Skip to content

Oshmem multiple contexts #6492

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

Merged
merged 8 commits into from
Mar 22, 2019
Merged

Oshmem multiple contexts #6492

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
b00209e
Revert "OMPI/OSHMEM: bug-fix: store mkeys for each oshmem ctx."
xinzhao3 Mar 8, 2019
79ba752
ompi/oshmem/spml/ucx: fix eps destroy in shmem_ctx_destroy().
xinzhao3 Mar 5, 2019
289595e
OMPI/OSHMEM: bug-fix: store mkeys for each oshmem ctx.
xinzhao3 Jan 4, 2019
9a06000
ompi/oshmem/spml/ucx: let shmem_finalize to clean up any ctx left
xinzhao3 Mar 7, 2019
48033ac
ompi/oshmem: add spml_context back to sshmem_type in memheap, to keep…
xinzhao3 Mar 7, 2019
e1c1ab0
ompi/oshmem/spml/ucx: defer clean up shmem_ctx to shmem_finalize
xinzhao3 Mar 8, 2019
e041400
ompi/oshmem/spml/ucx:delete oob path of getting rkeys in spml ucx
xinzhao3 Mar 8, 2019
9c3d00b
ompi/oshmem/spml/ucx: use lockfree array to optimize spml_ucx_progres…
xinzhao3 Mar 13, 2019
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
22 changes: 15 additions & 7 deletions opal/mca/common/ucx/common_ucx.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,6 +153,11 @@ void opal_common_ucx_mca_proc_added(void)
#endif
}

OPAL_DECLSPEC int opal_common_ucx_mca_pmix_fence_nb(int *fenced)
{
return opal_pmix.fence_nb(NULL, 0, opal_common_ucx_mca_fence_complete_cb, (void *)fenced);
}

OPAL_DECLSPEC int opal_common_ucx_mca_pmix_fence(ucp_worker_h worker)
{
volatile int fenced = 0;
Expand Down Expand Up @@ -182,9 +187,8 @@ static void opal_common_ucx_wait_all_requests(void **reqs, int count, ucp_worker
}
}

OPAL_DECLSPEC int opal_common_ucx_del_procs(opal_common_ucx_del_proc_t *procs, size_t count,
size_t my_rank, size_t max_disconnect, ucp_worker_h worker)
{
OPAL_DECLSPEC int opal_common_ucx_del_procs_nofence(opal_common_ucx_del_proc_t *procs, size_t count,
size_t my_rank, size_t max_disconnect, ucp_worker_h worker) {
size_t num_reqs;
size_t max_reqs;
void *dreq, **dreqs;
Expand Down Expand Up @@ -232,10 +236,14 @@ OPAL_DECLSPEC int opal_common_ucx_del_procs(opal_common_ucx_del_proc_t *procs, s
opal_common_ucx_wait_all_requests(dreqs, num_reqs, worker);
free(dreqs);

if (OPAL_SUCCESS != (ret = opal_common_ucx_mca_pmix_fence(worker))) {
return ret;
}

return OPAL_SUCCESS;
}

OPAL_DECLSPEC int opal_common_ucx_del_procs(opal_common_ucx_del_proc_t *procs, size_t count,
size_t my_rank, size_t max_disconnect, ucp_worker_h worker)
{
opal_common_ucx_del_procs_nofence(procs, count, my_rank, max_disconnect, worker);

return opal_common_ucx_mca_pmix_fence(worker);
}

3 changes: 3 additions & 0 deletions opal/mca/common/ucx/common_ucx.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,8 +105,11 @@ OPAL_DECLSPEC void opal_common_ucx_mca_deregister(void);
OPAL_DECLSPEC void opal_common_ucx_mca_proc_added(void);
OPAL_DECLSPEC void opal_common_ucx_empty_complete_cb(void *request, ucs_status_t status);
OPAL_DECLSPEC int opal_common_ucx_mca_pmix_fence(ucp_worker_h worker);
OPAL_DECLSPEC int opal_common_ucx_mca_pmix_fence_nb(int *fenced);
OPAL_DECLSPEC int opal_common_ucx_del_procs(opal_common_ucx_del_proc_t *procs, size_t count,
size_t my_rank, size_t max_disconnect, ucp_worker_h worker);
OPAL_DECLSPEC int opal_common_ucx_del_procs_nofence(opal_common_ucx_del_proc_t *procs, size_t count,
size_t my_rank, size_t max_disconnect, ucp_worker_h worker);
OPAL_DECLSPEC void opal_common_ucx_mca_var_register(const mca_base_component_t *component);

static inline
Expand Down
10 changes: 5 additions & 5 deletions oshmem/mca/spml/ikrit/spml_ikrit.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,7 +149,7 @@ int mca_spml_ikrit_put_simple(void* dst_addr,
void* src_addr,
int dst);

static void mca_spml_ikrit_cache_mkeys(shmem_ctx_t ctx, sshmem_mkey_t *, uint32_t seg, int remote_pe, int tr_id);
static void mca_spml_ikrit_cache_mkeys(sshmem_mkey_t *, uint32_t seg, int remote_pe, int tr_id);

static mxm_mem_key_t *mca_spml_ikrit_get_mkey_slow(int pe, void *va, int ptl_id, void **rva);

Expand Down Expand Up @@ -185,7 +185,7 @@ mca_spml_ikrit_t mca_spml_ikrit = {
mca_spml_ikrit_get_mkey_slow
};

static void mca_spml_ikrit_cache_mkeys(shmem_ctx_t ctx, sshmem_mkey_t *mkey, uint32_t seg, int dst_pe, int tr_id)
static void mca_spml_ikrit_cache_mkeys(sshmem_mkey_t *mkey, uint32_t seg, int dst_pe, int tr_id)
{
mxm_peer_t *peer;

Expand Down Expand Up @@ -504,7 +504,7 @@ sshmem_mkey_t *mca_spml_ikrit_register(void* addr,
my_rank, i, addr, (unsigned long long)size,
mca_spml_base_mkey2str(&mkeys[i]));

mca_spml_ikrit_cache_mkeys(NULL, &mkeys[i], memheap_find_segnum(addr), my_rank, i);
mca_spml_ikrit_cache_mkeys(&mkeys[i], memheap_find_segnum(addr), my_rank, i);
}
*count = MXM_PTL_LAST;

Expand Down Expand Up @@ -548,7 +548,7 @@ int mca_spml_ikrit_deregister(sshmem_mkey_t *mkeys)

}

int mca_spml_ikrit_oob_get_mkeys(shmem_ctx_t ctx, int pe, uint32_t seg, sshmem_mkey_t *mkeys)
int mca_spml_ikrit_oob_get_mkeys(int pe, uint32_t seg, sshmem_mkey_t *mkeys)
{
int ptl;

Expand All @@ -567,7 +567,7 @@ int mca_spml_ikrit_oob_get_mkeys(shmem_ctx_t ctx, int pe, uint32_t seg, sshmem_m
mkeys[ptl].len = 0;
mkeys[ptl].va_base = mca_memheap_seg2base_va(seg);
mkeys[ptl].u.key = MAP_SEGMENT_SHM_INVALID;
mca_spml_ikrit_cache_mkeys(NULL, &mkeys[ptl], seg, pe, ptl);
mca_spml_ikrit_cache_mkeys(&mkeys[ptl], seg, pe, ptl);
return OSHMEM_SUCCESS;
}

Expand Down
2 changes: 1 addition & 1 deletion oshmem/mca/spml/ikrit/spml_ikrit.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -182,7 +182,7 @@ extern sshmem_mkey_t *mca_spml_ikrit_register(void* addr,
uint64_t shmid,
int *count);
extern int mca_spml_ikrit_deregister(sshmem_mkey_t *mkeys);
extern int mca_spml_ikrit_oob_get_mkeys(shmem_ctx_t ctx, int pe,
extern int mca_spml_ikrit_oob_get_mkeys(int pe,
uint32_t segno,
sshmem_mkey_t *mkeys);

Expand Down
Loading