-
Notifications
You must be signed in to change notification settings - Fork 74k
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
[ROCm] Adding ROm support for "concat", "pack", "unpack" and "split" ops #28451
[ROCm] Adding ROm support for "concat", "pack", "unpack" and "split" ops #28451
Conversation
So there's good news and bad news. 👍 The good news is that everyone that needs to sign a CLA (the pull request submitter and all commit authors) have done so. Everything is all good there. 😕 The bad news is that it appears that one or more commits were authored or co-authored by someone other than the pull request submitter. We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that here in the pull request. Note to project maintainer: This is a terminal state, meaning the ℹ️ Googlers: Go here for more info. |
@chsigg , would you be able to take a look at this and PR 28450 as they are TF implementation changes? |
Please do not rename the symbols or macros as part of this PR. This is better done from our end, because of internal dependencies that you cannot update. |
9286815
to
476e554
Compare
CLAs look good, thanks! ℹ️ Googlers: Go here for more info. |
rebased this PR now that the updates related to Cuda* to Gpu* renaming are merged into master. This PR is now trivial, please review and merge. thanks |
The failures in the PR #28834 renames the This PR updates the the thanks |
output->dimension(1), output->data())); | ||
if (smem_usage < smem_max && smem_usage < kMaxSmemBytesPerformance) { | ||
TF_CHECK_OK(GpuLaunchKernel( | ||
(concat_variable_kernel<T, IntType, true>), dim3(config.block_count), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Remove extra parenthesis. Here and below.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
let me do that and push out a rebase
gpu_device.stream(), input_ptr, total_rows, | ||
total_cols, output_ptr_data)); | ||
TF_CHECK_OK( | ||
GpuLaunchKernel(SplitVOpKernel_fixed<T>, dim3(config.block_count), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does ROCm's dim3 not implicitly construct from int?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it does...the dim3 call does not need to be there...will remove it along with the extra parens
476e554
to
a62758b
Compare
Yes, we can address the dynamic shared memory in a separate change.
…On Thu, May 23, 2019, 20:59 Rajeshwar Reddy T ***@***.***> wrote:
@rthadur <https://github.com/rthadur> requested your review on: #28451
<#28451> [ROCm] Adding ROm
support for "concat", "pack", "unpack" and "split" ops.
—
You are receiving this because your review was requested.
Reply to this email directly, view it on GitHub
<#28451?email_source=notifications&email_token=ABZM5DRAVB5UBLKOAMVN2Q3PW3SPXA5CNFSM4HLDPJDKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGORTPWNTA#event-2363451084>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABZM5DVULTNHLG56LCHCHXDPW3SPXANCNFSM4HLDPJDA>
.
|
…plit_op PiperOrigin-RevId: 249878940
This PR add ROCm support for "concat", "pack", "unpack" and "split" ops
PR #28343 is a pre-req for this PR, and hence this PR includes commits from PR #28343
Only the last three commits in this PR is exclusive to this PR, and hence should be the only one reviewed.
Please ignore the cla check failure...that is getting triggered because this PR builds on top of PR #28343 (which is submitted by a different developer). Once that PR is merged, and I rebase this PR, the commit triggering the cla check failure will be gone from this commit
@tatianashp , @whchung : just FYI