
243fee518ebSJed Brown 
244*edd03b47SJacob Faibussowitsch PETSC_DEPRECATED_FUNCTION(3, 18, 0, "PetscLogObjectParent()", ) static inline PetscErrorCode PetscLogObjectParent(PetscObject o, PetscObject p)
245d71ae5a4SJacob Faibussowitsch {
2464f839811SJacob Faibussowitsch   (void)o;
2474f839811SJacob Faibussowitsch   (void)p;
2483ba16761SJacob Faibussowitsch   return PETSC_SUCCESS;
2494dfa11a4SJacob Faibussowitsch }
2504dfa11a4SJacob Faibussowitsch 
251*edd03b47SJacob Faibussowitsch PETSC_DEPRECATED_FUNCTION(3, 18, 0, "PetscLogObjectMemory()", ) static inline PetscErrorCode PetscLogObjectMemory(PetscObject o, PetscLogDouble m)
252d71ae5a4SJacob Faibussowitsch {
2534f839811SJacob Faibussowitsch   (void)o;
2544f839811SJacob Faibussowitsch   (void)m;
2553ba16761SJacob Faibussowitsch   return PETSC_SUCCESS;
2564dfa11a4SJacob Faibussowitsch }
2573bb1ff40SBarry Smith 
258aa482453SBarry Smith #if defined(PETSC_USE_LOG) /* --- Logging is turned on --------------------------------*/
259aeeda9f6SJohn Fettig PETSC_EXTERN PetscStageLog  petsc_stageLog;
260f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogGetStageLog(PetscStageLog *);
261f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetCurrent(PetscStageLog, int *);
262f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetEventPerfLog(PetscStageLog, int, PetscEventPerfLog *);
263614700edSBarry Smith 
264fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscGetFlops(PetscLogDouble *);
26577c4ece6SBarry Smith 
266aa482453SBarry Smith   #if defined(PETSC_HAVE_MPE)
267014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEBegin(void);
268014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEDump(const char[]);
26977c4ece6SBarry Smith   #endif
27077c4ece6SBarry Smith 
271014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLB)(PetscLogEvent, int, PetscObject, PetscObject, PetscObject, PetscObject);
272014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLE)(PetscLogEvent, int, PetscObject, PetscObject, PetscObject, PetscObject);
273014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHC)(PetscObject);
274014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHD)(PetscObject);
27577c4ece6SBarry Smith 
2769566063dSJacob Faibussowitsch   #define PetscLogObjectParents(p, n, d) PetscMacroReturnStandard(for (int _i = 0; _i < (n); ++_i) PetscCall(PetscLogObjectParent((PetscObject)(p), (PetscObject)(d)[_i]));)
2773ba16761SJacob Faibussowitsch   #define PetscLogObjectCreate(h)        ((PetscLogPHC) ? (*PetscLogPHC)((PetscObject)(h)) : PETSC_SUCCESS)
2783ba16761SJacob Faibussowitsch   #define PetscLogObjectDestroy(h)       ((PetscLogPHD) ? (*PetscLogPHD)((PetscObject)(h)) : PETSC_SUCCESS)
2793ca90d2dSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject, const char[], ...) PETSC_ATTRIBUTE_FORMAT(2, 3);
280fa2bb9feSLisandro Dalcin 
2818ba1e511SMatthew Knepley /* Initialization functions */
282bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogDefaultBegin(void);
283014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogAllBegin(void);
284bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogNestedBegin(void);
285014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogTraceBegin(FILE *);
286014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogActions(PetscBool);
287014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogObjects(PetscBool);
288fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogSetThreshold(PetscLogDouble, PetscLogDouble *);
2899371c9d4SSatish Balay PETSC_EXTERN PetscErrorCode PetscLogSet(PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject), PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject));
290fa2bb9feSLisandro Dalcin 
2918ba1e511SMatthew Knepley /* Output functions */
292014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogView(PetscViewer);
293f14045dbSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogViewFromOptions(void);
294014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogDump(const char[]);
29531d06abdSBarry Smith 
2964dd65854SConnor Ward /* Status checking functions */
2974dd65854SConnor Ward PETSC_EXTERN PetscErrorCode PetscLogIsActive(PetscBool *);
2984dd65854SConnor Ward 
299fa2bb9feSLisandro Dalcin /* Stage functions */
300014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageRegister(const char[], PetscLogStage *);
301014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePush(PetscLogStage);
302014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePop(void);
303014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetActive(PetscLogStage, PetscBool);
304014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetActive(PetscLogStage, PetscBool *);
305014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetVisible(PetscLogStage, PetscBool);
306014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetVisible(PetscLogStage, PetscBool *);
307014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetId(const char[], PetscLogStage *);
308f5d6ab90SLisandro Dalcin 
3098ba1e511SMatthew Knepley /* Event functions */
310014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventRegister(const char[], PetscClassId, PetscLogEvent *);
311217044c2SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSetCollective(PetscLogEvent, PetscBool);
312fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventIncludeClass(PetscClassId);
313fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventExcludeClass(PetscClassId);
314014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivate(PetscLogEvent);
315014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivate(PetscLogEvent);
316c00cb57fSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventDeactivatePush(PetscLogEvent);
317c00cb57fSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventDeactivatePop(PetscLogEvent);
318014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventSetActiveAll(PetscLogEvent, PetscBool);
319014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivateClass(PetscClassId);
320014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivateClass(PetscClassId);
3218c49a8d9SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventGetId(const char[], PetscLogEvent *);
32213230571SMichael Lange PETSC_EXTERN PetscErrorCode PetscLogEventGetPerfInfo(int, PetscLogEvent, PetscEventPerfInfo *);
323891e75beSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetDof(PetscLogEvent, PetscInt, PetscLogDouble);
3245d68e14cSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetError(PetscLogEvent, PetscInt, PetscLogDouble);
3257a101e5eSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogPushCurrentEvent_Internal(PetscLogEvent);
3267a101e5eSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogPopCurrentEvent_Internal(void);
3278ba1e511SMatthew Knepley 
328e3ed9ee7SBarry Smith PETSC_EXTERN PetscBool PetscLogMemory;
329e3ed9ee7SBarry Smith 
330fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscBool      PetscLogSyncOn; /* true if logging synchronization is enabled */
33162872c28SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSynchronize(PetscLogEvent, MPI_Comm);
33262872c28SLisandro Dalcin 
33362872c28SLisandro Dalcin   #define PetscLogEventSync(e, comm) \
3343ba16761SJacob 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)))
33577c4ece6SBarry Smith 
33652e6d16bSBarry Smith   #define PetscLogEventBegin(e, o1, o2, o3, o4) \
3373ba16761SJacob 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))
338043328b6SSatish Balay 
33952e6d16bSBarry Smith   #define PetscLogEventEnd(e, o1, o2, o3, o4) \
3403ba16761SJacob 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))
34177c4ece6SBarry Smith 
342014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventGetFlops(PetscLogEvent, PetscLogDouble *);
343014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventZeroFlops(PetscLogEvent);
344003131ecSBarry Smith 
345c708d6e3SStefano Zampini /* Global flop counter */
346c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_TotalFlops;
347c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_irecv_ct;
348c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_isend_ct;
349c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_recv_ct;
350c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_send_ct;
351c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_irecv_len;
352c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_isend_len;
353c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_recv_len;
354c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_send_len;
355c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_allreduce_ct;
356c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gather_ct;
357c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_scatter_ct;
358c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_ct;
359c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_any_ct;
360c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_wait_all_ct;
361c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_sum_of_waits_ct;
362c708d6e3SStefano Zampini 
363c708d6e3SStefano Zampini /* Thread local storage */
364c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_TotalFlops_th;
365c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_irecv_ct_th;
366c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_isend_ct_th;
367c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_recv_ct_th;
368c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_send_ct_th;
369c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_irecv_len_th;
370c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_isend_len_th;
371c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_recv_len_th;
372c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_send_len_th;
373c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_allreduce_ct_th;
374c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gather_ct_th;
375c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_scatter_ct_th;
376c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_ct_th;
377c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_any_ct_th;
378c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_wait_all_ct_th;
379c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_sum_of_waits_ct_th;
380c708d6e3SStefano Zampini 
381c708d6e3SStefano Zampini   /*
382c708d6e3SStefano Zampini    Flop counting:  We count each arithmetic operation (e.g., addition, multiplication) separately.
383c708d6e3SStefano Zampini 
384c708d6e3SStefano Zampini    For the complex numbers version, note that
385c708d6e3SStefano Zampini        1 complex addition = 2 flops
386c708d6e3SStefano Zampini        1 complex multiplication = 6 flops,
387c708d6e3SStefano Zampini    where we define 1 flop as that for a double precision scalar.  We roughly approximate
388c708d6e3SStefano Zampini    flop counting for complex numbers by multiplying the total flops by 4; this corresponds
389c708d6e3SStefano Zampini    to the assumption that we're counting mostly additions and multiplications -- and
390c708d6e3SStefano Zampini    roughly the same number of each.  More accurate counting could be done by distinguishing
391c708d6e3SStefano Zampini    among the various arithmetic operations.
392c708d6e3SStefano Zampini  */
393c708d6e3SStefano Zampini 
394c708d6e3SStefano Zampini   #if defined(PETSC_USE_COMPLEX)
395c708d6e3SStefano Zampini     #define PETSC_FLOPS_PER_OP 4.0
396c708d6e3SStefano Zampini   #else
397c708d6e3SStefano Zampini     #define PETSC_FLOPS_PER_OP 1.0
398c708d6e3SStefano Zampini   #endif
399c708d6e3SStefano Zampini 
400c708d6e3SStefano Zampini /*@C
401c708d6e3SStefano Zampini        PetscLogFlops - Log how many flops are performed in a calculation
402c708d6e3SStefano Zampini 
403c708d6e3SStefano Zampini    Input Parameter:
404c708d6e3SStefano Zampini .   flops - the number of flops
405c708d6e3SStefano Zampini 
406c708d6e3SStefano Zampini    Level: intermediate
407c708d6e3SStefano Zampini 
408c200f75eSSatish Balay    Note:
409c200f75eSSatish Balay      To limit the chance of integer overflow when multiplying by a constant, represent the constant as a double,
410c200f75eSSatish Balay      not an integer. Use `PetscLogFlops`(4.0*n) not `PetscLogFlops`(4*n)
411c200f75eSSatish Balay 
412c200f75eSSatish Balay .seealso: [](ch_profiling), `PetscLogView()`, `PetscLogGpuFlops()`
413c708d6e3SStefano Zampini @*/
414c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogFlops(PetscLogDouble n)
415c708d6e3SStefano Zampini {
416c708d6e3SStefano Zampini   PetscAssert(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Cannot log negative flops");
417c708d6e3SStefano Zampini   return PetscAddLogDouble(&petsc_TotalFlops, &petsc_TotalFlops_th, PETSC_FLOPS_PER_OP * n);
418c708d6e3SStefano Zampini }
419c708d6e3SStefano Zampini 
420ce85283eSBarry Smith   /*
421f621e05eSBarry Smith      These are used internally in the PETSc routines to keep a count of MPI messages and
422f621e05eSBarry Smith    their sizes.
423f621e05eSBarry Smith 
424af0996ceSBarry Smith      This does not work for MPI-Uni because our include/petsc/mpiuni/mpi.h file
425ce85283eSBarry Smith    uses macros to defined the MPI operations.
42615308354SBarry Smith 
42715308354SBarry Smith      It does not work correctly from HP-UX because it processes the
428bb4af37aSBarry Smith    macros in a way that sometimes it double counts, hence
429b6410449SSatish Balay    PETSC_HAVE_BROKEN_RECURSIVE_MACRO
4307c1e34a4SSatish Balay 
431f569fd43SBarry Smith      It does not work with Windows because winmpich lacks MPI_Type_size()
432ce85283eSBarry Smith */
433d5bc873cSPierre Jolivet   #if !defined(MPIUNI_H) && !defined(PETSC_HAVE_BROKEN_RECURSIVE_MACRO)
43477a39924SBarry Smith /*
43577a39924SBarry Smith    Logging of MPI activities
43677a39924SBarry Smith */
437c708d6e3SStefano Zampini static inline PetscErrorCode PetscMPITypeSize(PetscInt count, MPI_Datatype type, PetscLogDouble *length, PetscLogDouble *length_th)
438d71ae5a4SJacob Faibussowitsch {
439f12d10ccSJunchao Zhang   PetscMPIInt typesize;
4405f80ce2aSJacob Faibussowitsch 
4413ba16761SJacob Faibussowitsch   if (type == MPI_DATATYPE_NULL) return PETSC_SUCCESS;
4429566063dSJacob Faibussowitsch   PetscCallMPI(MPI_Type_size(type, &typesize));
443c708d6e3SStefano Zampini   return PetscAddLogDouble(length, length_th, (PetscLogDouble)(count * typesize));
444f95db71bSBarry Smith }
44577a39924SBarry Smith 
446c708d6e3SStefano Zampini static inline PetscErrorCode PetscMPITypeSizeComm(MPI_Comm comm, const PetscMPIInt *counts, MPI_Datatype type, PetscLogDouble *length, PetscLogDouble *length_th)
447d71ae5a4SJacob Faibussowitsch {
448f12d10ccSJunchao Zhang   PetscMPIInt    typesize, size, p;
449c708d6e3SStefano Zampini   PetscLogDouble l;
4503b9284c0SMatthew G Knepley 
4513ba16761SJacob Faibussowitsch   if (type == MPI_DATATYPE_NULL) return PETSC_SUCCESS;
4529566063dSJacob Faibussowitsch   PetscCallMPI(MPI_Comm_size(comm, &size));
4539566063dSJacob Faibussowitsch   PetscCallMPI(MPI_Type_size(type, &typesize));
454c708d6e3SStefano Zampini   for (p = 0, l = 0.0; p < size; ++p) l += (PetscLogDouble)(counts[p] * typesize);
455c708d6e3SStefano Zampini   return PetscAddLogDouble(length, length_th, l);
4563b9284c0SMatthew G Knepley }
4573b9284c0SMatthew G Knepley 
45872b11a25SBarry Smith /*
45972b11a25SBarry Smith     Returns 1 if the communicator is parallel else zero
46072b11a25SBarry Smith */
461d71ae5a4SJacob Faibussowitsch static inline int PetscMPIParallelComm(MPI_Comm comm)
462d71ae5a4SJacob Faibussowitsch {
4639371c9d4SSatish Balay   PetscMPIInt size;
4649371c9d4SSatish Balay   MPI_Comm_size(comm, &size);
4659371c9d4SSatish Balay   return size > 1;
46672b11a25SBarry Smith }
46772b11a25SBarry Smith 
468c708d6e3SStefano Zampini     #define MPI_Irecv(buf, count, datatype, source, tag, comm, request) \
469c708d6e3SStefano 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)))
47015308354SBarry Smith 
471c708d6e3SStefano Zampini     #define MPI_Irecv_c(buf, count, datatype, source, tag, comm, request) \
472c708d6e3SStefano 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)))
473c87b50c4SJunchao Zhang 
474c708d6e3SStefano Zampini     #define MPI_Isend(buf, count, datatype, dest, tag, comm, request) \
475c708d6e3SStefano 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)))
47615308354SBarry Smith 
477c708d6e3SStefano Zampini     #define MPI_Isend_c(buf, count, datatype, dest, tag, comm, request) \
478c708d6e3SStefano 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)))
479c87b50c4SJunchao Zhang 
480c708d6e3SStefano Zampini     #define MPI_Startall_irecv(count, datatype, number, requests) \
481c708d6e3SStefano 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))))
4820d4b0b6cSBarry Smith 
483c708d6e3SStefano Zampini     #define MPI_Startall_isend(count, datatype, number, requests) \
484c708d6e3SStefano 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))))
4850d4b0b6cSBarry Smith 
486c708d6e3SStefano 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)))
4870d4b0b6cSBarry Smith 
488c708d6e3SStefano Zampini     #define MPI_Recv(buf, count, datatype, source, tag, comm, status) \
489c708d6e3SStefano 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)))
49015308354SBarry Smith 
491c708d6e3SStefano Zampini     #define MPI_Recv_c(buf, count, datatype, source, tag, comm, status) \
492c708d6e3SStefano 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)))
493c87b50c4SJunchao Zhang 
494c708d6e3SStefano Zampini     #define MPI_Send(buf, count, datatype, dest, tag, comm) \
495c708d6e3SStefano 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)))
49677a39924SBarry Smith 
497c708d6e3SStefano Zampini     #define MPI_Send_c(buf, count, datatype, dest, tag, comm) \
498c708d6e3SStefano 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)))
499c87b50c4SJunchao Zhang 
500c708d6e3SStefano 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)))
50177a39924SBarry Smith 
502c708d6e3SStefano 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)))
50377a39924SBarry Smith 
504c708d6e3SStefano Zampini     #define MPI_Waitall(count, array_of_requests, array_of_statuses) \
505c708d6e3SStefano 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)))
50677a39924SBarry Smith 
507c708d6e3SStefano 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)))
5083914022bSBarry Smith 
509c708d6e3SStefano 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)))
51054fe5c21SBarry Smith 
511c708d6e3SStefano Zampini     #define MPI_Reduce_scatter_block(sendbuf, recvbuf, recvcount, datatype, op, comm) \
512c708d6e3SStefano Zampini       (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || MPI_Reduce_scatter_block((sendbuf), (recvbuf), (recvcount), (datatype), (op), (comm)))
513d6e4c47cSJed Brown 
5143b9284c0SMatthew G Knepley     #define MPI_Alltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm) \
515c708d6e3SStefano 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)))
5163b9284c0SMatthew G Knepley 
5173b9284c0SMatthew G Knepley     #define MPI_Alltoallv(sendbuf, sendcnts, sdispls, sendtype, recvbuf, recvcnts, rdispls, recvtype, comm) \
518c708d6e3SStefano 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)))
5193b9284c0SMatthew G Knepley 
520c708d6e3SStefano Zampini     #define MPI_Allgather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm) \
521c708d6e3SStefano Zampini       (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Allgather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm)))
52201faf4e4SMatthew Knepley 
52301faf4e4SMatthew Knepley     #define MPI_Allgatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, comm) \
524c708d6e3SStefano Zampini       (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Allgatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (comm)))
52501faf4e4SMatthew Knepley 
52601faf4e4SMatthew Knepley     #define MPI_Gather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm) \
527c708d6e3SStefano 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)))
52801faf4e4SMatthew Knepley 
52901faf4e4SMatthew Knepley     #define MPI_Gatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, root, comm) \
530c708d6e3SStefano 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)))
53101faf4e4SMatthew Knepley 
53201faf4e4SMatthew Knepley     #define MPI_Scatter(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm) \
533c708d6e3SStefano 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)))
53401faf4e4SMatthew Knepley 
53501faf4e4SMatthew Knepley     #define MPI_Scatterv(sendbuf, sendcount, displs, sendtype, recvbuf, recvcount, recvtype, root, comm) \
536c708d6e3SStefano 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)))
537f12d10ccSJunchao Zhang 
538f12d10ccSJunchao Zhang     #define MPI_Ialltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm, request) \
539c708d6e3SStefano 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)))
540f12d10ccSJunchao Zhang 
541f12d10ccSJunchao Zhang     #define MPI_Ialltoallv(sendbuf, sendcnts, sdispls, sendtype, recvbuf, recvcnts, rdispls, recvtype, comm, request) \
542c708d6e3SStefano 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)))
543f12d10ccSJunchao Zhang 
544f12d10ccSJunchao Zhang     #define MPI_Iallgather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, comm, request) \
545c708d6e3SStefano Zampini       (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Iallgather((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (recvtype), (comm), (request)))
546f12d10ccSJunchao Zhang 
547f12d10ccSJunchao Zhang     #define MPI_Iallgatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, comm, request) \
548c708d6e3SStefano Zampini       (PetscAddLogDouble(&petsc_gather_ct, &petsc_gather_ct_th, PetscMPIParallelComm(comm)) || MPI_Iallgatherv((sendbuf), (sendcount), (sendtype), (recvbuf), (recvcount), (displs), (recvtype), (comm), (request)))
549f12d10ccSJunchao Zhang 
550f12d10ccSJunchao Zhang     #define MPI_Igather(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \
551c708d6e3SStefano 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)))
552f12d10ccSJunchao Zhang 
553f12d10ccSJunchao Zhang     #define MPI_Igatherv(sendbuf, sendcount, sendtype, recvbuf, recvcount, displs, recvtype, root, comm, request) \
554c708d6e3SStefano 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)))
555f12d10ccSJunchao Zhang 
556f12d10ccSJunchao Zhang     #define MPI_Iscatter(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \
557c708d6e3SStefano 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)))
558f12d10ccSJunchao Zhang 
559f12d10ccSJunchao Zhang     #define MPI_Iscatterv(sendbuf, sendcount, displs, sendtype, recvbuf, recvcount, recvtype, root, comm, request) \
560c708d6e3SStefano 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)))
561f12d10ccSJunchao Zhang 
562715b587bSJunchao Zhang     #define MPIX_Send_enqueue(buf, count, datatype, dest, tag, comm) \
563715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_send_ct, &petsc_send_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_send_len), (&petsc_send_len_th)) || MPIX_Send_enqueue((buf), (count), (datatype), (dest), (tag), (comm)))
564715b587bSJunchao Zhang 
565715b587bSJunchao Zhang     #define MPIX_Recv_enqueue(buf, count, datatype, source, tag, comm, status) \
566715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_recv_ct, &petsc_recv_ct_th, 1) || PetscMPITypeSize((count), (datatype), (&petsc_recv_len), (&petsc_recv_len_th)) || MPIX_Recv_enqueue((buf), (count), (datatype), (source), (tag), (comm), (status)))
567715b587bSJunchao Zhang 
568715b587bSJunchao Zhang     #define MPIX_Isend_enqueue(buf, count, datatype, dest, tag, comm, request) \
569715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_isend_ct, &petsc_isend_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_isend_len), &(petsc_isend_len_th)) || MPIX_Isend_enqueue((buf), (count), (datatype), (dest), (tag), (comm), (request)))
570715b587bSJunchao Zhang 
571715b587bSJunchao Zhang     #define MPIX_Irecv_enqueue(buf, count, datatype, source, tag, comm, request) \
572715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_irecv_ct, &petsc_irecv_ct_th, 1) || PetscMPITypeSize((count), (datatype), &(petsc_irecv_len), &(petsc_irecv_len_th)) || MPIX_Irecv_enqueue((buf), (count), (datatype), (source), (tag), (comm), (request)))
573715b587bSJunchao Zhang 
574715b587bSJunchao Zhang     #define MPIX_Allreduce_enqueue(sendbuf, recvbuf, count, datatype, op, comm) \
575715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_allreduce_ct, &petsc_allreduce_ct_th, PetscMPIParallelComm(comm)) || MPIX_Allreduce_enqueue((sendbuf), (recvbuf), (count), (datatype), (op), (comm)))
576715b587bSJunchao Zhang 
577715b587bSJunchao Zhang     #define MPIX_Wait_enqueue(request, status) (PetscAddLogDouble(&petsc_wait_ct, &petsc_wait_ct_th, 1) || PetscAddLogDouble(&petsc_sum_of_waits_ct, &petsc_sum_of_waits_ct_th, 1) || MPIX_Wait_enqueue((request), (status)))
578715b587bSJunchao Zhang 
579715b587bSJunchao Zhang     #define MPIX_Waitall_enqueue(count, array_of_requests, array_of_statuses) \
580715b587bSJunchao Zhang       (PetscAddLogDouble(&petsc_wait_all_ct, &petsc_wait_all_ct_th, 1) || PetscAddLogDouble(&petsc_sum_of_waits_ct, &petsc_sum_of_waits_ct_th, count) || MPIX_Waitall_enqueue((count), (array_of_requests), (array_of_statuses)))
5810d4b0b6cSBarry Smith   #else
5820d4b0b6cSBarry Smith 
5839371c9d4SSatish Balay     #define MPI_Startall_irecv(count, datatype, number, requests) ((number) && MPI_Startall((number), (requests)))
5840d4b0b6cSBarry Smith 
5859371c9d4SSatish Balay     #define MPI_Startall_isend(count, datatype, number, requests) ((number) && MPI_Startall((number), (requests)))
5860d4b0b6cSBarry Smith 
5879371c9d4SSatish Balay     #define MPI_Start_isend(count, datatype, requests) (MPI_Start((requests)))
5880d4b0b6cSBarry Smith 
589c9b973beSBarry Smith   #endif /* !MPIUNI_H && ! PETSC_HAVE_BROKEN_RECURSIVE_MACRO */
590614700edSBarry Smith 
591df8cf0b5SBarry Smith #else /* ---Logging is turned off --------------------------------------------*/
592614700edSBarry Smith 
593e3ed9ee7SBarry Smith   #define PetscLogMemory PETSC_FALSE
594e3ed9ee7SBarry Smith 
5953ba16761SJacob Faibussowitsch   #define PetscLogFlops(n) ((void)(n), PETSC_SUCCESS)
5963ba16761SJacob Faibussowitsch   #define PetscGetFlops(a) (*(a) = 0.0, PETSC_SUCCESS)
597614700edSBarry Smith 
5983ba16761SJacob Faibussowitsch   #define PetscLogStageRegister(a, b)   PETSC_SUCCESS
5993ba16761SJacob Faibussowitsch   #define PetscLogStagePush(a)          PETSC_SUCCESS
6003ba16761SJacob Faibussowitsch   #define PetscLogStagePop()            PETSC_SUCCESS
6013ba16761SJacob Faibussowitsch   #define PetscLogStageSetActive(a, b)  PETSC_SUCCESS
6023ba16761SJacob Faibussowitsch   #define PetscLogStageGetActive(a, b)  PETSC_SUCCESS
6033ba16761SJacob Faibussowitsch   #define PetscLogStageGetVisible(a, b) PETSC_SUCCESS
6043ba16761SJacob Faibussowitsch   #define PetscLogStageSetVisible(a, b) PETSC_SUCCESS
6053ba16761SJacob Faibussowitsch   #define PetscLogStageGetId(a, b)      (*(b) = 0, PETSC_SUCCESS)
606f5d6ab90SLisandro Dalcin 
6073ba16761SJacob Faibussowitsch   #define PetscLogEventRegister(a, b, c)    PETSC_SUCCESS
6083ba16761SJacob Faibussowitsch   #define PetscLogEventSetCollective(a, b)  PETSC_SUCCESS
6093ba16761SJacob Faibussowitsch   #define PetscLogEventIncludeClass(a)      PETSC_SUCCESS
6103ba16761SJacob Faibussowitsch   #define PetscLogEventExcludeClass(a)      PETSC_SUCCESS
6113ba16761SJacob Faibussowitsch   #define PetscLogEventActivate(a)          PETSC_SUCCESS
6123ba16761SJacob Faibussowitsch   #define PetscLogEventDeactivate(a)        PETSC_SUCCESS
6133ba16761SJacob Faibussowitsch   #define PetscLogEventDeactivatePush(a)    PETSC_SUCCESS
6143ba16761SJacob Faibussowitsch   #define PetscLogEventDeactivatePop(a)     PETSC_SUCCESS
6153ba16761SJacob Faibussowitsch   #define PetscLogEventActivateClass(a)     PETSC_SUCCESS
6163ba16761SJacob Faibussowitsch   #define PetscLogEventDeactivateClass(a)   PETSC_SUCCESS
6173ba16761SJacob Faibussowitsch   #define PetscLogEventSetActiveAll(a, b)   PETSC_SUCCESS
6183ba16761SJacob Faibussowitsch   #define PetscLogEventGetId(a, b)          (*(b) = 0, PETSC_SUCCESS)
6193ba16761SJacob Faibussowitsch   #define PetscLogEventGetPerfInfo(a, b, c) PETSC_SUCCESS
6203ba16761SJacob Faibussowitsch   #define PetscLogEventSetDof(a, b, c)      PETSC_SUCCESS
6213ba16761SJacob Faibussowitsch   #define PetscLogEventSetError(a, b, c)    PETSC_SUCCESS
62277c4ece6SBarry Smith 
6233ba16761SJacob Faibussowitsch   #define PetscLogPLB PETSC_SUCCESS
6243ba16761SJacob Faibussowitsch   #define PetscLogPLE PETSC_SUCCESS
6253ba16761SJacob Faibussowitsch   #define PetscLogPHC PETSC_SUCCESS
6263ba16761SJacob Faibussowitsch   #define PetscLogPHD PETSC_SUCCESS
627f5d6ab90SLisandro Dalcin 
6283ba16761SJacob Faibussowitsch   #define PetscLogObjectParents(p, n, c) PETSC_SUCCESS
6293ba16761SJacob Faibussowitsch   #define PetscLogObjectCreate(h)        PETSC_SUCCESS
6303ba16761SJacob Faibussowitsch   #define PetscLogObjectDestroy(h)       PETSC_SUCCESS
6313ca90d2dSJacob Faibussowitsch PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject, const char[], ...) PETSC_ATTRIBUTE_FORMAT(2, 3);
632fa2bb9feSLisandro Dalcin 
6333ba16761SJacob Faibussowitsch   #define PetscLogDefaultBegin()     PETSC_SUCCESS
6343ba16761SJacob Faibussowitsch   #define PetscLogAllBegin()         PETSC_SUCCESS
6353ba16761SJacob Faibussowitsch   #define PetscLogNestedBegin()      PETSC_SUCCESS
6363ba16761SJacob Faibussowitsch   #define PetscLogTraceBegin(file)   PETSC_SUCCESS
6373ba16761SJacob Faibussowitsch   #define PetscLogActions(a)         PETSC_SUCCESS
6383ba16761SJacob Faibussowitsch   #define PetscLogObjects(a)         PETSC_SUCCESS
6393ba16761SJacob Faibussowitsch   #define PetscLogSetThreshold(a, b) PETSC_SUCCESS
6403ba16761SJacob Faibussowitsch   #define PetscLogSet(lb, le)        PETSC_SUCCESS
6413ba16761SJacob Faibussowitsch   #define PetscLogIsActive(flag)     (*(flag) = PETSC_FALSE, PETSC_SUCCESS)
642fa2bb9feSLisandro Dalcin 
6433ba16761SJacob Faibussowitsch   #define PetscLogView(viewer)      PETSC_SUCCESS
6443ba16761SJacob Faibussowitsch   #define PetscLogViewFromOptions() PETSC_SUCCESS
6453ba16761SJacob Faibussowitsch   #define PetscLogDump(c)           PETSC_SUCCESS
646fa2bb9feSLisandro Dalcin 
6473ba16761SJacob Faibussowitsch   #define PetscLogEventSync(e, comm)                            PETSC_SUCCESS
6483ba16761SJacob Faibussowitsch   #define PetscLogEventBegin(e, o1, o2, o3, o4)                 PETSC_SUCCESS
6493ba16761SJacob Faibussowitsch   #define PetscLogEventEnd(e, o1, o2, o3, o4)                   PETSC_SUCCESS
650ce6058e1SBarry Smith 
65123c86b1bSSatish Balay   /* If PETSC_USE_LOG is NOT defined, these still need to be! */
65223c86b1bSSatish Balay   #define MPI_Startall_irecv(count, datatype, number, requests) ((number) && MPI_Startall(number, requests))
65323c86b1bSSatish Balay   #define MPI_Startall_isend(count, datatype, number, requests) ((number) && MPI_Startall(number, requests))
65423c86b1bSSatish Balay   #define MPI_Start_isend(count, datatype, requests)            MPI_Start(requests)
65523c86b1bSSatish Balay 
65623c86b1bSSatish Balay #endif /* PETSC_USE_LOG */
65723c86b1bSSatish Balay 
658204a6943SJed Brown #define PetscPreLoadBegin(flag, name) \
659204a6943SJed Brown   do { \
660204a6943SJed Brown     PetscBool     PetscPreLoading = flag; \
661204a6943SJed Brown     int           PetscPreLoadMax, PetscPreLoadIt; \
662166c7f25SBarry Smith     PetscLogStage _stageNum; \
6639566063dSJacob Faibussowitsch     PetscCall(PetscOptionsGetBool(NULL, NULL, "-preload", &PetscPreLoading, NULL)); \
664204a6943SJed Brown     PetscPreLoadMax     = (int)(PetscPreLoading); \
665204a6943SJed Brown     PetscPreLoadingUsed = PetscPreLoading ? PETSC_TRUE : PetscPreLoadingUsed; \
666204a6943SJed Brown     for (PetscPreLoadIt = 0; PetscPreLoadIt <= PetscPreLoadMax; PetscPreLoadIt++) { \
667204a6943SJed Brown       PetscPreLoadingOn = PetscPreLoading; \
6689566063dSJacob Faibussowitsch       PetscCall(PetscBarrier(NULL)); \
6699566063dSJacob Faibussowitsch       if (PetscPreLoadIt > 0) PetscCall(PetscLogStageGetId(name, &_stageNum)); \
6709566063dSJacob Faibussowitsch       else PetscCall(PetscLogStageRegister(name, &_stageNum)); \
6719566063dSJacob Faibussowitsch       PetscCall(PetscLogStageSetActive(_stageNum, (PetscBool)(!PetscPreLoadMax || PetscPreLoadIt))); \
6729566063dSJacob Faibussowitsch       PetscCall(PetscLogStagePush(_stageNum));
6738e58c17dSMatthew Knepley 
674204a6943SJed Brown #define PetscPreLoadEnd() \
6759566063dSJacob Faibussowitsch   PetscCall(PetscLogStagePop()); \
676204a6943SJed Brown   PetscPreLoading = PETSC_FALSE; \
677043328b6SSatish Balay   } \
6789371c9d4SSatish Balay   } \
6799371c9d4SSatish Balay   while (0)
6808e58c17dSMatthew Knepley 
6819371c9d4SSatish Balay #define PetscPreLoadStage(name) \
6829371c9d4SSatish Balay   do { \
6839566063dSJacob Faibussowitsch     PetscCall(PetscLogStagePop()); \
6849566063dSJacob Faibussowitsch     if (PetscPreLoadIt > 0) PetscCall(PetscLogStageGetId(name, &_stageNum)); \
6859566063dSJacob Faibussowitsch     else PetscCall(PetscLogStageRegister(name, &_stageNum)); \
6869566063dSJacob Faibussowitsch     PetscCall(PetscLogStageSetActive(_stageNum, (PetscBool)(!PetscPreLoadMax || PetscPreLoadIt))); \
6879566063dSJacob Faibussowitsch     PetscCall(PetscLogStagePush(_stageNum)); \
688204a6943SJed Brown   } while (0)
689e9fa29b7SSatish Balay 
690ad14c47eSJed Brown /* some vars for logging */
691ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingUsed; /* true if we are or have done preloading */
692ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingOn;   /* true if we are currently in a preloading calculation */
693ad14c47eSJed Brown 
694c708d6e3SStefano Zampini #if defined(PETSC_USE_LOG) && defined(PETSC_HAVE_DEVICE)
695c708d6e3SStefano Zampini 
696c708d6e3SStefano Zampini /* Global GPU counters */
697c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_ct;
698c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_ct;
699c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_sz;
700c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_sz;
701c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_ct_scalar;
702c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_ct_scalar;
703c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_ctog_sz_scalar;
704c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtoc_sz_scalar;
705c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gflops;
706c708d6e3SStefano Zampini PETSC_EXTERN PetscLogDouble petsc_gtime;
707c708d6e3SStefano Zampini 
708c708d6e3SStefano Zampini /* Thread local storage */
709c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_ct_th;
710c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_ct_th;
711c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_sz_th;
712c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_sz_th;
713c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_ct_scalar_th;
714c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_ct_scalar_th;
715c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_ctog_sz_scalar_th;
716c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtoc_sz_scalar_th;
717c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gflops_th;
718c708d6e3SStefano Zampini PETSC_EXTERN_TLS PetscLogDouble petsc_gtime_th;
719c708d6e3SStefano Zampini 
720c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTime(void);
721c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTimeBegin(void);
722c708d6e3SStefano Zampini PETSC_EXTERN PetscErrorCode PetscLogGpuTimeEnd(void);
723c708d6e3SStefano Zampini 
724c708d6e3SStefano Zampini /*@C
725c708d6e3SStefano Zampini        PetscLogGpuFlops - Log how many flops are performed in a calculation on the device
726c708d6e3SStefano Zampini 
727c708d6e3SStefano Zampini    Input Parameter:
728c708d6e3SStefano Zampini .   flops - the number of flops
729c708d6e3SStefano Zampini 
730c200f75eSSatish Balay    Level: intermediate
731c200f75eSSatish Balay 
732c708d6e3SStefano Zampini    Notes:
733c708d6e3SStefano Zampini      To limit the chance of integer overflow when multiplying by a constant, represent the constant as a double,
734c200f75eSSatish Balay      not an integer. Use `PetscLogFlops`(4.0*n) not `PetscLogFlops`(4*n)
735c708d6e3SStefano Zampini 
736c708d6e3SStefano 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
737c708d6e3SStefano Zampini      just on the CPU would be the value from set from `PetscLogFlops()` minus the value set from `PetscLogGpuFlops()`
738c708d6e3SStefano Zampini 
739c200f75eSSatish Balay .seealso: [](ch_profiling), `PetscLogView()`, `PetscLogFlops()`, `PetscLogGpuTimeBegin()`, `PetscLogGpuTimeEnd()`
740c708d6e3SStefano Zampini @*/
741c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuFlops(PetscLogDouble n)
742c708d6e3SStefano Zampini {
743c708d6e3SStefano Zampini   PetscAssert(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Cannot log negative flops");
7443ba16761SJacob Faibussowitsch   PetscCall(PetscAddLogDouble(&petsc_TotalFlops, &petsc_TotalFlops_th, PETSC_FLOPS_PER_OP * n));
7453ba16761SJacob Faibussowitsch   PetscCall(PetscAddLogDouble(&petsc_gflops, &petsc_gflops_th, PETSC_FLOPS_PER_OP * n));
7463ba16761SJacob Faibussowitsch   return PETSC_SUCCESS;
747c708d6e3SStefano Zampini }
748c708d6e3SStefano Zampini 
749c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuTimeAdd(PetscLogDouble t)
750c708d6e3SStefano Zampini {
7513ba16761SJacob Faibussowitsch   return PetscAddLogDouble(&petsc_gtime, &petsc_gtime_th, t);
752c708d6e3SStefano Zampini }
753c708d6e3SStefano Zampini 
754c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogCpuToGpu(PetscLogDouble size)
755c708d6e3SStefano Zampini {
7563ba16761SJacob Faibussowitsch   return PetscAddLogDoubleCnt(&petsc_ctog_ct, &petsc_ctog_sz, &petsc_ctog_ct_th, &petsc_ctog_sz_th, size);
757c708d6e3SStefano Zampini }
758c708d6e3SStefano Zampini 
759c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuToCpu(PetscLogDouble size)
760c708d6e3SStefano Zampini {
7613ba16761SJacob Faibussowitsch   return PetscAddLogDoubleCnt(&petsc_gtoc_ct, &petsc_gtoc_sz, &petsc_gtoc_ct_th, &petsc_gtoc_sz_th, size);
762c708d6e3SStefano Zampini }
763c708d6e3SStefano Zampini 
764c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogCpuToGpuScalar(PetscLogDouble size)
765c708d6e3SStefano Zampini {
7663ba16761SJacob Faibussowitsch   return PetscAddLogDoubleCnt(&petsc_ctog_ct_scalar, &petsc_ctog_sz_scalar, &petsc_ctog_ct_scalar_th, &petsc_ctog_sz_scalar_th, size);
767c708d6e3SStefano Zampini }
768c708d6e3SStefano Zampini 
769c708d6e3SStefano Zampini static inline PetscErrorCode PetscLogGpuToCpuScalar(PetscLogDouble size)
770c708d6e3SStefano Zampini {
7713ba16761SJacob Faibussowitsch   return PetscAddLogDoubleCnt(&petsc_gtoc_ct_scalar, &petsc_gtoc_sz_scalar, &petsc_gtoc_ct_scalar_th, &petsc_gtoc_sz_scalar_th, size);
772c708d6e3SStefano Zampini }
773c708d6e3SStefano Zampini #else
774c708d6e3SStefano Zampini 
7753ba16761SJacob Faibussowitsch   #define PetscLogCpuToGpu(a)       PETSC_SUCCESS
7763ba16761SJacob Faibussowitsch   #define PetscLogGpuToCpu(a)       PETSC_SUCCESS
7773ba16761SJacob Faibussowitsch   #define PetscLogCpuToGpuScalar(a) PETSC_SUCCESS
7783ba16761SJacob Faibussowitsch   #define PetscLogGpuToCpuScalar(a) PETSC_SUCCESS
7793ba16761SJacob Faibussowitsch   #define PetscLogGpuFlops(a)       PETSC_SUCCESS
7803ba16761SJacob Faibussowitsch   #define PetscLogGpuTime()         PETSC_SUCCESS
7813ba16761SJacob Faibussowitsch   #define PetscLogGpuTimeAdd(a)     PETSC_SUCCESS
7823ba16761SJacob Faibussowitsch   #define PetscLogGpuTimeBegin()    PETSC_SUCCESS
7833ba16761SJacob Faibussowitsch   #define PetscLogGpuTimeEnd()      PETSC_SUCCESS
784c708d6e3SStefano Zampini 
785c708d6e3SStefano Zampini #endif /* PETSC_USE_LOG && PETSC_HAVE_DEVICE */
786c708d6e3SStefano Zampini 
787c708d6e3SStefano Zampini /* remove TLS defines */
788c708d6e3SStefano Zampini #undef PETSC_EXTERN_TLS
789c708d6e3SStefano Zampini #undef PETSC_TLS
790c708d6e3SStefano Zampini 
79197bb86f7SLois Curfman McInnes #endif
792