Optimized the scanning process in PE.

cocoa/base/en.lproj/core.strings

@@ -19,7 +19,7 @@
 
 "Analyzed %d/%d pictures" = "Analyzed %d/%d pictures";
 "Preparing for matching" = "Preparing for matching";
-"Matched %d/%d pictures" = "Matched %d/%d pictures";
+"Performed %d/%d chunk matches" = "Performed %d/%d chunk matches";
 "Verified %d/%d matches" = "Verified %d/%d matches";
 
 "Removing dead tracks from your iTunes Library" = "Removing dead tracks from your iTunes Library";

cocoa/base/fr.lproj/core.strings

@@ -18,7 +18,7 @@
 
 "Analyzed %d/%d pictures" = "Analyzé %d/%d images";
 "Preparing for matching" = "Préparation pour la comparaison";
-"Matched %d/%d pictures" = "Comparé %d/%d images";
+"Performed %d/%d chunk matches" = "%d/%d blocs d'images comparés";
 "Verified %d/%d matches" = "Vérifié %d/%d paires";
 
 "Removing dead tracks from your iTunes Library" = "Retrait des tracks mortes de votre librairie iTunes";

core_pe/matchbase.py

@@ -8,8 +8,9 @@
 
 import logging
 import multiprocessing
-from collections import defaultdict, deque
+from itertools import combinations
 
+from hscommon.util import extract
 from hscommon.trans import tr
 from jobprogress import job
 
@@ -17,15 +18,32 @@ from core.engine import Match
 from .block import avgdiff, DifferentBlockCountError, NoBlocksError
 from .cache import Cache
 
+# OPTIMIZATION NOTES:
+# The bottleneck of the matching phase is CPU, which is why we use multiprocessing. However, another
+# bottleneck that shows up when a lot of pictures are involved is Disk IO's because blocks
+# constantly have to be read from disks by subprocesses. This problem is especially big on CPUs
+# with a lot of cores. Therefore, we must minimize Disk IOs. The best way to achieve that is to
+# separate the files to scan in "chunks" and it's by chunk that blocks are read in memory and
+# compared to each other. Each file in a chunk has to be compared to each other, of course, but also
+# to files in other chunks. So chunkifying doesn't save us any actual comparison, but the advantage
+# is that instead of reading blocks from disk number_of_files**2 times, we read it
+# number_of_files*number_of_chunks times.
+# Determining the right chunk size is tricky, bceause if it's too big, too many blocks will be in
+# memory at the same time and we might end up with memory trashing, which is awfully slow. So,
+# because our *real* bottleneck is CPU, the chunk size must simply be enough so that the CPU isn't
+# starved by Disk IOs.
+
 MIN_ITERATIONS = 3
 BLOCK_COUNT_PER_SIDE = 15
+DEFAULT_CHUNK_SIZE = 1000
+MIN_CHUNK_SIZE = 100
 
 # Enough so that we're sure that the main thread will not wait after a result.get() call
-# cpucount*2 should be enough to be sure that the spawned process will not wait after the results
+# cpucount+1 should be enough to be sure that the spawned process will not wait after the results
 # collection made by the main process.
-RESULTS_QUEUE_LIMIT = multiprocessing.cpu_count() * 2
+RESULTS_QUEUE_LIMIT = multiprocessing.cpu_count() + 1
 
-def prepare_pictures(pictures, cache_path, j=job.nulljob):
+def prepare_pictures(pictures, cache_path, with_dimensions, j=job.nulljob):
     # The MemoryError handlers in there use logging without first caring about whether or not
     # there is enough memory left to carry on the operation because it is assumed that the
     # MemoryError happens when trying to read an image file, which is freed from memory by the
@@ -36,6 +54,7 @@ def prepare_pictures(pictures, cache_path, j=job.nulljob):
         for picture in j.iter_with_progress(pictures, tr("Analyzed %d/%d pictures")):
             picture.unicode_path = str(picture.path)
             logging.debug("Analyzing picture at {}".format(picture.unicode_path))
+            if with_dimensions:
                 picture.dimensions # pre-read dimensions
             try:
                 if picture.unicode_path not in cache:
@@ -53,18 +72,42 @@ def prepare_pictures(pictures, cache_path, j=job.nulljob):
     cache.close()
     return prepared
 
+def get_chunks(pictures):
+    min_chunk_count = multiprocessing.cpu_count() * 2 # have enough chunks to feed all subprocesses
+    chunk_count = len(pictures) // DEFAULT_CHUNK_SIZE
+    chunk_count = max(min_chunk_count, chunk_count)
+    chunk_size = (len(pictures) // chunk_count) + 1
+    chunk_size = max(MIN_CHUNK_SIZE, chunk_size)
+    logging.info("Creating {} chunks with a chunk size of {} for {} pictures".format(
+        chunk_count, chunk_size, len(pictures)))
+    chunks = [pictures[i:i+chunk_size] for i in range(0, len(pictures), chunk_size)]
+    return chunks
+
 def get_match(first, second, percentage):
     if percentage < 0:
         percentage = 0
     return Match(first, second, percentage)
 
-def async_compare(ref_id, other_ids, dbname, threshold):
+def async_compare(ref_ids, other_ids, dbname, threshold, picinfo):
+    # The list of ids in ref_ids have to be compared to the list of ids in other_ids. other_ids
+    # can be None. In this case, ref_ids has to be compared with itself
+    # picinfo is a dictionary {pic_id: (dimensions, is_ref)}
     cache = Cache(dbname)
     limit = 100 - threshold
-    ref_blocks = cache[ref_id]
+    ref_pairs = list(cache.get_multiple(ref_ids))
-    pairs = cache.get_multiple(other_ids)
+    if other_ids is not None:
+        other_pairs = list(cache.get_multiple(other_ids))
+        comparisons_to_do = [(r, o) for r in ref_pairs for o in other_pairs]
+    else:
+        comparisons_to_do = list(combinations(ref_pairs, 2))
     results = []
-    for other_id, other_blocks in pairs:
+    for (ref_id, ref_blocks), (other_id, other_blocks) in comparisons_to_do:
+        ref_dimensions, ref_is_ref = picinfo[ref_id]
+        other_dimensions, other_is_ref = picinfo[other_id]
+        if ref_is_ref and other_is_ref:
+            continue
+        if ref_dimensions != other_dimensions:
+            continue
         try:
             diff = avgdiff(ref_blocks, other_blocks, limit, MIN_ITERATIONS)
             percentage = 100 - diff
@@ -76,48 +119,60 @@ def async_compare(ref_id, other_ids, dbname, threshold):
     return results
     
 def getmatches(pictures, cache_path, threshold=75, match_scaled=False, j=job.nulljob):
+    def get_picinfo(p):
+        if match_scaled:
+            return (None, p.is_ref)
+        else:
+            return (p.dimensions, p.is_ref)
+    
+    def collect_results(collect_all=False):
+        # collect results and wait until the queue is small enough to accomodate a new results.
+        nonlocal async_results, matches, comparison_count
+        limit = 0 if collect_all else RESULTS_QUEUE_LIMIT
+        while len(async_results) > limit:
+            ready, working = extract(lambda r: r.ready(), async_results)
+            for result in ready:
+                matches += result.get()
+                async_results.remove(result)
+            comparison_count += 1
+        progress_msg = tr("Performed %d/%d chunk matches") % (comparison_count, len(comparisons_to_do))
+        j.set_progress(comparison_count, progress_msg)
+    
     j = j.start_subjob([3, 7])
-    pictures = prepare_pictures(pictures, cache_path, j)
+    pictures = prepare_pictures(pictures, cache_path, with_dimensions=not match_scaled, j=j)
     j = j.start_subjob([9, 1], tr("Preparing for matching"))
     cache = Cache(cache_path)
     id2picture = {}
-    dimensions2pictures = defaultdict(set)
     for picture in pictures:
         try:
             picture.cache_id = cache.get_id(picture.unicode_path)
             id2picture[picture.cache_id] = picture
-            if not match_scaled:
-                dimensions2pictures[picture.dimensions].add(picture)
         except ValueError:
             pass
     cache.close()
     pictures = [p for p in pictures if hasattr(p, 'cache_id')]
     pool = multiprocessing.Pool()
-    async_results = deque()
+    async_results = []
     matches = []
-    pictures_copy = set(pictures)
+    chunks = get_chunks(pictures)
-    for ref in j.iter_with_progress(pictures, tr("Matched %d/%d pictures")):
+    # We add a None element at the end of the chunk list because each chunk has to be compared
-        others = pictures_copy if match_scaled else dimensions2pictures[ref.dimensions]
+    # with itself. Thus, each chunk will show up as a ref_chunk having other_chunk set to None once.
-        others.remove(ref)
+    comparisons_to_do = list(combinations(chunks + [None], 2))
-        if ref.is_ref:
+    comparison_count = 0
-            # Don't spend time comparing two ref pics together.
+    j.start_job(len(comparisons_to_do))
-            others = [pic for pic in others if not pic.is_ref]
+    for ref_chunk, other_chunk in comparisons_to_do:
-        if others:
+        picinfo = {p.cache_id: get_picinfo(p) for p in ref_chunk}
-            cache_ids = [f.cache_id for f in others]
+        ref_ids = [p.cache_id for p in ref_chunk]
-            # We limit the number of cache_ids we send for multi-processing because otherwise, we
+        if other_chunk is not None:
-            # might get an error saying "String or BLOB exceeded size limit"
+            other_ids = [p.cache_id for p in other_chunk]
-            ARG_LIMIT = 1000
+            picinfo.update({p.cache_id: get_picinfo(p) for p in other_chunk})
-            while cache_ids:
+        else:
-                args = (ref.cache_id, cache_ids[:ARG_LIMIT], cache_path, threshold)
+            other_ids = None
+        args = (ref_ids, other_ids, cache_path, threshold, picinfo)
         async_results.append(pool.apply_async(async_compare, args))
-                cache_ids = cache_ids[ARG_LIMIT:]
+        collect_results()
-        # We use a while here because it's possible that more than one result has been added if
+    collect_results(collect_all=True)
-        # ARG_LIMIT has been reached.
+    pool.close()
-        while len(async_results) > RESULTS_QUEUE_LIMIT:
-            result = async_results.popleft()
-            matches.extend(result.get())
-    for result in async_results: # process the rest of the results
-        matches.extend(result.get())
     
     result = []
     for ref_id, other_id, percentage in j.iter_with_progress(matches, tr("Verified %d/%d matches"), every=10):
@@ -126,6 +181,8 @@ def getmatches(pictures, cache_path, threshold=75, match_scaled=False, j=job.nul
         if percentage == 100 and ref.md5 != other.md5:
             percentage = 99
         if percentage >= threshold:
+            ref.dimensions # pre-read dimensions for display in results
+            other.dimensions
             result.append(get_match(ref, other, percentage))
     return result
 

qt/lang/fr.ts

@@ -84,7 +84,7 @@
 </message>
 <message>
     <source>Matched %d/%d pictures</source>
-    <translation>Comparé %d/%d images</translation>
+    <translation>%d/%d blocs d'images comparés</translation>
 </message>
 <message>
     <source>Verified %d/%d matches</source>