CL/HIER: Add allgatherv #1050
Open
CL/HIER: Add allgatherv #1050
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
nsarka
force-pushed
the
nsarka/cl-hier-allgatherv
branch
from
9e34573 to
c126e93
Compare
nsarka
force-pushed
the
nsarka/cl-hier-allgatherv
branch
2 times, most recently
from
198da51 to
862cad6
Compare
Closed
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
commented
Dec 4, 2024
nsarka
force-pushed
the
nsarka/cl-hier-allgatherv
branch
2 times, most recently
from
0774618 to
1298163
Compare
|
Can one of the admins verify this patch? |
|
@aamirshafi Please review |
nsarka
force-pushed
the
nsarka/cl-hier-allgatherv
branch
from
584b1b4 to
0305f14
Compare
nsarka
force-pushed
the
nsarka/cl-hier-allgatherv
branch
from
0305f14 to
30f0ce1
Compare
| if (ucc_unlikely(cl_team->node_leaders == NULL)) { | ||
| cl_team->node_leaders = ucc_malloc(sizeof(ucc_rank_t) * team_size); | ||
| if (!cl_team->node_leaders) { | ||
| cl_error(team->context->lib, "Could not allocate node_leaders array"); |
| } | ||
| cl_team->leader_list = ucc_malloc(sizeof(ucc_rank_t) * ldr_sbgp_size); | ||
| if (!cl_team->node_leaders) { | ||
| cl_error(team->context->lib, "Could not allocate leader_list array"); |
| args.args.src.info.mem_type = UCC_MEMORY_TYPE_HOST; | ||
| args.args.dst.info_v.displacements = leader_disps; | ||
| args.args.dst.info_v.counts = leader_counts; | ||
| args.args.dst.info_v.buffer = args_old.args.dst.info_v.buffer; |
There was a problem hiding this comment.
here dst.Info_v.buffer size might be not enough to hold data because of gaps
| cl_team->top_sbgp != UCC_HIER_SBGP_NODE) { | ||
| args = args_old; | ||
| args.args.coll_type = UCC_COLL_TYPE_BCAST; | ||
| args.args.root = 0; |
There was a problem hiding this comment.
change to leader rank id
| 0, UCC_MSG_MAX, UCC_CL_HIER_DEFAULT_SCORE, | ||
| ucc_cl_hier_allgatherv_init, cl_team); | ||
| if (UCC_OK != status) { | ||
| cl_error(lib, "faild to add range to score_t"); |
| args->dst.info_v.datatype); | ||
| ucc_ee_executor_t *exec; | ||
| ucc_status_t status; | ||
| int i; |
| } | ||
|
|
||
| schedule->super.status = UCC_INPROGRESS; | ||
| schedule->super.super.status = UCC_INPROGRESS; |
There was a problem hiding this comment.
no need to set super.super.status
|
|
||
| out: | ||
| schedule->super.status = st; | ||
| schedule->super.super.status = st; |
There was a problem hiding this comment.
don't set it, it may lead to races with multiple threads
| ucc_rank_t i; | ||
|
|
||
| for (i = 0; i < *n_tasks; i++) { | ||
| ucc_ee_executor_task_t *etask = tasks[i]; |
There was a problem hiding this comment.
define variable at the beginning of function
Comment on lines
+11
to
+13
| ucc_schedule_t *schedule = ucc_derived_of(task, ucc_schedule_t); | ||
| ucc_cl_hier_schedule_t *cl_schedule = ucc_derived_of(schedule, | ||
| ucc_cl_hier_schedule_t); |
There was a problem hiding this comment.
Suggested change
| ucc_schedule_t *schedule = ucc_derived_of(task, ucc_schedule_t); | |
| ucc_cl_hier_schedule_t *cl_schedule = ucc_derived_of(schedule, | |
| ucc_cl_hier_schedule_t); | |
| ucc_cl_hier_schedule_t *cl_schedule = ucc_derived_of(task, | |
| ucc_cl_hier_schedule_t); |
janjust
requested changes
Jan 24, 2025
| @@ -1,48 +1,55 @@ | |||
| # | |||
| # Copyright (c) 2020-2022, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| # Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| return cl_team->node_leaders[team_rank]; | ||
| } | ||
|
|
||
| UCC_CL_HIER_PROFILE_FUNC(ucc_status_t, ucc_cl_hier_allgatherv_init, |
There was a problem hiding this comment.
@Sergei-Lebedev can you elaborate, is this something we simply don't support in allgatherv?
| @@ -0,0 +1,15 @@ | |||
| /** | |||
| * Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| @@ -1,5 +1,5 @@ | |||
| /** | |||
| * Copyright (c) 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| * Copyright (c) 2020-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| @@ -1,5 +1,5 @@ | |||
| /** | |||
| * Copyright (c) 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| * Copyright (c) 2020-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| @@ -1,5 +1,5 @@ | |||
| /** | |||
| * Copyright (c) 2020-2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
| * Copyright (c) 2020-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved. | |||
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds CL/HIER allgatherv and updates the allgatherv gtest to test non-contiguous dst buffers. It is meant to be used in conjunction with my TL/SHM gatherv implementation.
The algorithm is:
Comparison Data
2 nodes, 32 PPN on lego-grace cg (1 socket, 72 cores/socket)