def search_cuda_fast(search_filename):
BLOCK_SIZE = 100000000
GRID_SIZE = 5000
block_search = cuda_block_search()
patterns, pattern_starts = encode_strings(
[bytes(string, 'utf-8') for string in searched_words()])
patterns_gpu = gpuarray.to_gpu(patterns)
pattern_starts_gpu = gpuarray.to_gpu(pattern_starts)
matches = np.zeros(GRID_SIZE,
dtype=np.int32) # Must be higher than max id in kernel
matches_gpu = gpuarray.to_gpu(matches)
matches_count = 0
keyword_count = len(searched_words())
with open(search_filename, "rb") as search_file:
while True:
block = search_file.read(BLOCK_SIZE)
if not block:
break
last_newline = block.rfind(b'\n')
final_block = block
final_block_gpu = gpuarray.to_gpu(np.array(final_block))
if len(block) == BLOCK_SIZE:
search_file.seek(last_newline - len(block), 1)
block_search(final_block_gpu,
np.int32(last_newline),
patterns_gpu,
pattern_starts_gpu,
np.int32(len(searched_words())),
np.int32(BLOCK_SIZE / GRID_SIZE),
matches_gpu,
block=(keyword_count, 1, 1),
grid=(GRID_SIZE, 1))
result = pycuda.gpuarray.sum(matches_gpu).get()
print("File done")
return result
def search_file(search_filename):
keywords = searched_words()
matches = 0
with open(search_filename, "r") as search_file:
for line in tqdm.tqdm(search_file):
matches += int(is_match(line, keywords))
return matches
def search_file(search_filename):
keywords = searched_words()
regex = re.compile('|'.join(keywords), re.UNICODE)
matches = 0
with open(search_filename, "r") as search_file:
for line in tqdm.tqdm(search_file):
matches += int(is_match_regex(line, regex))
return matches
def search_cuda_fast(search_filename):
BLOCK_SIZE = 100000000
GRID_SIZE = 5000
patterns, pattern_starts = encode_strings(
[bytes(string, 'utf-8') for string in searched_words()])
pattern_gpu = cuda.to_device(patterns)
pattern_starts_gpu = cuda.to_device(pattern_starts)
matches = np.zeros(GRID_SIZE, dtype=np.int32)
matches_gpu = cuda.to_device(matches)
keyword_count = len(searched_words())
with open(search_filename, "rb") as search_file:
while True:
block = search_file.read(BLOCK_SIZE)
if not block:
break
last_newline = block.rfind(b'\n')
final_block = block
final_block_gpu = cuda.to_device(
np.frombuffer(final_block, dtype=np.ubyte))
if len(block) == BLOCK_SIZE:
search_file.seek(last_newline - len(block), 1)
block_search[GRID_SIZE, keyword_count](
final_block_gpu,
cuda.to_device(np.array([last_newline], dtype=np.int32)),
pattern_gpu, pattern_starts_gpu,
cuda.to_device(np.array([keyword_count], dtype=np.int32)),
cuda.to_device(
np.array([BLOCK_SIZE / GRID_SIZE],
dtype=np.int32)), matches_gpu)
result_array = matches_gpu.copy_to_host()
result = result_array.sum()
print("File done")
return result
def search_file(search_filename):
keywords = searched_words()
flashtext_processor = KeywordProcessor()
for keyword in keywords:
flashtext_processor.add_keyword(keyword)
matches = 0
with open(search_filename, "r") as search_file:
for line in tqdm.tqdm(search_file):
matches += int(is_match_flashtext(line, flashtext_processor))
return matches
def searching_worker(text_queue, file_count):
global files_read
global global_matches
keywords = searched_words()
matches = 0
while files_read != file_count: # if some files are still not read till the end, continue
while True:
try:
batch = text_queue.get(True, 5.0)
except queue.Empty as e:
break
for text in batch:
if is_match(text, keywords):
matches += 1
with lock:
global_matches += matches
def searching_process(text_queue, matches_shared, files_read, file_count):
keywords = searched_words()
matches = 0
while files_read.value != file_count:
while True:
try:
batch = text_queue.get(True, 5.0)
except queue.Empty as e:
break
for text in batch:
if is_match(text, keywords):
matches += 1
with matches_shared.get_lock():
matches_shared.value += matches
print("Searching process finished")
