197bb86f7SLois Curfman McInnes /* 27588ac45SBarry Smith Defines profile/logging in PETSc. 397bb86f7SLois Curfman McInnes */ 497bb86f7SLois Curfman McInnes 526bd1501SBarry Smith #if !defined(PETSCLOG_H) 626bd1501SBarry Smith #define PETSCLOG_H 7aaa7dc30SBarry Smith #include <petscsys.h> 8958c4211Shannah_mairs #include <petsctime.h> 9c8d78d4dSSatish Balay 10fa2bb9feSLisandro Dalcin /* General logging of information; different from event logging */ 11fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscInfo_Private(const char[],void*,const char[],...); 12fa2bb9feSLisandro Dalcin #if defined(PETSC_USE_INFO) 13fa2bb9feSLisandro Dalcin #define PetscInfo(A,S) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S) 14fa2bb9feSLisandro Dalcin #define PetscInfo1(A,S,a1) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1) 15fa2bb9feSLisandro Dalcin #define PetscInfo2(A,S,a1,a2) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2) 16fa2bb9feSLisandro Dalcin #define PetscInfo3(A,S,a1,a2,a3) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2,a3) 17fa2bb9feSLisandro Dalcin #define PetscInfo4(A,S,a1,a2,a3,a4) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2,a3,a4) 18fa2bb9feSLisandro Dalcin #define PetscInfo5(A,S,a1,a2,a3,a4,a5) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2,a3,a4,a5) 19fa2bb9feSLisandro Dalcin #define PetscInfo6(A,S,a1,a2,a3,a4,a5,a6) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2,a3,a4,a5,a6) 20fa2bb9feSLisandro Dalcin #define PetscInfo7(A,S,a1,a2,a3,a4,a5,a6,a7) PetscInfo_Private(PETSC_FUNCTION_NAME,A,S,a1,a2,a3,a4,a5,a6,a7) 21fa2bb9feSLisandro Dalcin #else 22fa2bb9feSLisandro Dalcin #define PetscInfo(A,S) 0 23fa2bb9feSLisandro Dalcin #define PetscInfo1(A,S,a1) 0 24fa2bb9feSLisandro Dalcin #define PetscInfo2(A,S,a1,a2) 0 25fa2bb9feSLisandro Dalcin #define PetscInfo3(A,S,a1,a2,a3) 0 26fa2bb9feSLisandro Dalcin #define PetscInfo4(A,S,a1,a2,a3,a4) 0 27fa2bb9feSLisandro Dalcin #define PetscInfo5(A,S,a1,a2,a3,a4,a5) 0 28fa2bb9feSLisandro Dalcin #define PetscInfo6(A,S,a1,a2,a3,a4,a5,a6) 0 29fa2bb9feSLisandro Dalcin #define PetscInfo7(A,S,a1,a2,a3,a4,a5,a6,a7) 0 30fa2bb9feSLisandro Dalcin #endif 31fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscInfoDeactivateClass(PetscClassId); 32fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscInfoActivateClass(PetscClassId); 33fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscBool PetscLogPrintInfo; /* if true, indicates PetscInfo() is turned on */ 34fa2bb9feSLisandro Dalcin 35b859824cSBarry Smith /*MC 36b859824cSBarry Smith PetscLogEvent - id used to identify PETSc or user events which timed portions (blocks of executable) 37b859824cSBarry Smith code. 38b859824cSBarry Smith 39b859824cSBarry Smith Level: intermediate 40b859824cSBarry Smith 41b859824cSBarry Smith .seealso: PetscLogEventRegister(), PetscLogEventBegin(), PetscLogEventEnd(), PetscLogStage 42b859824cSBarry Smith M*/ 43b859824cSBarry Smith typedef int PetscLogEvent; 44b859824cSBarry Smith 45b859824cSBarry Smith /*MC 46b859824cSBarry Smith PetscLogStage - id used to identify user stages (phases, sections) of runs - for logging 47b859824cSBarry Smith 48b859824cSBarry Smith Level: intermediate 49b859824cSBarry Smith 50522ce795SJed Brown .seealso: PetscLogStageRegister(), PetscLogStagePush(), PetscLogStagePop(), PetscLogEvent 51b859824cSBarry Smith M*/ 52b859824cSBarry Smith typedef int PetscLogStage; 53b859824cSBarry Smith 548ba1e511SMatthew Knepley #define PETSC_EVENT 1311311 55014dd563SJed Brown PETSC_EXTERN PetscLogEvent PETSC_LARGEST_EVENT; 5697bb86f7SLois Curfman McInnes 5719b02663SBarry Smith /* Global flop counter */ 58014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_TotalFlops; 59014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_tmp_flops; 6019b02663SBarry Smith 61bec0b493Shannah_mairs /* Global GPU counters */ 62bec0b493Shannah_mairs #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) 63bec0b493Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_ctog_ct; 64bec0b493Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_gtoc_ct; 65bec0b493Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_ctog_sz; 66bec0b493Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_gtoc_sz; 67958c4211Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_gflops; 68958c4211Shannah_mairs PETSC_EXTERN PetscLogDouble petsc_gtime; 69bec0b493Shannah_mairs #endif 70bec0b493Shannah_mairs 7131d06abdSBarry Smith /* We must make the following structures available to access the event 7231d06abdSBarry Smith activation flags in the PetscLogEventBegin/End() macros. These are not part of the PETSc public 7331d06abdSBarry Smith API and are not intended to be used by other parts of PETSc or by users. 74d49d4b11SBarry Smith 75fa2bb9feSLisandro Dalcin The code that manipulates these structures is in src/sys/logging/utils. 7631d06abdSBarry Smith */ 77aa213bdcSJed Brown typedef struct _n_PetscIntStack *PetscIntStack; 786a6a9b46SSatish Balay 79217044c2SLisandro Dalcin /* -----------------------------------------------------------------------------------------------------*/ 80c60ec953SBarry Smith /* 81aa213bdcSJed Brown PetscClassRegInfo, PetscClassPerfInfo - Each class has two data structures associated with it. The first has 82c60ec953SBarry Smith static information about it, the second collects statistics on how many objects of the class are created, 83c60ec953SBarry Smith how much memory they use, etc. 84c60ec953SBarry Smith 85aa213bdcSJed Brown PetscClassRegLog, PetscClassPerfLog - arrays of the PetscClassRegInfo and PetscClassPerfInfo for all classes. 86c60ec953SBarry Smith */ 876a6a9b46SSatish Balay typedef struct { 88c60ec953SBarry Smith char *name; /* The class name */ 89c60ec953SBarry Smith PetscClassId classid; /* The integer identifying this class */ 90aa213bdcSJed Brown } PetscClassRegInfo; 916a6a9b46SSatish Balay 926a6a9b46SSatish Balay typedef struct { 930700a824SBarry Smith PetscClassId id; /* The integer identifying this class */ 946a6a9b46SSatish Balay int creations; /* The number of objects of this class created */ 956a6a9b46SSatish Balay int destructions; /* The number of objects of this class destroyed */ 966a6a9b46SSatish Balay PetscLogDouble mem; /* The total memory allocated by objects of this class */ 976a6a9b46SSatish Balay PetscLogDouble descMem; /* The total memory allocated by descendents of these objects */ 98aa213bdcSJed Brown } PetscClassPerfInfo; 996a6a9b46SSatish Balay 100aa213bdcSJed Brown typedef struct _n_PetscClassRegLog *PetscClassRegLog; 101aa213bdcSJed Brown struct _n_PetscClassRegLog { 102c60ec953SBarry Smith int numClasses; /* The number of classes registered */ 103c60ec953SBarry Smith int maxClasses; /* The maximum number of classes */ 104aa213bdcSJed Brown PetscClassRegInfo *classInfo; /* The structure for class information (classids are monotonicly increasing) */ 105c60ec953SBarry Smith }; 1066a6a9b46SSatish Balay 107aa213bdcSJed Brown typedef struct _n_PetscClassPerfLog *PetscClassPerfLog; 108aa213bdcSJed Brown struct _n_PetscClassPerfLog { 109c60ec953SBarry Smith int numClasses; /* The number of logging classes */ 110c60ec953SBarry Smith int maxClasses; /* The maximum number of classes */ 111aa213bdcSJed Brown PetscClassPerfInfo *classInfo; /* The structure for class information (classids are monotonicly increasing) */ 112c60ec953SBarry Smith }; 113c60ec953SBarry Smith /* -----------------------------------------------------------------------------------------------------*/ 114c60ec953SBarry Smith /* 115aa213bdcSJed Brown PetscEventRegInfo, PetscEventPerfInfo - Each event has two data structures associated with it. The first has 116c60ec953SBarry Smith static information about it, the second collects statistics on how many times the event is used, how 117c60ec953SBarry Smith much time it takes, etc. 118c60ec953SBarry Smith 119aa213bdcSJed Brown PetscEventRegLog, PetscEventPerfLog - an array of all PetscEventRegInfo and PetscEventPerfInfo for all events. There is one 120c60ec953SBarry Smith of these for each stage. 121c60ec953SBarry Smith 122c60ec953SBarry Smith */ 1236a6a9b46SSatish Balay typedef struct { 1246a6a9b46SSatish Balay char *name; /* The name of this event */ 125c60ec953SBarry Smith PetscClassId classid; /* The class the event is associated with */ 126217044c2SLisandro Dalcin PetscBool collective; /* Flag this event as collective */ 1276a6a9b46SSatish Balay #if defined (PETSC_HAVE_MPE) 1286a6a9b46SSatish Balay int mpe_id_begin; /* MPE IDs that define the event */ 1296a6a9b46SSatish Balay int mpe_id_end; 1306a6a9b46SSatish Balay #endif 131aa213bdcSJed Brown } PetscEventRegInfo; 1326a6a9b46SSatish Balay 133c60ec953SBarry Smith typedef struct { 134c60ec953SBarry Smith int id; /* The integer identifying this event */ 135ace3abfcSBarry Smith PetscBool active; /* The flag to activate logging */ 136ace3abfcSBarry Smith PetscBool visible; /* The flag to print info in summary */ 137c60ec953SBarry Smith int depth; /* The nesting depth of the event call */ 138c60ec953SBarry Smith int count; /* The number of times this event was executed */ 139a42e5b54SMatthew G Knepley PetscLogDouble flops, flops2, flopsTmp; /* The flops and flops^2 used in this event */ 140a42e5b54SMatthew G Knepley PetscLogDouble time, time2, timeTmp; /* The time and time^2 taken for this event */ 141217044c2SLisandro Dalcin PetscLogDouble syncTime; /* The synchronization barrier time */ 142891e75beSMatthew G. Knepley PetscLogDouble dof[8]; /* The number of degrees of freedom associated with this event */ 1435d68e14cSMatthew G. Knepley PetscLogDouble errors[8]; /* The errors (user-defined) associated with this event */ 144c60ec953SBarry Smith PetscLogDouble numMessages; /* The number of messages in this event */ 145c60ec953SBarry Smith PetscLogDouble messageLength; /* The total message lengths in this event */ 146c60ec953SBarry Smith PetscLogDouble numReductions; /* The number of reductions in this event */ 147e3ed9ee7SBarry Smith PetscLogDouble memIncrease; /* How much the resident memory has increased in this event */ 148e3ed9ee7SBarry Smith PetscLogDouble mallocIncrease;/* How much the maximum malloced space has increased in this event */ 149e3ed9ee7SBarry Smith PetscLogDouble mallocSpace; /* How much the space was malloced and kept during this event */ 150e3ed9ee7SBarry Smith PetscLogDouble mallocIncreaseEvent; /* Maximum of the high water mark with in event minus memory available at the end of the event */ 151bec0b493Shannah_mairs #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) 152bec0b493Shannah_mairs PetscLogDouble CpuToGpuCount; /* The total number of CPU to GPU copies */ 153bec0b493Shannah_mairs PetscLogDouble GpuToCpuCount; /* The total number of GPU to CPU copies */ 154bec0b493Shannah_mairs PetscLogDouble CpuToGpuSize; /* The total size of CPU to GPU copies */ 155bec0b493Shannah_mairs PetscLogDouble GpuToCpuSize; /* The total size of GPU to CPU copies */ 156958c4211Shannah_mairs PetscLogDouble GpuFlops; /* The flops done on a GPU in this event */ 157958c4211Shannah_mairs PetscLogDouble GpuTime; /* The time spent on a GPU in this event */ 158bec0b493Shannah_mairs #endif 159aa213bdcSJed Brown } PetscEventPerfInfo; 160c60ec953SBarry Smith 161aa213bdcSJed Brown typedef struct _n_PetscEventRegLog *PetscEventRegLog; 162aa213bdcSJed Brown struct _n_PetscEventRegLog { 1636a6a9b46SSatish Balay int numEvents; /* The number of registered events */ 1646a6a9b46SSatish Balay int maxEvents; /* The maximum number of events */ 165aa213bdcSJed Brown PetscEventRegInfo *eventInfo; /* The registration information for each event */ 1666a6a9b46SSatish Balay }; 1676a6a9b46SSatish Balay 168aa213bdcSJed Brown typedef struct _n_PetscEventPerfLog *PetscEventPerfLog; 169aa213bdcSJed Brown struct _n_PetscEventPerfLog { 1706a6a9b46SSatish Balay int numEvents; /* The number of logging events */ 1716a6a9b46SSatish Balay int maxEvents; /* The maximum number of events */ 172aa213bdcSJed Brown PetscEventPerfInfo *eventInfo; /* The performance information for each event */ 1736a6a9b46SSatish Balay }; 174c60ec953SBarry Smith /* ------------------------------------------------------------------------------------------------------------*/ 175c60ec953SBarry Smith /* 176aa213bdcSJed Brown PetscStageInfo - Contains all the information about a particular stage. 1776a6a9b46SSatish Balay 178aa213bdcSJed Brown PetscStageLog - An array of PetscStageInfo for each registered stage. There is a single one of these in the code. 179c60ec953SBarry Smith */ 180aa213bdcSJed Brown typedef struct _PetscStageInfo { 1816a6a9b46SSatish Balay char *name; /* The stage name */ 182ace3abfcSBarry Smith PetscBool used; /* The stage was pushed on this processor */ 183aa213bdcSJed Brown PetscEventPerfInfo perfInfo; /* The stage performance information */ 184aa213bdcSJed Brown PetscEventPerfLog eventLog; /* The event information for this stage */ 185aa213bdcSJed Brown PetscClassPerfLog classLog; /* The class information for this stage */ 186aa213bdcSJed Brown } PetscStageInfo; 1876a6a9b46SSatish Balay 188aa213bdcSJed Brown typedef struct _n_PetscStageLog *PetscStageLog; 189aa213bdcSJed Brown struct _n_PetscStageLog { 1906a6a9b46SSatish Balay int numStages; /* The number of registered stages */ 1916a6a9b46SSatish Balay int maxStages; /* The maximum number of stages */ 192aa213bdcSJed Brown PetscIntStack stack; /* The stack for active stages */ 193aa213bdcSJed Brown int curStage; /* The current stage (only used in macros so we don't call PetscIntStackTop) */ 194aa213bdcSJed Brown PetscStageInfo *stageInfo; /* The information for each stage */ 195aa213bdcSJed Brown PetscEventRegLog eventLog; /* The registered events */ 196aa213bdcSJed Brown PetscClassRegLog classLog; /* The registered classes */ 1976a6a9b46SSatish Balay }; 198217044c2SLisandro Dalcin /* -----------------------------------------------------------------------------------------------------*/ 199fee518ebSJed Brown 2003bb1ff40SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogObjectParent(PetscObject,PetscObject); 2013bb1ff40SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogObjectMemory(PetscObject,PetscLogDouble); 2023bb1ff40SBarry Smith 203aa482453SBarry Smith #if defined(PETSC_USE_LOG) /* --- Logging is turned on --------------------------------*/ 204aeeda9f6SJohn Fettig PETSC_EXTERN PetscStageLog petsc_stageLog; 205f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogGetStageLog(PetscStageLog*); 206f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetCurrent(PetscStageLog,int*); 207f5d6ab90SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscStageLogGetEventPerfLog(PetscStageLog,int,PetscEventPerfLog*); 208614700edSBarry Smith 209da63de55SLois Curfman McInnes /* 210da63de55SLois Curfman McInnes Flop counting: We count each arithmetic operation (e.g., addition, multiplication) separately. 211da63de55SLois Curfman McInnes 212da63de55SLois Curfman McInnes For the complex numbers version, note that 213da63de55SLois Curfman McInnes 1 complex addition = 2 flops 214da63de55SLois Curfman McInnes 1 complex multiplication = 6 flops, 215da63de55SLois Curfman McInnes where we define 1 flop as that for a double precision scalar. We roughly approximate 216da63de55SLois Curfman McInnes flop counting for complex numbers by multiplying the total flops by 4; this corresponds 217da63de55SLois Curfman McInnes to the assumption that we're counting mostly additions and multiplications -- and 218da63de55SLois Curfman McInnes roughly the same number of each. More accurate counting could be done by distinguishing 219da63de55SLois Curfman McInnes among the various arithmetic operations. 220da63de55SLois Curfman McInnes */ 221da63de55SLois Curfman McInnes 222aa482453SBarry Smith #if defined(PETSC_USE_COMPLEX) 223542d4b3fSSatish Balay #define PETSC_FLOPS_PER_OP 4.0 224da63de55SLois Curfman McInnes #else 225542d4b3fSSatish Balay #define PETSC_FLOPS_PER_OP 1.0 226bf3909cdSBarry Smith #endif 227bf3909cdSBarry Smith 228dd19b90bSBarry Smith PETSC_STATIC_INLINE PetscErrorCode PetscLogFlops(PetscLogDouble n) 229dd19b90bSBarry Smith { 230dd19b90bSBarry Smith PetscFunctionBegin; 231bf3909cdSBarry Smith #if defined(PETSC_USE_DEBUG) 232dd19b90bSBarry Smith if (n < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Cannot log negative flops"); 233da63de55SLois Curfman McInnes #endif 234dd19b90bSBarry Smith petsc_TotalFlops += PETSC_FLOPS_PER_OP*n; 235dd19b90bSBarry Smith PetscFunctionReturn(0); 236dd19b90bSBarry Smith } 237fa2bb9feSLisandro Dalcin 238bec0b493Shannah_mairs #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) 239*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogCpuToGpu(PetscLogDouble size) 240*4db21f4bSStefano Zampini { 241958c4211Shannah_mairs PetscFunctionBegin; 242bec0b493Shannah_mairs petsc_ctog_ct += 1; 243bec0b493Shannah_mairs petsc_ctog_sz += size; 244bec0b493Shannah_mairs PetscFunctionReturn(0); 245bec0b493Shannah_mairs } 246*4db21f4bSStefano Zampini 247*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogGpuToCpu(PetscLogDouble size) 248*4db21f4bSStefano Zampini { 249958c4211Shannah_mairs PetscFunctionBegin; 250bec0b493Shannah_mairs petsc_gtoc_ct += 1; 251bec0b493Shannah_mairs petsc_gtoc_sz += size; 252bec0b493Shannah_mairs PetscFunctionReturn(0); 253bec0b493Shannah_mairs } 254*4db21f4bSStefano Zampini 255*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogGpuFlops(PetscLogDouble n) 256*4db21f4bSStefano Zampini { 257958c4211Shannah_mairs PetscFunctionBegin; 258958c4211Shannah_mairs #if defined(PETSC_USE_DEBUG) 259958c4211Shannah_mairs if (n < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Cannot log negative flops"); 260958c4211Shannah_mairs #endif 261958c4211Shannah_mairs petsc_TotalFlops += PETSC_FLOPS_PER_OP*n; 262958c4211Shannah_mairs petsc_gflops += PETSC_FLOPS_PER_OP*n; 263958c4211Shannah_mairs PetscFunctionReturn(0); 264958c4211Shannah_mairs } 265*4db21f4bSStefano Zampini 266*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogGpuTimeBegin() 267*4db21f4bSStefano Zampini { 268958c4211Shannah_mairs PetscErrorCode ierr; 269958c4211Shannah_mairs PetscFunctionBegin; 270958c4211Shannah_mairs ierr = PetscTimeSubtract(&petsc_gtime);CHKERRQ(ierr); 271958c4211Shannah_mairs PetscFunctionReturn(0); 272958c4211Shannah_mairs } 273*4db21f4bSStefano Zampini 274*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogGpuTimeEnd() 275*4db21f4bSStefano Zampini { 276958c4211Shannah_mairs PetscErrorCode ierr; 277958c4211Shannah_mairs PetscFunctionBegin; 278958c4211Shannah_mairs ierr = PetscTimeAdd(&petsc_gtime);CHKERRQ(ierr); 279958c4211Shannah_mairs PetscFunctionReturn(0); 280958c4211Shannah_mairs } 281*4db21f4bSStefano Zampini 282*4db21f4bSStefano Zampini PETSC_STATIC_INLINE PetscErrorCode PetscLogGpuTimeAdd(PetscLogDouble t) 283*4db21f4bSStefano Zampini { 284*4db21f4bSStefano Zampini PetscFunctionBegin; 285*4db21f4bSStefano Zampini petsc_gtime += t; 286*4db21f4bSStefano Zampini PetscFunctionReturn(0); 287*4db21f4bSStefano Zampini } 288bec0b493Shannah_mairs #endif 289bec0b493Shannah_mairs 290fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscGetFlops(PetscLogDouble *); 29177c4ece6SBarry Smith 292aa482453SBarry Smith #if defined (PETSC_HAVE_MPE) 293014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEBegin(void); 294014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogMPEDump(const char[]); 29577c4ece6SBarry Smith #endif 29677c4ece6SBarry Smith 297014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLB)(PetscLogEvent,int,PetscObject,PetscObject,PetscObject,PetscObject); 298014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPLE)(PetscLogEvent,int,PetscObject,PetscObject,PetscObject,PetscObject); 299014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHC)(PetscObject); 300014dd563SJed Brown PETSC_EXTERN PetscErrorCode (*PetscLogPHD)(PetscObject); 30177c4ece6SBarry Smith 302e83a5d19SLisandro Dalcin #define PetscLogObjectParents(p,n,d) 0;do{int _i; for (_i=0; _i<(n); _i++) {ierr = PetscLogObjectParent((PetscObject)(p),(PetscObject)(d)[_i]);CHKERRQ(ierr);}}while(0) 303fa2bb9feSLisandro Dalcin #define PetscLogObjectCreate(h) ((PetscLogPHC) ? (*PetscLogPHC)((PetscObject)(h)) : 0) 304fa2bb9feSLisandro Dalcin #define PetscLogObjectDestroy(h) ((PetscLogPHD) ? (*PetscLogPHD)((PetscObject)(h)) : 0) 305fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject, const char[], ...); 306fa2bb9feSLisandro Dalcin 3078ba1e511SMatthew Knepley /* Initialization functions */ 308bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogDefaultBegin(void); 309014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogAllBegin(void); 310bb1d7374SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogNestedBegin(void); 311014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogTraceBegin(FILE *); 312014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogActions(PetscBool); 313014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogObjects(PetscBool); 314fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogSetThreshold(PetscLogDouble,PetscLogDouble*); 315014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogSet(PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject), 3166849ba73SBarry Smith PetscErrorCode (*)(int, int, PetscObject, PetscObject, PetscObject, PetscObject)); 317fa2bb9feSLisandro Dalcin 3188ba1e511SMatthew Knepley /* Output functions */ 319014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogView(PetscViewer); 320f14045dbSBarry Smith PETSC_EXTERN PetscErrorCode PetscLogViewFromOptions(void); 321014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogDump(const char[]); 32231d06abdSBarry Smith 323fa2bb9feSLisandro Dalcin /* Stage functions */ 324014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageRegister(const char[],PetscLogStage*); 325014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePush(PetscLogStage); 326014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStagePop(void); 327014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetActive(PetscLogStage,PetscBool); 328014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetActive(PetscLogStage,PetscBool*); 329014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageSetVisible(PetscLogStage,PetscBool); 330014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetVisible(PetscLogStage,PetscBool*); 331014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogStageGetId(const char[],PetscLogStage*); 332f5d6ab90SLisandro Dalcin 3338ba1e511SMatthew Knepley /* Event functions */ 334014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventRegister(const char[],PetscClassId,PetscLogEvent*); 335217044c2SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSetCollective(PetscLogEvent,PetscBool); 336fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventIncludeClass(PetscClassId); 337fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventExcludeClass(PetscClassId); 338014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivate(PetscLogEvent); 339014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivate(PetscLogEvent); 340014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventSetActiveAll(PetscLogEvent,PetscBool); 341014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventActivateClass(PetscClassId); 342014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventDeactivateClass(PetscClassId); 3438c49a8d9SBarry Smith PETSC_EXTERN PetscErrorCode PetscLogEventGetId(const char[],PetscLogEvent*); 34413230571SMichael Lange PETSC_EXTERN PetscErrorCode PetscLogEventGetPerfInfo(int,PetscLogEvent,PetscEventPerfInfo*); 345891e75beSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetDof(PetscLogEvent, PetscInt, PetscLogDouble); 3465d68e14cSMatthew G. Knepley PETSC_EXTERN PetscErrorCode PetscLogEventSetError(PetscLogEvent, PetscInt, PetscLogDouble); 3478ba1e511SMatthew Knepley 3488ba1e511SMatthew Knepley /* Global counters */ 349014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_irecv_ct; 350014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_isend_ct; 351014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_recv_ct; 352014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_send_ct; 353014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_irecv_len; 354014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_isend_len; 355014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_recv_len; 356014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_send_len; 357014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_allreduce_ct; 358014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_gather_ct; 359014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_scatter_ct; 360014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_wait_ct; 361014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_wait_any_ct; 362014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_wait_all_ct; 363014dd563SJed Brown PETSC_EXTERN PetscLogDouble petsc_sum_of_waits_ct; 3648ba1e511SMatthew Knepley 365e3ed9ee7SBarry Smith PETSC_EXTERN PetscBool PetscLogMemory; 366e3ed9ee7SBarry Smith 367fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscBool PetscLogSyncOn; /* true if logging synchronization is enabled */ 36862872c28SLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogEventSynchronize(PetscLogEvent, MPI_Comm); 36962872c28SLisandro Dalcin 37062872c28SLisandro Dalcin #define PetscLogEventSync(e,comm) \ 37162872c28SLisandro Dalcin (((PetscLogPLB && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? \ 37262872c28SLisandro Dalcin PetscLogEventSynchronize((e),(comm)) : 0 )) 37377c4ece6SBarry Smith 37452e6d16bSBarry Smith #define PetscLogEventBegin(e,o1,o2,o3,o4) \ 3758b5db460SBarry Smith (((PetscLogPLB && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? \ 376495fc317SBarry Smith (*PetscLogPLB)((e),0,(PetscObject)(o1),(PetscObject)(o2),(PetscObject)(o3),(PetscObject)(o4)) : 0 )) 377043328b6SSatish Balay 37852e6d16bSBarry Smith #define PetscLogEventEnd(e,o1,o2,o3,o4) \ 3798b5db460SBarry Smith (((PetscLogPLE && petsc_stageLog->stageInfo[petsc_stageLog->curStage].perfInfo.active && petsc_stageLog->stageInfo[petsc_stageLog->curStage].eventLog->eventInfo[e].active) ? \ 380495fc317SBarry Smith (*PetscLogPLE)((e),0,(PetscObject)(o1),(PetscObject)(o2),(PetscObject)(o3),(PetscObject)(o4)) : 0 )) 38177c4ece6SBarry Smith 382014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventGetFlops(PetscLogEvent,PetscLogDouble*); 383014dd563SJed Brown PETSC_EXTERN PetscErrorCode PetscLogEventZeroFlops(PetscLogEvent); 384003131ecSBarry Smith 385ce85283eSBarry Smith /* 386f621e05eSBarry Smith These are used internally in the PETSc routines to keep a count of MPI messages and 387f621e05eSBarry Smith their sizes. 388f621e05eSBarry Smith 389af0996ceSBarry Smith This does not work for MPI-Uni because our include/petsc/mpiuni/mpi.h file 390ce85283eSBarry Smith uses macros to defined the MPI operations. 39115308354SBarry Smith 39215308354SBarry Smith It does not work correctly from HP-UX because it processes the 393bb4af37aSBarry Smith macros in a way that sometimes it double counts, hence 394b6410449SSatish Balay PETSC_HAVE_BROKEN_RECURSIVE_MACRO 3957c1e34a4SSatish Balay 396f569fd43SBarry Smith It does not work with Windows because winmpich lacks MPI_Type_size() 397ce85283eSBarry Smith */ 398c9b973beSBarry Smith #if !defined(MPIUNI_H) && !defined(PETSC_HAVE_BROKEN_RECURSIVE_MACRO) && !defined (PETSC_HAVE_MPI_MISSING_TYPESIZE) 39977a39924SBarry Smith /* 40077a39924SBarry Smith Logging of MPI activities 40177a39924SBarry Smith */ 402b23bfdefSJunchao Zhang PETSC_STATIC_INLINE PetscErrorCode PetscMPITypeSize(PetscInt count,MPI_Datatype type,PetscLogDouble *length) 403f95db71bSBarry Smith { 404f12d10ccSJunchao Zhang PetscMPIInt typesize; 405f12d10ccSJunchao Zhang PetscErrorCode ierr; 4067d30fa4aSBarry Smith if (type == MPI_DATATYPE_NULL) return 0; 407f12d10ccSJunchao Zhang ierr = MPI_Type_size(type,&typesize);CHKERRQ(ierr); 408f12d10ccSJunchao Zhang *length += (PetscLogDouble) (count*typesize); 4099c72ff1eSSatish Balay return 0; 410f95db71bSBarry Smith } 41177a39924SBarry Smith 412f12d10ccSJunchao Zhang PETSC_STATIC_INLINE PetscErrorCode PetscMPITypeSizeComm(MPI_Comm comm,const PetscMPIInt *counts,MPI_Datatype type,PetscLogDouble *length) 4133b9284c0SMatthew G Knepley { 414f12d10ccSJunchao Zhang PetscMPIInt typesize,size,p; 415f12d10ccSJunchao Zhang PetscErrorCode ierr; 4163b9284c0SMatthew G Knepley 4177d30fa4aSBarry Smith if (type == MPI_DATATYPE_NULL) return 0; 418f12d10ccSJunchao Zhang ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 419f12d10ccSJunchao Zhang ierr = MPI_Type_size(type,&typesize);CHKERRQ(ierr); 420f12d10ccSJunchao Zhang for (p=0; p<size; ++p) { 421f12d10ccSJunchao Zhang *length += (PetscLogDouble) (counts[p]*typesize); 422f12d10ccSJunchao Zhang } 423f12d10ccSJunchao Zhang return 0; 424f12d10ccSJunchao Zhang } 425f12d10ccSJunchao Zhang 426f12d10ccSJunchao Zhang PETSC_STATIC_INLINE PetscErrorCode PetscMPITypeSizeCount(PetscInt n,const PetscMPIInt *counts,MPI_Datatype type,PetscLogDouble *length) 427f12d10ccSJunchao Zhang { 428f12d10ccSJunchao Zhang PetscMPIInt typesize,p; 429f12d10ccSJunchao Zhang PetscErrorCode ierr; 430f12d10ccSJunchao Zhang 431f12d10ccSJunchao Zhang if (type == MPI_DATATYPE_NULL) return 0; 432f12d10ccSJunchao Zhang ierr = MPI_Type_size(type,&typesize);CHKERRQ(ierr); 433f12d10ccSJunchao Zhang for (p=0; p<n; ++p) { 434f12d10ccSJunchao Zhang *length += (PetscLogDouble) (counts[p]*typesize); 4353b9284c0SMatthew G Knepley } 4363b9284c0SMatthew G Knepley return 0; 4373b9284c0SMatthew G Knepley } 4383b9284c0SMatthew G Knepley 43972b11a25SBarry Smith /* 44072b11a25SBarry Smith Returns 1 if the communicator is parallel else zero 44172b11a25SBarry Smith */ 44272b11a25SBarry Smith PETSC_STATIC_INLINE int PetscMPIParallelComm(MPI_Comm comm) 44372b11a25SBarry Smith { 44472b11a25SBarry Smith PetscMPIInt size; MPI_Comm_size(comm,&size); return size > 1; 44572b11a25SBarry Smith } 44672b11a25SBarry Smith 44777a39924SBarry Smith #define MPI_Irecv(buf,count,datatype,source,tag,comm,request) \ 448f12d10ccSJunchao Zhang ((petsc_irecv_ct++,0) || PetscMPITypeSize((count),(datatype),&(petsc_irecv_len)) || MPI_Irecv((buf),(count),(datatype),(source),(tag),(comm),(request))) 44915308354SBarry Smith 45077a39924SBarry Smith #define MPI_Isend(buf,count,datatype,dest,tag,comm,request) \ 451f12d10ccSJunchao Zhang ((petsc_isend_ct++,0) || PetscMPITypeSize((count),(datatype),&(petsc_isend_len)) || MPI_Isend((buf),(count),(datatype),(dest),(tag),(comm),(request))) 45215308354SBarry Smith 453457b14b1SJunchao Zhang #define MPI_Startall_irecv(count,datatype,number,requests) \ 454f12d10ccSJunchao Zhang ((petsc_irecv_ct += (PetscLogDouble)(number),0) || PetscMPITypeSize((count),(datatype),&(petsc_irecv_len)) || ((number) && MPI_Startall((number),(requests)))) 4550d4b0b6cSBarry Smith 456457b14b1SJunchao Zhang #define MPI_Startall_isend(count,datatype,number,requests) \ 457f12d10ccSJunchao Zhang ((petsc_isend_ct += (PetscLogDouble)(number),0) || PetscMPITypeSize((count),(datatype),&(petsc_isend_len)) || ((number) && MPI_Startall((number),(requests)))) 4580d4b0b6cSBarry Smith 459457b14b1SJunchao Zhang #define MPI_Start_isend(count,datatype,requests) \ 460f12d10ccSJunchao Zhang ((petsc_isend_ct++,0) || PetscMPITypeSize((count),(datatype),(&petsc_isend_len)) || MPI_Start((requests))) 4610d4b0b6cSBarry Smith 462ce85283eSBarry Smith #define MPI_Recv(buf,count,datatype,source,tag,comm,status) \ 463f12d10ccSJunchao Zhang ((petsc_recv_ct++,0) || PetscMPITypeSize((count),(datatype),(&petsc_recv_len)) || MPI_Recv((buf),(count),(datatype),(source),(tag),(comm),(status))) 46415308354SBarry Smith 46577a39924SBarry Smith #define MPI_Send(buf,count,datatype,dest,tag,comm) \ 466f12d10ccSJunchao Zhang ((petsc_send_ct++,0) || PetscMPITypeSize((count),(datatype),(&petsc_send_len)) || MPI_Send((buf),(count),(datatype),(dest),(tag),(comm))) 46777a39924SBarry Smith 46877a39924SBarry Smith #define MPI_Wait(request,status) \ 4695fa69175SJed Brown ((petsc_wait_ct++,petsc_sum_of_waits_ct++,0) || MPI_Wait((request),(status))) 47077a39924SBarry Smith 47177a39924SBarry Smith #define MPI_Waitany(a,b,c,d) \ 4725fa69175SJed Brown ((petsc_wait_any_ct++,petsc_sum_of_waits_ct++,0) || MPI_Waitany((a),(b),(c),(d))) 47377a39924SBarry Smith 47477a39924SBarry Smith #define MPI_Waitall(count,array_of_requests,array_of_statuses) \ 4755fa69175SJed Brown ((petsc_wait_all_ct++,petsc_sum_of_waits_ct += (PetscLogDouble) (count),0) || MPI_Waitall((count),(array_of_requests),(array_of_statuses))) 47677a39924SBarry Smith 47777a39924SBarry Smith #define MPI_Allreduce(sendbuf,recvbuf,count,datatype,op,comm) \ 4785fa69175SJed Brown ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || MPI_Allreduce((sendbuf),(recvbuf),(count),(datatype),(op),(comm))) 4793914022bSBarry Smith 48054fe5c21SBarry Smith #define MPI_Bcast(buffer,count,datatype,root,comm) \ 4815fa69175SJed Brown ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || MPI_Bcast((buffer),(count),(datatype),(root),(comm))) 48254fe5c21SBarry Smith 483d6e4c47cSJed Brown #define MPI_Reduce_scatter_block(sendbuf,recvbuf,recvcount,datatype,op,comm) \ 4845fa69175SJed Brown ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || MPI_Reduce_scatter_block((sendbuf),(recvbuf),(recvcount),(datatype),(op),(comm))) 485d6e4c47cSJed Brown 4863b9284c0SMatthew G Knepley #define MPI_Alltoall(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,comm) \ 487f12d10ccSJunchao Zhang ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Alltoall((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(comm))) 4883b9284c0SMatthew G Knepley 4893b9284c0SMatthew G Knepley #define MPI_Alltoallv(sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm) \ 490f12d10ccSJunchao Zhang ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || PetscMPITypeSizeComm((comm),(sendcnts),(sendtype),(&petsc_send_len)) || MPI_Alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm))) 4913b9284c0SMatthew G Knepley 49201faf4e4SMatthew Knepley #define MPI_Allgather(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,comm) \ 4935fa69175SJed Brown ((petsc_gather_ct += PetscMPIParallelComm((comm)),0) || MPI_Allgather((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(comm))) 49401faf4e4SMatthew Knepley 49501faf4e4SMatthew Knepley #define MPI_Allgatherv(sendbuf,sendcount,sendtype,recvbuf,recvcount,displs,recvtype,comm) \ 4965fa69175SJed Brown ((petsc_gather_ct += PetscMPIParallelComm((comm)),0) || MPI_Allgatherv((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(displs),(recvtype),(comm))) 49701faf4e4SMatthew Knepley 49801faf4e4SMatthew Knepley #define MPI_Gather(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,root,comm) \ 499f12d10ccSJunchao Zhang ((petsc_gather_ct++,0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Gather((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm))) 50001faf4e4SMatthew Knepley 50101faf4e4SMatthew Knepley #define MPI_Gatherv(sendbuf,sendcount,sendtype,recvbuf,recvcount,displs,recvtype,root,comm) \ 502f12d10ccSJunchao Zhang ((petsc_gather_ct++,0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Gatherv((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(displs),(recvtype),(root),(comm))) 50301faf4e4SMatthew Knepley 50401faf4e4SMatthew Knepley #define MPI_Scatter(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,root,comm) \ 505f12d10ccSJunchao Zhang ((petsc_scatter_ct++,0) || PetscMPITypeSize((recvcount),(recvtype),(&petsc_recv_len)) || MPI_Scatter((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm))) 50601faf4e4SMatthew Knepley 50701faf4e4SMatthew Knepley #define MPI_Scatterv(sendbuf,sendcount,displs,sendtype,recvbuf,recvcount,recvtype,root,comm) \ 508f12d10ccSJunchao Zhang ((petsc_scatter_ct++,0) || PetscMPITypeSize((recvcount),(recvtype),(&petsc_recv_len)) || MPI_Scatterv((sendbuf),(sendcount),(displs),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm))) 509f12d10ccSJunchao Zhang 510f12d10ccSJunchao Zhang #define MPI_Ialltoall(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,comm,request) \ 511f12d10ccSJunchao Zhang ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Ialltoall((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(comm),(request))) 512f12d10ccSJunchao Zhang 513f12d10ccSJunchao Zhang #define MPI_Ialltoallv(sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm,request) \ 514f12d10ccSJunchao Zhang ((petsc_allreduce_ct += PetscMPIParallelComm((comm)),0) || PetscMPITypeSizeComm((comm),(sendcnts),(sendtype),(&petsc_send_len)) || MPI_Ialltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm),(request))) 515f12d10ccSJunchao Zhang 516f12d10ccSJunchao Zhang #define MPI_Iallgather(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,comm,request) \ 517f12d10ccSJunchao Zhang ((petsc_gather_ct += PetscMPIParallelComm((comm)),0) || MPI_Iallgather((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(comm),(request))) 518f12d10ccSJunchao Zhang 519f12d10ccSJunchao Zhang #define MPI_Iallgatherv(sendbuf,sendcount,sendtype,recvbuf,recvcount,displs,recvtype,comm,request) \ 520f12d10ccSJunchao Zhang ((petsc_gather_ct += PetscMPIParallelComm((comm)),0) || MPI_Iallgatherv((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(displs),(recvtype),(comm),(request))) 521f12d10ccSJunchao Zhang 522f12d10ccSJunchao Zhang #define MPI_Igather(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,root,comm,request) \ 523f12d10ccSJunchao Zhang ((petsc_gather_ct++,0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Igather((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm),(request))) 524f12d10ccSJunchao Zhang 525f12d10ccSJunchao Zhang #define MPI_Igatherv(sendbuf,sendcount,sendtype,recvbuf,recvcount,displs,recvtype,root,comm,request) \ 526f12d10ccSJunchao Zhang ((petsc_gather_ct++,0) || PetscMPITypeSize((sendcount),(sendtype),(&petsc_send_len)) || MPI_Igatherv((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(displs),(recvtype),(root),(comm),(request))) 527f12d10ccSJunchao Zhang 528f12d10ccSJunchao Zhang #define MPI_Iscatter(sendbuf,sendcount,sendtype,recvbuf,recvcount,recvtype,root,comm,request) \ 529f12d10ccSJunchao Zhang ((petsc_scatter_ct++,0) || PetscMPITypeSize((recvcount),(recvtype),(&petsc_recv_len)) || MPI_Iscatter((sendbuf),(sendcount),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm),(request))) 530f12d10ccSJunchao Zhang 531f12d10ccSJunchao Zhang #define MPI_Iscatterv(sendbuf,sendcount,displs,sendtype,recvbuf,recvcount,recvtype,root,comm,request) \ 532f12d10ccSJunchao Zhang ((petsc_scatter_ct++,0) || PetscMPITypeSize((recvcount),(recvtype),(&petsc_recv_len)) || MPI_Iscatterv((sendbuf),(sendcount),(displs),(sendtype),(recvbuf),(recvcount),(recvtype),(root),(comm),(request))) 533f12d10ccSJunchao Zhang 534f12d10ccSJunchao Zhang /* We treat MPI_Ineighbor_alltoallv as a set of isend/irecv instead of a traditional MPI collective. 535f12d10ccSJunchao Zhang OpenMPI-3.0 ran into error with outdegree = indegree = 0, so we use ((outdegree) || (indegree)) as a workaround. 536f12d10ccSJunchao Zhang */ 537f12d10ccSJunchao Zhang #define MPI_Start_ineighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm,request) \ 538f12d10ccSJunchao Zhang ((petsc_isend_ct += (PetscLogDouble)(outdegree),0) || (petsc_irecv_ct += (PetscLogDouble)(indegree),0) || PetscMPITypeSizeCount((outdegree),(sendcnts),(sendtype),(&petsc_isend_len)) || PetscMPITypeSizeCount((indegree),(recvcnts),(recvtype),(&petsc_irecv_len)) || (((outdegree) || (indegree)) && MPI_Ineighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm),(request)))) 539f12d10ccSJunchao Zhang 540f12d10ccSJunchao Zhang #define MPI_Start_neighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm) \ 541f12d10ccSJunchao Zhang ((petsc_isend_ct += (PetscLogDouble)(outdegree),0) || (petsc_irecv_ct += (PetscLogDouble)(indegree),0) || PetscMPITypeSizeCount((outdegree),(sendcnts),(sendtype),(&petsc_isend_len)) || PetscMPITypeSizeCount((indegree),(recvcnts),(recvtype),(&petsc_irecv_len)) || (((outdegree) || (indegree)) && MPI_Neighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm)))) 54201faf4e4SMatthew Knepley 5430d4b0b6cSBarry Smith #else 5440d4b0b6cSBarry Smith 545457b14b1SJunchao Zhang #define MPI_Startall_irecv(count,datatype,number,requests) \ 546ea1c87f7SJunchao Zhang ((number) && MPI_Startall((number),(requests))) 5470d4b0b6cSBarry Smith 548457b14b1SJunchao Zhang #define MPI_Startall_isend(count,datatype,number,requests) \ 549ea1c87f7SJunchao Zhang ((number) && MPI_Startall((number),(requests))) 5500d4b0b6cSBarry Smith 551457b14b1SJunchao Zhang #define MPI_Start_isend(count,datatype,requests) \ 5525fa69175SJed Brown (MPI_Start((requests))) 5530d4b0b6cSBarry Smith 554f12d10ccSJunchao Zhang #define MPI_Start_ineighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm,request) \ 555f12d10ccSJunchao Zhang (((outdegree) || (indegree)) && MPI_Ineighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm),(request))) 556f12d10ccSJunchao Zhang 557f12d10ccSJunchao Zhang #define MPI_Start_neighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm) \ 558f12d10ccSJunchao Zhang (((outdegree) || (indegree)) && MPI_Neighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm))) 559c9b973beSBarry Smith #endif /* !MPIUNI_H && ! PETSC_HAVE_BROKEN_RECURSIVE_MACRO */ 560614700edSBarry Smith 561df8cf0b5SBarry Smith #else /* ---Logging is turned off --------------------------------------------*/ 562614700edSBarry Smith 563e3ed9ee7SBarry Smith #define PetscLogMemory PETSC_FALSE 564e3ed9ee7SBarry Smith 565b0a32e0cSBarry Smith #define PetscLogFlops(n) 0 566fa2bb9feSLisandro Dalcin #define PetscGetFlops(a) (*(a) = 0.0,0) 567614700edSBarry Smith 568fa2bb9feSLisandro Dalcin #define PetscLogStageRegister(a,b) 0 569fa2bb9feSLisandro Dalcin #define PetscLogStagePush(a) 0 570fa2bb9feSLisandro Dalcin #define PetscLogStagePop() 0 571f5d6ab90SLisandro Dalcin #define PetscLogStageSetActive(a,b) 0 572f5d6ab90SLisandro Dalcin #define PetscLogStageGetActive(a,b) 0 573f5d6ab90SLisandro Dalcin #define PetscLogStageGetVisible(a,b) 0 574f5d6ab90SLisandro Dalcin #define PetscLogStageSetVisible(a,b) 0 575f5d6ab90SLisandro Dalcin #define PetscLogStageGetId(a,b) (*(b)=0,0) 576f5d6ab90SLisandro Dalcin 577fa2bb9feSLisandro Dalcin #define PetscLogEventRegister(a,b,c) 0 578fa2bb9feSLisandro Dalcin #define PetscLogEventSetCollective(a,b) 0 579fa2bb9feSLisandro Dalcin #define PetscLogEventIncludeClass(a) 0 580fa2bb9feSLisandro Dalcin #define PetscLogEventExcludeClass(a) 0 581b0a32e0cSBarry Smith #define PetscLogEventActivate(a) 0 582b0a32e0cSBarry Smith #define PetscLogEventDeactivate(a) 0 583b0a32e0cSBarry Smith #define PetscLogEventActivateClass(a) 0 584b0a32e0cSBarry Smith #define PetscLogEventDeactivateClass(a) 0 585e8e7597cSSatish Balay #define PetscLogEventSetActiveAll(a,b) 0 58646eb3cd7SSatish Balay #define PetscLogEventGetId(a,b) (*(b)=0,0) 587561b2a67SSatish Balay #define PetscLogEventGetPerfInfo(a,b,c) 0 5889c702cfcSMatthew G. Knepley #define PetscLogEventSetDof(a,b,c) 0 5899c702cfcSMatthew G. Knepley #define PetscLogEventSetError(a,b,c) 0 59077c4ece6SBarry Smith 5918b5db460SBarry Smith #define PetscLogPLB 0 5928b5db460SBarry Smith #define PetscLogPLE 0 5938b5db460SBarry Smith #define PetscLogPHC 0 5948b5db460SBarry Smith #define PetscLogPHD 0 595f5d6ab90SLisandro Dalcin 596efee365bSSatish Balay #define PetscLogObjectParents(p,n,c) 0 59752e6d16bSBarry Smith #define PetscLogObjectCreate(h) 0 59852e6d16bSBarry Smith #define PetscLogObjectDestroy(h) 0 599fa2bb9feSLisandro Dalcin PETSC_EXTERN PetscErrorCode PetscLogObjectState(PetscObject,const char[],...); 600fa2bb9feSLisandro Dalcin 601bb1d7374SBarry Smith #define PetscLogDefaultBegin() 0 602b0a32e0cSBarry Smith #define PetscLogAllBegin() 0 603bb1d7374SBarry Smith #define PetscLogNestedBegin() 0 604fa2bb9feSLisandro Dalcin #define PetscLogTraceBegin(file) 0 60573fda44aSBarry Smith #define PetscLogActions(a) 0 606fa2bb9feSLisandro Dalcin #define PetscLogObjects(a) 0 607fa2bb9feSLisandro Dalcin #define PetscLogSetThreshold(a,b) 0 608fa2bb9feSLisandro Dalcin #define PetscLogSet(lb,le) 0 609fa2bb9feSLisandro Dalcin 610fa2bb9feSLisandro Dalcin #define PetscLogView(viewer) 0 611fa2bb9feSLisandro Dalcin #define PetscLogViewFromOptions() 0 612fa2bb9feSLisandro Dalcin #define PetscLogDump(c) 0 613fa2bb9feSLisandro Dalcin 61462872c28SLisandro Dalcin #define PetscLogEventSync(e,comm) 0 615fa2bb9feSLisandro Dalcin #define PetscLogEventBegin(e,o1,o2,o3,o4) 0 616fa2bb9feSLisandro Dalcin #define PetscLogEventEnd(e,o1,o2,o3,o4) 0 617ce6058e1SBarry Smith 618aa482453SBarry Smith /* If PETSC_USE_LOG is NOT defined, these still need to be! */ 619ea1c87f7SJunchao Zhang #define MPI_Startall_irecv(count,datatype,number,requests) ((number) && MPI_Startall(number,requests)) 620ea1c87f7SJunchao Zhang #define MPI_Startall_isend(count,datatype,number,requests) ((number) && MPI_Startall(number,requests)) 621457b14b1SJunchao Zhang #define MPI_Start_isend(count,datatype,requests) MPI_Start(requests) 622f12d10ccSJunchao Zhang #define MPI_Start_ineighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm,request) \ 623f12d10ccSJunchao Zhang (((outdegree) || (indegree)) && MPI_Ineighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm),(request))) 624f12d10ccSJunchao Zhang #define MPI_Start_neighbor_alltoallv(outdegree,indegree,sendbuf,sendcnts,sdispls,sendtype,recvbuf,recvcnts,rdispls,recvtype,comm) \ 625f12d10ccSJunchao Zhang (((outdegree) || (indegree)) && MPI_Neighbor_alltoallv((sendbuf),(sendcnts),(sdispls),(sendtype),(recvbuf),(recvcnts),(rdispls),(recvtype),(comm))) 626f141ce34SMatthew Knepley 627aa482453SBarry Smith #endif /* PETSC_USE_LOG */ 6286daaf66cSBarry Smith 629204a6943SJed Brown #define PetscPreLoadBegin(flag,name) \ 630204a6943SJed Brown do {\ 631204a6943SJed Brown PetscBool PetscPreLoading = flag;\ 632204a6943SJed Brown int PetscPreLoadMax,PetscPreLoadIt;\ 633166c7f25SBarry Smith PetscLogStage _stageNum;\ 6348cbcd9ccSBarry Smith PetscErrorCode _3_ierr; \ 635c5929fdfSBarry Smith _3_ierr = PetscOptionsGetBool(NULL,NULL,"-preload",&PetscPreLoading,NULL);CHKERRQ(_3_ierr); \ 636204a6943SJed Brown PetscPreLoadMax = (int)(PetscPreLoading);\ 637204a6943SJed Brown PetscPreLoadingUsed = PetscPreLoading ? PETSC_TRUE : PetscPreLoadingUsed;\ 638204a6943SJed Brown for (PetscPreLoadIt=0; PetscPreLoadIt<=PetscPreLoadMax; PetscPreLoadIt++) {\ 639204a6943SJed Brown PetscPreLoadingOn = PetscPreLoading;\ 6400298fd71SBarry Smith _3_ierr = PetscBarrier(NULL);CHKERRQ(_3_ierr);\ 641204a6943SJed Brown if (PetscPreLoadIt>0) {\ 6428e58c17dSMatthew Knepley _3_ierr = PetscLogStageGetId(name,&_stageNum);CHKERRQ(_3_ierr);\ 6438e58c17dSMatthew Knepley } else {\ 644a3bc4eb9SBarry Smith _3_ierr = PetscLogStageRegister(name,&_stageNum);CHKERRQ(_3_ierr); \ 6458e58c17dSMatthew Knepley }\ 646204a6943SJed Brown _3_ierr = PetscLogStageSetActive(_stageNum,(PetscBool)(!PetscPreLoadMax || PetscPreLoadIt));\ 6478e58c17dSMatthew Knepley _3_ierr = PetscLogStagePush(_stageNum);CHKERRQ(_3_ierr); 6488e58c17dSMatthew Knepley 649204a6943SJed Brown #define PetscPreLoadEnd() \ 650043328b6SSatish Balay _3_ierr = PetscLogStagePop();CHKERRQ(_3_ierr);\ 651204a6943SJed Brown PetscPreLoading = PETSC_FALSE;\ 652043328b6SSatish Balay }\ 653204a6943SJed Brown } while (0) 6548e58c17dSMatthew Knepley 655204a6943SJed Brown #define PetscPreLoadStage(name) do { \ 656043328b6SSatish Balay _3_ierr = PetscLogStagePop();CHKERRQ(_3_ierr); \ 657204a6943SJed Brown if (PetscPreLoadIt>0) { \ 6588e58c17dSMatthew Knepley _3_ierr = PetscLogStageGetId(name,&_stageNum);CHKERRQ(_3_ierr); \ 6598e58c17dSMatthew Knepley } else { \ 660a3bc4eb9SBarry Smith _3_ierr = PetscLogStageRegister(name,&_stageNum);CHKERRQ(_3_ierr); \ 6618e58c17dSMatthew Knepley } \ 662204a6943SJed Brown _3_ierr = PetscLogStageSetActive(_stageNum,(PetscBool)(!PetscPreLoadMax || PetscPreLoadIt)); \ 663204a6943SJed Brown _3_ierr = PetscLogStagePush(_stageNum);CHKERRQ(_3_ierr); \ 664204a6943SJed Brown } while (0) 665e9fa29b7SSatish Balay 666ad14c47eSJed Brown /* some vars for logging */ 667ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingUsed; /* true if we are or have done preloading */ 668ad14c47eSJed Brown PETSC_EXTERN PetscBool PetscPreLoadingOn; /* true if we are currently in a preloading calculation */ 669ad14c47eSJed Brown 67097bb86f7SLois Curfman McInnes #endif 671