Fix MaxPool Block / Width Sharding with Large Kernels / Wide Reductions #14531
Conversation
This file contains 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
…s and wide reductions
wransom-TT
requested review from
ayerofieiev-tt,
dmakoviichuk-tt,
rfurko-tt,
cfjchu,
TT-BrianLiu,
razorback3,
dongjin-na,
mywoodstock,
shwetankTT,
sankarmanoj-tt and
pavlejosipovic
as code owners
wransom-TT
force-pushed
the
wransom/LargeKernel
branch
from
8b1f490 to
ecc2bd0
Compare
mywoodstock
reviewed
Nov 5, 2024
ttnn/cpp/ttnn/operations/pool/maxpool/device/kernels/compute/max_pool_multi_core.cpp
Show resolved
Hide resolved
.../device/kernels/dataflow/reader_max_pool_2d_multi_core_sharded_with_halo_large_kernel_v2.cpp
Outdated
Show resolved
Hide resolved
...ool/maxpool/device/kernels/dataflow/reader_max_pool_2d_multi_core_sharded_with_halo_wide.cpp
Outdated
Show resolved
Hide resolved
ttnn/cpp/ttnn/operations/pool/maxpool/device/max_pool2d_multi_core_program_factory.cpp
Outdated
Show resolved
Hide resolved
nkpatel-tt
reviewed
Nov 5, 2024
| if (is_large_kernel) { | ||
| uint32_t max_pool_partials_cb_id = tt::CB::c_intermed1; // max_pool partials | ||
| uint32_t max_pool_partials_cb_pagesize = in_cb_sz; | ||
| uint32_t max_pool_partials_cb_pagesize = out_cb_pagesize; |
There was a problem hiding this comment.
Minor comment, This variable should be min(out_cb_pagesize, TILE_SIZE8nbytes).
There was a problem hiding this comment.
Hey looks like a small typo, I updated it to min(out_cb_pagesize, TILE_SIZE * 8 * out_nbytes);, let me know if this isn't what you meant!
nkpatel-tt
reviewed
Nov 5, 2024
...cpp/ttnn/operations/pool/maxpool/device/kernels/compute/max_pool_multi_core_large_kernel.cpp
Outdated
Show resolved
Hide resolved
nkpatel-tt
reviewed
Nov 5, 2024
.../device/kernels/dataflow/reader_max_pool_2d_multi_core_sharded_with_halo_large_kernel_v2.cpp
Show resolved
Hide resolved
mywoodstock
approved these changes
Nov 6, 2024
TT-BrianLiu
approved these changes
Nov 8, 2024
ct-clmsn
pushed a commit
to ct-clmsn/tt-metal
that referenced
this pull request
Nov 12, 2024
…ns (tenstorrent#14531) * tenstorrent#14249: Fixed bug for width and block sharding with large kernel sizes and wide reductions
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
For Block and Width Sharding with large kernels, Max Pool produced incorrect results. This has been fixed, plus support has been added for large kernels with wide reductions which was previously untested.
Ticket
Link to Github Issue
Problem description
When block and width sharding functionality was introduced the large reader and compute kernels were not updated and these cases were not covered by test cases. Additionally, the large kernels did not support wide reductions.
What's changed
Checklist