[ET-VK][qconv][ez] Make q8ta_im2col shader support stride_w != 1#17386
Merged
meta-codesync[bot] merged 8 commits intogh/SS-JIA/417/basefrom Feb 13, 2026
Merged
[ET-VK][qconv][ez] Make q8ta_im2col shader support stride_w != 1#17386meta-codesync[bot] merged 8 commits intogh/SS-JIA/417/basefrom
meta-codesync[bot] merged 8 commits intogh/SS-JIA/417/basefrom
Conversation
The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/17386
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit db64852 with merge base dcfd12d ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
This was referenced Feb 11, 2026
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
manuelcandales
approved these changes
Feb 12, 2026
added 4 commits
February 12, 2026 14:23
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
…w != 1" The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/) [ghstack-poisoned]
meta-codesync
Bot
merged commit 445c84c
into
gh/SS-JIA/417/base
196 of 197 checks passed
SS-JIA
pushed a commit
that referenced
this pull request
Feb 13, 2026
Pull Request resolved: #17386 The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. ghstack-source-id: 340983072 @exported-using-ghexport Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/)
SS-JIA
pushed a commit
that referenced
this pull request
Feb 13, 2026
Pull Request resolved: #17386 The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. ghstack-source-id: 340983072 @exported-using-ghexport Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/)
chizkiyahu
pushed a commit
to chizkiyahu/executorch
that referenced
this pull request
Feb 23, 2026
Pull Request resolved: pytorch#17386 The im2col path previously had two shaders: the generic `q8ta_im2col` which loaded input elements one int8 at a time and supported all strides, and `q8ta_im2col_4w4c` which loaded packed int32s (4 channels at once) but was restricted to stride_w == 1 because it assumed consecutive width positions in the input (i.e. `input_x_base + i`). This diff replaces both shaders with a single `q8ta_im2col` that uses the efficient packed loading approach from `q8ta_im2col_4w4c` while generalizing the width offset to `input_x_base + i * stride_x`. This removes the stride_w == 1 restriction and deletes the separate `q8ta_im2col_4w4c` shader entirely. The C++ dispatch is updated to reference the unified shader name, and the test gate on `stride.w == 1` for the im2col path is removed. Authored with Claude. ghstack-source-id: 340983072 @exported-using-ghexport Differential Revision: [D93000161](https://our.internmc.facebook.com/intern/diff/D93000161/)
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):
The im2col path previously had two shaders: the generic
q8ta_im2colwhich loaded input elements one int8 at a time and supported all strides,
and
q8ta_im2col_4w4cwhich loaded packed int32s (4 channels at once)but was restricted to stride_w == 1 because it assumed consecutive
width positions in the input (i.e.
input_x_base + i).This diff replaces both shaders with a single
q8ta_im2colthat usesthe efficient packed loading approach from
q8ta_im2col_4w4cwhilegeneralizing the width offset to
input_x_base + i * stride_x. Thisremoves the stride_w == 1 restriction and deletes the separate
q8ta_im2col_4w4cshader entirely.The C++ dispatch is updated to reference the unified shader name, and
the test gate on
stride.w == 1for the im2col path is removed.Authored with Claude.
Differential Revision: D93000161