Skip to content

Commit cd3fce9

Browse files
authored
DOCS: Updating NEWS for v1.1 (port to v1.1) (#647)
* DOCS: Updating NEWS for v1.1 * DOCS: Updating NEWS about RCCL TL support
1 parent 12e454d commit cd3fce9

File tree

1 file changed

+63
-1
lines changed

1 file changed

+63
-1
lines changed

NEWS

Lines changed: 63 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,69 @@
33
*
44
* See file LICENSE for terms.
55
*/
6-
## 1.0.0 (TBD)
6+
7+
## Current
8+
9+
## 1.1.0 (TBD)
10+
11+
## Features
12+
13+
## API
14+
- Added float 128 and float 32, 64, 128 (complex) data types
15+
- Added Active Sets based collectives to support dynamic groups as well as
16+
point-to-point messaging
17+
- Added ucc_team_get_attr interface
18+
19+
## Core
20+
- Config file support
21+
- Fixed component search
22+
23+
## CL
24+
25+
- Added split rail allreduce collective implementation
26+
- Enable hierarchical alltoallv and barrier
27+
- Fixed cleanup bugs
28+
29+
30+
## TL
31+
- Added SELF TL supporting team size one
32+
33+
### UCP
34+
35+
- Added service broadcast
36+
- Added reduce_scatterv ring algorithm
37+
- Added k-nomial based gather collective implementation
38+
- Added one-sided get based algorithms
39+
40+
### SHARP
41+
- Fixed SHARP OOB
42+
- Added SHARP broadcast
43+
44+
45+
46+
### GPU Collectives (CUDA, NCCL TL and RCCL TL)
47+
- Added RCCL TL to support RCCL collectives
48+
- Added support for CUDA TL (intranode collectives for NVIDIA GPUs)
49+
- Added multiring allgatherv, alltoall, reduce-scatter, and reduce-scatterv
50+
multiring in CUDA TL
51+
- Added topo based ring construction in CUDA TL to maximize bandwidth
52+
- Added NCCL gather, scatter and its vector variant
53+
- Enable using multiple streams for collectives
54+
- Added support for RCCL gather (v), scatter (v), broadcast, allgather (v),
55+
barrier, alltoall (v) and all reduce collectives
56+
- Added ROCm memory component
57+
- Adapted all GPU collectives to executor design
58+
59+
60+
### Tests
61+
- Added tests for triggered collectives in perftests
62+
- Fixed bugs in multi-threading tests
63+
64+
### Utils
65+
- Added CPU model and vendor detection
66+
- Several bug fixes in all components
67+
68+
## 1.0.0 (April 19th, 2022)
769

870
### Features
971

0 commit comments

Comments
 (0)