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/SHARP: SHARP OOB fixes #746

Merged
merged 4 commits into from
Mar 15, 2023
Merged

TL/SHARP: SHARP OOB fixes #746

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
0f1fd69
TL/SHARP: SHARP OOB fixes
bureddy Feb 22, 2023
b6ebc8a
Merge branch 'master' into sharp-oob
bureddy Mar 13, 2023
b9313fa
Merge branch 'master' into sharp-oob
bureddy Mar 15, 2023
6a35346
Merge branch 'master' into sharp-oob
manjugv Mar 15, 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
2 changes: 2 additions & 0 deletions config/m4/sharp.m4
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,8 @@ AS_IF([test "x$with_sharp" != "xno"],
[
AC_SUBST(SHARP_CPPFLAGS, "-I$check_sharp_dir/include/ ")
AC_SUBST(SHARP_LDFLAGS, "-lsharp_coll -L$check_sharp_dir/lib")
AC_CHECK_DECLS([SHARP_COLL_HIDE_ERRORS], [], [], [[#include <sharp/api/sharp_coll.h>]])
AC_CHECK_DECLS([SHARP_COLL_DISABLE_LAZY_GROUP_RESOURCE_ALLOC], [], [], [[#include <sharp/api/sharp_coll.h>]])
],
[
AS_IF([test "x$with_sharp" != "xguess"],
Expand Down
9 changes: 8 additions & 1 deletion src/components/tl/sharp/tl_sharp.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ static ucc_config_field_t ucc_tl_sharp_context_config_table[] = {
{"", "", NULL, ucc_offsetof(ucc_tl_sharp_context_config_t, super),
UCC_CONFIG_TYPE_TABLE(ucc_tl_context_config_table)},

{"DEVICES", "mlx5_0:1",
{"DEVICES", "",
Copy link
Contributor

Choose a reason for hiding this comment

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

how device will be selected for previous SHARP releases?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

it was sharplib user's responsibility to specify the device.. there is no default device selection

Copy link
Contributor

Choose a reason for hiding this comment

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

right, this is my question, if one builds UCC 1.2 with SHARP 3.0 how it will work without providing device?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

the default selection was fixed in libsharp in the last HPCX release i. it will fail if we use older than that. that should be ok

"SHARP device list",
ucc_offsetof(ucc_tl_sharp_context_config_t, dev_list),
UCC_CONFIG_TYPE_STRING},
Expand All @@ -55,6 +55,13 @@ static ucc_config_field_t ucc_tl_sharp_context_config_table[] = {
ucc_offsetof(ucc_tl_sharp_context_config_t, context_per_team),
UCC_CONFIG_TYPE_BOOL},

#if HAVE_DECL_SHARP_COLL_DISABLE_LAZY_GROUP_RESOURCE_ALLOC
{"ENABLE_LAZY_GROUP_ALLOC", "n",
"Enable lazy group resource allocation",
ucc_offsetof(ucc_tl_sharp_context_config_t, enable_lazy_group_alloc),
UCC_CONFIG_TYPE_BOOL},
#endif

{"RAND_SEED", "0",
"Seed for random sharp job ID. (0 - use default).",
ucc_offsetof(ucc_tl_sharp_context_config_t, rand_seed),
Expand Down
1 change: 1 addition & 0 deletions src/components/tl/sharp/tl_sharp.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@ typedef struct ucc_tl_sharp_context_config {
unsigned int rand_seed;
unsigned int uprogress_num_polls;
int context_per_team;
int enable_lazy_group_alloc;
} ucc_tl_sharp_context_config_t;

typedef struct ucc_tl_sharp_lib {
Expand Down
11 changes: 11 additions & 0 deletions src/components/tl/sharp/tl_sharp_context.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -300,6 +300,17 @@ ucc_status_t ucc_tl_sharp_context_init(ucc_tl_sharp_context_t *sharp_ctx,
init_spec.config = sharp_coll_default_config;
init_spec.config.user_progress_num_polls = sharp_ctx->cfg.uprogress_num_polls;
init_spec.config.ib_dev_list = sharp_ctx->cfg.dev_list;
#if HAVE_DECL_SHARP_COLL_HIDE_ERRORS
if (lib->super.super.log_component.log_level < UCC_LOG_LEVEL_DEBUG) {
init_spec.config.flags |= SHARP_COLL_HIDE_ERRORS;
}
#endif
#if HAVE_DECL_SHARP_COLL_DISABLE_LAZY_GROUP_RESOURCE_ALLOC
if(!sharp_ctx->cfg.enable_lazy_group_alloc) {
init_spec.config.flags |= SHARP_COLL_DISABLE_LAZY_GROUP_RESOURCE_ALLOC;
}
#endif
Copy link
Contributor

Choose a reason for hiding this comment

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

print warning if disabling lazy group alloc is not supported by sharp and user has changed default value

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

maybe we can expose the user option also only when flag is there


init_spec.job_id = ((getpid() ^ pthread_self())
^ rand_r(&sharp_ctx->cfg.rand_seed));
init_spec.enable_thread_support =
Expand Down