Use tfhe_cuda_backend insteaof of concrete_cuda #1064

BourgerieQuentin · 2024-09-24T12:39:06Z

cla-bot · 2024-09-24T12:39:09Z

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Ubuntu.
This is most likely caused by a git client misconfiguration; please make sure to:

check if your git client is configured with an email to sign commits git config --list | grep email
If not, set it up using git config --global user.email [email protected]
Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

bcm-at-zama · 2024-09-24T12:43:58Z

99 files! it's a lot of changes

cla-bot · 2024-09-24T12:59:31Z

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Ubuntu.
This is most likely caused by a git client misconfiguration; please make sure to:

check if your git client is configured with an email to sign commits git config --list | grep email
If not, set it up using git config --global user.email [email protected]
Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

cla-bot · 2024-09-24T13:00:18Z

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Ubuntu.
This is most likely caused by a git client misconfiguration; please make sure to:

check if your git client is configured with an email to sign commits git config --list | grep email
If not, set it up using git config --global user.email [email protected]
Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

antoniupop · 2024-09-24T13:37:40Z

compilers/concrete-compiler/compiler/lib/Runtime/GPUDFG.cpp

          p->output_lwe_dim.val, p->base_log.val, p->level.val, num_samples);
+      cuda_drop_async(indexes_gpu, s, loc);
+      free(indexes);


If this is freed here, the CUDA runtime will likely try reading this array much later, once it can actually do the memory transfer. This needs to be deallocated after a synchronization point - see below.

antoniupop · 2024-09-24T13:38:38Z

compilers/concrete-compiler/compiler/lib/Runtime/GPUDFG.cpp

      cuda_drop_async(test_vector_idxes_gpu, s, loc);
      cuda_drop_async(glwe_ct_gpu, s, loc);
+      cuda_drop_async(indexes_gpu, s, loc);
+      free(indexes);


Same as above.

antoniupop · 2024-09-24T13:40:35Z

compilers/concrete-compiler/compiler/lib/Runtime/GPUDFG.cpp

      cuda_drop_async(test_vector_idxes_gpu, s, loc);
      cuda_drop_async(glwe_ct_gpu, s, loc);
+      cuda_drop_async(indexes_gpu, s, loc);
+      free(indexes);
      Dependence *dep =
          new Dependence(loc, out, out_gpu, false, false, d0->chunk_id);
      // As streams are not synchronized, we can only free this vector


I can''t add comment at right line - just 4 lines below is the way to deallocate data after CUDA synch points - it registers the pointer with the runtime then will deallocate after the next cuda synch.

cla-bot · 2024-09-24T13:44:53Z

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Ubuntu.
This is most likely caused by a git client misconfiguration; please make sure to:

check if your git client is configured with an email to sign commits git config --list | grep email
If not, set it up using git config --global user.email [email protected]
Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

cla-bot · 2024-09-24T14:13:30Z

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Ubuntu.
This is most likely caused by a git client misconfiguration; please make sure to:

check if your git client is configured with an email to sign commits git config --list | grep email
If not, set it up using git config --global user.email [email protected]
Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

github-actions · 2024-12-10T10:17:36Z

⚠️ This PR contains unsigned commits. To get your PR merged, please sign those commits (git rebase --exec 'git commit -S --amend --no-edit -n' @{upstream}) and force push them to this branch (git push --force-with-lease).

If you're new to commit signing, there are different ways to set it up:

Sign commits with gpg

Follow the steps below to set up commit signing with gpg:

Sign commits with ssh-agent

Follow the steps below to set up commit signing with ssh-agent:

Sign commits with 1Password

You can also sign commits using 1Password, which lets you sign commits with biometrics without the signing key leaving the local 1Password process.

Learn how to use 1Password to sign your commits.

github-actions

Benchmark

Benchmark suite	Current: `7fc2663`	Previous: `2a1f449`	Ratio
`v0 PBS table generation`	`59009422` ns/iter (`± 1243379`)	`58930979` ns/iter (`± 314241`)	`1.00`
`v0 PBS simulate dag table generation`	`37890203` ns/iter (`± 414241`)	`37554649` ns/iter (`± 469151`)	`1.01`
`v0 WoP-PBS table generation`	`52298219` ns/iter (`± 447515`)	`52447087` ns/iter (`± 585670`)	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

bcm-at-zama changed the title ~~Use tfhe_cuda_backend insteof od concrete_cuda~~ Use tfhe_cuda_backend insteaof of concrete_cuda Sep 24, 2024

BourgerieQuentin force-pushed the cuda-tfhe-rs branch from d5584ef to eb8c467 Compare September 24, 2024 13:00

antoniupop requested changes Sep 24, 2024

View reviewed changes

perf(backend-cuda): Update concrete_cuda to tfhe_cuda_backend

aa6563e

BourgerieQuentin force-pushed the cuda-tfhe-rs branch from 5014daa to aa6563e Compare September 24, 2024 15:06

cla-bot bot added the cla-signed label Sep 24, 2024

test(compiler): Add latest cifar benchmark

ad106f7

BourgerieQuentin force-pushed the cuda-tfhe-rs branch from 6a52c4e to ad106f7 Compare September 25, 2024 13:05

BourgerieQuentin requested a review from pdroalves September 25, 2024 13:52

github-actions bot reviewed Dec 10, 2024

View reviewed changes

BourgerieQuentin force-pushed the cuda-tfhe-rs branch 3 times, most recently from e51a9c5 to b9886d1 Compare December 10, 2024 13:05

try std bootstrap

7fc2663

BourgerieQuentin force-pushed the cuda-tfhe-rs branch from b9886d1 to 7fc2663 Compare December 10, 2024 13:15

BourgerieQuentin closed this Dec 10, 2024

BourgerieQuentin deleted the cuda-tfhe-rs branch December 10, 2024 16:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use tfhe_cuda_backend insteaof of concrete_cuda #1064

Use tfhe_cuda_backend insteaof of concrete_cuda #1064

BourgerieQuentin commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

bcm-at-zama commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

antoniupop Sep 24, 2024

antoniupop Sep 24, 2024

antoniupop Sep 24, 2024

cla-bot bot commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

github-actions bot commented Dec 10, 2024

github-actions bot left a comment •

edited

Loading

Use tfhe_cuda_backend insteaof of concrete_cuda #1064

Use tfhe_cuda_backend insteaof of concrete_cuda #1064

Conversation

BourgerieQuentin commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

bcm-at-zama commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

antoniupop Sep 24, 2024

Choose a reason for hiding this comment

antoniupop Sep 24, 2024

Choose a reason for hiding this comment

antoniupop Sep 24, 2024

Choose a reason for hiding this comment

cla-bot bot commented Sep 24, 2024

cla-bot bot commented Sep 24, 2024

github-actions bot commented Dec 10, 2024

github-actions bot left a comment • edited Loading

Choose a reason for hiding this comment

Benchmark

github-actions bot left a comment •

edited

Loading