Merge pull request #6695 from bosilca/fix/vector_stride_0

A big refresh of the datatype engine

ompi/datatype/ompi_datatype.h

@@ -1,6 +1,6 @@
 /* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
 /*
- * Copyright (c) 2009-2013 The University of Tennessee and The University
+ * Copyright (c) 2009-2019 The University of Tennessee and The University
  *                         of Tennessee Research Foundation.  All rights
  *                         reserved.
  * Copyright (c) 2009      Oak Ridge National Labs.  All rights reserved.
@@ -76,7 +76,7 @@ struct ompi_datatype_t {
     struct opal_hash_table_t *d_keyhash;         /**< Attribute fields */
 
     void*              args;                     /**< Data description for the user */
-    opal_atomic_intptr_t packed_description;       /**< Packed description of the datatype */
+    opal_atomic_intptr_t packed_description;     /**< Packed description of the datatype */
     uint64_t           pml_data;                 /**< PML-specific information */
     /* --- cacheline 6 boundary (384 bytes) --- */
     char               name[MPI_MAX_OBJECT_NAME];/**< Externally visible name */

ompi/datatype/ompi_datatype_create_indexed.c

@@ -87,10 +87,10 @@ int32_t ompi_datatype_create_hindexed( int count, const int* pBlockLength, const
         return ompi_datatype_duplicate( &ompi_mpi_datatype_null.dt, newType);
     }
 
+    ompi_datatype_type_extent( oldType, &extent );
     disp = pDisp[i];
     dLength = pBlockLength[i];
     endat = disp + dLength * extent;
-    ompi_datatype_type_extent( oldType, &extent );
 
     pdt = ompi_datatype_create( (count - i) * (2 + oldType->super.desc.used) );
     for( i += 1; i < count; i++ ) {
@@ -162,17 +162,17 @@ int32_t ompi_datatype_create_hindexed_block( int count, int bLength, const ptrdi
     pdt = ompi_datatype_create( count * (2 + oldType->super.desc.used) );
     disp = pDisp[0];
     dLength = bLength;
-    endat = disp + dLength;
+    endat = disp + dLength * extent;
     for( i = 1; i < count; i++ ) {
         if( endat == pDisp[i] ) {
             /* contiguous with the previsious */
             dLength += bLength;
-            endat += bLength;
+            endat += bLength * extent;
         } else {
             ompi_datatype_add( pdt, oldType, dLength, disp, extent );
             disp = pDisp[i];
             dLength = bLength;
-            endat = disp + bLength;
+            endat = disp + bLength * extent;
         }
     }
     ompi_datatype_add( pdt, oldType, dLength, disp, extent );

ompi/datatype/ompi_datatype_external.c

@@ -3,7 +3,7 @@
  * Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
  *                         University Research and Technology
  *                         Corporation.  All rights reserved.
- * Copyright (c) 2004-2016 The University of Tennessee and The University
+ * Copyright (c) 2004-2019 The University of Tennessee and The University
  *                         of Tennessee Research Foundation.  All rights
  *                         reserved.
  * Copyright (c) 2004-2008 High Performance Computing Center Stuttgart,
@@ -26,7 +26,6 @@
 #include <stdio.h>
 
 #include "ompi/runtime/params.h"
-#include "ompi/communicator/communicator.h"
 #include "ompi/datatype/ompi_datatype.h"
 #include "opal/datatype/opal_convertor.h"
 

opal/datatype/opal_convertor.c

@@ -3,7 +3,7 @@
  * Copyright (c) 2004-2006 The Trustees of Indiana University and Indiana
  *                         University Research and Technology
  *                         Corporation.  All rights reserved.
- * Copyright (c) 2004-2018 The University of Tennessee and The University
+ * Copyright (c) 2004-2019 The University of Tennessee and The University
  *                         of Tennessee Research Foundation.  All rights
  *                         reserved.
  * Copyright (c) 2004-2006 High Performance Computing Center Stuttgart,
@@ -324,8 +324,9 @@ int32_t opal_convertor_unpack( opal_convertor_t* pConv,
     return pConv->fAdvance( pConv, iov, out_size, max_data );
 }
 
-static inline int opal_convertor_create_stack_with_pos_contig( opal_convertor_t* pConvertor,
-                                                               size_t starting_point, const size_t* sizes )
+static inline int
+opal_convertor_create_stack_with_pos_contig( opal_convertor_t* pConvertor,
+                                             size_t starting_point, const size_t* sizes )
 {
     dt_stack_t* pStack;   /* pointer to the position on the stack */
     const opal_datatype_t* pData = pConvertor->pDesc;
@@ -349,14 +350,14 @@ static inline int opal_convertor_create_stack_with_pos_contig( opal_convertor_t*
     pStack[0].disp     = count * extent;
 
     /* now compute the number of pending bytes */
-    count = starting_point - count * pData->size;
+    count = starting_point % pData->size;
     /**
      * We save the current displacement starting from the begining
      * of this data.
      */
     if( OPAL_LIKELY(0 == count) ) {
         pStack[1].type     = pElems->elem.common.type;
-        pStack[1].count    = pElems->elem.count;
+        pStack[1].count    = pElems->elem.blocklen;
     } else {
         pStack[1].type  = OPAL_DATATYPE_UINT1;
         pStack[1].count = pData->size - count;
@@ -370,9 +371,9 @@ static inline int opal_convertor_create_stack_with_pos_contig( opal_convertor_t*
     return OPAL_SUCCESS;
 }
 
-static inline
-int opal_convertor_create_stack_at_begining( opal_convertor_t* convertor,
-                                             const size_t* sizes )
+static inline int
+opal_convertor_create_stack_at_begining( opal_convertor_t* convertor,
+                                         const size_t* sizes )
 {
     dt_stack_t* pStack = convertor->pStack;
     dt_elem_desc_t* pElems;
@@ -402,7 +403,7 @@ int opal_convertor_create_stack_at_begining( opal_convertor_t* convertor,
         pStack[1].count = pElems[0].loop.loops;
         pStack[1].type  = OPAL_DATATYPE_LOOP;
     } else {
-        pStack[1].count = pElems[0].elem.count;
+        pStack[1].count = pElems[0].elem.count * pElems[0].elem.blocklen;
         pStack[1].type  = pElems[0].elem.common.type;
     }
     return OPAL_SUCCESS;
@@ -578,8 +579,9 @@ int32_t opal_convertor_prepare_for_recv( opal_convertor_t* convertor,
     assert(! (convertor->flags & CONVERTOR_SEND));
     OPAL_CONVERTOR_PREPARE( convertor, datatype, count, pUserBuf );
 
-    if( convertor->flags & CONVERTOR_WITH_CHECKSUM ) {
-        if( !(convertor->flags & CONVERTOR_HOMOGENEOUS) ) {
+#if defined(CHECKSUM)
+    if( OPAL_UNLIKELY(convertor->flags & CONVERTOR_WITH_CHECKSUM) ) {
+        if( OPAL_UNLIKELY(!(convertor->flags & CONVERTOR_HOMOGENEOUS)) ) {
             convertor->fAdvance = opal_unpack_general_checksum;
         } else {
             if( convertor->pDesc->flags & OPAL_DATATYPE_FLAG_CONTIGUOUS ) {
@@ -588,8 +590,9 @@ int32_t opal_convertor_prepare_for_recv( opal_convertor_t* convertor,
                 convertor->fAdvance = opal_generic_simple_unpack_checksum;
             }
         }
-    } else {
-        if( !(convertor->flags & CONVERTOR_HOMOGENEOUS) ) {
+    } else
+#endif  /* defined(CHECKSUM) */
+        if( OPAL_UNLIKELY(!(convertor->flags & CONVERTOR_HOMOGENEOUS)) ) {
             convertor->fAdvance = opal_unpack_general;
         } else {
             if( convertor->pDesc->flags & OPAL_DATATYPE_FLAG_CONTIGUOUS ) {
@@ -598,7 +601,6 @@ int32_t opal_convertor_prepare_for_recv( opal_convertor_t* convertor,
                 convertor->fAdvance = opal_generic_simple_unpack;
             }
         }
-    }
     return OPAL_SUCCESS;
 }
 
@@ -617,6 +619,7 @@ int32_t opal_convertor_prepare_for_send( opal_convertor_t* convertor,
 
     OPAL_CONVERTOR_PREPARE( convertor, datatype, count, pUserBuf );
 
+#if defined(CHECKSUM)
     if( convertor->flags & CONVERTOR_WITH_CHECKSUM ) {
         if( CONVERTOR_SEND_CONVERSION == (convertor->flags & (CONVERTOR_SEND_CONVERSION|CONVERTOR_HOMOGENEOUS)) ) {
             convertor->fAdvance = opal_pack_general_checksum;
@@ -631,7 +634,8 @@ int32_t opal_convertor_prepare_for_send( opal_convertor_t* convertor,
                 convertor->fAdvance = opal_generic_simple_pack_checksum;
             }
         }
-    } else {
+    } else
+#endif  /* defined(CHECKSUM) */
         if( CONVERTOR_SEND_CONVERSION == (convertor->flags & (CONVERTOR_SEND_CONVERSION|CONVERTOR_HOMOGENEOUS)) ) {
             convertor->fAdvance = opal_pack_general;
         } else {
@@ -645,7 +649,6 @@ int32_t opal_convertor_prepare_for_send( opal_convertor_t* convertor,
                 convertor->fAdvance = opal_generic_simple_pack;
             }
         }
-    }
     return OPAL_SUCCESS;
 }
 

opal/datatype/opal_convertor.h

@@ -332,8 +332,10 @@ opal_convertor_set_position( opal_convertor_t* convertor,
     /* Remove the completed flag if it's already set */
     convertor->flags &= ~CONVERTOR_COMPLETED;
 
-    if( !(convertor->flags & CONVERTOR_WITH_CHECKSUM) &&
-        (convertor->flags & OPAL_DATATYPE_FLAG_NO_GAPS) &&
+    if( (convertor->flags & OPAL_DATATYPE_FLAG_NO_GAPS) &&
+#if defined(CHECKSUM)
+        !(convertor->flags & CONVERTOR_WITH_CHECKSUM) &&
+#endif  /* defined(CHECKSUM) */
         (convertor->flags & (CONVERTOR_SEND | CONVERTOR_HOMOGENEOUS)) ) {
         /* Contiguous and no checkpoint and no homogeneous unpack */
         convertor->bConverted = *position;