void biosal_input_stream_receive(struct thorium_actor *actor, struct thorium_message *message)
{
int tag;
int source;
uint64_t count;
struct biosal_input_stream *concrete_self;
int i;
int has_sequence;
int sequences;
int sequence_index;
int buffer_size;
char *buffer;
char *read_buffer;
struct biosal_mega_block mega_block;
char *file_name_in_buffer;
if (thorium_actor_take_action(actor, message)) {
return;
}
concrete_self = thorium_actor_concrete_actor(actor);
tag = thorium_message_action(message);
source = thorium_message_source(message);
buffer = (char *)thorium_message_buffer(message);
/* Do nothing if there is an error.
* has_error returns the error to the source.
*/
/*
if (biosal_input_stream_has_error(actor, message)) {
return;
}
*/
if (tag == ACTION_INPUT_OPEN) {
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG ACTION_INPUT_OPEN\n");
#endif
if (concrete_self->open) {
concrete_self->error = BIOSAL_INPUT_ERROR_ALREADY_OPEN;
thorium_actor_send_reply_int(actor, ACTION_INPUT_OPEN_REPLY, concrete_self->error);
thorium_actor_send_to_self_empty(actor, ACTION_ASK_TO_STOP);
return;
}
concrete_self->open = 1;
/* TODO: find out the maximum read length in some way */
concrete_self->maximum_sequence_length = BIOSAL_INPUT_MAXIMUM_SEQUENCE_LENGTH;
concrete_self->buffer_for_sequence = (char *)core_memory_allocate(concrete_self->maximum_sequence_length, MEMORY_INPUT_STREAM);
/*biosal_input_stream_init(actor);*/
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG biosal_input_stream_receive open %s\n",
buffer);
#endif
/*core_memory_copy(&concrete_self->file_index, buffer, sizeof(concrete_self->file_index));*/
file_name_in_buffer = buffer;
printf("stream/%d (node/%d) opens file %s offset %" PRIu64 "\n", thorium_actor_name(actor),
thorium_actor_node_name(actor), file_name_in_buffer,
concrete_self->starting_offset);
#ifdef DEBUG_ISSUE_594
thorium_message_print(message);
printf("Buffer %s\n", buffer);
#endif
concrete_self->file_name = core_memory_allocate(strlen(file_name_in_buffer) + 1, MEMORY_INPUT_STREAM);
strcpy(concrete_self->file_name, file_name_in_buffer);
biosal_input_proxy_init(&concrete_self->proxy, concrete_self->file_name,
concrete_self->starting_offset, concrete_self->ending_offset);
concrete_self->proxy_ready = 1;
/* Die if there is an error...
*/
if (biosal_input_stream_has_error(actor, message)) {
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG has error\n");
#endif
thorium_actor_send_reply_int(actor, ACTION_INPUT_OPEN_REPLY, concrete_self->error);
thorium_actor_send_to_self_empty(actor, ACTION_ASK_TO_STOP);
return;
}
concrete_self->controller = source;
/* no error here... */
thorium_actor_send_reply_int(actor, ACTION_INPUT_OPEN_REPLY, concrete_self->error);
} else if (tag == ACTION_INPUT_COUNT) {
/* count a little bit and yield the worker */
if (concrete_self->count_customer == THORIUM_ACTOR_NOBODY) {
concrete_self->count_customer = source;
}
if (biosal_input_stream_check_open_error(actor, message)) {
thorium_actor_send_reply_int64_t(actor, ACTION_INPUT_COUNT_REPLY, concrete_self->error);
thorium_actor_send_to_self_empty(actor, ACTION_ASK_TO_STOP);
return;
}
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG ACTION_INPUT_COUNT received...\n");
#endif
i = 0;
/* continue counting ... */
has_sequence = 1;
while (i < concrete_self->granularity && has_sequence) {
has_sequence = biosal_input_proxy_get_sequence(&concrete_self->proxy,
concrete_self->buffer_for_sequence);
#if 0
printf("Sequence= %s\n", concrete_self->buffer_for_sequence);
#endif
i++;
}
sequences = biosal_input_proxy_size(&concrete_self->proxy);
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG ACTION_INPUT_COUNT sequences %d...\n", sequences);
#endif
if (!has_sequence || sequences % concrete_self->mega_block_size == 0) {
biosal_mega_block_init(&mega_block, -1, concrete_self->last_offset,
sequences - concrete_self->last_entries, sequences);
core_vector_push_back(&concrete_self->mega_blocks, &mega_block);
concrete_self->last_entries = sequences;
concrete_self->last_offset = biosal_input_proxy_offset(&concrete_self->proxy);
thorium_actor_send_int64_t(actor, concrete_self->controller, ACTION_INPUT_COUNT_PROGRESS, sequences);
}
if (has_sequence) {
/*printf("DEBUG yield\n");*/
thorium_actor_send_to_self_empty(actor, ACTION_YIELD);
/* notify the controller of our progress...
*/
} else {
thorium_actor_send_to_self_empty(actor, ACTION_INPUT_COUNT_READY);
}
} else if (tag == ACTION_YIELD_REPLY) {
if (biosal_input_stream_check_open_error(actor, message)) {
/*
* it is not clear that there can be an error when receiving YIELD.
error = concrete_self->error;
thorium_actor_send_reply_int(actor, ACTION_INPUT_COUNT_REPLY, error);
thorium_actor_send_to_self(actor, ACTION_ASK_TO_STOP);
*/
return;
}
thorium_actor_send_to_self_empty(actor, ACTION_INPUT_COUNT);
} else if (tag == ACTION_INPUT_COUNT_READY) {
if (biosal_input_stream_check_open_error(actor, message)) {
return;
}
count = biosal_input_proxy_size(&concrete_self->proxy);
thorium_actor_send_vector(actor, concrete_self->count_customer, ACTION_INPUT_COUNT_REPLY,
&concrete_self->mega_blocks);
printf("input_stream/%d on node/%d counted entries in %s, %" PRIu64 "\n",
thorium_actor_name(actor), thorium_actor_node_name(actor),
concrete_self->file_name, count);
} else if (tag == ACTION_INPUT_CLOSE) {
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG destroy proxy\n");
#endif
concrete_self->error = BIOSAL_INPUT_ERROR_NO_ERROR;
if (biosal_input_stream_check_open_error(actor, message)) {
concrete_self->error = BIOSAL_INPUT_ERROR_FILE_NOT_OPEN;
thorium_message_init(message, ACTION_INPUT_CLOSE_REPLY, sizeof(concrete_self->error),
&concrete_self->error);
thorium_actor_send(actor, source, message);
thorium_actor_send_to_self_empty(actor, ACTION_ASK_TO_STOP);
return;
}
thorium_message_init(message, ACTION_INPUT_CLOSE_REPLY, sizeof(concrete_self->error),
&concrete_self->error);
thorium_actor_send(actor, source, message);
thorium_actor_send_to_self_empty(actor, ACTION_ASK_TO_STOP);
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("actor %d sending ACTION_INPUT_CLOSE_REPLY to %d\n",
thorium_actor_name(actor), source);
#endif
} else if (tag == ACTION_INPUT_GET_SEQUENCE) {
if (biosal_input_stream_check_open_error(actor, message)) {
/* the error management could be better. */
concrete_self->error = BIOSAL_INPUT_ERROR_FILE_NOT_OPEN;
thorium_message_init(message, ACTION_INPUT_GET_SEQUENCE_REPLY, sizeof(concrete_self->error),
&concrete_self->error);
thorium_actor_send(actor, source, message);
return;
}
sequence_index = biosal_input_proxy_size(&concrete_self->proxy);
/* TODO it would be clearer to use a struct to pack a int and a char []
* then to use the code below.
*/
read_buffer = concrete_self->buffer_for_sequence + sizeof(sequence_index);
has_sequence = biosal_input_proxy_get_sequence(&concrete_self->proxy,
read_buffer);
if (!has_sequence) {
thorium_message_init(message, ACTION_INPUT_GET_SEQUENCE_END, 0, NULL);
thorium_actor_send(actor, source, message);
return;
}
buffer_size = sizeof(sequence_index) + strlen(read_buffer) + 1;
core_memory_copy(concrete_self->buffer_for_sequence, &sequence_index, sizeof(sequence_index));
thorium_message_init(message, ACTION_INPUT_GET_SEQUENCE_REPLY,
buffer_size, concrete_self->buffer_for_sequence);
thorium_actor_send(actor, source, message);
} else if (tag == ACTION_INPUT_PUSH_SEQUENCES) {
biosal_input_stream_push_sequences(actor, message);
} else if (tag == ACTION_ASK_TO_STOP) {
thorium_actor_send_to_self_empty(actor, ACTION_STOP);
thorium_actor_send_range_empty(actor, &concrete_self->parallel_streams,
ACTION_ASK_TO_STOP);
thorium_actor_send_reply_empty(actor, ACTION_ASK_TO_STOP_REPLY);
} else if (tag == ACTION_INPUT_STREAM_RESET) {
/* fail silently
*/
if (!concrete_self->open) {
thorium_actor_send_reply_empty(actor, ACTION_INPUT_STREAM_RESET_REPLY);
return;
}
#ifdef BIOSAL_INPUT_STREAM_DEBUG
printf("DEBUG ACTION_INPUT_STREAM_RESET\n");
#endif
biosal_input_proxy_destroy(&concrete_self->proxy);
biosal_input_proxy_init(&concrete_self->proxy, concrete_self->file_name,
concrete_self->starting_offset,
concrete_self->ending_offset);
thorium_actor_send_reply_empty(actor, ACTION_INPUT_STREAM_RESET_REPLY);
} else if (tag == ACTION_PUSH_SEQUENCE_DATA_BLOCK_REPLY) {
thorium_actor_send_to_supervisor_int(actor, ACTION_INPUT_PUSH_SEQUENCES_REPLY,
source);
}
}
void biosal_input_stream_init(struct thorium_actor *actor)
{
struct biosal_input_stream *input;
struct biosal_input_stream *concrete_self;
input = (struct biosal_input_stream *)thorium_actor_concrete_actor(actor);
concrete_self = input;
concrete_self->proxy_ready = 0;
concrete_self->buffer_for_sequence = NULL;
concrete_self->maximum_sequence_length = 0;
concrete_self->open = 0;
concrete_self->error = BIOSAL_INPUT_ERROR_NO_ERROR;
concrete_self->file_name = NULL;
biosal_dna_codec_init(&concrete_self->codec);
if (biosal_dna_codec_must_use_two_bit_encoding(&concrete_self->codec,
thorium_actor_get_node_count(actor))) {
biosal_dna_codec_enable_two_bit_encoding(&concrete_self->codec);
}
/*concrete_self->mega_block_size = 2097152*/
/*
* This is the mega block size in number of sequences.
*/
concrete_self->mega_block_size = 2097152;
concrete_self->granularity = 1024;
concrete_self->last_offset = 0;
concrete_self->last_entries = 0;
concrete_self->starting_offset = 0;
/*
* Use a large ending offset.
*/
concrete_self->ending_offset = 0;
--concrete_self->ending_offset;
core_vector_init(&concrete_self->mega_blocks, sizeof(struct biosal_mega_block));
thorium_actor_add_action(actor, ACTION_INPUT_STREAM_SET_START_OFFSET, biosal_input_stream_set_start_offset);
thorium_actor_add_action(actor, ACTION_INPUT_STREAM_SET_END_OFFSET, biosal_input_stream_set_end_offset);
#ifdef ENABLE_PARALLEL_COUNT
thorium_actor_add_action(actor, ACTION_INPUT_COUNT_IN_PARALLEL, biosal_input_stream_count_in_parallel);
thorium_actor_add_action(actor, ACTION_INPUT_COUNT_REPLY, biosal_input_stream_count_reply);
#else
thorium_actor_add_action(actor, ACTION_INPUT_COUNT_IN_PARALLEL, biosal_input_stream_count_in_parallel_mock);
thorium_actor_add_action(actor, ACTION_INPUT_COUNT_REPLY, biosal_input_stream_count_reply_mock);
#endif
concrete_self->count_customer = THORIUM_ACTOR_NOBODY;
#if 0
core_string_init(&concrete_self->file_for_parallel_counting, NULL);
#endif
/*
* Parallel counting.
*/
concrete_self->total_entries = 0;
/*
* Disable parallel counting.
*/
concrete_self->finished_parallel_stream_count = 0;
core_vector_init(&concrete_self->spawners, sizeof(int));
core_vector_init(&concrete_self->parallel_streams, sizeof(int));
core_vector_init(&concrete_self->start_offsets, sizeof(uint64_t));
core_vector_init(&concrete_self->end_offsets, sizeof(uint64_t));
core_vector_init(&concrete_self->parallel_mega_blocks, sizeof(struct core_vector));
printf("%s/%d is now online on node %d\n",
thorium_actor_script_name(actor),
thorium_actor_name(actor),
thorium_actor_node_name(actor));
}
static void source_receive(struct thorium_actor *self, struct thorium_message *message)
{
int action;
void *buffer;
int leader;
int source;
int count;
struct source *concrete_self;
int name;
concrete_self = (struct source *)thorium_actor_concrete_actor(self);
action = thorium_message_action(message);
buffer = thorium_message_buffer(message);
source = thorium_message_source(message);
name = thorium_actor_name(self);
count = thorium_message_count(message);
if (action == ACTION_ASK_TO_STOP) {
thorium_actor_log(self, "sent %d ACTION_PING messages\n",
concrete_self->message_count);
thorium_actor_send_to_self_empty(self, ACTION_STOP);
} else if (action == ACTION_NOTIFY) {
#ifdef LATENCY_PROBE_USE_MULTIPLEXER
thorium_actor_send_to_self_empty(self, ACTION_ENABLE_MULTIPLEXER);
#endif
CORE_DEBUGGER_ASSERT(core_vector_empty(&concrete_self->targets));
core_vector_unpack(&concrete_self->targets, buffer);
if (source_is_important(self)) {
printf("%d (node %d worker %d) has %d targets\n", thorium_actor_name(self),
thorium_actor_node_name(self),
thorium_actor_worker_name(self),
(int)core_vector_size(&concrete_self->targets));
}
concrete_self->leader = source;
source_send_ping(self);
} else if (action == ACTION_PING_REPLY) {
CORE_DEBUGGER_ASSERT(count == 0);
++concrete_self->message_count;
CORE_DEBUGGER_ASSERT_IS_EQUAL_INT(count, 0);
CORE_DEBUGGER_ASSERT_IS_NULL(buffer);
if (concrete_self->message_count % PERIOD == 0 || concrete_self->event_count < 500) {
if (source_is_important(self)) {
printf("progress %d %d/%d\n",
name, concrete_self->message_count, concrete_self->event_count);
}
}
if (concrete_self->message_count == concrete_self->event_count) {
leader = concrete_self->leader;
thorium_actor_send_empty(self, leader, ACTION_NOTIFY_REPLY);
if (source_is_important(self))
printf("%d (ACTION_PING sent: %d)"
" sends ACTION_NOTIFY_REPLY to %d\n", thorium_actor_name(self),
concrete_self->message_count,
leader);
} else {
source_send_ping(self);
}
}
}
