Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Modifies converse schedeuler to prioritize NodeGroup messages #3676

Open
wants to merge 16 commits into
base: main
Choose a base branch
from
+31 −17
Open

Modifies converse schedeuler to prioritize NodeGroup messages #3676

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
f436743
Modifies converse schedeuler's getNextMessage so nodeGroup messages c…
Dec 3, 2022
800aa6b
fixed syntax error (missing right parent).
lvkale Dec 4, 2022
ed823ff
Merge branch 'main' into nodegrpPrio
ZwFink Apr 9, 2024
62b0569
Addressing changes
ZwFink Apr 9, 2024
028cb95
Merge remote-tracking branch 'origin/main' into nodegrpPrio
ZwFink Apr 11, 2024
7b3b414
Merge branch 'main' into nodegrpPrio
ZwFink Apr 29, 2024
df3e677
Merge branch 'main' into nodegrpPrio
ZwFink May 2, 2024
0f31dd0
initialize scheduler iter
ZwFink May 3, 2024
5f3e5b2
remove extra whitespace
ZwFink May 3, 2024
7342221
Remove text after the parenthetical
ZwFink May 3, 2024
f39d307
Merge branch 'main' into nodegrpPrio
ZwFink May 28, 2024
22fdc83
Remove long comment
ZwFink May 28, 2024
83e904f
Re-organize, fix whitespace
ZwFink May 28, 2024
f3eb123
fix merge conflict
ZwFink May 28, 2024
bc5f37c
Normalize indentation
ZwFink May 29, 2024
3b0ebbd
More whitespace
ZwFink May 31, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 29 additions & 17 deletions src/conv-core/convcore.C
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1657,6 +1657,8 @@ void CsdSchedulerState_new(CsdSchedulerState_t *s)
#if CMK_NODE_QUEUE_AVAILABLE
s->nodeQ=CsvAccess(CsdNodeQueue);
s->nodeLock=CsvAccess(CsdNodeQueueLock);
ZwFink marked this conversation as resolved.
Show resolved Hide resolved
s->nodeGrpFreq=4;
s->iter = 0;
#endif
#if CMK_GRID_QUEUE_AVAILABLE
s->gridQ=CpvAccess(CsdGridQueue);
Expand Down Expand Up @@ -1720,29 +1722,39 @@ void CsdSchedulerState_new(CsdSchedulerState_t *s)
*/
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The long expository comment preceeding the CsdNextMessage definition should be updated to match the new behavior, and any other behavior that it doesn't describe, such as csdLocalMax querying the PE onnode FIFO (localQ) and scheduler queues first.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Even better, break up the exposition so that the explanations are adjacent to the code they are intended to describe.

void *CsdNextMessage(CsdSchedulerState_t *s) {
void *msg;
if((*(s->localCounter))-- >0)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The new code should be below the localCounter branch, right above the CmiGetNonLocal() call. The default CsdLocalMax is 0 so normally it won't matter, but if the user says to prioritize local messages then they should come first.

{
/* This avoids a race condition with migration detected by megatest*/
msg=CdsFifo_Dequeue(s->localQ);
if (msg!=NULL)
{

if ((*(s->localCounter))-- > 0) {
/* This avoids a race condition with migration detected by megatest*/
msg = CdsFifo_Dequeue(s->localQ);
if (msg != NULL) {
#if CMI_QD
CpvAccess(cQdState)->mProcessed++;
CpvAccess(cQdState)->mProcessed++;
#endif
return msg;
}
CqsDequeue(s->schedQ,(void **)&msg);
if (msg!=NULL) return msg;
}

return msg;
}
CqsDequeue(s->schedQ, (void**)&msg);
if (msg != NULL)
return msg;
}

#if CMK_NODE_QUEUE_AVAILABLE
s->iter++;
// we use nodeGrpFreq == 0 to mean
// don't check NodeQ with high priority
if (s->nodeGrpFreq && (0 == (s->iter % s->nodeGrpFreq))) {
msg = CmiGetNonLocalNodeQ();
if (NULL != msg) return msg;
}
#endif

*(s->localCounter)=CsdLocalMax;
if ( NULL!=(msg=CmiGetNonLocal()) ||
if ( NULL!=(msg=CmiGetNonLocal()) ||
NULL!=(msg=CdsFifo_Dequeue(s->localQ)) ) {
#if CMI_QD
CpvAccess(cQdState)->mProcessed++;
CpvAccess(cQdState)->mProcessed++;
#endif
return msg;
}
return msg;
}
#if CMK_GRID_QUEUE_AVAILABLE
/*#warning "CsdNextMessage: CMK_GRID_QUEUE_AVAILABLE" */
CqsDequeue (s->gridQ, (void **) &msg);
Expand Down
2 changes: 2 additions & 0 deletions src/conv-core/converse.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1129,6 +1129,8 @@ typedef struct {
void *localQ;
Queue nodeQ;
Queue schedQ;
unsigned short iter;
unsigned short nodeGrpFreq;
int *localCounter;
#if CMK_OBJECT_QUEUE_AVAILABLE
Queue objQ;
Expand Down
Loading