Improve bounds_check_indices for VBE#3388
Closed
sryap wants to merge 1 commit into
Closed
Conversation
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 18, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
ae31edc to
47a0923
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 22, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
47a0923 to
e00a881
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
force-pushed
the
export-D66084041
branch
from
e00a881 to
8e8360f
Compare
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 25, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 25, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
8e8360f to
00498a0
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
1 similar comment
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 25, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
00498a0 to
fcfad1a
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 25, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
fcfad1a to
faf3e7b
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 26, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
faf3e7b to
e081512
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
force-pushed
the
export-D66084041
branch
from
e081512 to
9b93585
Compare
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 26, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 29, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
9b93585 to
b80f25d
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 30, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
b80f25d to
c4436bd
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
sryap
force-pushed
the
export-D66084041
branch
from
c4436bd to
dc2de02
Compare
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Nov 30, 2024
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
Summary: X-link: facebookresearch/FBGEMM#477 Instead of over launching thread blocks, use `b_t_map` to launch only necessary thread blocks to increase occupancy for the VBE case Note that `b_t_map` is necessary for the TBE look for the VBE case. It is generated during the TBE forward pass. In this diff, we call `generate_vbe_metdata` twice (before bounds check and before forward look up). These two calls can be fused into one. We will clean this up in the subsequent diffs. Reviewed By: Fiery Differential Revision: D66084041
sryap
force-pushed
the
export-D66084041
branch
from
dc2de02 to
84e9352
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66084041 |
Contributor
|
This pull request has been merged in a821f59. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary:
Instead of over launching thread blocks, use
b_t_mapto launch onlynecessary thread blocks to increase occupancy for the VBE case
Note that
b_t_mapis necessary for the TBE look for the VBE case. Itis generated during the TBE forward pass. In this diff, we call
generate_vbe_metdatatwice (before bounds check and before forwardlook up). These two calls can be fused into one. We will clean this
up in the subsequent diffs.
Differential Revision: D66084041