Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
S
Slurm
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Package Registry
Model registry
Operate
Environments
Terraform modules
Monitor
Incidents
Service Desk
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Terms and privacy
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
tud-zih-energy
Slurm
Commits
573fede2
Commit
573fede2
authored
18 years ago
by
Christopher J. Morrone
Browse files
Options
Downloads
Patches
Plain Diff
Code cleanup
parent
9dc5e55c
No related branches found
Branches containing commit
No related tags found
No related merge requests found
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
src/slaunch/slaunch.c
+27
-21
27 additions, 21 deletions
src/slaunch/slaunch.c
with
27 additions
and
21 deletions
src/slaunch/slaunch.c
+
27
−
21
View file @
573fede2
...
@@ -109,7 +109,7 @@ static void _run_srun_epilog (void);
...
@@ -109,7 +109,7 @@ static void _run_srun_epilog (void);
static int _run_srun_script (char *script);
static int _run_srun_script (char *script);
#endif
#endif
static
void
_setup_local_fds
(
slurm_step_io_fds_t
*
cio_fds
,
int
jobid
,
static
void
_setup_local_fds
(
slurm_step_io_fds_t
*
cio_fds
,
int
jobid
,
int
stepid
,
uint32_t
*
hostids
);
int
stepid
,
slurm_step_layout_t
*
step_layout
);
static
void
_task_start
(
launch_tasks_response_msg_t
*
msg
);
static
void
_task_start
(
launch_tasks_response_msg_t
*
msg
);
static
void
_task_finish
(
task_exit_msg_t
*
msg
);
static
void
_task_finish
(
task_exit_msg_t
*
msg
);
static
void
_task_state_struct_init
(
int
num_tasks
);
static
void
_task_state_struct_init
(
int
num_tasks
);
...
@@ -128,11 +128,9 @@ int slaunch(int argc, char **argv)
...
@@ -128,11 +128,9 @@ int slaunch(int argc, char **argv)
job_step_create_request_msg_t
step_req
;
job_step_create_request_msg_t
step_req
;
slurm_job_step_launch_t
params
;
slurm_job_step_launch_t
params
;
slurm_job_step_launch_callbacks_t
callbacks
;
slurm_job_step_launch_callbacks_t
callbacks
;
int
rc
;
uint32_t
*
hostids
=
NULL
;
log_init
(
xbasename
(
argv
[
0
]),
logopt
,
0
,
NULL
);
char
**
env
;
char
**
env
;
int
i
,
j
;
log_init
(
xbasename
(
argv
[
0
]),
logopt
,
0
,
NULL
);
xsignal
(
SIGQUIT
,
_ignore_signal
);
xsignal
(
SIGQUIT
,
_ignore_signal
);
xsignal
(
SIGPIPE
,
_ignore_signal
);
xsignal
(
SIGPIPE
,
_ignore_signal
);
...
@@ -246,26 +244,17 @@ int slaunch(int argc, char **argv)
...
@@ -246,26 +244,17 @@ int slaunch(int argc, char **argv)
params
.
task_prolog
=
opt
.
task_prolog
;
params
.
task_prolog
=
opt
.
task_prolog
;
params
.
task_epilog
=
opt
.
task_epilog
;
params
.
task_epilog
=
opt
.
task_epilog
;
/* set up the hostids */
hostids
=
xmalloc
(
sizeof
(
uint32_t
)
*
step_ctx
->
step_resp
->
step_layout
->
task_cnt
);
for
(
i
=
0
;
i
<
step_ctx
->
step_resp
->
step_layout
->
node_cnt
;
i
++
)
for
(
j
=
0
;
j
<
step_ctx
->
step_resp
->
step_layout
->
tasks
[
i
];
j
++
)
hostids
[
step_ctx
->
step_resp
->
step_layout
->
tids
[
i
][
j
]]
=
i
;
/* FIXME - don't peek into the step context, that's cheating! */
/* FIXME - don't peek into the step context, that's cheating! */
_setup_local_fds
(
&
params
.
local_fds
,
(
int
)
step_ctx
->
job_id
,
_setup_local_fds
(
&
params
.
local_fds
,
(
int
)
step_ctx
->
job_id
,
(
int
)
step_ctx
->
step_resp
->
job_step_id
,
hostids
);
(
int
)
step_ctx
->
step_resp
->
job_step_id
,
step_ctx
->
step_resp
->
step_layout
);
params
.
parallel_debug
=
opt
.
parallel_debug
?
true
:
false
;
params
.
parallel_debug
=
opt
.
parallel_debug
?
true
:
false
;
callbacks
.
task_start
=
_task_start
;
callbacks
.
task_start
=
_task_start
;
callbacks
.
task_finish
=
_task_finish
;
callbacks
.
task_finish
=
_task_finish
;
_mpir_init
(
step_req
.
num_tasks
);
_mpir_init
(
step_req
.
num_tasks
);
rc
=
slurm_step_launch
(
step_ctx
,
&
params
,
&
callbacks
);
if
(
slurm_step_launch
(
step_ctx
,
&
params
,
&
callbacks
)
!=
SLURM_SUCCESS
)
{
if
(
rc
!=
SLURM_SUCCESS
)
{
error
(
"Application launch failed: %m"
);
error
(
"Application launch failed: %m"
);
goto
cleanup
;
goto
cleanup
;
}
}
...
@@ -481,9 +470,27 @@ static int _run_srun_script (char *script)
...
@@ -481,9 +470,27 @@ static int _run_srun_script (char *script)
}
}
#endif
#endif
static
int
_taskid_to_nodeid
(
slurm_step_layout_t
*
layout
,
int
taskid
)
{
int
i
,
nodeid
;
for
(
nodeid
=
0
;
nodeid
<
layout
->
node_cnt
;
nodeid
++
)
{
for
(
i
=
0
;
i
<
layout
->
tasks
[
nodeid
];
i
++
)
{
if
(
layout
->
tids
[
nodeid
][
i
]
==
taskid
)
{
debug3
(
"task %d is on node %d"
,
taskid
,
nodeid
);
return
nodeid
;
}
}
}
return
-
1
;
/* node ID not found */
}
static
void
static
void
_setup_local_fds
(
slurm_step_io_fds_t
*
cio_fds
,
int
jobid
,
int
stepid
,
_setup_local_fds
(
slurm_step_io_fds_t
*
cio_fds
,
int
jobid
,
int
stepid
,
uint32_t
*
hostids
)
slurm_step_layout_t
*
step_layout
)
{
{
bool
err_shares_out
=
false
;
bool
err_shares_out
=
false
;
fname_t
*
ifname
,
*
ofname
,
*
efname
;
fname_t
*
ifname
,
*
ofname
,
*
efname
;
...
@@ -499,9 +506,8 @@ _setup_local_fds(slurm_step_io_fds_t *cio_fds, int jobid, int stepid,
...
@@ -499,9 +506,8 @@ _setup_local_fds(slurm_step_io_fds_t *cio_fds, int jobid, int stepid,
cio_fds
->
in
.
fd
=
STDIN_FILENO
;
cio_fds
->
in
.
fd
=
STDIN_FILENO
;
}
else
if
(
ifname
->
type
==
IO_ONE
)
{
}
else
if
(
ifname
->
type
==
IO_ONE
)
{
cio_fds
->
in
.
taskid
=
ifname
->
taskid
;
cio_fds
->
in
.
taskid
=
ifname
->
taskid
;
cio_fds
->
in
.
nodeid
=
hostids
[
ifname
->
taskid
];
cio_fds
->
in
.
nodeid
=
_taskid_to_nodeid
(
step_layout
,
// step_layout_host_id(
ifname
->
taskid
);
// step_layout, ifname->taskid);
}
else
{
}
else
{
cio_fds
->
in
.
fd
=
open
(
ifname
->
name
,
O_RDONLY
);
cio_fds
->
in
.
fd
=
open
(
ifname
->
name
,
O_RDONLY
);
if
(
cio_fds
->
in
.
fd
==
-
1
)
if
(
cio_fds
->
in
.
fd
==
-
1
)
...
...
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment