diff options
author | Quincey Koziol <koziol@lbl.gov> | 2017-03-13 19:04:33 (GMT) |
---|---|---|
committer | Quincey Koziol <koziol@lbl.gov> | 2017-03-13 19:04:33 (GMT) |
commit | 847d675f2743ec420ef6c9efdd6e52ae93b4fe44 (patch) | |
tree | 3e74669322711c4ea14f8fd80ecf14da984c6212 /src/H5Cmpio.c | |
parent | 56e5b4ed05fb3bad51c226ec5b567871a45e0bce (diff) | |
parent | 4fad103187db9095afc8eab90b5f5544feb1d19b (diff) | |
download | hdf5-847d675f2743ec420ef6c9efdd6e52ae93b4fe44.zip hdf5-847d675f2743ec420ef6c9efdd6e52ae93b4fe44.tar.gz hdf5-847d675f2743ec420ef6c9efdd6e52ae93b4fe44.tar.bz2 |
Merge pull request #333 in HDFFV/hdf5 from merge_page_buffering_07 to develop
* commit '4fad103187db9095afc8eab90b5f5544feb1d19b':
Bring changes to I/O parameters from page_buffering branch.
Merge in reentrency changes to "make space in cache" from page_buffering branch.
Minor cleanups and bring over "prefetched dirty" fixes for entries loaded from a cache image.
Remove some usage of "prefetched_dirty" flag (which hasn't been merged from the page_buffering branch yet. Also, bring over improvements to flush candidate entries for parallel code.
Align with incoming page buffering changes: minor cleanups, centralize removing entries from collective metadata read list
Diffstat (limited to 'src/H5Cmpio.c')
-rw-r--r-- | src/H5Cmpio.c | 990 |
1 files changed, 577 insertions, 413 deletions
diff --git a/src/H5Cmpio.c b/src/H5Cmpio.c index ebb98b3..1f43866 100644 --- a/src/H5Cmpio.c +++ b/src/H5Cmpio.c @@ -65,6 +65,11 @@ /* Local Prototypes */ /********************/ static herr_t H5C__collective_write(H5F_t *f, hid_t dxpl_id); +static herr_t H5C__flush_candidate_entries(H5F_t *f, hid_t dxpl_id, + unsigned entries_to_flush[H5C_RING_NTYPES], + unsigned entries_to_clear[H5C_RING_NTYPES]); +static herr_t H5C__flush_candidates_in_ring(H5F_t *f, hid_t dxpl_id, + H5C_ring_t ring, unsigned entries_to_flush, unsigned entries_to_clear); /*********************/ @@ -175,31 +180,18 @@ H5C_apply_candidate_list(H5F_t * f, int mpi_rank, int mpi_size) { - hbool_t restart_scan; - hbool_t prev_is_dirty; int i; int m; int n; unsigned first_entry_to_flush; unsigned last_entry_to_flush; - unsigned entries_to_clear = 0; - unsigned entries_to_flush = 0; - unsigned entries_to_flush_or_clear_last = 0; - unsigned entries_to_flush_collectively = 0; - unsigned entries_cleared = 0; - unsigned entries_flushed = 0; - unsigned entries_delayed = 0; - unsigned entries_flushed_or_cleared_last = 0; - unsigned entries_flushed_collectively = 0; - unsigned entries_examined = 0; - unsigned initial_list_len; + unsigned total_entries_to_clear = 0; + unsigned total_entries_to_flush = 0; int * candidate_assignment_table = NULL; + unsigned entries_to_flush[H5C_RING_NTYPES]; + unsigned entries_to_clear[H5C_RING_NTYPES]; haddr_t addr; - H5C_cache_entry_t * clear_ptr = NULL; - H5C_cache_entry_t * next_ptr = NULL; H5C_cache_entry_t * entry_ptr = NULL; - H5C_cache_entry_t * flush_ptr = NULL; - H5C_cache_entry_t * delayed_ptr = NULL; #if H5C_DO_SANITY_CHECKS haddr_t last_addr; #endif /* H5C_DO_SANITY_CHECKS */ @@ -220,6 +212,10 @@ H5C_apply_candidate_list(H5F_t * f, HDassert(0 <= mpi_rank); HDassert(mpi_rank < mpi_size); + /* Initialize the entries_to_flush and entries_to_clear arrays */ + HDmemset(entries_to_flush, 0, sizeof(entries_to_flush)); + HDmemset(entries_to_clear, 0, sizeof(entries_to_clear)); + #if H5C_APPLY_CANDIDATE_LIST__DEBUG HDfprintf(stdout, "%s:%d: setting up candidate assignment table.\n", FUNC, mpi_rank); @@ -330,17 +326,26 @@ H5C_apply_candidate_list(H5F_t * f, */ HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "Listed entry is protected?!?!?") + /* Sanity checks */ + HDassert(entry_ptr->magic == H5C__H5C_CACHE_ENTRY_T_MAGIC); + HDassert(entry_ptr->ring >= H5C_RING_USER); + HDassert(entry_ptr->ring <= H5C_RING_SB); + HDassert(!entry_ptr->flush_immediately); + HDassert(!entry_ptr->clear_on_unprotect); + /* Determine whether the entry is to be cleared or flushed, * and mark it accordingly. We will scan the protected and * pinned list shortly, and clear or flush according to these * markings. */ if(u >= first_entry_to_flush && u <= last_entry_to_flush) { - entries_to_flush++; + total_entries_to_flush++; + entries_to_flush[entry_ptr->ring]++; entry_ptr->flush_immediately = TRUE; } /* end if */ else { - entries_to_clear++; + total_entries_to_clear++; + entries_to_clear[entry_ptr->ring]++; entry_ptr->clear_on_unprotect = TRUE; } /* end else */ @@ -356,372 +361,36 @@ H5C_apply_candidate_list(H5F_t * f, } /* end if */ } /* end for */ +#if H5C_DO_SANITY_CHECKS + m = 0; + n = 0; + for(i = 0; i < H5C_RING_NTYPES; i++) { + m += (int)entries_to_flush[i]; + n += (int)entries_to_clear[i]; + } /* end if */ + + HDassert((unsigned)m == total_entries_to_flush); + HDassert((unsigned)n == total_entries_to_clear); +#endif /* H5C_DO_SANITY_CHECKS */ + #if H5C_APPLY_CANDIDATE_LIST__DEBUG HDfprintf(stdout, "%s:%d: num candidates/to clear/to flush = %u/%u/%u.\n", - FUNC, mpi_rank, num_candidates, entries_to_clear, - entries_to_flush); + FUNC, mpi_rank, num_candidates, total_entries_to_clear, + total_entries_to_flush); #endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ /* We have now marked all the entries on the candidate list for * either flush or clear -- now scan the LRU and the pinned list - * for these entries and do the deed. + * for these entries and do the deed. Do this via a call to + * H5C__flush_candidate_entries(). * * Note that we are doing things in this round about manner so as * to preserve the order of the LRU list to the best of our ability. * If we don't do this, my experiments indicate that we will have a * noticably poorer hit ratio as a result. */ - -#if H5C_APPLY_CANDIDATE_LIST__DEBUG - HDfprintf(stdout, "%s:%d: scanning LRU list. len = %d.\n", FUNC, mpi_rank, - (int)(cache_ptr->LRU_list_len)); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* ===================================================================== * - * Now scan the LRU and PEL lists, flushing or clearing entries as - * needed. - * - * The flush_me_last flag may dictate how or - * when some entries can be flushed, and should be addressed here. - * However, in their initial implementation, these flags only apply to the - * superblock, so there's only a relatively small change to this function - * to account for this one case where they come into play. If these flags - * are ever expanded upon, this function and the following flushing steps - * should be reworked to account for additional cases. - * ===================================================================== */ - - HDassert(entries_to_flush >= 0); - - restart_scan = FALSE; - entries_examined = 0; - initial_list_len = cache_ptr->LRU_list_len; - entry_ptr = cache_ptr->LRU_tail_ptr; - - /* Examine each entry in the LRU list */ - while ( ( entry_ptr != NULL ) - && - ( entries_examined <= (entries_to_flush + 1) * initial_list_len ) - && - ( (entries_cleared + entries_flushed) < num_candidates ) ) { - - if ( entry_ptr->prev != NULL ) - prev_is_dirty = entry_ptr->prev->is_dirty; - - /* If this process needs to clear this entry. */ - if(entry_ptr->clear_on_unprotect) { - - HDassert(entry_ptr->is_dirty); - - next_ptr = entry_ptr->next; - entry_ptr->clear_on_unprotect = FALSE; - clear_ptr = entry_ptr; - entry_ptr = entry_ptr->prev; - entries_cleared++; - -#if ( H5C_APPLY_CANDIDATE_LIST__DEBUG > 1 ) - HDfprintf(stdout, "%s:%d: clearing 0x%llx.\n", FUNC, mpi_rank, - (long long)clear_ptr->addr); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* reset entries_removed_counter and - * last_entry_removed_ptr prior to the call to - * H5C__flush_single_entry() so that we can spot - * unexpected removals of entries from the cache, - * and set the restart_scan flag if proceeding - * would be likely to cause us to scan an entry - * that is no longer in the cache. - * - * Note that as of this writing (April 2015) this - * case cannot occur in the parallel case. However - * Quincey is making noises about changing this, hence - * the insertion of this test. - * - * Note also that there is no test code to verify - * that this code actually works (although similar code - * in the serial version exists and is tested). - * - * Implementing a test will likely require implementing - * flush op like facilities in the parallel tests. At - * a guess this will not be terribly painful, but it - * will take a bit of time. - */ - cache_ptr->entries_removed_counter = 0; - cache_ptr->last_entry_removed_ptr = NULL; - - if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__GENERATE_IMAGE_FLAG | H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't clear entry.") - - if((cache_ptr->entries_removed_counter > 1) || - (cache_ptr->last_entry_removed_ptr == entry_ptr)) - restart_scan = TRUE; - } /* end if */ - - /* Else, if this process needs to flush this entry. */ - else if (entry_ptr->flush_immediately) { - - HDassert(entry_ptr->is_dirty); - - next_ptr = entry_ptr->next; - entry_ptr->flush_immediately = FALSE; - flush_ptr = entry_ptr; - entry_ptr = entry_ptr->prev; - entries_flushed++; - -#if ( H5C_APPLY_CANDIDATE_LIST__DEBUG > 1 ) - HDfprintf(stdout, "%s:%d: flushing 0x%llx.\n", FUNC, mpi_rank, - (long long)flush_ptr->addr); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* reset entries_removed_counter and - * last_entry_removed_ptr prior to the call to - * H5C__flush_single_entry() so that we can spot - * unexpected removals of entries from the cache, - * and set the restart_scan flag if proceeding - * would be likely to cause us to scan an entry - * that is no longer in the cache. - * - * Note that as of this writing (April 2015) this - * case cannot occur in the parallel case. However - * Quincey is making noises about changing this, hence - * the insertion of this test. - * - * Note also that there is no test code to verify - * that this code actually works (although similar code - * in the serial version exists and is tested). - * - * Implementing a test will likely require implementing - * flush op like facilities in the parallel tests. At - * a guess this will not be terribly painful, but it - * will take a bit of time. - */ - cache_ptr->entries_removed_counter = 0; - cache_ptr->last_entry_removed_ptr = NULL; - - /* Add this entry to the list of entries to collectively write */ - if(H5C__flush_single_entry(f, dxpl_id, flush_ptr, H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't flush entry.") - - if((cache_ptr->entries_removed_counter > 1) || - (cache_ptr->last_entry_removed_ptr == entry_ptr)) - restart_scan = TRUE; - } /* end else-if */ - - /* Otherwise, no action to be taken on this entry. Grab the next. */ - else { - entry_ptr = entry_ptr->prev; - - if ( entry_ptr != NULL ) - next_ptr = entry_ptr->next; - - } /* end else */ - - if ( ( entry_ptr != NULL ) - && - ( ( restart_scan ) - || - ( entry_ptr->is_dirty != prev_is_dirty ) - || - ( entry_ptr->next != next_ptr ) - || - ( entry_ptr->is_protected ) - || - ( entry_ptr->is_pinned ) - ) - ) { - - /* something has happened to the LRU -- start over - * from the tail. - * - * Recall that this code should be un-reachable at present, - * as all the operations by entries on flush that could cause - * it to be reachable are disallowed in the parallel case at - * present. Hence the following assertion which should be - * removed if the above changes. - */ - - HDassert( ! restart_scan ); - HDassert( entry_ptr->is_dirty == prev_is_dirty ); - HDassert( entry_ptr->next == next_ptr ); - HDassert( ! entry_ptr->is_protected ); - HDassert( ! entry_ptr->is_pinned ); - - HDassert(FALSE); /* see comment above */ - - restart_scan = FALSE; - entry_ptr = cache_ptr->LRU_tail_ptr; -/* - H5C__UPDATE_STATS_FOR_LRU_SCAN_RESTART(cache_ptr) -*/ - } - - entries_examined++; - } /* end while */ - -#if H5C_APPLY_CANDIDATE_LIST__DEBUG - HDfprintf(stdout, "%s:%d: entries examined/cleared/flushed = %u/%u/%u.\n", - FUNC, mpi_rank, entries_examined, - entries_cleared, entries_flushed); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* It is also possible that some of the cleared entries are on the - * pinned list. Must scan that also. - * - * WARNING: - * - * As we now allow unpinning, and removal of other entries as a side - * effect of flushing an entry, it is possible that the next entry - * in a PEL scan could either be no longer pinned, or no longer in - * the cache by the time we get to it. - * - * At present, this is not possible in this case, as we disallow such - * operations in the parallel version of the library. However, Quincey - * has been making noises about relaxing this. If and when he does, - * we have a potential problem here. - * - * The same issue exists in the serial cache, and there are tests - * to detect this problem when it occurs, and adjust to it. As seen - * above in the LRU scan, I have ported such tests to the parallel - * code where a close cognate exists in the serial code. - * - * I haven't done so here, as there are no PEL scans where the problem - * can occur in the serial code. Needless to say, this will have to - * be repaired if the constraints on pre_serialize and serialize - * callbacks are relaxed in the parallel version of the metadata cache. - * - * JRM -- 4/1/15 - */ - -#if H5C_APPLY_CANDIDATE_LIST__DEBUG - HDfprintf(stdout, "%s:%d: scanning pinned entry list. len = %d\n", - FUNC, mpi_rank, (int)(cache_ptr->pel_len)); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - entry_ptr = cache_ptr->pel_head_ptr; - while((entry_ptr != NULL) && - ((entries_cleared + entries_flushed + entries_delayed) - < num_candidates)) { - - /* If entry is marked for flush or for clear */ - if((entry_ptr->clear_on_unprotect||entry_ptr->flush_immediately)) { - - /* If this entry needs to be flushed last */ - if (entry_ptr->flush_me_last) { - - /* At this time, only the superblock supports being - flushed last. Conveniently, it also happens to be the only - entry that supports being flushed collectively, as well. Also - conveniently, it's always pinned, so we only need to check - for it while scanning the PEL here. Finally, it's never - included in a candidate list that excludes other dirty - entries in a cache, so we can handle this relatively simple - case here. - - For now, this function asserts this and saves the entry - to flush it after scanning the rest of the PEL list. - - If there are ever more entries that either need to be - flushed last and/or flushed collectively, this whole routine - will need to be reworked to handle all additional cases. As - it is the simple case of a single pinned entry needing - flushed last and collectively is just a minor addition to - this routine, but signficantly buffing up the usage of - flush_me_last will require a more - intense rework of this function and potentially the function - of candidate lists as a whole. */ - - entries_to_flush_or_clear_last++; - entries_to_flush_collectively++; - HDassert(entries_to_flush_or_clear_last == 1); - HDassert(entries_to_flush_collectively == 1); - - /* Delay the entry. It will be flushed later. */ - delayed_ptr = entry_ptr; - entries_delayed++; - HDassert(entries_delayed == 1); - - } /* end if */ - - /* Else, this process needs to clear this entry. */ - else if (entry_ptr->clear_on_unprotect) { - HDassert(!entry_ptr->flush_immediately); - entry_ptr->clear_on_unprotect = FALSE; - clear_ptr = entry_ptr; - entry_ptr = entry_ptr->next; - entries_cleared++; - -#if ( H5C_APPLY_CANDIDATE_LIST__DEBUG > 1 ) - HDfprintf(stdout, "%s:%d: clearing 0x%llx.\n", FUNC, mpi_rank, - (long long)clear_ptr->addr); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__GENERATE_IMAGE_FLAG | H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't clear entry.") - } /* end else-if */ - - /* Else, if this process needs to independently flush this entry. */ - else if (entry_ptr->flush_immediately) { - entry_ptr->flush_immediately = FALSE; - flush_ptr = entry_ptr; - entry_ptr = entry_ptr->next; - entries_flushed++; - -#if ( H5C_APPLY_CANDIDATE_LIST__DEBUG > 1 ) - HDfprintf(stdout, "%s:%d: flushing 0x%llx.\n", FUNC, mpi_rank, - (long long)flush_ptr->addr); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* Add this entry to the list of entries to collectively write */ - if(H5C__flush_single_entry(f, dxpl_id, flush_ptr, H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't clear entry.") - } /* end else-if */ - } /* end if */ - - /* Otherwise, this entry is not marked for flush or clear. Grab the next. */ - else { - entry_ptr = entry_ptr->next; - } /* end else */ - - } /* end while */ - -#if H5C_APPLY_CANDIDATE_LIST__DEBUG - HDfprintf(stdout, - "%s:%d: pel entries examined/cleared/flushed = %u/%u/%u.\n", - FUNC, mpi_rank, entries_examined, - entries_cleared, entries_flushed); - HDfprintf(stdout, "%s:%d: done.\n", FUNC, mpi_rank); - - HDfsync(stdout); -#endif /* H5C_APPLY_CANDIDATE_LIST__DEBUG */ - - /* ====================================================================== * - * Now, handle all delayed entries. * - * * - * This can *only* be the superblock at this time, so it's relatively * - * easy to deal with. We're collectively flushing the entry saved from * - * above. This will need to be handled differently if there are ever more * - * than one entry needing this special treatment.) * - * ====================================================================== */ - - if (delayed_ptr) { - - if (delayed_ptr->clear_on_unprotect) { - if(H5C__flush_single_entry(f, dxpl_id, delayed_ptr, H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__GENERATE_IMAGE_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't flush entry.") - - entry_ptr->clear_on_unprotect = FALSE; - entries_cleared++; - } else if (delayed_ptr->flush_immediately) { - /* Add this entry to the list of entries to collectively write */ - if(H5C__flush_single_entry(f, dxpl_id, delayed_ptr, H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't flush entry collectively.") - - entry_ptr->flush_immediately = FALSE; - entries_flushed++; - } /* end if */ - - entries_flushed_collectively++; - entries_flushed_or_cleared_last++; - } /* end if */ + if(H5C__flush_candidate_entries(f, dxpl_id, entries_to_flush, entries_to_clear) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "flush candidates failed") /* If we've deferred writing to do it collectively, take care of that now */ if(f->coll_md_write) { @@ -733,21 +402,6 @@ H5C_apply_candidate_list(H5F_t * f, HGOTO_ERROR(H5E_CACHE, H5E_WRITEERROR, FAIL, "can't write metadata collectively") } /* end if */ - /* ====================================================================== * - * Finished flushing everything. * - * ====================================================================== */ - - HDassert((entries_flushed == entries_to_flush)); - HDassert((entries_cleared == entries_to_clear)); - HDassert((entries_flushed_or_cleared_last == entries_to_flush_or_clear_last)); - HDassert((entries_flushed_collectively == entries_to_flush_collectively)); - - if((entries_flushed != entries_to_flush) || - (entries_cleared != entries_to_clear) || - (entries_flushed_or_cleared_last != entries_to_flush_or_clear_last) || - (entries_flushed_collectively != entries_to_flush_collectively)) - HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "entry count mismatch") - done: if(candidate_assignment_table != NULL) candidate_assignment_table = (int *)H5MM_xfree((void *)candidate_assignment_table); @@ -998,11 +652,13 @@ H5C_mark_entries_as_clean(H5F_t * f, { H5C_t * cache_ptr; unsigned entries_cleared; + unsigned pinned_entries_cleared; + hbool_t progress; unsigned entries_examined; unsigned initial_list_len; haddr_t addr; -#if H5C_DO_SANITY_CHECKS unsigned pinned_entries_marked = 0; +#if H5C_DO_SANITY_CHECKS unsigned protected_entries_marked = 0; unsigned other_entries_marked = 0; haddr_t last_addr; @@ -1086,19 +742,13 @@ H5C_mark_entries_as_clean(H5F_t * f, } /* end if */ entry_ptr->clear_on_unprotect = TRUE; + if(entry_ptr->is_pinned) + pinned_entries_marked++; #if H5C_DO_SANITY_CHECKS - if ( entry_ptr->is_protected ) { - + else if(entry_ptr->is_protected) protected_entries_marked++; - - } else if ( entry_ptr->is_pinned ) { - - pinned_entries_marked++; - - } else { - + else other_entries_marked++; - } #endif /* H5C_DO_SANITY_CHECKS */ } } @@ -1140,7 +790,8 @@ H5C_mark_entries_as_clean(H5F_t * f, entry_ptr = entry_ptr->prev; entries_cleared++; - if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) + if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, + (H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__GENERATE_IMAGE_FLAG)) < 0) HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "can't clear entry") } /* end if */ else @@ -1155,19 +806,27 @@ H5C_mark_entries_as_clean(H5F_t * f, /* It is also possible that some of the cleared entries are on the * pinned list. Must scan that also. */ - entry_ptr = cache_ptr->pel_head_ptr; - while(entry_ptr != NULL) { - if(entry_ptr->clear_on_unprotect) { - entry_ptr->clear_on_unprotect = FALSE; - clear_ptr = entry_ptr; - entry_ptr = entry_ptr->next; - entries_cleared++; + pinned_entries_cleared = 0; + progress = TRUE; + while((pinned_entries_cleared < pinned_entries_marked) && progress) { + progress = FALSE; + entry_ptr = cache_ptr->pel_head_ptr; + while(entry_ptr != NULL) { + if(entry_ptr->clear_on_unprotect && entry_ptr->flush_dep_ndirty_children == 0) { + entry_ptr->clear_on_unprotect = FALSE; + clear_ptr = entry_ptr; + entry_ptr = entry_ptr->next; + entries_cleared++; + pinned_entries_cleared++; + progress = TRUE; - if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__DEL_FROM_SLIST_ON_DESTROY_FLAG) < 0) - HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "Can't clear entry") - } /* end if */ - else - entry_ptr = entry_ptr->next; + if(H5C__flush_single_entry(f, dxpl_id, clear_ptr, + (H5C__FLUSH_CLEAR_ONLY_FLAG | H5C__GENERATE_IMAGE_FLAG)) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "can't clear entry") + } /* end if */ + else + entry_ptr = entry_ptr->next; + } /* end while */ } /* end while */ #if H5C_DO_SANITY_CHECKS @@ -1297,7 +956,6 @@ H5C__collective_write(H5F_t *f, hid_t dxpl_id) /* Get number of entries in collective write list */ count = (int)H5SL_count(cache_ptr->coll_write_list); - if(count > 0) { H5FD_mpio_xfer_t xfer_mode = H5FD_MPIO_COLLECTIVE; H5SL_node_t *node; @@ -1410,4 +1068,510 @@ done: FUNC_LEAVE_NOAPI(ret_value); } /* end H5C__collective_write() */ + + +/*------------------------------------------------------------------------- + * Function: H5C__flush_candidate_entries + * + * Purpose: Flush or clear (as indicated) the candidate entries that + * have been marked in the metadata cache. In so doing, + * observe rings and flush dependencies. + * + * Note that this function presumes that: + * + * 1) no candidate entries are protected, + * + * 2) all candidate entries are dirty, and + * + * 3) if a candidate entry has a dirty flush dependency + * child, that child is also a candidate entry. + * + * The function will fail if any of these preconditions are + * not met. + * + * Candidate entries are marked by setting either the + * flush_immediately or the clear_on_unprotect flags in the + * cache entry (but not both). Entries marked flush_immediately + * will be flushed, those marked clear_on_unprotect will be + * cleared. + * + * Note that this function is a modified version of + * H5C_flush_cache() -- any changes there may need to be + * reflected here and vise versa. + * + * Return: Non-negative on success/Negative on failure. + * + * Programmer: John Mainzer + * 2/10/17 + * + * Changes: None. + * + *------------------------------------------------------------------------- + */ +static herr_t +H5C__flush_candidate_entries(H5F_t *f, hid_t dxpl_id, + unsigned entries_to_flush[H5C_RING_NTYPES], + unsigned entries_to_clear[H5C_RING_NTYPES]) +{ +#if H5C_DO_SANITY_CHECKS + int i; + uint32_t index_len = 0; + size_t index_size = (size_t)0; + size_t clean_index_size = (size_t)0; + size_t dirty_index_size = (size_t)0; + size_t slist_size = (size_t)0; + uint32_t slist_len = 0; +#endif /* H5C_DO_SANITY_CHECKS */ + H5C_ring_t ring; + H5C_t * cache_ptr; + herr_t ret_value = SUCCEED; + + FUNC_ENTER_STATIC + + HDassert(f); + HDassert(f->shared); + + cache_ptr = f->shared->cache; + + HDassert(cache_ptr); + HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); + HDassert(cache_ptr->slist_ptr); + + HDassert(entries_to_flush[H5C_RING_UNDEFINED] == 0); + HDassert(entries_to_clear[H5C_RING_UNDEFINED] == 0); + +#if H5C_DO_SANITY_CHECKS + HDassert(cache_ptr->index_ring_len[H5C_RING_UNDEFINED] == 0); + HDassert(cache_ptr->index_ring_size[H5C_RING_UNDEFINED] == (size_t)0); + HDassert(cache_ptr->clean_index_ring_size[H5C_RING_UNDEFINED] == (size_t)0); + HDassert(cache_ptr->dirty_index_ring_size[H5C_RING_UNDEFINED] == (size_t)0); + HDassert(cache_ptr->slist_ring_len[H5C_RING_UNDEFINED] == 0); + HDassert(cache_ptr->slist_ring_size[H5C_RING_UNDEFINED] == (size_t)0); + + for(i = H5C_RING_USER; i < H5C_RING_NTYPES; i++) { + index_len += cache_ptr->index_ring_len[i]; + index_size += cache_ptr->index_ring_size[i]; + clean_index_size += cache_ptr->clean_index_ring_size[i]; + dirty_index_size += cache_ptr->dirty_index_ring_size[i]; + + slist_len += cache_ptr->slist_ring_len[i]; + slist_size += cache_ptr->slist_ring_size[i]; + } /* end for */ + + HDassert(cache_ptr->index_len == index_len); + HDassert(cache_ptr->index_size == index_size); + HDassert(cache_ptr->clean_index_size == clean_index_size); + HDassert(cache_ptr->dirty_index_size == dirty_index_size); + HDassert(cache_ptr->slist_len == slist_len); + HDassert(cache_ptr->slist_size == slist_size); +#endif /* H5C_DO_SANITY_CHECKS */ + +#if H5C_DO_EXTREME_SANITY_CHECKS + if(H5C_validate_protected_entry_list(cache_ptr) < 0 + || H5C_validate_pinned_entry_list(cache_ptr) < 0 + || H5C_validate_lru_list(cache_ptr) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "an extreme sanity check failed on entry") +#endif /* H5C_DO_EXTREME_SANITY_CHECKS */ + + cache_ptr->flush_in_progress = TRUE; + + /* flush each ring, starting from the outermost ring and + * working inward. + */ + ring = H5C_RING_USER; + while(ring < H5C_RING_NTYPES) { + if(H5C__flush_candidates_in_ring(f, dxpl_id, ring, entries_to_flush[ring], entries_to_clear[ring]) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "flush candidates in ring failed") + + ring++; + } /* end while */ + +done: + cache_ptr->flush_in_progress = FALSE; + + FUNC_LEAVE_NOAPI(ret_value) +} /* H5C__flush_candidate_entries() */ + + +/*------------------------------------------------------------------------- + * Function: H5C__flush_candidates_in_ring + * + * Purpose: Flush or clear (as indicated) the candidate entries + * contained in the specified cache and ring. All candidate + * entries in rings outside the specified ring must have been + * flushed (or cleared) on entry. + * + * Note that this function presumes that: + * + * 1) no candidate entries are protected, + * + * 2) all candidate entries are dirty, and + * + * 3) if a candidate entry has a dirty flush dependency + * child, that child is also a candidate entry. + * + * The function will fail if any of these preconditions are + * not met. + * + * Candidate entries are marked by setting either the + * flush_immediately or the clear_on_unprotect flags in the + * cache entry (but not both). Entries marked flush_immediately + * will be flushed, those marked clear_on_unprotect will be + * cleared. + * + * Candidate entries residing in the LRU must be flushed + * (or cleared) in LRU order to avoid performance issues. + * + * Return: Non-negative on success/Negative on failure. + * + * Programmer: John Mainzer + * 2/10/17 + * + *------------------------------------------------------------------------- + */ +static herr_t +H5C__flush_candidates_in_ring(H5F_t *f, hid_t dxpl_id, H5C_ring_t ring, + unsigned entries_to_flush, unsigned entries_to_clear) +{ + H5C_t * cache_ptr; + hbool_t progress; + hbool_t restart_scan = FALSE; + unsigned entries_flushed = 0; + unsigned entries_cleared = 0; +#if H5C_DO_SANITY_CHECKS + unsigned init_index_len; +#endif /* H5C_DO_SANITY_CHECKS */ + unsigned clear_flags = H5C__FLUSH_CLEAR_ONLY_FLAG | + H5C__GENERATE_IMAGE_FLAG; + unsigned flush_flags = H5C__NO_FLAGS_SET; + unsigned op_flags; + H5C_cache_entry_t *op_ptr; + H5C_cache_entry_t *entry_ptr; + herr_t ret_value = SUCCEED; + + FUNC_ENTER_STATIC + + /* Sanity checks */ + HDassert(f); + HDassert(f->shared); + cache_ptr = f->shared->cache; + HDassert(cache_ptr); + HDassert(cache_ptr->magic == H5C__H5C_T_MAGIC); + HDassert(cache_ptr->slist_ptr); + HDassert(ring > H5C_RING_UNDEFINED); + HDassert(ring < H5C_RING_NTYPES); + +#if H5C_DO_EXTREME_SANITY_CHECKS + if((H5C_validate_protected_entry_list(cache_ptr) < 0) || + (H5C_validate_pinned_entry_list(cache_ptr) < 0) || + (H5C_validate_lru_list(cache_ptr) < 0)) + HGOTO_ERROR(H5E_CACHE, H5E_SYSTEM, FAIL, "an extreme sanity check failed on entry") +#endif /* H5C_DO_EXTREME_SANITY_CHECKS */ + +#if H5C_DO_SANITY_CHECKS + /* index len should not change */ + init_index_len = cache_ptr->index_len; +#endif /* H5C_DO_SANITY_CHECKS */ + + /* Examine entries in the LRU list, and flush or clear all entries + * so marked in the target ring. + * + * With the current implementation of flush dependencies, no entry + * in the LRU can have flush dependency children -- thus one pass + * through the LRU will be sufficient. + * + * It is possible that this will change -- hence the assertion. + */ + restart_scan = FALSE; + entry_ptr = cache_ptr->LRU_tail_ptr; + while(((entries_flushed < entries_to_flush) || (entries_cleared < entries_to_clear)) + && (entry_ptr != NULL)) { + hbool_t prev_is_dirty = FALSE; + H5C_cache_entry_t *next_ptr; + + /* Entries in the LRU must not have flush dependency children */ + HDassert(entry_ptr->flush_dep_nchildren == 0); + + /* Remember dirty state of entry to advance to */ + if(entry_ptr->prev != NULL) + prev_is_dirty = entry_ptr->prev->is_dirty; + + /* If the entry is in the ring */ + if(entry_ptr->ring == ring) { + /* If this process needs to clear this entry. */ + if(entry_ptr->clear_on_unprotect) { + HDassert(entry_ptr->is_dirty); + + /* Set entry and flags for operation */ + op_ptr = entry_ptr; + op_flags = clear_flags; + + /* Set next entry appropriately */ + next_ptr = entry_ptr->next; + + /* Reset entry flag */ + entry_ptr->clear_on_unprotect = FALSE; + entries_cleared++; + } /* end if */ + else if(entry_ptr->flush_immediately) { + HDassert(entry_ptr->is_dirty); + + /* Set entry and flags for operation */ + op_ptr = entry_ptr; + op_flags = flush_flags; + + /* Set next entry appropriately */ + next_ptr = entry_ptr->next; + + /* Reset entry flag */ + entry_ptr->flush_immediately = FALSE; + entries_flushed++; + } /* end else-if */ + else { + /* No operation for this entry */ + op_ptr = NULL; + + /* Set next entry appropriately */ + next_ptr = entry_ptr; + } /* end else */ + + /* Advance to next entry */ + entry_ptr = entry_ptr->prev; + + /* Check for operation */ + if(op_ptr) { + /* reset entries_removed_counter and + * last_entry_removed_ptr prior to the call to + * H5C__flush_single_entry() so that we can spot + * unexpected removals of entries from the cache, + * and set the restart_scan flag if proceeding + * would be likely to cause us to scan an entry + * that is no longer in the cache. + * + * Note that as of this writing, this + * case cannot occur in the parallel case. + * + * Note also that there is no test code to verify + * that this code actually works (although similar code + * in the serial version exists and is tested). + */ + cache_ptr->entries_removed_counter = 0; + cache_ptr->last_entry_removed_ptr = NULL; + + if(H5C__flush_single_entry(f, dxpl_id, op_ptr, op_flags) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "can't flush entry") + + if(cache_ptr->entries_removed_counter != 0 + || cache_ptr->last_entry_removed_ptr != NULL) + restart_scan = TRUE; + } /* end if */ + } /* end if */ + else { + /* Remember "next" pointer (after advancing entries) */ + next_ptr = entry_ptr; + + /* Advance to next entry */ + entry_ptr = entry_ptr->prev; + } /* end else */ + + /* Check for restarts, etc. */ + if((entry_ptr != NULL) && + (restart_scan || (entry_ptr->is_dirty != prev_is_dirty) + || (entry_ptr->next != next_ptr) || entry_ptr->is_protected + || entry_ptr->is_pinned)) { + + /* Something has happened to the LRU -- start over + * from the tail. + * + * Recall that this code should be un-reachable at present, + * as all the operations by entries on flush that could cause + * it to be reachable are disallowed in the parallel case at + * present. Hence the following assertion which should be + * removed if the above changes. + */ + HDassert(!restart_scan); + HDassert(entry_ptr->is_dirty == prev_is_dirty); + HDassert(entry_ptr->next == next_ptr); + HDassert(!entry_ptr->is_protected); + HDassert(!entry_ptr->is_pinned); + + HDassert(FALSE); /* see comment above */ + + restart_scan = FALSE; + entry_ptr = cache_ptr->LRU_tail_ptr; + + H5C__UPDATE_STATS_FOR_LRU_SCAN_RESTART(cache_ptr) + } /* end if */ + } /* end while */ + + /* It is also possible that some of the cleared entries are on the + * pinned list. Must scan that also. + * + * Observe that in the case of the pinned entry list, most of the + * entries will have flush dependency children. As entries with + * flush dependency children may not be flushed until all of their + * children are clean, multiple passes throguh the pinned entry list + * may be required. + * + * WARNING: + * + * As we now allow unpinning, and removal of other entries as a side + * effect of flushing an entry, it is possible that the next entry + * in a PEL scan could either be no longer pinned, or no longer in + * the cache by the time we get to it. + * + * At present, this should not be possible in this case, as we disallow + * such operations in the parallel version of the library. However, + * this may change, and to that end, I have included code to detect + * such changes and cause this function to fail if they are detected. + */ + progress = TRUE; + while(progress && ((entries_flushed < entries_to_flush) || (entries_cleared < entries_to_clear))) { + progress = FALSE; + entry_ptr = cache_ptr->pel_head_ptr; + while((entry_ptr != NULL) && + ((entries_flushed < entries_to_flush) || (entries_cleared < entries_to_clear))) { + H5C_cache_entry_t *prev_ptr; + hbool_t next_is_dirty = FALSE; + + HDassert(entry_ptr->is_pinned); + + /* Remember dirty state of entry to advance to */ + if(entry_ptr->next != NULL) + next_is_dirty = entry_ptr->next->is_dirty; + + if(entry_ptr->ring == ring && entry_ptr->flush_dep_ndirty_children == 0) { + if(entry_ptr->clear_on_unprotect) { + HDassert(entry_ptr->is_dirty); + + /* Set entry and flags for operation */ + op_ptr = entry_ptr; + op_flags = clear_flags; + + /* Reset entry flag */ + entry_ptr->clear_on_unprotect = FALSE; + entries_cleared++; + progress = TRUE; + } /* end if */ + else if(entry_ptr->flush_immediately) { + HDassert(entry_ptr->is_dirty); + + /* Set entry and flags for operation */ + op_ptr = entry_ptr; + op_flags = flush_flags; + + /* Reset entry flag */ + entry_ptr->flush_immediately = FALSE; + entries_flushed++; + progress = TRUE; + } /* end else-if */ + else + /* No operation for this entry */ + op_ptr = NULL; + + /* Check for operation */ + if(op_ptr) { + /* reset entries_removed_counter and + * last_entry_removed_ptr prior to the call to + * H5C__flush_single_entry() so that we can spot + * unexpected removals of entries from the cache, + * and set the restart_scan flag if proceeding + * would be likely to cause us to scan an entry + * that is no longer in the cache. + * + * Note that as of this writing, this + * case cannot occur in the parallel case. + * + * Note also that there is no test code to verify + * that this code actually works (although similar code + * in the serial version exists and is tested). + */ + cache_ptr->entries_removed_counter = 0; + cache_ptr->last_entry_removed_ptr = NULL; + + /* Add this entry to the list of entries to collectively write + * + * This comment is misleading -- the entry will be added to the + * collective write list only if said list exists. + * + * JRM -- 2/9/17 + */ + if(H5C__flush_single_entry(f, dxpl_id, op_ptr, op_flags) < 0) + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "can't flush entry") + + if(cache_ptr->entries_removed_counter != 0 + || cache_ptr->last_entry_removed_ptr != NULL) + restart_scan = TRUE; + } /* end if */ + } /* end if */ + + /* Remember "previous" pointer (after advancing entries) */ + prev_ptr = entry_ptr; + + /* Advance to next entry */ + entry_ptr = entry_ptr->next; + + /* Check for restarts, etc. */ + if((entry_ptr != NULL) && + (restart_scan || (entry_ptr->is_dirty != next_is_dirty) + || (entry_ptr->prev != prev_ptr) || entry_ptr->is_protected + || !entry_ptr->is_pinned)) { + /* Something has happened to the pinned entry list -- start + * over from the head. + * + * Recall that this code should be un-reachable at present, + * as all the operations by entries on flush that could cause + * it to be reachable are disallowed in the parallel case at + * present. Hence the following assertion which should be + * removed if the above changes. + */ + + HDassert(!restart_scan); + HDassert(entry_ptr->is_dirty == next_is_dirty); + HDassert(entry_ptr->prev == prev_ptr); + HDassert(!entry_ptr->is_protected); + HDassert(entry_ptr->is_pinned); + + HDassert(FALSE); /* see comment above */ + + restart_scan = FALSE; + + entry_ptr = cache_ptr->pel_head_ptr; + + /* we don't keeps stats for pinned entry list scan + * restarts. If this code ever becomes reachable, + * define the necessary field, and implement the + * the following macro: + * + * H5C__UPDATE_STATS_FOR_PEL_SCAN_RESTART(cache_ptr) + */ + } /* end if */ + } /* end while ( ( entry_ptr != NULL ) && + * ( ( entries_flushed > entries_to_flush ) || + * ( entries_cleared > entries_to_clear ) ) ) + */ + } /* end while ( ( ( entries_flushed > entries_to_flush ) || + * ( entries_cleared > entries_to_clear ) ) && + * ( progress ) ) + */ + +#if H5C_DO_SANITY_CHECKS + HDassert(init_index_len == cache_ptr->index_len); +#endif /* H5C_DO_SANITY_CHECKS */ + + if(entries_flushed != entries_to_flush || entries_cleared != entries_to_clear) { + entry_ptr = cache_ptr->il_head; + while(entry_ptr != NULL) { + HDassert(!entry_ptr->clear_on_unprotect || (entry_ptr->ring > ring)); + HDassert(!entry_ptr->flush_immediately || (entry_ptr->ring > ring)); + entry_ptr = entry_ptr->il_next; + } /* end while */ + + HGOTO_ERROR(H5E_CACHE, H5E_CANTFLUSH, FAIL, "can't flush/clear all entries") + } /* end if */ + +done: + FUNC_LEAVE_NOAPI(ret_value) +} /* H5C__flush_candidates_in_ring() */ #endif /* H5_HAVE_PARALLEL */ + |