TL/MLX5: multicast design: adding helper.h, progress.h, and mcast.h

[MamziB]

TL/MLX5: multicast design: adding helper.h, progress.h, and mcast.h

[Sergei-Lebedev]

please fix CI

[MamziB]

@Sergei-Lebedev I fixed the comment. Thanks for the feedback. I see that all the errors in testing system comes from failure in building UCX, which has nothing to do with this PR. @manjugv FYI

core/ucp_mm.c:980:74: error: too few arguments to function call, expected 8, have 7
                          UCS_BIT(md_index), UCT_MD_MEM_ACCESS_ALL, &memh);
                                                                         ^
core/ucp_mm.inl:35:1: note: 'ucp_memh_get' declared here
ucp_memh_get(ucp_context_h context, void *address, size_t length,
^
1 error generated.
make[1]: *** [Makefile:1199: core/libucp_la-ucp_mm.lo] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/tmp/ucx/src/ucp'
make: *** [Makefile:782: install-recursive] Error 1

[janjust]

@Sergei-Lebedev I fixed the comment. Thanks for the feedback. I see that all the errors in testing system comes from failure in building UCX, which has nothing to do with this PR. @manjugv FYI

core/ucp_mm.c:980:74: error: too few arguments to function call, expected 8, have 7
                          UCS_BIT(md_index), UCT_MD_MEM_ACCESS_ALL, &memh);
                                                                         ^
core/ucp_mm.inl:35:1: note: 'ucp_memh_get' declared here
ucp_memh_get(ucp_context_h context, void *address, size_t length,
^
1 error generated.
make[1]: *** [Makefile:1199: core/libucp_la-ucp_mm.lo] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/tmp/ucx/src/ucp'
make: *** [Makefile:782: install-recursive] Error 1

This was fixed in UCX @Sergei-Lebedev @manjugv I really think the UCC CI should build against UCX release branch, and not master, we faced the same issue briefly in OMPI.

[manjugv]

@Sergei-Lebedev I fixed the comment. Thanks for the feedback. I see that all the errors in testing system comes from failure in building UCX, which has nothing to do with this PR. @manjugv FYI

core/ucp_mm.c:980:74: error: too few arguments to function call, expected 8, have 7
                          UCS_BIT(md_index), UCT_MD_MEM_ACCESS_ALL, &memh);
                                                                         ^
core/ucp_mm.inl:35:1: note: 'ucp_memh_get' declared here
ucp_memh_get(ucp_context_h context, void *address, size_t length,
^
1 error generated.
make[1]: *** [Makefile:1199: core/libucp_la-ucp_mm.lo] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/tmp/ucx/src/ucp'
make: *** [Makefile:782: install-recursive] Error 1

This was fixed in UCX @Sergei-Lebedev @manjugv I really think the UCC CI should build against UCX release branch, and not master, we faced the same issue briefly in OMPI.

Regarding CI, let's convey that to the CI team. Between, there is a commit title error as well.

[MamziB]

@samnordmann @Sergei-Lebedev Thank you so much guys for the constructive comments. I have resolved/responded to your comments and questions. Then I pushed a single commit.

[MamziB]

Thanks @samnordmann for the new comments. I have resolved them. I updated the commit. clang-tidy failure is not related to this PR:

dpkg: error processing archive /tmp/apt-dpkg-install-D3JNOM/89-rocthrust-dev_2.18.0.50700-63~20.04_amd64.deb (--unpack):
 cannot copy extracted data for './opt/rocm-5.7.0/include/thrust/detail/preprocessor.h' to '/opt/rocm-5.7.0/include/thrust/detail/preprocessor.h.dpkg-new': failed to write (No space left on device)
No apport report written because the error message indicates a disk full error
dpkg-deb: error: paste subprocess was killed by signal (Broken pipe)
Selecting previously unselected package rocwmma-dev.
Preparing to unpack .../90-rocwmma-dev_1.2.0.50700-63~20.04_amd64.deb ...
Unpacking rocwmma-dev (1.2.0.50700-63~20.04) ...
Selecting previously unselected package rocm-hip-sdk.
Preparing to unpack .../91-rocm-hip-sdk_5.7.0.50700-63~20.04_amd64.deb ...
Unpacking rocm-hip-sdk (5.7.0.50700-63~20.04) ...
dpkg: error: failed to write status database record about 'libjson-perl' to '/var/lib/dpkg/status': No space left on device
E: Sub-process /usr/bin/dpkg returned an error code (2)
Error: Process completed with exit code 100.

[MamziB]

@samnordmann Thanks for the detailed comments. I have resolved them, please take a look. I have also updated the commit.

[MamziB]

@samnordmann Thank you for the new comments. Please take a look. I also updated the commit.

[MamziB]

@manjugv any thoughts why ucc checks are failing?

ucc — Test FAILed. No test results found.

[MamziB]

@Sergei-Lebedev thanks for the constructive comments. I have resolved them and updated the PR's commit.

[MamziB]

@Sergei-Lebedev Thank you very much for the detailed and constructive comments. Please see my answers. I have also updated the commit. Thank you.

[Sergei-Lebedev]

@MamziB pls rebase

[MamziB]

@Sergei-Lebedev Thanks for approving the PR. I have rebased the branch on the top of the laster master branch.