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

TL/MLX5: a2a part2 -- full collective #802

Merged
merged 15 commits into from
Jul 20, 2023
Merged

TL/MLX5: a2a part2 -- full collective #802

Show file tree
Hide file tree
Changes from 10 commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
bcab357
TL/MLX5: a2a part 2 -- full coll
samnordmann Jun 28, 2023
c35498b
REVIEW: minor comments
samnordmann Jun 29, 2023
5818458
REVIEW: minor comments
samnordmann Jul 2, 2023
724b731
REVIEW: fix linter and align
samnordmann Jul 3, 2023
b132bba
TL/MLX5: minor changes
samnordmann Jul 9, 2023
57ccf4c
TL/MLX5: handle msgsize=0
samnordmann Jul 10, 2023
6796e52
TL/MLX5: fix ib_ctx cleanup
samnordmann Jul 10, 2023
b7a6cf7
TL/MLX5: fix team's dm alloc status
samnordmann Jul 10, 2023
0bd113e
TEST: minor change in gtest
samnordmann Jul 10, 2023
5918b09
CODESTYLE: clang format
samnordmann Jul 10, 2023
a8cda6d
TL/MLX5: fix multiple outstanding ops
samnordmann Jul 11, 2023
e06055b
TL/MLX5: cleanup
samnordmann Jul 11, 2023
8d37b72
CODESTYLE: clang-format
samnordmann Jul 11, 2023
33c1ee5
TL/MLX5: fix msgsize computation
samnordmann Jul 13, 2023
f61437c
Merge branch 'master' into tl_mlx5/a2a_part2
samnordmann Jul 20, 2023
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
12 changes: 7 additions & 5 deletions src/components/tl/mlx5/Makefile.am
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,11 +4,13 @@

if TL_MLX5_ENABLED

alltoall = \
alltoall/alltoall.h \
alltoall/alltoall.c \
alltoall/alltoall_mkeys.h \
alltoall/alltoall_mkeys.c
alltoall = \
alltoall/alltoall.h \
alltoall/alltoall.c \
alltoall/alltoall_mkeys.h \
alltoall/alltoall_mkeys.c \
alltoall/alltoall_inline.h \
alltoall/alltoall_coll.c

mcast = \
mcast/tl_mlx5_mcast_context.c \
Expand Down
7 changes: 4 additions & 3 deletions src/components/tl/mlx5/alltoall/alltoall.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,7 @@ static ucc_status_t build_rank_map(ucc_tl_mlx5_alltoall_t *a2a,
return UCC_OK;
}

ucc_status_t ucc_tl_mlx5_alltoall_init_start(ucc_tl_mlx5_team_t *team)
ucc_status_t ucc_tl_mlx5_team_alltoall_init_start(ucc_tl_mlx5_team_t *team)
{
ucc_tl_mlx5_context_t *ctx = UCC_TL_MLX5_TEAM_CTX(team);
ucc_tl_mlx5_alltoall_t *a2a = NULL;
Expand Down Expand Up @@ -204,7 +204,7 @@ static ucc_status_t ucc_tl_mlx5_alltoall_atomic_alloc(ucc_tl_mlx5_team_t *team)
#endif

if (!a2a->net.atomic.counters) {
tl_error(UCC_TL_TEAM_LIB(team),
tl_debug(UCC_TL_TEAM_LIB(team),
"failed to allocate %zd bytes for atomic counters array",
size);
return UCC_ERR_NO_MEMORY;
Expand Down Expand Up @@ -270,7 +270,8 @@ static ucc_status_t ucc_tl_mlx5_alltoall_barrier_alloc(ucc_tl_mlx5_team_t *team)
return UCC_OK;
}

ucc_status_t ucc_tl_mlx5_alltoall_init_progress(ucc_tl_mlx5_team_t *tl_team)
ucc_status_t
ucc_tl_mlx5_team_alltoall_init_progress(ucc_tl_mlx5_team_t *tl_team)
{
ucc_tl_mlx5_team_t *team = ucc_derived_of(tl_team,
ucc_tl_mlx5_team_t);
Expand Down
7 changes: 5 additions & 2 deletions src/components/tl/mlx5/alltoall/alltoall.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,8 +138,11 @@ typedef struct ucc_tl_mlx5_alltoall {
ucc_tl_mlx5_a2a_bcast_data_t bcast_data;
} ucc_tl_mlx5_alltoall_t;

ucc_status_t ucc_tl_mlx5_alltoall_init_start(ucc_tl_mlx5_team_t *team);
ucc_status_t ucc_tl_mlx5_alltoall_init_progress(ucc_tl_mlx5_team_t *team);
ucc_status_t ucc_tl_mlx5_team_alltoall_init_start(ucc_tl_mlx5_team_t *team);
ucc_status_t ucc_tl_mlx5_team_alltoall_init_progress(ucc_tl_mlx5_team_t *team);
ucc_status_t ucc_tl_mlx5_alltoall_init(ucc_base_coll_args_t *coll_args,
ucc_base_team_t * team,
ucc_coll_task_t ** task_h);
void ucc_tl_mlx5_alltoall_cleanup(ucc_tl_mlx5_team_t *team);

static inline ucc_tl_mlx5_alltoall_ctrl_t*
Expand Down
Loading