17ed1e67fa
Fix build error
...
[ROCm/rccl-tests commit: 4c0ec7347c
2020-09-25 18:06:09 +00:00
52c50c1653
Add test for alltoallv API
...
[ROCm/rccl-tests commit: 3f1dfacc95
2020-09-24 14:24:52 -04:00
300d3ede81
Prioritize NCCL_HOME
...
[ROCm/rccl-tests commit: 0d1940e18e
2020-09-24 14:18:23 -04:00
ead006c5d7
Add alltoallv test
...
[ROCm/rccl-tests commit: 58dcd35af2
2020-09-01 18:55:25 +00:00
e73a11283b
Add cumask option
...
[ROCm/rccl-tests commit: 3d63a84d97
2020-08-21 21:34:55 +00:00
99d8ddc145
Change scatter and gather bandwidth calculation to match alltoall ( #7 )
...
[ROCm/rccl-tests commit: 346cb16442
2020-07-06 17:12:50 -07:00
075041a9fa
Merge remote-tracking branch 'nccl/master' into HEAD
...
[ROCm/rccl-tests commit: 5361dd8177
2020-07-06 17:54:31 +00:00
8b83a414c5
Fix typo in src/Makefile
...
[ROCm/rccl-tests commit: afdaf59b3b
2020-06-24 14:39:22 -07:00
0624d2cede
Add gencode for CUDA11
...
[ROCm/rccl-tests commit: b2603a2e85
2020-06-23 18:16:46 -07:00
12d86bd58f
Change all_gather/reduce_scatter algbw to match the documentation.
...
Fix #45 : All_gather and reduce_scatter algorithm bandwidth was
computed as time/count*(nranks-1) which is not consistent with the
way we compute it for other collectives.
This change makes algbw higher; busbw is unchanged.
[ROCm/rccl-tests commit: ec1b5e22e6
2020-06-19 10:42:19 -07:00
fcaaf2c4a1
Fix #47 : compilation error on NCCL<2.7
...
Return an error when trying to run alltoall test when compiled
against NCCL<2.7.
[ROCm/rccl-tests commit: 07ac716c1a
2020-06-18 15:02:51 -07:00
cf70df2498
Merge pull request #46 from NVIDIA/p2p
...
Add alltoall perf test
[ROCm/rccl-tests commit: a7b304dde5
2020-06-17 10:45:29 -07:00
3a6293b748
Fix some memory leaks
...
[ROCm/rccl-tests commit: af4fa0f4cf
2020-06-17 10:44:32 -07:00
0dfae3da28
Remove sm_30
...
[ROCm/rccl-tests commit: 7a833631b2
2020-06-15 08:54:21 -07:00
0b235b136a
Correct szie when init sendbuff
...
[ROCm/rccl-tests commit: 83b846cf4f
2020-06-10 23:24:08 +00:00
279011312a
Update rccl_bfloat16.h to match rocBLAS
...
[ROCm/rccl-tests commit: a698b55cf5
2020-05-11 15:21:19 -07:00
0f10b6ceeb
Add option to use alltoall, gather and scatter API
...
These APIs launche RCCL kernel implementation by default. If environmental
variable RCCL_ALLTOALL_KERNEL_DISABLE=1, then the APIs use wrapper around
ncclSend and ncclRecv.
[ROCm/rccl-tests commit: 2813c96826
2020-05-11 15:21:19 -07:00
a6c3e6c011
Add sendrecv benchmark
...
[ROCm/rccl-tests commit: 0d7c4db33e
2020-04-10 14:45:17 -07:00
41b53c0974
Improve makefile to avoid LD_LIBRARY_PATH
...
[ROCm/rccl-tests commit: 07b8876277
2020-04-10 14:45:17 -07:00
af08abefb0
Add gather and scatter test
...
[ROCm/rccl-tests commit: 6932a583e7
2020-04-10 14:45:10 -07:00
b132257d16
Merge remote-tracking branch 'nccl/p2p' into p2p
...
[ROCm/rccl-tests commit: c2dc7b613b
2020-04-02 15:06:36 -07:00
c633de20d6
Add alltoall perf test
...
[ROCm/rccl-tests commit: 119a0ecf60
2020-03-17 12:00:19 -07:00
c947fe45ff
Add option for stress testing
...
[ROCm/rccl-tests commit: 6e9e05972b
2020-03-11 13:40:17 -07:00
c76094c704
Add -L$(MPI_HOME)/lib64 to NVLDFLAGS
...
In some cases, the MPI library is not in $(MPI_HOME)/lib but
in $(MPI_HOME)/lib64. For example, on RedHat like Linux system
(CentOS, Amazon Linux), and MPI is installed by yum or rpm.
Under such circumstance, the current make file will cause failure.
This patch address this issue by adding -L$(MPI_HOME)/lib64 to
NVLDFLAGS in src/Makefile.
Signed-off-by: Wei Zhang <wzam@amazon.com >
[ROCm/rccl-tests commit: 0f173234bb
2019-12-16 16:18:22 -08:00
e1d118e925
Fix build error with hip-clang
...
[ROCm/rccl-tests commit: bd53e98df3
2019-11-22 10:31:31 -08:00
907877df59
Add bf16 support in rccl-tests
...
[ROCm/rccl-tests commit: 1cda2f52b6
2019-11-18 13:18:34 -08:00
123ea97750
Fix build with RCCL 1.x API
...
[ROCm/rccl-tests commit: 32399955af
2019-11-12 23:04:39 +00:00
9c89b0f8bd
Fix incorrect print out when data size is greater than 4GB
...
[ROCm/rccl-tests commit: aa0f02bee0
2019-11-11 11:39:59 -08:00
55749e6e97
Init data for throughput iterations to avoid all zero data
...
[ROCm/rccl-tests commit: 13d0ddd12e
2019-09-27 09:12:00 -07:00
b74c00f548
Fix memory leak and possible buffer overrun
...
[ROCm/rccl-tests commit: ab82f1af6f
2019-08-09 10:22:14 -07:00
5b5cf00cbe
Allow call ncclCommAbort on Ctrl+C
...
[ROCm/rccl-tests commit: 23c374475f
2019-07-26 00:15:16 +00:00
0288eeef75
Checking that number of requested GPUs is not more than number of available GPUs
...
[ROCm/rccl-tests commit: 043eef6999
2019-07-11 15:36:21 +00:00
30d348de0e
Adding unit tests and files for CI ( #4 )
...
* Adding initial unit test and Jenkins code.
Fixing scope of unit tests
Adding unit tests and files for CI
Fixing Jenkinsfile
* Removing typos from Jenkinsfile
* Making some fixes to the Jenkins file; temporarily disabling MPI
* Making corrections to Jenkinsfile
* Correcting dockerNodes entry in Jenkinsfile
* Fixed Jenkinsfile for CI
* Correcting Jenkinsfile for CI
* Updating README to include instructions on how to run unit tests.
[ROCm/rccl-tests commit: 924521ff57
2019-06-28 08:52:44 -07:00
75b5e43633
Fix missing space in Makefile
...
[ROCm/rccl-tests commit: 3fac1d679b
2019-05-16 13:18:23 -07:00
514e912b49
enable building with mpich
...
Use following command to build: MPICH=1 make
[ROCm/rccl-tests commit: 86f053be84
2019-05-13 23:45:28 +00:00
b0e3a2ac64
allow using different memory types for input and output buffers
...
[ROCm/rccl-tests commit: 3f89175af5
2019-05-01 12:58:04 -07:00
852651be37
fix build with 1.0 library
...
[ROCm/rccl-tests commit: 7e80ea9d3a
2019-04-26 17:06:05 -07:00
08a59332af
Adding way to specify a custom RCCL shared library file to link against
...
[ROCm/rccl-tests commit: 10e1572f72
2019-04-22 10:25:49 -07:00
8ff34620fb
workaround weak symbol issue
...
hcc prints "error: alias must point to a defined variable or function"
[ROCm/rccl-tests commit: 4474fe168d
2019-04-18 10:34:55 -07:00
aac7cfb64f
Adding AMD copyright notices
...
[ROCm/rccl-tests commit: 71e663e62d
2019-04-10 15:28:40 -07:00
3c8cfb2d6e
hipify nccl-tests to become rccl-tests
...
[ROCm/rccl-tests commit: a15f771cb2
2019-04-10 13:43:58 -07:00
18902f40a7
Resync all tests with test code from NCCL 2.4
...
Major rework to merge most of the changes from the NCCL internal
tests into the public ones
Added "-m <agg_iters>" operation aggregation option.
Data integrity checking is now much more performant at scale.
Startup times at scale are improved.
Test latency units are now displayed in usec.
[ROCm/rccl-tests commit: cbe7f65400
2019-04-05 13:42:15 -07:00
3958cc347e
Print NCCL version at start
...
[ROCm/rccl-tests commit: 925a70576e
2017-12-21 15:10:09 -08:00
224eddfc15
Fix NCCL_HOME to be consistent with README
...
[ROCm/rccl-tests commit: 25016c8eeb
2017-08-09 10:41:31 -07:00
4cb47ccb21
Initial commit
...
[ROCm/rccl-tests commit: b188a15299
2017-08-08 16:18:34 -07:00
Wenkai Du
Wenkai Du
Luke Yeager
Sylvain Jeaugey
Luke Yeager
Sourav Chakraborty
Wei Zhang
Gilbert Lee
Stanley Tsang
David Addison