diff options
Diffstat (limited to 'src/H5Cimage.c')
| -rw-r--r-- | src/H5Cimage.c | 538 |
1 files changed, 269 insertions, 269 deletions
diff --git a/src/H5Cimage.c b/src/H5Cimage.c index db44c7a..ee286d9 100644 --- a/src/H5Cimage.c +++ b/src/H5Cimage.c @@ -102,11 +102,11 @@ static herr_t H5C__decode_cache_image_header(const H5F_t *f, static herr_t H5C__decode_cache_image_entry(const H5F_t *f, const H5C_t *cache_ptr, const uint8_t **buf, unsigned entry_num); #endif /* NDEBUG */ /* only used in assertions */ -static herr_t H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, +static herr_t H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, H5C_cache_entry_t *pf_entry_ptr, H5C_cache_entry_t **fd_children); static herr_t H5C__encode_cache_image_header(const H5F_t *f, const H5C_t *cache_ptr, uint8_t **buf); -static herr_t H5C__encode_cache_image_entry(H5F_t *f, H5C_t *cache_ptr, +static herr_t H5C__encode_cache_image_entry(H5F_t *f, H5C_t *cache_ptr, uint8_t **buf, unsigned entry_num); static herr_t H5C__prep_for_file_close__compute_fd_heights(const H5C_t *cache_ptr); static void H5C__prep_for_file_close__compute_fd_heights_real( @@ -146,8 +146,8 @@ H5FL_DEFINE(H5C_cache_entry_t); * * Function: H5C_cache_image_pending() * - * Purpose: Tests to see if the load of a metadata cache image - * load is pending (i.e. will be executed on the next + * Purpose: Tests to see if the load of a metadata cache image + * load is pending (i.e. will be executed on the next * protect or insert) * * Returns TRUE if a cache image load is pending, and FALSE @@ -179,16 +179,16 @@ H5C_cache_image_pending(const H5C_t *cache_ptr) /*------------------------------------------------------------------------- * Function: H5C_cache_image_status() * - * Purpose: Examine the metadata cache associated with the supplied - * instance of H5F_t to determine whether the load of a - * cache image has either been queued or executed, and if + * Purpose: Examine the metadata cache associated with the supplied + * instance of H5F_t to determine whether the load of a + * cache image has either been queued or executed, and if * construction of a cache image has been requested. * * This done, it set *load_ci_ptr to TRUE if a cache image * has either been loaded or a load has been requested, and * to FALSE otherwise. * - * Similarly, set *write_ci_ptr to TRUE if construction of + * Similarly, set *write_ci_ptr to TRUE if construction of * a cache image has been requested, and to FALSE otherwise. * * Return: SUCCEED on success, and FAIL on failure. @@ -213,7 +213,7 @@ H5C_cache_image_status(H5F_t * f, hbool_t *load_ci_ptr, hbool_t *write_ci_ptr) HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); HDassert(load_ci_ptr); HDassert(write_ci_ptr); - + *load_ci_ptr = cache_ptr->load_image || cache_ptr->image_loaded; *write_ci_ptr = cache_ptr->image_ctl.generate_image; @@ -224,7 +224,7 @@ H5C_cache_image_status(H5F_t * f, hbool_t *load_ci_ptr, hbool_t *write_ci_ptr) /*------------------------------------------------------------------------- * Function: H5C__construct_cache_image_buffer() * - * Purpose: Allocate a buffer of size cache_ptr->image_len, and + * Purpose: Allocate a buffer of size cache_ptr->image_len, and * load it with an image of the metadata cache image block. * * Note that by the time this function is called, the cache @@ -330,7 +330,7 @@ H5C__construct_cache_image_buffer(H5F_t * f, H5C_t *cache_ptr) HDassert((cache_ptr->image_entries)[u].type_id == (fake_cache_ptr->image_entries)[u].type_id); HDassert((cache_ptr->image_entries)[u].lru_rank == (fake_cache_ptr->image_entries)[u].lru_rank); HDassert((cache_ptr->image_entries)[u].is_dirty == (fake_cache_ptr->image_entries)[u].is_dirty); - /* don't check image_fd_height as it is not stored in + /* don't check image_fd_height as it is not stored in * the metadata cache image block. */ HDassert((cache_ptr->image_entries)[u].fd_child_count == (fake_cache_ptr->image_entries)[u].fd_child_count); @@ -346,7 +346,7 @@ H5C__construct_cache_image_buffer(H5F_t * f, H5C_t *cache_ptr) (fake_cache_ptr->image_entries)[u].fd_parent_addrs = (haddr_t *)H5MM_xfree((fake_cache_ptr->image_entries)[u].fd_parent_addrs); (fake_cache_ptr->image_entries)[u].fd_parent_count = 0; } /* end if */ - else + else HDassert((fake_cache_ptr->image_entries)[u].fd_parent_count == 0); HDassert((cache_ptr->image_entries)[u].image_ptr); @@ -431,28 +431,28 @@ done: * Function: H5C__deserialize_prefetched_entry() * * Purpose: Deserialize the supplied prefetched entry entry, and return - * a pointer to the deserialized entry in *entry_ptr_ptr. + * a pointer to the deserialized entry in *entry_ptr_ptr. * If successful, remove the prefetched entry from the cache, * and free it. Insert the deserialized entry into the cache. * - * Note that the on disk image of the entry is not freed -- + * Note that the on disk image of the entry is not freed -- * a pointer to it is stored in the deserialized entries' * image_ptr field, and its image_up_to_date field is set to * TRUE unless the entry is dirtied by the deserialize call. * * If the prefetched entry is a flush dependency child, - * destroy that flush dependency prior to calling the + * destroy that flush dependency prior to calling the * deserialize callback. If appropriate, the flush dependency * relationship will be recreated by the cache client. * * If the prefetched entry is a flush dependency parent, - * destroy the flush dependency relationship with all its + * destroy the flush dependency relationship with all its * children. As all these children must be prefetched entries, - * recreate these flush dependency relationships with + * recreate these flush dependency relationships with * deserialized entry after it is inserted in the cache. * - * Since deserializing a prefetched entry is semantically - * equivalent to a load, issue an entry loaded nofification + * Since deserializing a prefetched entry is semantically + * equivalent to a load, issue an entry loaded nofification * if the notify callback is defined. * * Return: SUCCEED on success, and FAIL on failure. @@ -468,22 +468,22 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, H5C_cache_entry_t **entry_ptr_ptr, const H5C_class_t *type, haddr_t addr, void *udata) { - hbool_t dirty = FALSE; /* Flag indicating whether thing was - * dirtied during deserialize + hbool_t dirty = FALSE; /* Flag indicating whether thing was + * dirtied during deserialize */ size_t len; /* Size of image in file */ void * thing = NULL; /* Pointer to thing loaded */ H5C_cache_entry_t * pf_entry_ptr; /* pointer to the prefetched entry */ /* supplied in *entry_ptr_ptr. */ - H5C_cache_entry_t * ds_entry_ptr; /* Alias for thing loaded, as cache - * entry + H5C_cache_entry_t * ds_entry_ptr; /* Alias for thing loaded, as cache + * entry */ H5C_cache_entry_t** fd_children = NULL; /* Pointer to a dynamically */ /* allocated array of pointers to */ /* the flush dependency children of */ /* the prefetched entry, or NULL if */ /* that array does not exist. */ - unsigned flush_flags = (H5C__FLUSH_INVALIDATE_FLAG | + unsigned flush_flags = (H5C__FLUSH_INVALIDATE_FLAG | H5C__FLUSH_CLEAR_ONLY_FLAG); int i; herr_t ret_value = SUCCEED; /* Return value */ @@ -513,8 +513,8 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, /* verify absence of prohibited or unsupported type flag combinations */ HDassert(!(type->flags & H5C__CLASS_SKIP_READS)); - - /* Can't see how skip reads could be usefully combined with + + /* Can't see how skip reads could be usefully combined with * either the speculative read flag. Hence disallow. */ HDassert(!((type->flags & H5C__CLASS_SKIP_READS) && @@ -535,7 +535,7 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, HDassert(pf_entry_ptr->flush_dep_parent[i]->flush_dep_nchildren > 0); HDassert(pf_entry_ptr->fd_parent_addrs); HDassert(pf_entry_ptr->flush_dep_parent[i]->addr == pf_entry_ptr->fd_parent_addrs[i]); - + if(H5C_destroy_flush_dependency(pf_entry_ptr->flush_dep_parent[i], pf_entry_ptr) < 0) HGOTO_ERROR(H5E_CACHE, H5E_CANTUNDEPEND, FAIL, "can't destroy pf entry parent flush dependency") @@ -543,11 +543,11 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, } /* end for */ HDassert(pf_entry_ptr->flush_dep_nparents == 0); - /* If *pf_entry_ptr is a flush dependency parent, destroy its flush - * dependency relationships with all its children (which must be + /* If *pf_entry_ptr is a flush dependency parent, destroy its flush + * dependency relationships with all its children (which must be * prefetched entries as well). * - * These flush dependency relationships will have to be restored + * These flush dependency relationships will have to be restored * after the deserialized entry is inserted into the cache in order * to transfer these relationships to the new entry. Hence save the * pointers to the flush dependency children of *pf_enty_ptr for later @@ -561,16 +561,16 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, HGOTO_ERROR(H5E_CACHE, H5E_CANTUNDEPEND, FAIL, "can't destroy pf entry child flush dependency(s).") } /* end if */ - /* Since the size of the on disk image is known exactly, there is - * no need for either a call to the get_initial_load_size() callback, + /* Since the size of the on disk image is known exactly, there is + * no need for either a call to the get_initial_load_size() callback, * or retries if the H5C__CLASS_SPECULATIVE_LOAD_FLAG flag is set. * Similarly, there is no need to clamp possible reads beyond * EOF. */ len = pf_entry_ptr->size; - /* Deserialize the prefetched on-disk image of the entry into the - * native memory form + /* Deserialize the prefetched on-disk image of the entry into the + * native memory form */ if(NULL == (thing = type->deserialize(pf_entry_ptr->image_ptr, len, udata, &dirty))) HGOTO_ERROR(H5E_CACHE, H5E_CANTLOAD, FAIL, "Can't deserialize image") @@ -587,14 +587,14 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, * * HDassert( ( dirty == FALSE ) || ( type->id == 5 || type->id == 6 ) ); * - * note that type ids 5 & 6 are associated with object headers in the + * note that type ids 5 & 6 are associated with object headers in the * metadata cache. * * When we get to using H5C for other purposes, we may wish to * tighten up the assert so that the loophole only applies to the * metadata cache. * - * Note that at present, dirty can't be set to true with prefetched + * Note that at present, dirty can't be set to true with prefetched * entries. However this may change, so include this functionality * against that posibility. * @@ -682,8 +682,8 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, /* We have successfully deserialized the prefetched entry. * * Before we return a pointer to the deserialized entry, we must remove - * the prefetched entry from the cache, discard it, and replace it with - * the deserialized entry. Note that we do not free the prefetched + * the prefetched entry from the cache, discard it, and replace it with + * the deserialized entry. Note that we do not free the prefetched * entries image, as that has been transferred to the deserialized * entry. * @@ -695,11 +695,11 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, * 1) Set pf_entry_ptr->image_ptr to NULL. Since we have already * transferred the buffer containing the image to *ds_entry_ptr, * this is not a memory leak. - * + * * 2) Call H5C__flush_single_entry() with the H5C__FLUSH_INVALIDATE_FLAG * and H5C__FLUSH_CLEAR_ONLY_FLAG flags set. */ - pf_entry_ptr->image_ptr = NULL; + pf_entry_ptr->image_ptr = NULL; if(pf_entry_ptr->is_dirty) { HDassert(pf_entry_ptr->in_slist); flush_flags |= H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG; @@ -723,7 +723,7 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, H5C__UPDATE_RP_FOR_INSERTION(cache_ptr, ds_entry_ptr, FAIL) - /* Deserializing a prefetched entry is the conceptual equivalent of + /* Deserializing a prefetched entry is the conceptual equivalent of * loading it from file. If the deserialized entry has a notify callback, * send an "after load" notice now that the deserialized entry is fully * integrated into the cache. @@ -732,9 +732,9 @@ H5C__deserialize_prefetched_entry(H5F_t *f, H5C_t *cache_ptr, (ds_entry_ptr->type->notify)(H5C_NOTIFY_ACTION_AFTER_LOAD, ds_entry_ptr) < 0) HGOTO_ERROR(H5E_CACHE, H5E_CANTNOTIFY, FAIL, "can't notify client about entry loaded into cache") - /* Restore flush dependencies with the flush dependency children of - * of the prefetched entry. Note that we must protect *ds_entry_ptr - * before the call to avoid triggering sanity check failures, and + /* Restore flush dependencies with the flush dependency children of + * of the prefetched entry. Note that we must protect *ds_entry_ptr + * before the call to avoid triggering sanity check failures, and * then unprotect it afterwards. */ i = 0; @@ -798,8 +798,8 @@ done: /*------------------------------------------------------------------------- * Function: H5C__free_image_entries_array * - * Purpose: If the image entries array exists, free the image - * associated with each entry, and then free the image + * Purpose: If the image entries array exists, free the image + * associated with each entry, and then free the image * entries array proper. * * Note that by the time this function is called, the cache @@ -834,7 +834,7 @@ H5C__free_image_entries_array(H5C_t * cache_ptr) /* Get pointer to image entry */ ie_ptr = &((cache_ptr->image_entries)[u]); - /* Sanity checks */ + /* Sanity checks */ HDassert(ie_ptr); HDassert(ie_ptr->magic == H5C_IMAGE_ENTRY_T_MAGIC); HDassert(ie_ptr->image_ptr); @@ -866,14 +866,14 @@ H5C__free_image_entries_array(H5C_t * cache_ptr) /*------------------------------------------------------------------------- * Function: H5C_force_cache_image_load() * - * Purpose: On rare occasions, it is necessary to run + * Purpose: On rare occasions, it is necessary to run * H5MF_tidy_self_referential_fsm_hack() prior to the first - * metadata cache access. This is a problem as if there is a - * cache image at the end of the file, that routine will + * metadata cache access. This is a problem as if there is a + * cache image at the end of the file, that routine will * discard it. * * We solve this issue by calling this function, which will - * load the cache image and then call + * load the cache image and then call * H5MF_tidy_self_referential_fsm_hack() to discard it. * * Return: SUCCEED on success, and FAIL on failure. @@ -1053,12 +1053,12 @@ H5C__read_cache_image(H5F_t *f, H5C_t *cache_ptr) if ( ( NULL == aux_ptr ) || ( aux_ptr->mpi_rank == 0 ) ) { - HDassert((NULL == aux_ptr) || + HDassert((NULL == aux_ptr) || (aux_ptr->magic == H5AC__H5AC_AUX_T_MAGIC)); #endif /* H5_HAVE_PARALLEL */ /* Read the buffer (if serial access, or rank 0 of parallel access) */ - if(H5F_block_read(f, H5FD_MEM_SUPER, cache_ptr->image_addr, + if(H5F_block_read(f, H5FD_MEM_SUPER, cache_ptr->image_addr, cache_ptr->image_len, cache_ptr->image_buffer) < 0) HGOTO_ERROR(H5E_CACHE, H5E_READERROR, FAIL, "Can't read metadata cache image block") @@ -1068,9 +1068,9 @@ H5C__read_cache_image(H5F_t *f, H5C_t *cache_ptr) if ( aux_ptr ) { /* Broadcast cache image */ - if ( MPI_SUCCESS != - (mpi_result = MPI_Bcast(cache_ptr->image_buffer, - (int)cache_ptr->image_len, MPI_BYTE, + if ( MPI_SUCCESS != + (mpi_result = MPI_Bcast(cache_ptr->image_buffer, + (int)cache_ptr->image_len, MPI_BYTE, 0, aux_ptr->mpi_comm)) ) HMPI_GOTO_ERROR(FAIL, "MPI_Bcast failed", mpi_result) @@ -1080,9 +1080,9 @@ H5C__read_cache_image(H5F_t *f, H5C_t *cache_ptr) else if ( aux_ptr ) { /* Retrieve the contents of the metadata cache image from process 0 */ - if ( MPI_SUCCESS != - (mpi_result = MPI_Bcast(cache_ptr->image_buffer, - (int)cache_ptr->image_len, MPI_BYTE, + if ( MPI_SUCCESS != + (mpi_result = MPI_Bcast(cache_ptr->image_buffer, + (int)cache_ptr->image_len, MPI_BYTE, 0, aux_ptr->mpi_comm)) ) HMPI_GOTO_ERROR(FAIL, "can't receive cache image MPI_Bcast", \ @@ -1128,14 +1128,14 @@ H5C__load_cache_image(H5F_t *f) HDassert(cache_ptr); HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); - /* If the image address is defined, load the image, decode it, - * and insert its contents into the metadata cache. + /* If the image address is defined, load the image, decode it, + * and insert its contents into the metadata cache. * - * Note that under normal operating conditions, it is an error if the + * Note that under normal operating conditions, it is an error if the * image address is HADDR_UNDEF. However, to facilitate testing, * we allow this special value of the image address which means that - * no image exists, and that the load operation should be skipped - * silently. + * no image exists, and that the load operation should be skipped + * silently. */ if(H5F_addr_defined(cache_ptr->image_addr)) { /* Sanity checks */ @@ -1184,25 +1184,25 @@ done: /*------------------------------------------------------------------------- * Function: H5C_load_cache_image_on_next_protect() * - * Purpose: Note the fact that a metadata cache image superblock + * Purpose: Note the fact that a metadata cache image superblock * extension message exists, along with the base address * and length of the metadata cache image block. * - * Once this notification is received the metadata cache - * image block must be read, decoded, and loaded into the + * Once this notification is received the metadata cache + * image block must be read, decoded, and loaded into the * cache on the next call to H5C_protect(). * - * Further, if the file is opened R/W, the metadata cache - * image superblock extension message must be deleted from + * Further, if the file is opened R/W, the metadata cache + * image superblock extension message must be deleted from * the superblock extension and the image block freed * * Contrawise, if the file is openened R/O, the metadata * cache image superblock extension message and image block * must be left as is. Further, any dirty entries in the - * cache image block must be marked as clean to avoid + * cache image block must be marked as clean to avoid * attempts to write them on file close. * - * Return: SUCCEED + * Return: SUCCEED * * Programmer: John Mainzer * 7/6/15 @@ -1291,41 +1291,41 @@ H5C__image_entry_cmp(const void *_entry1, const void *_entry2) /*------------------------------------------------------------------------- * Function: H5C__prep_image_for_file_close * - * Purpose: The objective of the call is to allow the metadata cache - * to do any preparatory work prior to generation of a + * Purpose: The objective of the call is to allow the metadata cache + * to do any preparatory work prior to generation of a * cache image. * - * In particular, the cache must + * In particular, the cache must * * 1) serialize all its entries, * - * 2) compute the size of the metadata cache image, + * 2) compute the size of the metadata cache image, * * 3) allocate space for the metadata cache image, and * * 4) setup the metadata cache image superblock extension - * message with the address and size of the metadata + * message with the address and size of the metadata * cache image. * - * The parallel case is complicated by the fact that - * while all metadata caches must contain the same set of - * dirty entries, there is no such requirement for clean + * The parallel case is complicated by the fact that + * while all metadata caches must contain the same set of + * dirty entries, there is no such requirement for clean * entries or the order that entries appear in the LRU. * * Thus, there is no requirement that different processes * will construct cache images of the same size. * - * This is not a major issue as long as all processes include - * the same set of dirty entries in the cache -- as they - * currently do (note that this will change when we implement - * the ageout feature). Since only the process zero cache - * writes the cache image, all that is necessary is to - * broadcast the process zero cache size for use in the - * superblock extension messages and cache image block + * This is not a major issue as long as all processes include + * the same set of dirty entries in the cache -- as they + * currently do (note that this will change when we implement + * the ageout feature). Since only the process zero cache + * writes the cache image, all that is necessary is to + * broadcast the process zero cache size for use in the + * superblock extension messages and cache image block * allocations. * - * Note: At present, cache image is disabled in the - * parallel case as the new collective metadata write + * Note: At present, cache image is disabled in the + * parallel case as the new collective metadata write * code must be modified to support cache image. * * Return: Non-negative on success/Negative on failure @@ -1354,8 +1354,8 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); HDassert(image_generated); - /* If the file is opened and closed without any access to - * any group or data set, it is possible that the cache image (if + /* If the file is opened and closed without any access to + * any group or data set, it is possible that the cache image (if * it exists) has not been read yet. Do this now if required. */ if(cache_ptr->load_image) { @@ -1365,15 +1365,15 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) } /* end if */ /* Before we start to generate the cache image (if requested), verify - * that the superblock supports superblock extension messages, and + * that the superblock supports superblock extension messages, and * silently cancel any request for a cache image if it does not. * * Ideally, we would do this when the cache image is requested, - * but the necessary information is not necessary available at that + * but the necessary information is not necessary available at that * time -- hence this last minute check. * - * Note that under some error conditions, the superblock will be - * undefined in this case as well -- if so, assume that the + * Note that under some error conditions, the superblock will be + * undefined in this case as well -- if so, assume that the * superblock does not support superblock extension messages. * Also verify that the file's high_bound is at least release * 1.10.x, otherwise cancel the request for a cache image @@ -1390,17 +1390,17 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) /* Generate the cache image, if requested */ if(cache_ptr->image_ctl.generate_image) { /* Create the cache image super block extension message. - * + * * Note that the base address and length of the metadata cache * image are undefined at this point, and thus will have to be * updated later. * - * Create the super block extension message now so that space + * Create the super block extension message now so that space * is allocated for it (if necessary) before we allocate space * for the cache image block. * - * To simplify testing, do this only if the - * H5C_CI__GEN_MDCI_SBE_MESG bit is set in + * To simplify testing, do this only if the + * H5C_CI__GEN_MDCI_SBE_MESG bit is set in * cache_ptr->image_ctl.flags. */ if(cache_ptr->image_ctl.flags & H5C_CI__GEN_MDCI_SBE_MESG) @@ -1411,20 +1411,20 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) if(H5C__serialize_cache(f) < 0) HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "serialization of the cache failed") - /* Scan the cache and record data needed to construct the + /* Scan the cache and record data needed to construct the * cache image. In particular, for each entry we must record: * * 1) rank in LRU (if entry is in LRU) * - * 2) Whether the entry is dirty prior to flush of + * 2) Whether the entry is dirty prior to flush of * cache just prior to close. * * 3) Addresses of flush dependency parents (if any). * - * 4) Number of flush dependency children (if any). + * 4) Number of flush dependency children (if any). * - * In passing, also compute the size of the metadata cache - * image. With the recent modifications of the free space + * In passing, also compute the size of the metadata cache + * image. With the recent modifications of the free space * manager code, this size should be correct. */ if(H5C__prep_for_file_close__scan_entries(f, cache_ptr) < 0) @@ -1432,7 +1432,7 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) HDassert(HADDR_UNDEF == cache_ptr->image_addr); #ifdef H5_HAVE_PARALLEL - /* In the parallel case, overwrite the image_len with the + /* In the parallel case, overwrite the image_len with the * value computed by process 0. */ if(cache_ptr->aux_ptr) { /* we have multiple processes */ @@ -1453,17 +1453,17 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) else { if(MPI_SUCCESS != (mpi_result = MPI_Bcast(&p0_image_len, 1, MPI_UNSIGNED, 0, aux_ptr->mpi_comm))) HMPI_GOTO_ERROR(FAIL, "MPI_Bcast failed", mpi_result) - + aux_ptr->p0_image_len = p0_image_len; } /* end else */ - /* Allocate space for a cache image of size equal to that - * computed by the process 0. This may be different from + /* Allocate space for a cache image of size equal to that + * computed by the process 0. This may be different from * cache_ptr->image_data_len if mpi_rank != 0. However, since - * cache image write is suppressed on all processes other than + * cache image write is suppressed on all processes other than * process 0, this doesn't matter. * - * Note that we allocate the cache image directly from the file + * Note that we allocate the cache image directly from the file * driver so as to avoid unsettling the free space managers. */ if(HADDR_UNDEF == (cache_ptr->image_addr = H5FD_alloc(f->shared->lf, H5FD_MEM_SUPER, f, @@ -1472,8 +1472,8 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) } /* end if */ else #endif /* H5_HAVE_PARALLEL */ - /* Allocate the cache image block. Note that we allocate this - * this space directly from the file driver so as to avoid + /* Allocate the cache image block. Note that we allocate this + * this space directly from the file driver so as to avoid * unsettling the free space managers. */ if(HADDR_UNDEF == (cache_ptr->image_addr = H5FD_alloc(f->shared->lf, H5FD_MEM_SUPER, f, @@ -1497,25 +1497,25 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) * image block to the next alignment boundary, and then setting * the image_data_len to the actual size of the cache_image. * - * On the off chance that there is some other way to get a + * On the off chance that there is some other way to get a * a fragment on a cache image allocation, leave the following * assertion in the code so we will find out. */ HDassert((eoa_frag_size == 0) || (f->shared->alignment != 1)); /* Eventually it will be possible for the length of the cache image - * block on file to be greater than the size of the data it - * contains. However, for now they must be the same. Set + * block on file to be greater than the size of the data it + * contains. However, for now they must be the same. Set * cache_ptr->image_len accordingly. */ cache_ptr->image_len = cache_ptr->image_data_len; - /* update the metadata cache image superblock extension - * message with the new cache image block base address and + /* update the metadata cache image superblock extension + * message with the new cache image block base address and * length. * - * to simplify testing, do this only if the - * H5C_CI__GEN_MDC_IMAGE_BLK bit is set in + * to simplify testing, do this only if the + * H5C_CI__GEN_MDC_IMAGE_BLK bit is set in * cache_ptr->image_ctl.flags. */ if(cache_ptr->image_ctl.flags & H5C_CI__GEN_MDC_IMAGE_BLK) @@ -1525,18 +1525,18 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) /* At this point: * * 1) space in the file for the metadata cache image - * is allocated, + * is allocated, * - * 2) the metadata cache image superblock extension - * message exists and (if so configured) contains + * 2) the metadata cache image superblock extension + * message exists and (if so configured) contains * the correct data, * - * 3) All entries in the cache that will appear in the + * 3) All entries in the cache that will appear in the * cache image are serialized with up to date images. * * Since we just updated the cache image message, * the super block extension message is dirty. However, - * since the superblock and the superblock extension + * since the superblock and the superblock extension * can't be included in the cache image, this is a non- * issue. * @@ -1544,16 +1544,16 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) * the cache are marked as such, and we have a count * of same. * - * 5) Flush dependency heights are calculated for all + * 5) Flush dependency heights are calculated for all * entries that will be included in the cache image. * * If there are any entries to be included in the metadata cache - * image, allocate, populate, and sort the image_entries array. + * image, allocate, populate, and sort the image_entries array. * - * If the metadata cache image will be empty, delete the - * metadata cache image superblock extension message, set + * If the metadata cache image will be empty, delete the + * metadata cache image superblock extension message, set * cache_ptr->image_ctl.generate_image to FALSE. This will - * allow the file close to continue normally without the + * allow the file close to continue normally without the * unnecessary generation of the metadata cache image. */ if(cache_ptr->num_entries_in_image > 0) { @@ -1567,9 +1567,9 @@ H5C__prep_image_for_file_close(H5F_t *f, hbool_t *image_generated) else { /* cancel creation of metadata cache image */ HDassert(cache_ptr->image_entries == NULL); - /* To avoid breaking the control flow tests, only delete - * the mdci superblock extension message if the - * H5C_CI__GEN_MDC_IMAGE_BLK flag is set in + /* To avoid breaking the control flow tests, only delete + * the mdci superblock extension message if the + * H5C_CI__GEN_MDC_IMAGE_BLK flag is set in * cache_ptr->image_ctl.flags. */ if(cache_ptr->image_ctl.flags & H5C_CI__GEN_MDC_IMAGE_BLK) @@ -1591,8 +1591,8 @@ done: /*------------------------------------------------------------------------- * Function: H5C_set_cache_image_config * - * Purpose: If *config_ptr contains valid data, copy it into the - * image_ctl field of *cache_ptr. Make adjustments for + * Purpose: If *config_ptr contains valid data, copy it into the + * image_ctl field of *cache_ptr. Make adjustments for * changes in configuration as required. * * If the file is open read only, silently @@ -1601,9 +1601,9 @@ done: * * Note that in addition to being inapplicable in the * read only case, cache image is also inapplicable if - * the superblock does not support superblock extension - * messages. Unfortunately, this information need not - * be available at this point. Thus we check for this + * the superblock does not support superblock extension + * messages. Unfortunately, this information need not + * be available at this point. Thus we check for this * later, in H5C_prep_for_file_close() and cancel the * cache image request if appropriate. * @@ -1638,7 +1638,7 @@ H5C_set_cache_image_config(const H5F_t *f, H5C_t *cache_ptr, HGOTO_ERROR(H5E_ARGS, H5E_BADRANGE, FAIL, "invalid cache image configuration") #ifdef H5_HAVE_PARALLEL - /* The collective metadata write code is not currently compatible + /* The collective metadata write code is not currently compatible * with cache image. Until this is fixed, suppress cache image silently * if there is more than one process. * JRM -- 11/8/16 @@ -1652,15 +1652,15 @@ H5C_set_cache_image_config(const H5F_t *f, H5C_t *cache_ptr, else { #endif /* H5_HAVE_PARALLEL */ /* A cache image can only be generated if the file is opened read / write - * and the superblock supports superblock extension messages. + * and the superblock supports superblock extension messages. * - * However, the superblock version is not available at this point -- + * However, the superblock version is not available at this point -- * hence we can only check the former requirement now. Do the latter - * check just before we construct the image.. + * check just before we construct the image.. * * If the file is opened read / write, apply the supplied configuration. * - * If it is not, set the image configuration to the default, which has + * If it is not, set the image configuration to the default, which has * the effect of silently disabling the cache image if it was requested. */ if(H5F_INTENT(f) & H5F_ACC_RDWR) @@ -1683,7 +1683,7 @@ done: /*------------------------------------------------------------------------- * Function: H5C_validate_cache_image_config() * - * Purpose: Run a sanity check on the provided instance of struct + * Purpose: Run a sanity check on the provided instance of struct * H5AC_cache_image_config_t. * * Do nothing and return SUCCEED if no errors are detected, @@ -1715,8 +1715,8 @@ H5C_validate_cache_image_config(H5C_cache_image_ctl_t * ctl_ptr) if(ctl_ptr->save_resize_status != FALSE) HGOTO_ERROR(H5E_CACHE, H5E_BADVALUE, FAIL, "unexpected value in save_resize_status field") - /* At present, we do not support prefetched entry ageouts. Thus - * the entry_ageout field must be set to + /* At present, we do not support prefetched entry ageouts. Thus + * the entry_ageout field must be set to * H5AC__CACHE_IMAGE__ENTRY_AGEOUT__NONE. */ if(ctl_ptr->entry_ageout != H5AC__CACHE_IMAGE__ENTRY_AGEOUT__NONE) @@ -1740,7 +1740,7 @@ done: * Purpose: Compute the size of the header of the metadata cache * image block, and return the value. * - * Return: Size of the header section of the metadata cache image + * Return: Size of the header section of the metadata cache image * block in bytes. * * Programmer: John Mainzer @@ -1777,7 +1777,7 @@ H5C__cache_image_block_entry_header_size(const H5F_t * f) * Purpose: Compute the size of the header of the metadata cache * image block, and return the value. * - * Return: Size of the header section of the metadata cache image + * Return: Size of the header section of the metadata cache image * block in bytes. * * Programmer: John Mainzer @@ -1806,9 +1806,9 @@ H5C__cache_image_block_header_size(const H5F_t * f) /*------------------------------------------------------------------------- * Function: H5C__decode_cache_image_header() * - * Purpose: Decode the metadata cache image buffer header from the + * Purpose: Decode the metadata cache image buffer header from the * supplied buffer and load the data into the supplied instance - * of H5C_t. Advances the buffer pointer to the first byte + * of H5C_t. Advances the buffer pointer to the first byte * after the header image, or unchanged on failure. * * Return: Non-negative on success/Negative on failure @@ -1853,7 +1853,7 @@ H5C__decode_cache_image_header(const H5F_t *f, H5C_t *cache_ptr, /* Decode flags */ flags = *p++; - if(flags & H5C__MDCI_HEADER_HAVE_RESIZE_STATUS) + if(flags & H5C__MDCI_HEADER_HAVE_RESIZE_STATUS) have_resize_status = TRUE; if(have_resize_status) HGOTO_ERROR(H5E_CACHE, H5E_BADVALUE, FAIL, "MDC resize status not yet supported") @@ -1867,7 +1867,7 @@ H5C__decode_cache_image_header(const H5F_t *f, H5C_t *cache_ptr, /* Read num entries */ UINT32DECODE(p, cache_ptr->num_entries_in_image); - if(cache_ptr->num_entries_in_image == 0) + if(cache_ptr->num_entries_in_image == 0) HGOTO_ERROR(H5E_CACHE, H5E_BADVALUE, FAIL, "Bad metadata cache entry count") /* Verify expected length of header */ @@ -1888,13 +1888,13 @@ done: /*------------------------------------------------------------------------- * Function: H5C__decode_cache_image_entry() * - * Purpose: Decode the metadata cache image entry from the supplied + * Purpose: Decode the metadata cache image entry from the supplied * buffer into the supplied instance of H5C_image_entry_t. * This includes allocating a buffer for the entry image, - * loading it, and seting ie_ptr->image_ptr to point to + * loading it, and seting ie_ptr->image_ptr to point to * the buffer. * - * Advances the buffer pointer to the first byte + * Advances the buffer pointer to the first byte * after the entry, or unchanged on failure. * * Return: Non-negative on success/Negative on failure @@ -1998,8 +1998,8 @@ H5C__decode_cache_image_entry(const H5F_t *f, const H5C_t *cache_ptr, /* Verify expected length of entry image */ if((size_t)(p - *buf) != H5C__cache_image_block_entry_header_size(f)) HGOTO_ERROR(H5E_CACHE, H5E_BADSIZE, FAIL, "Bad entry image len") - - /* If parent count greater than zero, allocate array for parent + + /* If parent count greater than zero, allocate array for parent * addresses, and decode addresses into the array. */ if(fd_parent_count > 0) { @@ -2053,9 +2053,9 @@ done: /*------------------------------------------------------------------------- * Function: H5C__destroy_pf_entry_child_flush_deps() * - * Purpose: Destroy all flush dependencies in this the supplied + * Purpose: Destroy all flush dependencies in this the supplied * prefetched entry is the parent. Note that the children - * in these flush dependencies must be prefetched entries as + * in these flush dependencies must be prefetched entries as * well. * * As this action is part of the process of transferring all @@ -2063,8 +2063,8 @@ done: * prefetched entry, ensure that the data necessary to complete * the transfer is retained. * - * Note: The current implementation of this function is - * quite inefficient -- mostly due to the current + * Note: The current implementation of this function is + * quite inefficient -- mostly due to the current * implementation of flush dependencies. This should * be fixed at some point. * @@ -2076,7 +2076,7 @@ done: *------------------------------------------------------------------------- */ static herr_t -H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, +H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, H5C_cache_entry_t *pf_entry_ptr, H5C_cache_entry_t **fd_children) { H5C_cache_entry_t * entry_ptr; @@ -2103,8 +2103,8 @@ H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, while(entry_ptr != NULL) { HDassert(entry_ptr->magic == H5C__H5C_CACHE_ENTRY_T_MAGIC); - /* Here we look at entry_ptr->flush_dep_nparents and not - * entry_ptr->fd_parent_count as it is possible that some + /* Here we look at entry_ptr->flush_dep_nparents and not + * entry_ptr->fd_parent_count as it is possible that some * or all of the prefetched flush dependency child relationships * have already been destroyed. */ @@ -2145,9 +2145,9 @@ H5C__destroy_pf_entry_child_flush_deps(H5C_t *cache_ptr, HGOTO_ERROR(H5E_CACHE, H5E_CANTUNDEPEND, FAIL, "can't destroy pf entry child flush dependency") #ifndef NDEBUG - /* Sanity check -- verify that the address of the parent + /* Sanity check -- verify that the address of the parent * appears in entry_ptr->fd_parent_addrs. Must do a search, - * as with flush dependency creates and destroys, + * as with flush dependency creates and destroys, * entry_ptr->fd_parent_addrs and entry_ptr->flush_dep_parent * can list parents in different order. */ @@ -2181,8 +2181,8 @@ done: /*------------------------------------------------------------------------- * Function: H5C__encode_cache_image_header() * - * Purpose: Encode the metadata cache image buffer header in the - * supplied buffer. Updates buffer pointer to the first byte + * Purpose: Encode the metadata cache image buffer header in the + * supplied buffer. Updates buffer pointer to the first byte * after the header image in the buffer, or unchanged on failure. * * Return: Non-negative on success/Negative on failure @@ -2259,8 +2259,8 @@ done: /*------------------------------------------------------------------------- * Function: H5C__encode_cache_image_entry() * - * Purpose: Encode the metadata cache image buffer header in the - * supplied buffer. Updates buffer pointer to the first byte + * Purpose: Encode the metadata cache image buffer header in the + * supplied buffer. Updates buffer pointer to the first byte * after the entry in the buffer, or unchanged on failure. * * Return: Non-negative on success/Negative on failure @@ -2271,7 +2271,7 @@ done: *------------------------------------------------------------------------- */ static herr_t -H5C__encode_cache_image_entry(H5F_t *f, H5C_t *cache_ptr, uint8_t **buf, +H5C__encode_cache_image_entry(H5F_t *f, H5C_t *cache_ptr, uint8_t **buf, unsigned entry_num) { H5C_image_entry_t * ie_ptr; /* Pointer to entry to encode */ @@ -2306,13 +2306,13 @@ H5C__encode_cache_image_entry(H5F_t *f, H5C_t *cache_ptr, uint8_t **buf, *p++ = (uint8_t)(ie_ptr->type_id); /* Compose and encode flags */ - if(ie_ptr->is_dirty) + if(ie_ptr->is_dirty) flags |= H5C__MDCI_ENTRY_DIRTY_FLAG; - if(ie_ptr->lru_rank > 0) + if(ie_ptr->lru_rank > 0) flags |= H5C__MDCI_ENTRY_IN_LRU_FLAG; if(ie_ptr->fd_child_count > 0) flags |= H5C__MDCI_ENTRY_IS_FD_PARENT_FLAG; - if(ie_ptr->fd_parent_count > 0) + if(ie_ptr->fd_parent_count > 0) flags |= H5C__MDCI_ENTRY_IS_FD_CHILD_FLAG; *p++ = flags; @@ -2377,51 +2377,51 @@ done: * flush dependency children. (Recall that the flush dependency * height of an entry in a flush dependency relationship is the * length of the longest path from the entry to a leaf entry -- - * that is an entry with flush dependency parents, but no - * flush dependency children. With the introduction of the + * that is an entry with flush dependency parents, but no + * flush dependency children. With the introduction of the * possibility of multiple flush dependency parents, we have - * a flush partial dependency latice, not a flush dependency - * tree. But since the partial latice is acyclic, the concept + * a flush partial dependency latice, not a flush dependency + * tree. But since the partial latice is acyclic, the concept * of flush dependency height still makes sense. * - * The purpose of this function is to compute the flush + * The purpose of this function is to compute the flush * dependency height of all entries that appear in the cache - * image. + * image. * - * At present, entries are included or excluded from the + * At present, entries are included or excluded from the * cache image depending upon the ring in which they reside. * Thus there is no chance that one side of a flush dependency * will be in the cache image, and the other side not. * * However, once we start placing a limit on the size of the * cache image, or start excluding prefetched entries from - * the cache image if they haven't been accessed in some - * number of file close / open cycles, this will no longer - * be the case. + * the cache image if they haven't been accessed in some + * number of file close / open cycles, this will no longer + * be the case. * * In particular, if a flush dependency child is dirty, and * one of its flush dependency parents is dirty and not in * the cache image, then the flush dependency child cannot * be in the cache image without violating flush ordering. * - * Observe that a clean flush dependency child can be either - * in or out of the cache image without effect on flush + * Observe that a clean flush dependency child can be either + * in or out of the cache image without effect on flush * dependencies. * - * Similarly, a flush dependency parent can always be part - * of a cache image, regardless of whether it is clean or + * Similarly, a flush dependency parent can always be part + * of a cache image, regardless of whether it is clean or * dirty -- but remember that a flush dependency parent can * also be a flush dependency child. - * + * * Finally, note that for purposes of the cache image, flush - * dependency height ends when a flush dependecy relation + * dependency height ends when a flush dependecy relation * passes off the cache image. * - * On exit, the flush dependency height of each entry in the + * On exit, the flush dependency height of each entry in the * cache image should be calculated and stored in the cache * entry. Entries will be removed from the cache image if * necessary to maintain flush ordering. - * + * * Return: Non-negative on success/Negative on failure * * Programmer: John Mainzer @@ -2447,10 +2447,10 @@ H5C__prep_for_file_close__compute_fd_heights(const H5C_t *cache_ptr) HDassert(cache_ptr); HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); - /* Remove from the cache image all dirty entries that are + /* Remove from the cache image all dirty entries that are * flush dependency children of dirty entries that are not in the - * cache image. Must do this, as if we fail to do so, the parent - * will be written to file before the child. Since it is possible + * cache image. Must do this, as if we fail to do so, the parent + * will be written to file before the child. Since it is possible * that the child will have dirty children of its own, this may take * multiple passes through the index list. */ @@ -2484,17 +2484,17 @@ H5C__prep_for_file_close__compute_fd_heights(const H5C_t *cache_ptr) entry_ptr = entry_ptr->il_next; } /* while ( entry_ptr != NULL ) */ - } /* while ( ! done ) */ + } /* while ( ! done ) */ /* at present, entries are included in the cache image if they reside - * in a specified set of rings. Thus it should be impossible for - * entries_removed_from_image to be positive. Assert that this is - * so. Note that this will change when we start aging entries out + * in a specified set of rings. Thus it should be impossible for + * entries_removed_from_image to be positive. Assert that this is + * so. Note that this will change when we start aging entries out * of the cache image. */ HDassert(entries_removed_from_image == 0); - /* Next, remove from entries in the cache image, references to + /* Next, remove from entries in the cache image, references to * flush dependency parents or children that are not in the cache image. */ entry_ptr = cache_ptr->il_head; @@ -2575,14 +2575,14 @@ H5C__prep_for_file_close__compute_fd_heights(const H5C_t *cache_ptr) entry_ptr = entry_ptr->il_next; } /* while (entry_ptr != NULL) */ - /* At present, no extenal parent or child flush dependency links + /* At present, no extenal parent or child flush dependency links * should exist -- hence the following assertions. This will change * if we support ageout of entries in the cache image. */ HDassert(external_child_fd_refs_removed == 0); HDassert(external_parent_fd_refs_removed == 0); - /* At this point we should have removed all flush dependencies that + /* At this point we should have removed all flush dependencies that * cross cache image boundaries. Now compute the flush dependency * heights for all entries in the image. * @@ -2600,7 +2600,7 @@ H5C__prep_for_file_close__compute_fd_heights(const H5C_t *cache_ptr) parent_ptr = entry_ptr->flush_dep_parent[u]; HDassert(parent_ptr->magic == H5C__H5C_CACHE_ENTRY_T_MAGIC); - if(parent_ptr->include_in_image && parent_ptr->image_fd_height <= 0) + if(parent_ptr->include_in_image && parent_ptr->image_fd_height <= 0) H5C__prep_for_file_close__compute_fd_heights_real(parent_ptr, 1); } /* end for */ } /* end if */ @@ -2623,37 +2623,37 @@ done: * * The basic observation behind this function is as follows: * - * Suppose you have an entry E with a flush dependency - * height of X. Then the parents of E must all have + * Suppose you have an entry E with a flush dependency + * height of X. Then the parents of E must all have * flush dependency X + 1 or greater. * * Use this observation to compute flush dependency height * of all entries in the cache image via the following * recursive algorithm: * - * 1) On entry, set the flush dependency height of the + * 1) On entry, set the flush dependency height of the * supplied cache entry to the supplied value. * - * 2) Examine all the flush dependency parents of the - * supplied entry. + * 2) Examine all the flush dependency parents of the + * supplied entry. * - * If the parent is in the cache image, and has flush + * If the parent is in the cache image, and has flush * dependency height less than or equal to the flush - * dependency height of the current entry, call the + * dependency height of the current entry, call the * recursive routine on the parent with flush dependency - * height equal to the flush dependency height of the + * height equal to the flush dependency height of the * child plus 1. * * Otherwise do nothing. * * Observe that if the flush dependency height of all entries - * in the image is initialized to zero, and if this recursive - * function is called with flush dependency height 0 on all - * entries in the cache image with FD parents in the image, - * but without FD children in the image, the correct flush - * dependency height should be set for all entries in the + * in the image is initialized to zero, and if this recursive + * function is called with flush dependency height 0 on all + * entries in the cache image with FD parents in the image, + * but without FD children in the image, the correct flush + * dependency height should be set for all entries in the * cache image. - * + * * Return: void * * Programmer: John Mainzer @@ -2698,9 +2698,9 @@ H5C__prep_for_file_close__compute_fd_heights_real(H5C_cache_entry_t *entry_ptr, * Function: H5C__prep_for_file_close__setup_image_entries_array * * Purpose: Allocate space for the image_entries array, and load - * each instance of H5C_image_entry_t in the array with + * each instance of H5C_image_entry_t in the array with * the data necessary to construct the metadata cache image. - * + * * Return: Non-negative on success/Negative on failure * * Programmer: John Mainzer @@ -2758,14 +2758,14 @@ H5C__prep_for_file_close__setup_image_entries_array(H5C_t *cache_ptr) image_entries[u].ring = entry_ptr->ring; /* When a prefetched entry is included in the image, store - * its underlying type id in the image entry, not + * its underlying type id in the image entry, not * H5AC_PREFETCHED_ENTRY_ID. In passing, also increment * the age (up to H5AC__CACHE_IMAGE__ENTRY_AGEOUT__MAX). */ if(entry_ptr->type->id == H5AC_PREFETCHED_ENTRY_ID) { image_entries[u].type_id = entry_ptr->prefetch_type_id; image_entries[u].age = entry_ptr->age + 1; - + if(image_entries[u].age > H5AC__CACHE_IMAGE__ENTRY_AGEOUT__MAX) image_entries[u].age = H5AC__CACHE_IMAGE__ENTRY_AGEOUT__MAX; } /* end if */ @@ -2780,13 +2780,13 @@ H5C__prep_for_file_close__setup_image_entries_array(H5C_t *cache_ptr) image_entries[u].fd_parent_count = entry_ptr->fd_parent_count; image_entries[u].fd_parent_addrs = entry_ptr->fd_parent_addrs; image_entries[u].fd_child_count = entry_ptr->fd_child_count; - image_entries[u].fd_dirty_child_count = + image_entries[u].fd_dirty_child_count = entry_ptr->fd_dirty_child_count; image_entries[u].image_ptr = entry_ptr->image_ptr; - /* Null out entry_ptr->fd_parent_addrs and set + /* Null out entry_ptr->fd_parent_addrs and set * entry_ptr->fd_parent_count to zero so that ownership of the - * flush dependency parents address array is transferred to the + * flush dependency parents address array is transferred to the * image entry. */ entry_ptr->fd_parent_count = 0; @@ -2819,8 +2819,8 @@ done: /*------------------------------------------------------------------------- * Function: H5C__prep_for_file_close__scan_entries * - * Purpose: Scan all entries in the metadata cache, and store all - * entry specific data required for construction of the + * Purpose: Scan all entries in the metadata cache, and store all + * entry specific data required for construction of the * metadata cache image block and likely to be discarded * or modified during the cache flush on file close. * @@ -2838,7 +2838,7 @@ done: * * Finally, compute the size of the metadata cache image * block. - * + * * Return: Non-negative on success/Negative on failure * * Programmer: John Mainzer @@ -2891,7 +2891,7 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) HDassert(entry_ptr->image_ptr); /* Initially, we mark all entries in the rings included - * in the cache image as being included in the in the + * in the cache image as being included in the in the * image. Depending on circumstances, we may exclude some * of these entries later. */ @@ -2907,7 +2907,7 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) entry_ptr->image_fd_height = 0; /* will compute this later */ /* Initially, include all flush dependency parents in the - * the list of flush dependencies to be stored in the + * the list of flush dependencies to be stored in the * image. We may remove some or all of these later. */ if(entry_ptr->flush_dep_nparents > 0) { @@ -2915,7 +2915,7 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) * as needed. */ if(entry_ptr->flush_dep_nparents == entry_ptr->fd_parent_count ) { - /* parent addresses array should already be allocated + /* parent addresses array should already be allocated * and of the correct size. */ HDassert(entry_ptr->fd_parent_addrs); @@ -2947,8 +2947,8 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) HDassert(entry_ptr->fd_parent_addrs == NULL); /* Initially, all flush dependency children are included int - * the count of flush dependency child relationships to be - * represented in the cache image. Some or all of these + * the count of flush dependency child relationships to be + * represented in the cache image. Some or all of these * may be dropped from the image later. */ if(entry_ptr->flush_dep_nchildren > 0) { @@ -2970,23 +2970,23 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) /* Now compute the flush dependency heights of all flush dependency * relationships to be represented in the image. * - * If all entries in the target rings are included in the - * image, the flush dependency heights are simply the heights + * If all entries in the target rings are included in the + * image, the flush dependency heights are simply the heights * of all flush dependencies in the target rings. * - * However, if we restrict appearance in the cache image either - * by number of entries in the image, restrictions on the number - * of times a prefetched entry can appear in an image, or image + * However, if we restrict appearance in the cache image either + * by number of entries in the image, restrictions on the number + * of times a prefetched entry can appear in an image, or image * size, it is possible that flush dependency parents or children * of entries that are in the image may not be included in the - * the image. In this case, we must prune all flush dependency - * relationships that cross the image boundary, and all exclude - * from the image all dirty flush dependency children that have - * a dirty flush dependency parent that is not in the image. + * the image. In this case, we must prune all flush dependency + * relationships that cross the image boundary, and all exclude + * from the image all dirty flush dependency children that have + * a dirty flush dependency parent that is not in the image. * This is necessary to preserve the required flush ordering. - * - * These details are tended to by the following call to - * H5C__prep_for_file_close__compute_fd_heights(). Because the + * + * These details are tended to by the following call to + * H5C__prep_for_file_close__compute_fd_heights(). Because the * exact contents of the image cannot be known until after this * call, computation of the image size is delayed. */ @@ -2994,7 +2994,7 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "computation of flush dependency heights failed?!?") /* At this point, all entries that will appear in the cache - * image should be marked correctly. Compute the size of the + * image should be marked correctly. Compute the size of the * cache image. */ entries_visited = 0; @@ -3031,11 +3031,11 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) /* Now scan the LRU list to set the lru_rank fields of all entries * on the LRU. * - * Note that we start with rank 1, and increment by 1 with each - * entry on the LRU. + * Note that we start with rank 1, and increment by 1 with each + * entry on the LRU. * * Note that manually pinned entryies will have lru_rank -1, - * and no flush dependency. Putting these entries at the head of + * and no flush dependency. Putting these entries at the head of * the reconstructed LRU should be appropriate. */ entry_ptr = cache_ptr->LRU_head_ptr; @@ -3045,11 +3045,11 @@ H5C__prep_for_file_close__scan_entries(const H5F_t *f, H5C_t *cache_ptr) /* to avoid confusion, don't set lru_rank on epoch markers. * Note that we still increment the lru_rank, so that the holes - * in the sequence of entries on the LRU will indicate the - * locations of epoch markers (if any) when we reconstruct + * in the sequence of entries on the LRU will indicate the + * locations of epoch markers (if any) when we reconstruct * the LRU. * - * Do not set lru_rank or increment lru_rank for entries + * Do not set lru_rank or increment lru_rank for entries * that will not be included in the cache image. */ if(entry_ptr->type->id == H5AC_EPOCH_MARKER_ID) @@ -3076,9 +3076,9 @@ done: * Function: H5C__reconstruct_cache_contents() * * Purpose: Scan the image buffer, and create a prefetched - * cache entry for every entry in the buffer. Insert the - * prefetched entries in the index and the LRU, and - * reconstruct any flush dependencies. Order the entries + * cache entry for every entry in the buffer. Insert the + * prefetched entries in the index and the LRU, and + * reconstruct any flush dependencies. Order the entries * in the LRU as indicated by the stored lru_ranks. * * Return: SUCCEED on success, and FAIL on failure. @@ -3127,11 +3127,11 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) if(NULL == (pf_entry_ptr = H5C__reconstruct_cache_entry(f, cache_ptr, &p))) HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "reconstruction of cache entry failed") - /* Note that we make no checks on available cache space before + /* Note that we make no checks on available cache space before * inserting the reconstructed entry into the metadata cache. * * This is OK since the cache must be almost empty at the beginning - * of the process, and since we check cache size at the end of the + * of the process, and since we check cache size at the end of the * reconstruction process. */ @@ -3147,7 +3147,7 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) H5C__UPDATE_STATS_FOR_PREFETCH(cache_ptr, pf_entry_ptr->is_dirty) - /* If the prefetched entry is the child in one or more flush + /* If the prefetched entry is the child in one or more flush * dependency relationships, recreate those flush dependencies. */ for(v = 0; v < pf_entry_ptr->fd_parent_count; v++) { @@ -3171,7 +3171,7 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) */ H5C__UPDATE_RP_FOR_PROTECT(cache_ptr, parent_ptr, FAIL) parent_ptr->is_protected = TRUE; - + /* Setup the flush dependency */ if(H5C_create_flush_dependency(parent_ptr, pf_entry_ptr) < 0) HGOTO_ERROR(H5E_CACHE, H5E_CANTDEPEND, FAIL, "Can't restore flush dependency") @@ -3211,7 +3211,7 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) pf_entry_ptr = pf_entry_ptr->il_next; } /* end while */ - /* Scan the LRU, and verify the expected ordering of the + /* Scan the LRU, and verify the expected ordering of the * prefetched entries. */ { @@ -3233,7 +3233,7 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) HDassert((entry_ptr->lru_rank == -1) || (entry_ptr->lru_rank > i)); - if ( ( entry_ptr->lru_rank > 1 ) && + if ( ( entry_ptr->lru_rank > 1 ) && ( entry_ptr->lru_rank > i + 1 ) ) lru_rank_holes += entry_ptr->lru_rank - (i + 1); @@ -3245,9 +3245,9 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) entry_ptr = entry_ptr->next; } /* end while */ - /* Holes in the sequences of LRU ranks can appear due to epoch + /* Holes in the sequences of LRU ranks can appear due to epoch * markers. They are left in to allow re-insertion of the - * epoch markers on reconstruction of the cache -- thus + * epoch markers on reconstruction of the cache -- thus * the following sanity check will have to be revised when * we add code to store and restore adaptive resize status. */ @@ -3255,7 +3255,7 @@ H5C__reconstruct_cache_contents(H5F_t *f, H5C_t *cache_ptr) } /* end block */ #endif /* NDEBUG */ - /* Check to see if the cache is oversize, and evict entries as + /* Check to see if the cache is oversize, and evict entries as * necessary to remain within limits. */ if(cache_ptr->index_size >= cache_ptr->max_cache_size) { @@ -3289,7 +3289,7 @@ done: * Return a pointer to the newly allocated cache entry, * or NULL on failure. * - * Return: Pointer to the new instance of H5C_cache_entry on success, + * Return: Pointer to the new instance of H5C_cache_entry on success, * or NULL on failure. * * Programmer: John Mainzer @@ -3347,7 +3347,7 @@ H5C__reconstruct_cache_entry(const H5F_t *f, H5C_t *cache_ptr, is_fd_child = TRUE; #endif /* NDEBUG */ /* only used in assertions */ - /* Force dirty entries to clean if the file read only -- must do + /* Force dirty entries to clean if the file read only -- must do * this as otherwise the cache will attempt to write them on file * close. Since the file is R/O, the metadata cache image superblock * extension message and the cache image block will not be removed. @@ -3355,7 +3355,7 @@ H5C__reconstruct_cache_entry(const H5F_t *f, H5C_t *cache_ptr, * * However, if the dirty entry (marked clean for purposes of the R/O * file open) is evicted and then referred to, the cache will read - * either invalid or obsolete data from the file. Handle this by + * either invalid or obsolete data from the file. Handle this by * setting the prefetched_dirty field, and hiding such entries from * the eviction candidate selection algorithm. */ @@ -3375,7 +3375,7 @@ H5C__reconstruct_cache_entry(const H5F_t *f, H5C_t *cache_ptr, /* Decode dirty dependency child count */ UINT16DECODE(p, pf_entry_ptr->fd_dirty_child_count); - if(!file_is_rw) + if(!file_is_rw) pf_entry_ptr->fd_dirty_child_count = 0; if(pf_entry_ptr->fd_dirty_child_count > pf_entry_ptr->fd_child_count) HGOTO_ERROR(H5E_CACHE, H5E_BADVALUE, NULL, "invalid dirty flush dependency child count") @@ -3401,8 +3401,8 @@ H5C__reconstruct_cache_entry(const H5F_t *f, H5C_t *cache_ptr, /* Verify expected length of entry image */ if((size_t)(p - *buf) != H5C__cache_image_block_entry_header_size(f)) HGOTO_ERROR(H5E_CACHE, H5E_BADSIZE, NULL, "Bad entry image len") - - /* If parent count greater than zero, allocate array for parent + + /* If parent count greater than zero, allocate array for parent * addresses, and decode addresses into the array. */ if(pf_entry_ptr->fd_parent_count > 0) { @@ -3457,7 +3457,7 @@ done: /*------------------------------------------------------------------------- * Function: H5C__write_cache_image_superblock_msg * - * Purpose: Write the cache image superblock extension message, + * Purpose: Write the cache image superblock extension message, * creating if specified. * * In general, the size and location of the cache image block @@ -3560,7 +3560,7 @@ H5C__write_cache_image(H5F_t *f, const H5C_t *cache_ptr) } /* end if */ } /* end block */ #endif /* H5_HAVE_PARALLEL */ - + done: FUNC_LEAVE_NOAPI(ret_value) } /* H5C__write_cache_image() */ |