Skip to content

GitLab

Commit 6071b26b authored by Misbah Mubarak's avatar Misbah Mubarak
Browse files

Updating MPI Sim Layer to use quicklists

parent db6998f0
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 1261 additions and 114 deletions
src/models/Makefile.subdir
View file @ 6071b26b
......@@ -40,25 +40,18 @@ src_libcodes_net_a_SOURCES = \
src/models/networks/model-net/model-net-sched-impl.h \
src/models/networks/model-net/model-net-sched-impl.c \
src/models/network-workloads/model-net-mpi-wrklds.c \
src/models/network-workloads/model-net-mpi-replay.c \
src/models/network-workloads/model-net-synthetic.c \
src/models/network-workloads/model-net-dumpi-traces-dump.c
bin_PROGRAMS += src/models/network-workloads/model-net-mpi-replay
bin_PROGRAMS += src/models/network-workloads/model-net-mpi-wrklds
bin_PROGRAMS += src/models/network-workloads/model-net-dumpi-traces-dump
bin_PROGRAMS += src/models/network-workloads/model-net-synthetic
src_models_network_workloads_model_net_mpi_wrklds_SOURCES = src/models/network-workloads/model-net-mpi-wrklds.c
#src_models_network_workloads_model_net_mpi_wrklds_LDADD = $(testlib) $(CODES_BASE_LIBS)
#src_models_network_workloads_model_net_mpi_wrklds_LDFLAGS = $(CODES_BASE_LDFLAGS)
#src_models_network_workloads_model_net_mpi_wrklds_CFLAGS = ${CODES_BASE_CFLAGS}
src_models_network_workloads_model_net_mpi_replay_SOURCES = src/models/network-workloads/model-net-mpi-replay.c
src_models_network_workloads_model_net_mpi_wrklds_SOURCES = src/models/network-workloads/model-net-mpi-wrklds.c
src_models_network_workloads_model_net_synthetic_SOURCES = src/models/network-workloads/model-net-synthetic.c
#src_models_network_workloads_model_net_synthetic_LDADD = $(testlib) $(CODES_BASE_LIBS)
#src_models_network_workloads_model_net_synthetic_LDFLAGS = $(CODES_BASE_LDFLAGS)
#src_models_network_workloads_model_net_synthetic_CFLAGS = ${CODES_BASE_CFLAGS}
src_models_network_workloads_model_net_dumpi_traces_dump_SOURCES = src/models/network-workloads/model-net-dumpi-traces-dump.c
#src_models_network_workloads_model_net_dumpi_traces_dump_LDADD = $(testlib) $(CODES_BASE_LIBS)
#src_models_network_workloads_model_net_dumpi_traces_dump_LDFLAGS = $(CODES_BASE_LDFLAGS)
#src_models_network_workloads_model_net_dumpi_traces_dump_CFLAGS = ${CODES_BASE_CFLAGS}
src/models/network-workloads/conf/modelnet-mpi-test-dragonfly.conf
View file @ 6071b26b
......@@ -14,13 +14,13 @@ PARAMS
modelnet_order=( "dragonfly" );
# scheduler options
modelnet_scheduler="fcfs";
chunk_size="32";
chunk_size="64";
# modelnet_scheduler="round-robin";
num_vcs="1";
num_routers="4";
local_vc_size="65535";
global_vc_size="128000";
cn_vc_size="65536";
local_vc_size="32768";
global_vc_size="65536";
cn_vc_size="32768";
local_bandwidth="5.25";
global_bandwidth="4.7";
cn_bandwidth="5.25";
......
src/models/network-workloads/model-net-mpi-replay.c 0 → 100644
View file @ 6071b26b
This diff is collapsed.
src/models/network-workloads/model-net-mpi-wrklds.c
View file @ 6071b26b
......@@ -132,8 +132,6 @@ struct nw_state
struct nw_message
{
int msg_type;
/* for reverse computation */
struct codes_workload_op * op;
struct
{
......@@ -173,7 +171,7 @@ static void codes_exec_mpi_wait(
nw_state* s, tw_lp* lp, nw_message * m, struct codes_workload_op * mpi_op);
/* reverse of mpi wait function. */
static void codes_exec_mpi_wait_rc(
nw_state* s, tw_bf* bf, nw_message* m, tw_lp* lp);
nw_state* s, nw_message* m, tw_lp* lp, struct codes_workload_op * mpi_op);
/* executes MPI isend and send operations */
static void codes_exec_mpi_send(
nw_state* s, tw_lp* lp, struct codes_workload_op * mpi_op);
......@@ -182,7 +180,7 @@ static void codes_exec_mpi_recv(
nw_state* s, tw_lp* lp, nw_message * m, struct codes_workload_op * mpi_op);
/* reverse of mpi recv function. */
static void codes_exec_mpi_recv_rc(
nw_state* s, nw_message* m, tw_lp* lp);
nw_state* s, nw_message* m, tw_lp* lp, struct codes_workload_op * mpi_op);
/* execute the computational delay */
static void codes_exec_comp_delay(
nw_state* s, tw_lp* lp, struct codes_workload_op * mpi_op);
......@@ -458,8 +456,6 @@ static int notify_waits(nw_state* s, tw_bf* bf, tw_lp* lp, nw_message* m, dumpi_
for(i = 0; i < required_count; i++)
remove_req_id(&s->completed_reqs, wait_elem->mpi_op->u.waits.req_ids[i]);
//rc_stack_push(&lp, wait_elem->mpi_op, free, s->st);
codes_issue_next_event(lp); //wait completed
}
}
......@@ -467,7 +463,7 @@ static int notify_waits(nw_state* s, tw_bf* bf, tw_lp* lp, nw_message* m, dumpi_
}
/* reverse handler of MPI wait operation */
static void codes_exec_mpi_wait_rc(nw_state* s, tw_bf* bf, nw_message* m, tw_lp* lp)
static void codes_exec_mpi_wait_rc(nw_state* s, nw_message* m, tw_lp* lp, struct codes_workload_op * mpi_op)
{
if(s->pending_waits)
{
......@@ -476,8 +472,9 @@ static void codes_exec_mpi_wait_rc(nw_state* s, tw_bf* bf, nw_message* m, tw_lp*
}
else
{
mpi_completed_queue_insert_op(&s->completed_reqs, m->op->u.wait.req_id);
tw_rand_reverse_unif(lp->rng);
mpi_completed_queue_insert_op(&s->completed_reqs, mpi_op->u.wait.req_id);
tw_rand_reverse_unif(lp->rng);
rc_stack_pop(s->st);
}
}
......@@ -505,9 +502,11 @@ static void codes_exec_mpi_wait(nw_state* s, tw_lp* lp, nw_message * m, struct c
wait_op->num_completed = 0;
wait_op->start_time = tw_now(lp);
s->pending_waits = wait_op;
rc_stack_push(lp, wait_op, free, s->st);
}
static void codes_exec_mpi_wait_all_rc(nw_state* s, tw_bf* bf, nw_message* m, tw_lp* lp)
static void codes_exec_mpi_wait_all_rc(nw_state* s, nw_message* m, tw_lp* lp, struct codes_workload_op * mpi_op)
{
if(lp->gid == TRACE)
{
......@@ -517,12 +516,12 @@ static void codes_exec_mpi_wait_all_rc(nw_state* s, tw_bf* bf, nw_message* m, tw
if(m->u.rc.found_match)
{
int i;
int count = m->op->u.waits.count;
int count = mpi_op->u.waits.count;
dumpi_req_id req_id[count];
for( i = 0; i < count; i++)
{
req_id[i] = m->op->u.waits.req_ids[i];
req_id[i] = mpi_op->u.waits.req_ids[i];
mpi_completed_queue_insert_op(&s->completed_reqs, req_id[i]);
}
tw_rand_reverse_unif(lp->rng);
......@@ -530,7 +529,7 @@ static void codes_exec_mpi_wait_all_rc(nw_state* s, tw_bf* bf, nw_message* m, tw
else
{
struct pending_waits* wait_op = s->pending_waits;
free(wait_op);
rc_stack_pop(s->st);
s->pending_waits = NULL;
assert(!s->pending_waits);
if(lp->gid == TRACE)
......@@ -584,7 +583,8 @@ static void codes_exec_mpi_wait_all(
wait_op->mpi_op = mpi_op;
wait_op->num_completed = num_completed;
wait_op->start_time = tw_now(lp);
s->pending_waits = wait_op;
rc_stack_push(lp, wait_op, free, s->st);
s->pending_waits = wait_op;
}
}
......@@ -833,25 +833,24 @@ static void codes_exec_comp_delay(
tw_event_send(e);
rc_stack_push(&lp, mpi_op, free, s->st);
}
/* reverse computation operation for MPI irecv */
static void codes_exec_mpi_recv_rc(nw_state* s, nw_message* m, tw_lp* lp)
static void codes_exec_mpi_recv_rc(nw_state* s, nw_message* m, tw_lp* lp, struct codes_workload_op * mpi_op)
{
num_bytes_recvd -= m->op->u.recv.num_bytes;
num_bytes_recvd -= mpi_op->u.recv.num_bytes;
s->recv_time = m->u.rc.saved_recv_time;
if(m->u.rc.found_match >= 0)
{
s->recv_time = m->u.rc.saved_recv_time;
mpi_queue_update(s->arrival_queue, m->u.rc.ptr_match_op, m->u.rc.found_match);
remove_req_id(&s->completed_reqs, m->op->u.recv.req_id);
remove_req_id(&s->completed_reqs, mpi_op->u.recv.req_id);
tw_rand_reverse_unif(lp->rng);
}
else if(m->u.rc.found_match < 0)
{
mpi_queue_remove_tail(lp->gid, s->pending_recvs_queue);
if(m->op->op_type == CODES_WK_IRECV)
if(mpi_op->op_type == CODES_WK_IRECV)
tw_rand_reverse_unif(lp->rng);
}
}
......@@ -893,7 +892,6 @@ static void codes_exec_mpi_recv(nw_state* s, tw_lp* lp, nw_message * m, struct c
else
{
m->u.rc.found_match = found_matching_sends;
rc_stack_push(&lp, mpi_op, free, s->st);
codes_issue_next_event(lp);
}
}
......@@ -947,7 +945,6 @@ static void codes_exec_mpi_send(nw_state* s, tw_lp* lp, struct codes_workload_op
if(mpi_op->op_type == CODES_WK_ISEND)
codes_issue_next_event(lp);
rc_stack_push(&lp, mpi_op, free, s->st);
}
/* MPI collective operations */
......@@ -998,8 +995,9 @@ static void update_send_completion_queue(nw_state* s, tw_bf * bf, nw_message * m
static void update_arrival_queue_rc(nw_state* s, tw_bf * bf, nw_message * m, tw_lp * lp)
{
s->recv_time = m->u.rc.saved_recv_time;
codes_local_latency_reverse(lp);
codes_local_latency_reverse(lp);
rc_stack_pop(s->st);
if(m->u.rc.found_match >= 0)
{
......@@ -1054,11 +1052,12 @@ static void update_arrival_queue(nw_state* s, tw_bf * bf, nw_message * m, tw_lp
arrived_op->u.send.num_bytes = m->u.msg_info.num_bytes;
arrived_op->u.send.tag = m->u.msg_info.tag;
arrived_op->u.send.req_id = m->u.msg_info.req_id;
rc_stack_push(lp, arrived_op, free, s->st);
int found_matching_recv = mpi_queue_remove_matching_op(s, lp, m, s->pending_recvs_queue, arrived_op);
if(TRACE == lp->gid)
printf("\n %lf update arrival queue req id %d %d", tw_now(lp), arrived_op->u.send.req_id, m->op->u.send.source_rank);
printf("\n %lf update arrival queue req id %d %d", tw_now(lp), arrived_op->u.send.req_id, arrived_op->u.send.source_rank);
if(found_matching_recv < 0)
{
m->u.rc.found_match = -1;
......@@ -1067,7 +1066,6 @@ static void update_arrival_queue(nw_state* s, tw_bf * bf, nw_message * m, tw_lp
else
{
m->u.rc.found_match = found_matching_recv;
rc_stack_push(&lp, arrived_op, free, s->st);
notify_waits(s, bf, lp, m, m->u.rc.saved_matched_req);
}
}
......@@ -1165,29 +1163,33 @@ void nw_test_event_handler(nw_state* s, tw_bf * bf, nw_message * m, tw_lp * lp)
static void get_next_mpi_operation_rc(nw_state* s, tw_bf * bf, nw_message * m, tw_lp * lp)
{
codes_workload_get_next_rc(wrkld_id, 0, (int)s->nw_id, m->op);
if(m->op->op_type == CODES_WK_END)
struct codes_workload_op * mpi_op =
(struct codes_workload_op *)rc_stack_pop(s->st);
codes_workload_get_next_rc(wrkld_id, 0, (int)s->nw_id, mpi_op);
if(mpi_op->op_type == CODES_WK_END)
return;
switch(m->op->op_type)
switch(mpi_op->op_type)
{
case CODES_WK_SEND:
case CODES_WK_ISEND:
{
if(lp->gid == TRACE)
printf("\n %lf reverse send req %d ", tw_now(lp), (int)m->op->u.send.req_id);
model_net_event_rc(net_id, lp, m->op->u.send.num_bytes);
if(m->op->op_type == CODES_WK_ISEND)
printf("\n %lf reverse send req %d ", tw_now(lp), (int)mpi_op->u.send.req_id);
model_net_event_rc(net_id, lp, mpi_op->u.send.num_bytes);
if(mpi_op->op_type == CODES_WK_ISEND)
tw_rand_reverse_unif(lp->rng);
s->num_sends--;
num_bytes_sent -= m->op->u.send.num_bytes;
num_bytes_sent -= mpi_op->u.send.num_bytes;
}
break;
case CODES_WK_IRECV:
case CODES_WK_RECV:
{
codes_exec_mpi_recv_rc(s, m, lp);
codes_exec_mpi_recv_rc(s, m, lp, mpi_op);
s->num_recvs--;
}
break;
......@@ -1196,7 +1198,7 @@ static void get_next_mpi_operation_rc(nw_state* s, tw_bf * bf, nw_message * m, t
s->num_delays--;
if (!disable_delay) {
tw_rand_reverse_unif(lp->rng);
s->compute_time -= s_to_ns(m->op->u.delay.seconds);
s->compute_time -= s_to_ns(mpi_op->u.delay.seconds);
}
}
break;
......@@ -1217,13 +1219,13 @@ static void get_next_mpi_operation_rc(nw_state* s, tw_bf * bf, nw_message * m, t
case CODES_WK_WAIT:
{
s->num_wait--;
codes_exec_mpi_wait_rc(s, bf, m, lp);
codes_exec_mpi_wait_rc(s, m, lp, mpi_op);
}
break;
case CODES_WK_WAITALL:
{
s->num_waitall--;
codes_exec_mpi_wait_all_rc(s, bf, m, lp);
codes_exec_mpi_wait_all_rc(s, m, lp, mpi_op);
}
break;
case CODES_WK_WAITSOME:
......@@ -1234,7 +1236,7 @@ static void get_next_mpi_operation_rc(nw_state* s, tw_bf * bf, nw_message * m, t
}
break;
default:
printf("\n Invalid op type %d ", m->op->op_type);
printf("\n Invalid op type %d ", mpi_op->op_type);
}
}
......@@ -1242,8 +1244,7 @@ static void get_next_mpi_operation(nw_state* s, tw_bf * bf, nw_message * m, tw_l
{
struct codes_workload_op * mpi_op = malloc(sizeof(struct codes_workload_op));
codes_workload_get_next(wrkld_id, 0, (int)s->nw_id, mpi_op);
m->op = mpi_op;
rc_stack_push(lp, mpi_op, free, s->st);
if(mpi_op->op_type == CODES_WK_END)
{
......
src/models/networks/model-net/dragonfly.c
View file @ 6071b26b
......@@ -19,6 +19,7 @@
#include "codes/net/dragonfly.h"
#include "sys/file.h"
#include "codes/quickhash.h"
#include "codes/rc-stack.h"
#define CREDIT_SIZE 8
#define MEAN_PROCESS 1.0
......@@ -308,6 +309,12 @@ static int dragonfly_get_msg_sz(void)
return sizeof(terminal_message);
}
static void free_tmp(void * ptr)
{
struct dfly_qhash_entry * dfly = ptr;
free(dfly->remote_event_data);
free(dfly);
}
static void append_to_terminal_message_list(
terminal_message_list ** thisq,
terminal_message_list ** thistail,
......@@ -1225,19 +1232,15 @@ void packet_arrive_rc(terminal_state * s, tw_bf * bf, terminal_message * msg, tw
if(msg->path_type == NON_MINIMAL)
nonmin_count--;
uint64_t total_chunks = msg->total_size / s->params->chunk_size;
if(msg->total_size % s->params->chunk_size)
total_chunks++;
if(!total_chunks)
total_chunks = 1;
struct qhash_head * hash_link = NULL;
struct dfly_qhash_entry * tmp = NULL;
struct dfly_hash_key key;
key.message_id = msg->message_id;
key.sender_id = msg->sender_lp;
struct qhash_head * hash_link = NULL;
hash_link = qhash_search(s->rank_tbl, &key);
tmp = qhash_entry(hash_link, struct dfly_qhash_entry, hash_link);
if(bf->c1)
{
......@@ -1256,45 +1259,27 @@ void packet_arrive_rc(terminal_state * s, tw_bf * bf, terminal_message * msg, tw
dragonfly_max_latency = msg->saved_available_time;
if(msg->saved_completed_chunks >= total_chunks)
if(bf->c7)
{
s->finished_msgs--;
s->total_msg_time -= (tw_now(lp) - msg->msg_start_time);
s->total_msg_size -= msg->total_size;
//assert(!hash_link);
void *m_data_src = model_net_method_get_edata(DRAGONFLY, msg);
struct dfly_qhash_entry * d_entry = malloc(sizeof(struct dfly_qhash_entry));
d_entry->num_chunks = msg->saved_completed_chunks;
d_entry->key = key;
d_entry->remote_event_data = NULL;
d_entry->remote_event_size = 0;
struct dfly_qhash_entry * d_entry_pop = (struct dfly_qhash_entry*)rc_stack_pop(s->st);
qhash_add(s->rank_tbl, &key, &(d_entry_pop->hash_link));
s->rank_tbl_pop++;
if(msg->saved_remote_esize > 0)
{
d_entry->remote_event_data = (void*)malloc(msg->saved_remote_esize);
memcpy(d_entry->remote_event_data, m_data_src, msg->saved_remote_esize);
d_entry->remote_event_size = msg->saved_remote_esize;
}
qhash_add(s->rank_tbl, &key, &(d_entry->hash_link));
hash_link = &(d_entry_pop->hash_link);
tmp = d_entry_pop;
s->rank_tbl_pop++;
int net_id = model_net_get_id(LP_METHOD_NM);
if(bf->c4)
model_net_event_rc2(lp, &msg->event_rc);
}
hash_link = NULL;
hash_link = qhash_search(s->rank_tbl, &key);
assert(hash_link);
struct dfly_qhash_entry * tmp2 = NULL;
tmp2 = qhash_entry(hash_link, struct dfly_qhash_entry, hash_link);
assert(tmp2);
tmp2->num_chunks--;
assert(tmp);
tmp->num_chunks--;
return;
}
......@@ -1375,7 +1360,7 @@ void packet_arrive(terminal_state * s, tw_bf * bf, terminal_message * msg,
num_chunks++;
if(!num_chunks)
num_chunks = 1;
num_chunks = 1;
completed_packets++;
......@@ -1388,8 +1373,6 @@ void packet_arrive(terminal_state * s, tw_bf * bf, terminal_message * msg,
if(msg->path_type != MINIMAL && msg->path_type != NON_MINIMAL)
printf("\n Wrong message path type %d ", msg->path_type);
msg->saved_remote_esize = 0;
#if DEBUG == 1
if( msg->packet_ID == TRACK
&& msg->chunk_id == num_chunks-1
......@@ -1409,13 +1392,14 @@ void packet_arrive(terminal_state * s, tw_bf * bf, terminal_message * msg,
* them */
void *m_data_src = model_net_method_get_edata(DRAGONFLY, msg);
struct qhash_head *hash_link = NULL;
struct dfly_qhash_entry * tmp = NULL;
struct dfly_hash_key key;
key.message_id = msg->message_id;
key.sender_id = msg->sender_lp;
hash_link = qhash_search(s->rank_tbl, &key);
struct dfly_qhash_entry * tmp = NULL;
tmp = qhash_entry(hash_link, struct dfly_qhash_entry, hash_link);
/* If an entry does not exist then create one */
if(!hash_link)
......@@ -1427,10 +1411,11 @@ void packet_arrive(terminal_state * s, tw_bf * bf, terminal_message * msg,
d_entry->remote_event_data = NULL;
qhash_add(s->rank_tbl, &key, &(d_entry->hash_link));
s->rank_tbl_pop++;
hash_link = &(d_entry->hash_link);
tmp = d_entry;
}
hash_link = qhash_search(s->rank_tbl, &key);
tmp = qhash_entry(hash_link, struct dfly_qhash_entry, hash_link);
assert(tmp);
tmp->num_chunks++;
......@@ -1458,46 +1443,35 @@ void packet_arrive(terminal_state * s, tw_bf * bf, terminal_message * msg,
dragonfly_max_latency = tw_now( lp ) - msg->travel_start_time;
}
}
if(msg->remote_event_size_bytes > 0 && !is_last_msg)
if(msg->remote_event_size_bytes > 0 && !tmp->remote_event_data)
{
/* Retreive the remote event entry */
if(!tmp->remote_event_data)
{
tmp->remote_event_data = (void*)malloc(msg->remote_event_size_bytes);
assert(tmp->remote_event_data);
tmp->remote_event_size = msg->remote_event_size_bytes;
memcpy(tmp->remote_event_data, m_data_src, msg->remote_event_size_bytes);
}
tmp->remote_event_data = (void*)malloc(msg->remote_event_size_bytes);
assert(tmp->remote_event_data);
tmp->remote_event_size = msg->remote_event_size_bytes;
memcpy(tmp->remote_event_data, m_data_src, msg->remote_event_size_bytes);
}
/* If all chunks of a message have arrived then send a remote event to the
* callee*/
if(is_last_msg)
{
bf->c7 = 1;
s->finished_msgs++;
s->total_msg_time += (tw_now(lp) - msg->msg_start_time);
s->total_msg_size += msg->total_size;
if(msg->remote_event_size_bytes > 0)
{
char * remote_data = malloc(msg->remote_event_size_bytes);
memcpy(remote_data, m_data_src, msg->remote_event_size_bytes);
send_remote_event(s, msg, lp, bf, remote_data, msg->remote_event_size_bytes);
rc_stack_push(&lp, remote_data, free, s->st);
send_remote_event(s, msg, lp, bf, m_data_src, msg->remote_event_size_bytes);
}
else
{
void *m_data = model_net_method_get_edata(DRAGONFLY, msg);
send_remote_event(s, msg, lp, bf, tmp->remote_event_data, tmp->remote_event_size);
msg->saved_remote_esize = tmp->remote_event_size;
/* append remote event data to this message */
memcpy(m_data, tmp->remote_event_data, tmp->remote_event_size);
}
msg->saved_completed_chunks = tmp->num_chunks;
/* Remove the hash entry */
qhash_del(hash_link);
rc_stack_push(&lp, tmp->remote_event_data, free, s->st);
rc_stack_push(&lp, tmp, free, s->st);
rc_stack_push(lp, tmp, free_tmp, s->st);
s->rank_tbl_pop--;
}
return;
......
tests/conf/modelnet-test-dragonfly.conf
View file @ 6071b26b
......@@ -24,5 +24,5 @@ PARAMS
global_bandwidth="4.7";
cn_bandwidth="5.25";
message_size="336";
routing="adaptive";
routing="nonminimal";
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment