Skip to content

ch4/wait: Optimize waitall when all requests are local #7620

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

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

ch4/wait: Optimize waitall when all requests are local #7620

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8b4b5c9
ch4/progress: Improvements to global progress
raffenet Oct 8, 2025
527a225
ch4: Initialize request locality at creation
raffenet Oct 13, 2025
d1f41d8
ch4/ipc: Add missing locality setting
raffenet Oct 13, 2025
67241fe
ch4/wait: Optimize progress for local requests
raffenet Oct 10, 2025
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
1 change: 1 addition & 0 deletions src/mpid/ch4/include/mpidpost.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@ MPL_STATIC_INLINE_PREFIX void MPID_Request_create_hook(MPIR_Request * req)
req->dev.type = MPIDI_REQ_TYPE_NONE;
#ifndef MPIDI_CH4_DIRECT_NETMOD
req->dev.anysrc_partner = NULL;
req->dev.is_local = 0;
#endif

MPIR_FUNC_EXIT;
Expand Down
1 change: 1 addition & 0 deletions src/mpid/ch4/shm/ipc/src/ipc_p2p.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,6 +203,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_IPCI_send_lmt(const void *buf, MPI_Aint count
MPIDIG_REQUEST(sreq, u.ipc.peer_rank) = rank;
MPIDIG_REQUEST(sreq, u.ipc.peer_req) = NULL;
MPIDIG_REQUEST(sreq, u.ipc.src_dt_ptr) = NULL;
MPIDI_REQUEST_SET_LOCAL(sreq, 1, NULL);
MPIDI_SHM_REQUEST(sreq, ipc.ipc_type) = ipc_attr->ipc_type;

/* Allocate am_hdr and fill ipc_hdr */
Expand Down
9 changes: 4 additions & 5 deletions src/mpid/ch4/src/ch4_progress.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,8 +64,7 @@ extern MPL_TLS int no_progress_counter;

MPL_STATIC_INLINE_PREFIX int MPIDI_do_global_progress(void)
{
if ((MPIDI_global.n_vcis == 1 || !MPIR_CVAR_CH4_GLOBAL_PROGRESS) &&
!MPIR_CVAR_CH4_PROGRESS_THROTTLE) {
if (!MPIR_CVAR_CH4_GLOBAL_PROGRESS && !MPIR_CVAR_CH4_PROGRESS_THROTTLE) {
return 0;
} else {
global_vci_poll_count++;
Expand All @@ -88,7 +87,7 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_do_global_progress(void)
#ifdef MPIDI_CH4_DIRECT_NETMOD
#define MPIDI_PROGRESS(vci, is_global) \
do { \
if (state->flag & MPIDI_PROGRESS_NM && (is_global || !made_progress)) { \
if (is_global || (state->flag & MPIDI_PROGRESS_NM && !made_progress)) { \
MPIDI_THREAD_CS_ENTER_VCI_OPTIONAL(vci); \
mpi_errno = MPIDI_NM_progress(vci, &made_progress); \
MPIDI_THREAD_CS_EXIT_VCI_OPTIONAL(vci); \
Expand All @@ -99,13 +98,13 @@ MPL_STATIC_INLINE_PREFIX int MPIDI_do_global_progress(void)
#else
#define MPIDI_PROGRESS(vci, is_global) \
do { \
if (state->flag & MPIDI_PROGRESS_SHM && (is_global || !made_progress)) { \
if (is_global || (state->flag & MPIDI_PROGRESS_SHM && !made_progress)) { \
MPID_THREAD_CS_ENTER(VCI, MPIDI_VCI_LOCK(vci)); \
mpi_errno = MPIDI_SHM_progress(vci, &made_progress); \
MPID_THREAD_CS_EXIT(VCI, MPIDI_VCI_LOCK(vci)); \
MPIR_ERR_CHECK(mpi_errno); \
} \
if (state->flag & MPIDI_PROGRESS_NM && (is_global || !made_progress)) { \
if (is_global || (state->flag & MPIDI_PROGRESS_NM && !made_progress)) { \
MPIDI_THREAD_CS_ENTER_VCI_OPTIONAL(vci); \
mpi_errno = MPIDI_NM_progress(vci, &made_progress); \
MPIDI_THREAD_CS_EXIT_VCI_OPTIONAL(vci); \
Expand Down
18 changes: 16 additions & 2 deletions src/mpid/ch4/src/ch4_wait.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,18 +11,27 @@
MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci(MPIR_Request * req,
MPID_Progress_state * state)
{
state->flag = MPIDI_PROGRESS_ALL; /* TODO: check request is_local/anysource */
state->flag = MPIDI_PROGRESS_ALL;

int vci = MPIDI_Request_get_vci(req);

state->vci_count = 1;
state->vci[0] = vci;

#ifndef MPIDI_CH4_DIRECT_NETMOD
if (!req->dev.anysrc_partner && MPIDI_REQUEST(req, is_local)) {
state->flag &= ~MPIDI_PROGRESS_NM;
}
#endif
}

MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci_n(int n, MPIR_Request ** reqs,
MPID_Progress_state * state)
{
state->flag = MPIDI_PROGRESS_ALL; /* TODO: check request is_local/anysource */
state->flag = MPIDI_PROGRESS_ALL;
#ifndef MPIDI_CH4_DIRECT_NETMOD
state->flag &= ~MPIDI_PROGRESS_NM;
#endif

int idx = 0;
for (int i = 0; i < n; i++) {
Expand All @@ -45,6 +54,11 @@ MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci_n(int n, MPIR_Request ** re
if (!found) {
state->vci[idx++] = vci;
}
#ifndef MPIDI_CH4_DIRECT_NETMOD
if (!MPIDI_REQUEST(reqs[i], is_local) || reqs[i]->dev.anysrc_partner) {
state->flag |= MPIDI_PROGRESS_NM;
}
#endif
}
state->vci_count = idx;
}
Expand Down