Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

UCP: tm offload rndv support #1549

Merged
merged 5 commits into from
Jun 1, 2017

Conversation

alex-mikheev
Copy link
Contributor

@yosefe pls take a look

@yosefe yosefe added the Feature New feature label May 28, 2017
@yosefe yosefe added this to the v1.3 milestone May 28, 2017
@mellanox-github
Copy link
Contributor

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1753/ for details.

@mellanox-github
Copy link
Contributor

Test FAILed.
See http://hpc-master.lab.mtl.com:8080/job/hpc-ucx-pr/3691/ for details (Mellanox internal link).

@mellanox-github
Copy link
Contributor

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1757/ for details.

uint64_t remote_addr, size_t length,
const void *rkey_buf)
{
ucp_rndv_rts_hdr_t *rts = (ucp_rndv_rts_hdr_t*)(((ucp_tag_hdr_t*)hdr) - 1);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

isn't this illegal to access memory before 'hdr'?


ucs_status_t ucp_tag_offload_rndv_zcopy(uct_pending_req_t *self)
{
void *rndv_op;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pls move this variable to be last, so variables with initializers are first

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1758/ for details.

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://hpc-master.lab.mtl.com:8080/job/hpc-ucx-pr/3695/ for details (Mellanox internal link).

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1759/ for details.

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://hpc-master.lab.mtl.com:8080/job/hpc-ucx-pr/3696/ for details (Mellanox internal link).

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1767/ for details.

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://hpc-master.lab.mtl.com:8080/job/hpc-ucx-pr/3705/ for details (Mellanox internal link).

sizeof(*rndv_rts_hdr), recv_flags);
}

UCP_THREAD_CS_EXIT_CONDITIONAL(&context->mt_lock);
return status;
}

ucs_status_t ucp_rndv_rts_handler(void *arg, void *data, size_t length,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

need to move UCS_PROFILE_FUNC to this function instead of ucp_rndv_process_rts, and add UCS_PROFILE_FUNC to ucp_tag_offload_unexp_rndv. This is because we need to profile the entry points from UCT to UCP.

if (status != UCS_OK) {
ucs_error("Failed to cancel recv in the transport: %s",
ucs_status_string(status));
if (req->flags & UCP_REQUEST_FLAG_OFFLOADED) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if (!offloaded) return

ucp_worker_iface_t, queue);
ucp_rndv_rts_hdr_t rts;

rts.sreq = sreq->super;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pls add comment like "emulate rts without key"

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ucx-pr/1784/ for details.

@mellanox-github
Copy link
Contributor

Test PASSed.
See http://hpc-master.lab.mtl.com:8080/job/hpc-ucx-pr/3721/ for details (Mellanox internal link).

@yosefe yosefe merged commit 4962c52 into openucx:master Jun 1, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Feature New feature
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants