[ET-VK][conv1d] Implement height-packed depthwise conv1d operator#18333
Merged
meta-codesync[bot] merged 5 commits intogh/SS-JIA/495/basefrom Mar 27, 2026
Merged
[ET-VK][conv1d] Implement height-packed depthwise conv1d operator#18333meta-codesync[bot] merged 5 commits intogh/SS-JIA/495/basefrom
meta-codesync[bot] merged 5 commits intogh/SS-JIA/495/basefrom
Conversation
Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18333
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (2 Unrelated Failures)As of commit 1b33653 with merge base 7c79395 ( BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
This PR needs a
|
added 2 commits
March 19, 2026 15:48
…perator" Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/) [ghstack-poisoned]
…perator" Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/) [ghstack-poisoned]
trviv
approved these changes
Mar 26, 2026
added 2 commits
March 27, 2026 10:40
…perator" Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/) [ghstack-poisoned]
…perator" Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/) [ghstack-poisoned]
meta-codesync
bot
merged commit a65231b
into
gh/SS-JIA/495/base
157 of 162 checks passed
SS-JIA
pushed a commit
that referenced
this pull request
Mar 27, 2026
Pull Request resolved: #18333 Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s ghstack-source-id: 358903219 @exported-using-ghexport Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/)
SS-JIA
pushed a commit
that referenced
this pull request
Mar 27, 2026
Pull Request resolved: #18333 Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s ghstack-source-id: 358903219 @exported-using-ghexport Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/)
rascani
pushed a commit
to rascani/executorch
that referenced
this pull request
Apr 1, 2026
Pull Request resolved: pytorch#18333 Implement a depthwise conv1d operator using height-packed layout where channels are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter to each channel independently (groups=C), so 4 channels can be processed in parallel using element-wise vec4 FMA over kernel positions. Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel (4 channels at one spatial position). Inner loop iterates over kernel positions K with bounds-checked input access for padding. Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4 channels' weights at one kernel position. Supports both buffer and texture3d storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation. Registered as et_vk.conv1d_dw.default (standalone custom op). Performance on Adreno 750 (S24): - [1,128,4096] K=31 buffer f16: 231 GFLOP/s - [1,128,4096] K=31 buffer f32: 155 GFLOP/s - [1,512,2048] K=5 buffer f32: 66 GFLOP/s ghstack-source-id: 358903219 @exported-using-ghexport Differential Revision: [D97344091](https://our.internmc.facebook.com/intern/diff/D97344091/)
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.
Stack from ghstack (oldest at bottom):
Implement a depthwise conv1d operator using height-packed layout where channels
are the packed dimension (WHCN dim 1). Depthwise conv applies a separate filter
to each channel independently (groups=C), so 4 channels can be processed in
parallel using element-wise vec4 FMA over kernel positions.
Thread mapping: X=C/4, Y=L_out, Z=N. Each thread computes one output texel
(4 channels at one spatial position). Inner loop iterates over kernel positions
K with bounds-checked input access for padding.
Weight [C,1,K] is prepacked as channels-packed so each vec4 load gives 4
channels' weights at one kernel position. Supports both buffer and texture3d
storage, fp32/fp16, optional bias, and arbitrary stride/padding/dilation.
Registered as et_vk.conv1d_dw.default (standalone custom op).
Performance on Adreno 750 (S24):
Differential Revision: D97344091