ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Make sure everything is imediatly acked, even if not matched

Просмотр исходного кода 
Buffer first descriptor on the sendreq until postive ACK 
Set bytes delivered only after postive ACK, removed num_acks, etc, in general
trying to remove as much state as possible so that rolling things back isn't
such a nightmare 

This commit was SVN r9187.
Этот коммит содержится в:
Galen Shipman 2006-03-01 22:37:10 +00:00
родитель c2fded586a
Коммит 84d3055db5
8 изменённых файлов: 250 добавлений и 132 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

35
ompi/mca/pml/dr/pml_dr.h
Просмотреть файл

@ -251,5 +251,40 @@ extern int mca_pml_dr_start(
MCA_BML_BASE_BTL_DES_ALLOC(bml_btl, des, \
sizeof(mca_pml_dr_hdr_t) + (sizeof(mca_btl_base_segment_t) << 4), size)
/* ADLER_NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
#define ADLER_NMAX 5551
#define MOD_ADLER 65521
#define DO1(buf,i) {_a += buf[i]; _b += _a;}
#define DO2(buf,i) DO1(buf,i); DO1(buf,i+1);
#define DO4(buf,i) DO2(buf,i); DO2(buf,i+2);
#define DO8(buf,i) DO4(buf,i); DO4(buf,i+4);
#define DO16(buf) DO8(buf,0); DO8(buf,8);
#define COMPUTE_SPECIFIC_CHECKSUM( DATA, LENGTH, ADLER32) \
do { \
uint8_t *_data = (DATA); /* Pointer to the data to be summed */ \
size_t _len = (LENGTH); /* Length in bytes */ \
uint32_t _a = (ADLER32) & 0xffff, \
_b = ((ADLER32) >> 16) & 0xffff; \
\
while( _len > 0 ) { \
unsigned _tlen = _len > ADLER_NMAX ? ADLER_NMAX : _len; \
_len -= _tlen; \
while( _tlen >= 16 ) { \
DO16(_data); \
_data += 16; \
_tlen -= 16; \
} \
if( 0 != _tlen ) do { \
_a += *_data++; _b += _a; \
} while( --_tlen > 0 ); \
_a = _a % MOD_ADLER; \
_b = _b % MOD_ADLER; \
} \
(ADLER32) = _b << 16 | _a; \
} while(0)
#endif

1
ompi/mca/pml/dr/pml_dr_comm.c
Просмотреть файл

@ -36,6 +36,7 @@ static void mca_pml_dr_comm_proc_construct(mca_pml_dr_comm_proc_t* proc)
OBJ_CONSTRUCT(&proc->specific_receives, opal_list_t);
OBJ_CONSTRUCT(&proc->unexpected_frags, opal_list_t);
OBJ_CONSTRUCT(&proc->acked_vfrags, opal_list_t);
proc->acked_vfrags_ptr = NULL;
}

27
ompi/mca/pml/dr/pml_dr_comm.h
Просмотреть файл

@ -115,24 +115,19 @@ static inline void mca_pml_dr_comm_proc_set_acked(mca_pml_dr_comm_proc_t* proc,
int8_t direction = 0; /* 1 is next, -1 is previous */
mca_pml_dr_acked_item_t *new_item, *next_item, *prev_item;
while(true) {
if(NULL == item && direction == 0) {
if(NULL == item &&
(direction == 0 || direction == 1)) {
new_item = OBJ_NEW(mca_pml_dr_acked_item_t);
new_item->vfrag_id_low = new_item->vfrag_id_high = vfrag_id;
opal_list_append(&proc->acked_vfrags, (opal_list_item_t*) item);
proc->acked_vfrags_ptr = (opal_list_item_t*) item;
opal_list_append(&proc->acked_vfrags, (opal_list_item_t*) new_item);
proc->acked_vfrags_ptr = (opal_list_item_t*) new_item;
return;
} else if (NULL == item && direction == 1) {
opal_list_append(&proc->acked_vfrags, (opal_list_item_t*) item);
proc->acked_vfrags_ptr = (opal_list_item_t*) item;
return;
} else if (NULL == item && direction == -1) {
opal_list_prepend(&proc->acked_vfrags, (opal_list_item_t*) item);
proc->acked_vfrags_ptr = (opal_list_item_t*) item;
new_item = OBJ_NEW(mca_pml_dr_acked_item_t);
new_item->vfrag_id_low = new_item->vfrag_id_high = vfrag_id;
opal_list_prepend(&proc->acked_vfrags, (opal_list_item_t*) new_item);
proc->acked_vfrags_ptr = (opal_list_item_t*) new_item;
return;
} else if(item->vfrag_id_high >= vfrag_id && item->vfrag_id_low <= vfrag_id ) {
@ -177,7 +172,7 @@ static inline void mca_pml_dr_comm_proc_set_acked(mca_pml_dr_comm_proc_t* proc,
opal_list_insert_pos(&proc->acked_vfrags,
(opal_list_item_t*) item,
(opal_list_item_t*) new_item);
proc->acked_vfrags_ptr = (opal_list_item_t*) item;
proc->acked_vfrags_ptr = (opal_list_item_t*) new_item;
return;
} else {
direction = 1;
@ -195,7 +190,7 @@ static inline void mca_pml_dr_comm_proc_set_acked(mca_pml_dr_comm_proc_t* proc,
(opal_list_item_t*) next_item,
(opal_list_item_t*) new_item);
}
proc->acked_vfrags_ptr = (opal_list_item_t*) item;
proc->acked_vfrags_ptr = (opal_list_item_t*) new_item;
return;
} else {
direction = -1;

8
ompi/mca/pml/dr/pml_dr_hdr.h
Просмотреть файл

@ -36,11 +36,11 @@
#define MCA_PML_DR_HDR_TYPE_ACK 3
#define MCA_PML_DR_HDR_TYPE_FRAG 4
#define MCA_PML_DR_HDR_FLAGS_ACK 1 /* is an ack required */
#define MCA_PML_DR_HDR_FLAGS_NBO 2 /* is the hdr in network byte order */
#define MCA_PML_DR_HDR_FLAGS_NBO 1 /* is the hdr in network byte order */
#define MCA_PML_DR_HDR_FLAGS_VFRAG 2
#define MCA_PML_DR_HDR_FLAGS_MATCH 4 /* is the ack in response to a match */
#define MCA_PML_DR_HDR_FLAGS_RNDV 8 /* is the ack in response to a rndv */
#define MCA_PML_DR_HDR_FLAGS_BUFFERED 16
/**
* Common hdr attributes - must be first element in each hdr type
*/

92
ompi/mca/pml/dr/pml_dr_recvfrag.c
Просмотреть файл

@ -51,6 +51,12 @@ OBJ_CLASS_INSTANCE(
NULL
);
static void mca_pml_dr_recv_frag_unmatched_ack(
mca_pml_dr_match_hdr_t* hdr,
ompi_proc_t* ompi_proc,
uint8_t flags,
uint8_t mask);
/*
* Release resources.
*/
@ -437,7 +443,7 @@ bool mca_pml_dr_recv_frag_match(
opal_list_t additional_matches;
ompi_proc_t* ompi_proc;
int rc;
uint32_t csum = 0;
/* communicator pointer */
comm_ptr=ompi_comm_lookup(hdr->hdr_ctx);
comm=(mca_pml_dr_comm_t *)comm_ptr->c_pml_comm;
@ -517,7 +523,8 @@ rematch:
OPAL_THREAD_UNLOCK(&comm->matching_lock);
return rc;
}
MCA_PML_DR_RECV_FRAG_INIT(frag,proc->ompi_proc,hdr,segments,num_segments,btl);
MCA_PML_DR_RECV_FRAG_INIT(frag,proc->ompi_proc,hdr,segments,num_segments,btl);
opal_list_append( &proc->unexpected_frags, (opal_list_item_t *)frag );
}
@ -536,6 +543,7 @@ rematch:
* This message comes after the next expected, so it
* is ahead of sequence. Save it for later.
*/
mca_pml_dr_recv_frag_t* frag;
MCA_PML_DR_RECV_FRAG_ALLOC(frag, rc);
if(OMPI_SUCCESS != rc) {
@ -551,7 +559,27 @@ rematch:
/* release matching lock before processing fragment */
if(match != NULL) {
mca_pml_dr_recv_request_progress(match,btl,segments,num_segments);
}
} else if (hdr->hdr_common.hdr_type == MCA_PML_DR_HDR_TYPE_MATCH) {
COMPUTE_SPECIFIC_CHECKSUM(segments->seg_addr.pval +
sizeof(mca_pml_dr_match_hdr_t),
segments->seg_len - sizeof(mca_pml_dr_match_hdr_t),
csum);
mca_pml_dr_recv_frag_unmatched_ack(hdr,
ompi_proc,
MCA_PML_DR_HDR_FLAGS_MATCH,
csum == hdr->hdr_csum ? 1 : 0);
} else {
COMPUTE_SPECIFIC_CHECKSUM(segments->seg_addr.pval +
sizeof(mca_pml_dr_rendezvous_hdr_t),
segments->seg_len - sizeof(mca_pml_dr_rendezvous_hdr_t),
csum);
mca_pml_dr_recv_frag_unmatched_ack(hdr,
ompi_proc,
MCA_PML_DR_HDR_FLAGS_BUFFERED,
csum == hdr->hdr_csum ? : 0);
}
if(additional_match) {
opal_list_item_t* item;
while(NULL != (item = opal_list_remove_first(&additional_matches))) {
@ -564,6 +592,64 @@ rematch:
}
static void mca_pml_dr_recv_frag_unmatched_ack(
mca_pml_dr_match_hdr_t* hdr,
ompi_proc_t* ompi_proc,
uint8_t flags,
uint8_t mask)
{
mca_bml_base_endpoint_t* bml_endpoint = NULL;
mca_btl_base_descriptor_t* des;
mca_bml_base_btl_t* bml_btl;
mca_pml_dr_recv_frag_t* frag;
mca_pml_dr_ack_hdr_t* ack;
int rc;
bml_endpoint = (mca_bml_base_endpoint_t*) ompi_proc->proc_pml;
bml_btl = mca_bml_base_btl_array_get_next(&bml_endpoint->btl_eager);
/* allocate descriptor */
MCA_PML_DR_DES_ALLOC(bml_btl, des, sizeof(mca_pml_dr_ack_hdr_t));
if(NULL == des) {
goto retry;
}
/* fill out header */
ack = (mca_pml_dr_ack_hdr_t*)des->des_src->seg_addr.pval;
ack->hdr_common.hdr_type = MCA_PML_DR_HDR_TYPE_ACK;
ack->hdr_common.hdr_flags = flags;
ack->hdr_vid = hdr->hdr_vid;
ack->hdr_vmask = mask;
ack->hdr_src_req = hdr->hdr_src_req;
assert(ack->hdr_src_req.pval);
ack->hdr_dst_req.pval = NULL;
ack->hdr_common.hdr_csum = opal_csum(ack, sizeof(mca_pml_dr_ack_hdr_t));
/* initialize descriptor */
des->des_flags |= MCA_BTL_DES_FLAGS_PRIORITY;
des->des_cbfunc = mca_pml_dr_ctl_completion;
rc = mca_bml_base_send(bml_btl, des, MCA_BTL_TAG_PML);
if(rc != OMPI_SUCCESS) {
mca_bml_base_free(bml_btl, des);
goto retry;
}
/* mca_pml_dr_comm_proc_set_acked(comm_proc, ack->hdr_vid); */
return;
/* queue request to retry later */
retry:
MCA_PML_DR_RECV_FRAG_ALLOC(frag,rc);
frag->hdr.hdr_match = *hdr;
frag->num_segments = 0;
opal_list_append(&mca_pml_dr.acks_pending, (opal_list_item_t*)frag);
}
/**
* Scan the list of frags that came in ahead of time to see if any
* can be processed at this time. If they can, try and match the

28
ompi/mca/pml/dr/pml_dr_recvreq.c
Просмотреть файл

@ -139,7 +139,8 @@ static void mca_pml_dr_ctl_completion(
static void mca_pml_dr_recv_request_matched(
mca_pml_dr_recv_request_t* recvreq,
mca_pml_dr_rendezvous_hdr_t* hdr,
mca_pml_dr_match_hdr_t* hdr,
uint8_t flags,
uint8_t mask)
{
ompi_proc_t* proc = recvreq->req_proc;
@ -156,7 +157,7 @@ static void mca_pml_dr_recv_request_matched(
/* if this hasn't been initialized yet - this is a synchronous send */
if(NULL == proc) {
ompi_proc_t *ompi_proc = ompi_comm_peer_lookup(
recvreq->req_recv.req_base.req_comm, hdr->hdr_match.hdr_src);
recvreq->req_recv.req_base.req_comm, hdr->hdr_src);
proc = recvreq->req_proc = ompi_proc;
}
bml_endpoint = (mca_bml_base_endpoint_t*) proc->proc_pml;
@ -171,10 +172,10 @@ static void mca_pml_dr_recv_request_matched(
/* fill out header */
ack = (mca_pml_dr_ack_hdr_t*)des->des_src->seg_addr.pval;
ack->hdr_common.hdr_type = MCA_PML_DR_HDR_TYPE_ACK;
ack->hdr_common.hdr_flags = MCA_PML_DR_HDR_FLAGS_MATCH;
ack->hdr_vid = hdr->hdr_match.hdr_vid;
ack->hdr_common.hdr_flags = flags;
ack->hdr_vid = hdr->hdr_vid;
ack->hdr_vmask = mask;
ack->hdr_src_req = hdr->hdr_match.hdr_src_req;
ack->hdr_src_req = hdr->hdr_src_req;
assert(ack->hdr_src_req.pval);
ack->hdr_dst_req.pval = recvreq;
ack->hdr_common.hdr_csum = opal_csum(ack, sizeof(mca_pml_dr_ack_hdr_t));
@ -189,20 +190,21 @@ static void mca_pml_dr_recv_request_matched(
goto retry;
}
mca_pml_dr_comm_proc_set_acked(comm_proc, ack->hdr_vid);
mca_pml_dr_comm_proc_set_acked(comm_proc, ack->hdr_vid);
return;
/* queue request to retry later */
retry:
MCA_PML_DR_RECV_FRAG_ALLOC(frag,rc);
frag->hdr.hdr_rndv = *hdr;
frag->hdr.hdr_match = *hdr;
frag->num_segments = 0;
frag->request = recvreq;
opal_list_append(&mca_pml_dr.acks_pending, (opal_list_item_t*)frag);
}
/*
* Generate an ack w/ the current vfrag status.
*/
@ -235,7 +237,7 @@ static void mca_pml_dr_recv_request_vfrag_ack(
ack = (mca_pml_dr_ack_hdr_t*)des->des_src->seg_addr.pval;
ack->hdr_common.hdr_type = MCA_PML_DR_HDR_TYPE_ACK;
ack->hdr_common.hdr_flags = 0;
ack->hdr_common.hdr_flags = MCA_PML_DR_HDR_FLAGS_VFRAG;
ack->hdr_vid = vfrag->vf_id;
ack->hdr_vmask = vfrag->vf_ack;
ack->hdr_src_req = recvreq->req_vfrag0.vf_send;
@ -250,7 +252,7 @@ static void mca_pml_dr_recv_request_vfrag_ack(
if(rc != OMPI_SUCCESS) {
mca_bml_base_free(bml_btl, des);
}
mca_pml_dr_comm_proc_set_acked(comm_proc, ack->hdr_vid);
mca_pml_dr_comm_proc_set_acked(comm_proc, ack->hdr_vid);
}
@ -293,6 +295,7 @@ void mca_pml_dr_recv_request_progress(
bytes_received,
bytes_delivered,
csum);
break;
case MCA_PML_DR_HDR_TYPE_RNDV:
@ -315,9 +318,10 @@ void mca_pml_dr_recv_request_progress(
if(csum != hdr->hdr_match.hdr_csum) {
assert(0);
}
mca_pml_dr_recv_request_matched(recvreq, &hdr->hdr_rndv,
(csum == hdr->hdr_match.hdr_csum));
mca_pml_dr_recv_request_matched(recvreq, &hdr->hdr_match,
MCA_PML_DR_HDR_FLAGS_RNDV,
csum == hdr->hdr_match.hdr_csum ? 1 : 0);
break;
case MCA_PML_DR_HDR_TYPE_FRAG:

126
ompi/mca/pml/dr/pml_dr_sendreq.c
Просмотреть файл

@ -131,9 +131,7 @@ void mca_pml_dr_match_completion_cache(
/* signal request completion */
OPAL_THREAD_LOCK(&ompi_request_lock);
if(sendreq->req_num_acks == sendreq->req_num_vfrags) {
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
}
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
OPAL_THREAD_UNLOCK(&ompi_request_lock);
}
@ -157,15 +155,14 @@ void mca_pml_dr_match_completion_free(
orte_errmgr.abort();
}
/* we don't want to free the descriptor until we get a postive ACK */
/* free the descriptor */
mca_bml_base_free( bml_btl, descriptor );
/* mca_bml_base_free( bml_btl, descriptor ); */
/* signal request completion */
OPAL_THREAD_LOCK(&ompi_request_lock);
if(sendreq->req_num_acks == sendreq->req_num_vfrags) {
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
}
OPAL_THREAD_UNLOCK(&ompi_request_lock);
/* OPAL_THREAD_LOCK(&ompi_request_lock); */
/* MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq); */
/* OPAL_THREAD_UNLOCK(&ompi_request_lock); */
}
/*
@ -187,15 +184,12 @@ static void mca_pml_dr_rndv_completion(
orte_errmgr.abort();
}
/* count bytes of user data actually delivered */
OPAL_THREAD_LOCK(&ompi_request_lock);
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,descriptor,sizeof(mca_pml_dr_rendezvous_hdr_t));
OPAL_THREAD_UNLOCK(&ompi_request_lock);
/* return the descriptor */
mca_bml_base_free(bml_btl, descriptor);
/* mca_bml_base_free(bml_btl, descriptor); */
/* advance the request */
MCA_PML_DR_SEND_REQUEST_ADVANCE(sendreq);
/* check for pending requests */
@ -244,11 +238,8 @@ static void mca_pml_dr_frag_completion(
OPAL_THREAD_LOCK(&ompi_request_lock);
/* count bytes of user data actually delivered */
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,descriptor,sizeof(mca_pml_dr_frag_hdr_t));
if (OPAL_THREAD_ADD_SIZE_T(&sendreq->req_pipeline_depth,-1) == 0 &&
sendreq->req_bytes_delivered == sendreq->req_send.req_bytes_packed &&
sendreq->req_num_acks == sendreq->req_num_vfrags) {
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
(sendreq->req_send.req_bytes_packed - sendreq->req_send_offset) == 0) {
schedule = false;
} else {
schedule = true;
@ -291,8 +282,9 @@ int mca_pml_dr_send_request_start_buffered(
if(NULL == descriptor) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
sendreq->descriptor = descriptor; /* hang on to this for later */
segment = descriptor->des_src;
/* pack the data into the BTL supplied buffer */
iov.iov_base = (void*)((unsigned char*)segment->seg_addr.pval +
sizeof(mca_pml_dr_rendezvous_hdr_t));
@ -315,6 +307,7 @@ int mca_pml_dr_send_request_start_buffered(
sendreq->req_send_offset = max_data;
sendreq->req_vfrag0.vf_size = max_data;
sendreq->req_vfrag0.bml_btl = bml_btl;
sendreq->req_vfrag0.vf_mask = 1;
descriptor->des_cbfunc = mca_pml_dr_rndv_completion;
descriptor->des_flags |= MCA_BTL_DES_FLAGS_PRIORITY;
@ -394,12 +387,14 @@ int mca_pml_dr_send_request_start_copy(
size_t max_data;
int32_t free_after;
int rc;
/* allocate descriptor */
mca_bml_base_alloc(bml_btl, &descriptor, sizeof(mca_pml_dr_match_hdr_t) + size);
if(NULL == descriptor) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
}
sendreq->descriptor = descriptor; /* hang on to this for later */
segment = descriptor->des_src;
/* pack the data into the supplied buffer */
@ -438,6 +433,8 @@ int mca_pml_dr_send_request_start_copy(
sendreq->req_send_offset = max_data;
sendreq->req_vfrag0.vf_size = max_data;
sendreq->req_vfrag0.bml_btl = bml_btl;
sendreq->req_vfrag0.vf_mask = 1;
/* short message */
descriptor->des_cbfunc = mca_pml_dr_match_completion_free;
descriptor->des_flags |= MCA_BTL_DES_FLAGS_PRIORITY;
@ -452,7 +449,8 @@ int mca_pml_dr_send_request_start_copy(
rc = mca_bml_base_send(bml_btl, descriptor, MCA_BTL_TAG_PML);
if(OMPI_SUCCESS != rc) {
mca_bml_base_free(bml_btl, descriptor );
}
}
return rc;
}
@ -482,6 +480,7 @@ int mca_pml_dr_send_request_start_prepare(
if(NULL == descriptor) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
sendreq->descriptor = descriptor; /* hang on to this for later */
segment = descriptor->des_src;
/* build match header */
@ -532,6 +531,7 @@ int mca_pml_dr_send_request_start_rndv(
mca_pml_dr_hdr_t* hdr;
int rc;
/* prepare descriptor */
if(size == 0) {
mca_bml_base_alloc(
@ -552,8 +552,9 @@ int mca_pml_dr_send_request_start_rndv(
if(NULL == des) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
sendreq->descriptor = des; /* hang on to this for later */
segment = des->des_src;
/* build hdr */
hdr = (mca_pml_dr_hdr_t*)segment->seg_addr.pval;
hdr->hdr_common.hdr_flags = flags;
@ -575,6 +576,7 @@ int mca_pml_dr_send_request_start_rndv(
sendreq->req_send_offset = size;
sendreq->req_vfrag0.vf_size = size;
sendreq->req_vfrag0.bml_btl = bml_btl;
sendreq->req_vfrag0.vf_mask = 1;
/* send */
rc = mca_bml_base_send(bml_btl, des, MCA_BTL_TAG_PML);
@ -632,7 +634,7 @@ int mca_pml_dr_send_request_schedule(mca_pml_dr_send_request_t* sendreq)
vfrag->bml_btl = bml_btl;
vfrag->sendreq = sendreq;
offset = 0;
sendreq->req_num_vfrags++;
} else { /* always schedule the vfrag accross the same btl */
bml_btl = vfrag->bml_btl;
}
@ -813,41 +815,91 @@ void mca_pml_dr_send_request_acked(
mca_pml_dr_ack_hdr_t* ack)
{
mca_pml_dr_vfrag_t* vfrag;
mca_btl_base_descriptor_t* descriptor;
assert(sendreq);
descriptor = sendreq->descriptor;
MCA_PML_DR_SEND_REQUEST_VFRAG_PENDING(sendreq, ack, vfrag);
/* MCA_PML_DR_VFRAG_ACK_STOP(vfrag); */
if(ack->hdr_common.hdr_flags & MCA_PML_DR_HDR_FLAGS_MATCH) {
if(ack->hdr_vmask) {
if(ack->hdr_common.hdr_flags & MCA_PML_DR_HDR_FLAGS_BUFFERED) {
if((ack->hdr_vmask & vfrag->vf_mask) == vfrag->vf_mask) {
OPAL_THREAD_LOCK(&ompi_request_lock);
mca_bml_base_free( (mca_bml_base_btl_t*) descriptor->des_context, descriptor );
sendreq->descriptor = NULL;
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,
vfrag,
sizeof(mca_pml_dr_rendezvous_hdr_t));
OPAL_THREAD_UNLOCK(&ompi_request_lock);
} else {
assert(0);
}
} else if(ack->hdr_common.hdr_flags & MCA_PML_DR_HDR_FLAGS_RNDV) {
if((ack->hdr_vmask & vfrag->vf_mask) == vfrag->vf_mask) {
OPAL_THREAD_LOCK(&ompi_request_lock);
if(sendreq->descriptor) {
mca_bml_base_free( (mca_bml_base_btl_t*) descriptor->des_context, descriptor );
sendreq->descriptor = NULL;
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,
vfrag,
sizeof(mca_pml_dr_rendezvous_hdr_t));
}
OPAL_THREAD_UNLOCK(&ompi_request_lock);
sendreq->req_vfrag0.vf_recv = ack->hdr_dst_req;
MCA_PML_DR_SEND_REQUEST_ADVANCE(sendreq);
/* we know that we have more data, otherwise it would have been
an MCA_PML_DR_HDR_FLAGS_MATCH */
if(sendreq->req_bytes_delivered == sendreq->req_send.req_bytes_packed){
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
} else {
MCA_PML_DR_SEND_REQUEST_ADVANCE(sendreq);
}
}
else {
vfrag->vf_idx = 0;
vfrag->vf_ack = 0;
assert(0);
/* retransmit missing fragments */
/* OPAL_THREAD_LOCK(&sendreq->req_mutex); */
/* opal_list_append(&sendreq->req_retrans, (opal_list_item_t*)vfrag); */
/* OPAL_THREAD_UNLOCK(&sendreq->req_mutex); */
/* mca_pml_dr_send_request_schedule(sendreq); */
}
} else {
} else if(ack->hdr_common.hdr_flags & MCA_PML_DR_HDR_FLAGS_MATCH) {
if((ack->hdr_vmask & vfrag->vf_mask) == vfrag->vf_mask) {
mca_bml_base_free( (mca_bml_base_btl_t*) descriptor->des_context, descriptor );
sendreq->descriptor = NULL;
OPAL_THREAD_LOCK(&ompi_request_lock);
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,
vfrag,
sizeof(mca_pml_dr_match_hdr_t));
OPAL_THREAD_UNLOCK(&ompi_request_lock);
/* everything was in the match, mark complete */
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
}
} else if(ack->hdr_common.hdr_flags & MCA_PML_DR_HDR_FLAGS_VFRAG){
/* add in acknowledged fragments */
vfrag->vf_ack |= ack->hdr_vmask;
/* have all fragments w/in this vfrag been acked? */
if((vfrag->vf_ack & vfrag->vf_mask) == vfrag->vf_mask) {
OPAL_THREAD_LOCK(&ompi_request_lock);
MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq,
vfrag,
sizeof(mca_pml_dr_frag_hdr_t));
OPAL_THREAD_UNLOCK(&ompi_request_lock);
/* return vfrag */
if (vfrag != &sendreq->req_vfrag0) {
MCA_PML_DR_VFRAG_RETURN(vfrag);
}
/* are we done with this request */
/* are we done with this request ? */
OPAL_THREAD_LOCK(&ompi_request_lock);
sendreq->req_num_acks++;
if(sendreq->req_bytes_delivered == sendreq->req_send.req_bytes_packed &&
sendreq->req_num_acks == sendreq->req_num_vfrags) {
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
if(sendreq->req_bytes_delivered == sendreq->req_send.req_bytes_packed){
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq);
}
OPAL_THREAD_UNLOCK(&ompi_request_lock);

65
ompi/mca/pml/dr/pml_dr_sendreq.h
Просмотреть файл

@ -39,7 +39,6 @@
extern "C" {
#endif
struct mca_pml_dr_send_request_t {
mca_pml_base_send_request_t req_send;
/* ompi_proc_t* req_proc; */
@ -60,9 +59,7 @@ struct mca_pml_dr_send_request_t {
opal_list_t req_pending;
opal_list_t req_retrans;
opal_mutex_t req_mutex;
size_t req_num_acks;
size_t req_num_vfrags;
mca_btl_base_descriptor_t* descriptor; /* descriptor for first frag, retransmission */
};
typedef struct mca_pml_dr_send_request_t mca_pml_dr_send_request_t;
@ -146,8 +143,6 @@ do {
sendreq->req_endpoint = endpoint; \
sendreq->req_vfrag0.vf_id = OPAL_THREAD_ADD32(&proc->vfrag_id,1); \
sendreq->req_vfrag = &sendreq->req_vfrag0; \
sendreq->req_num_acks = 0; \
sendreq->req_num_vfrags = 0; \
\
/* select a btl */ \
bml_btl = mca_bml_base_btl_array_get_next(&endpoint->btl_eager); \
@ -161,48 +156,7 @@ do {
rc = mca_pml_dr_send_request_start_copy(sendreq, bml_btl, size); \
break; \
default: \
if(size == 0) { \
mca_btl_base_descriptor_t* descriptor; \
mca_btl_base_segment_t* segment; \
mca_pml_dr_hdr_t* hdr; \
/* allocate a descriptor */ \
MCA_PML_DR_DES_ALLOC(bml_btl, descriptor, sizeof(mca_pml_dr_match_hdr_t)); \
if(NULL == descriptor) { \
return OMPI_ERR_OUT_OF_RESOURCE; \
} \
segment = descriptor->des_src; \
/* setup vfrag */ \
sendreq->req_vfrag0.vf_size = 0; \
\
/* build hdr */ \
hdr = (mca_pml_dr_hdr_t*)segment->seg_addr.pval; \
hdr->hdr_common.hdr_flags = 0; \
hdr->hdr_common.hdr_type = MCA_PML_DR_HDR_TYPE_MATCH; \
hdr->hdr_match.hdr_ctx = sendreq->req_send.req_base.req_comm->c_contextid; \
hdr->hdr_match.hdr_src = sendreq->req_send.req_base.req_comm->c_my_rank; \
hdr->hdr_match.hdr_tag = sendreq->req_send.req_base.req_tag; \
hdr->hdr_match.hdr_seq = sendreq->req_send.req_base.req_sequence; \
hdr->hdr_match.hdr_src_req.pval = sendreq; \
hdr->hdr_match.hdr_vid = sendreq->req_vfrag0.vf_id; \
hdr->hdr_common.hdr_csum = opal_csum(hdr, sizeof(mca_pml_dr_match_hdr_t)); \
\
/* short message */ \
descriptor->des_cbfunc = mca_pml_dr_match_completion_cache; \
descriptor->des_flags |= MCA_BTL_DES_FLAGS_PRIORITY; \
descriptor->des_cbdata = sendreq; \
\
/* request is complete at mpi level */ \
OPAL_THREAD_LOCK(&ompi_request_lock); \
MCA_PML_DR_SEND_REQUEST_MPI_COMPLETE(sendreq); \
OPAL_THREAD_UNLOCK(&ompi_request_lock); \
\
/* send */ \
rc = mca_bml_base_send(bml_btl, descriptor, MCA_BTL_TAG_PML); \
if(OMPI_SUCCESS != rc) { \
mca_bml_base_free(bml_btl, descriptor ); \
} \
\
} else if (bml_btl->btl_flags & MCA_BTL_FLAGS_SEND_INPLACE) { \
if (bml_btl->btl_flags & MCA_BTL_FLAGS_SEND_INPLACE) { \
rc = mca_pml_dr_send_request_start_prepare(sendreq, bml_btl, size); \
} else { \
rc = mca_pml_dr_send_request_start_copy(sendreq, bml_btl, size); \
@ -282,9 +236,7 @@ do {
/* has an acknowledgment been received */ \
if(OPAL_THREAD_ADD32(&sendreq->req_state, 1) == 2) { \
OPAL_THREAD_LOCK(&ompi_request_lock); \
if(sendreq->req_bytes_delivered == sendreq->req_send.req_bytes_packed) { \
MCA_PML_DR_SEND_REQUEST_PML_COMPLETE(sendreq); \
} else { \
if(sendreq->req_send.req_bytes_packed - sendreq->req_send_offset) { \
schedule = true; \
} \
OPAL_THREAD_UNLOCK(&ompi_request_lock); \
@ -405,16 +357,9 @@ do { \
* Update bytes delivered on request based on supplied descriptor
*/
#define MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq, descriptor, hdrlen) \
#define MCA_PML_DR_SEND_REQUEST_SET_BYTES_DELIVERED(sendreq, vfrag, hdrlen) \
do { \
size_t i; \
mca_btl_base_segment_t* segments = descriptor->des_src; \
\
for(i=0; i<descriptor->des_src_cnt; i++) { \
sendreq->req_bytes_delivered += segments[i].seg_len; \
} \
sendreq->req_bytes_delivered -= hdrlen; \
\
sendreq->req_bytes_delivered += vfrag->vf_size; \
} while(0)
/*