Support int8 texture tensors without requiring int8 buffers #4485
Conversation
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/4485
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit c269cf9 with merge base 1090bcd ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D60536832 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D60536832 |
SS-JIA
force-pushed
the
export-D60536832
branch
from
ad47a9a to
e61fc08
Compare
SS-JIA
added a commit
to SS-JIA/executorch-1
that referenced
this pull request
Aug 1, 2024
…4485) Summary: Pull Request resolved: pytorch#4485 ## Context By default, storage buffers in Vulkan must contain 32 bit data types; using 8 bit and 16 bit data types in buffers can be enabled optionally by supporting the [VK_KHR_8bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_8bit_storage.html) extension or the [VK_KHR_16bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_16bit_storage.html) extension respectively. Previously, 8-bit and 16-bit tensors were enabled by using those extensions; however, this meant that 8-bit and 16-bit tensors could not be used if the Vulkan driver does not support the corresponding extension. This diff adds support for 8-bit texture-backed tensors without the need for the VK_KHR_8bit_storage extension. This is done by introducing shaders that manually pack and repack 4 8-bit integers into a single int32 value. Once the tensor data has been transferred to an image texture (which will use the `VK_FORMAT_R8G8B8A8_SINT` image format) the extension will no longer be required. Differential Revision: D60536832
jorgep31415
approved these changes
Aug 2, 2024
SS-JIA
added a commit
to SS-JIA/executorch-1
that referenced
this pull request
Aug 2, 2024
…4485) Summary: Pull Request resolved: pytorch#4485 ## Context By default, storage buffers in Vulkan must contain 32 bit data types; using 8 bit and 16 bit data types in buffers can be enabled optionally by supporting the [VK_KHR_8bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_8bit_storage.html) extension or the [VK_KHR_16bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_16bit_storage.html) extension respectively. Previously, 8-bit and 16-bit tensors were enabled by using those extensions; however, this meant that 8-bit and 16-bit tensors could not be used if the Vulkan driver does not support the corresponding extension. This diff adds support for 8-bit texture-backed tensors without the need for the VK_KHR_8bit_storage extension. This is done by introducing shaders that manually pack and repack 4 8-bit integers into a single int32 value. Once the tensor data has been transferred to an image texture (which will use the `VK_FORMAT_R8G8B8A8_SINT` image format) the extension will no longer be required. Differential Revision: D60536832
|
This pull request was exported from Phabricator. Differential Revision: D60536832 |
SS-JIA
force-pushed
the
export-D60536832
branch
from
e61fc08 to
c08a603
Compare
…4485) Summary: Pull Request resolved: pytorch#4485 ## Context By default, storage buffers in Vulkan must contain 32 bit data types; using 8 bit and 16 bit data types in buffers can be enabled optionally by supporting the [VK_KHR_8bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_8bit_storage.html) extension or the [VK_KHR_16bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_16bit_storage.html) extension respectively. Previously, 8-bit and 16-bit tensors were enabled by using those extensions; however, this meant that 8-bit and 16-bit tensors could not be used if the Vulkan driver does not support the corresponding extension. This diff adds support for 8-bit texture-backed tensors without the need for the VK_KHR_8bit_storage extension. This is done by introducing shaders that manually pack and repack 4 8-bit integers into a single int32 value. Once the tensor data has been transferred to an image texture (which will use the `VK_FORMAT_R8G8B8A8_SINT` image format) the extension will no longer be required. Reviewed By: jorgep31415 Differential Revision: D60536832
|
This pull request was exported from Phabricator. Differential Revision: D60536832 |
SS-JIA
force-pushed
the
export-D60536832
branch
from
c08a603 to
c269cf9
Compare
|
This pull request has been merged in 448c7d3. |
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Changes following from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Changes following from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) ghstack-source-id: 246607065 Pull Request resolved: #5934
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
…ture" Changes following from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Changes following from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Pull Request resolved: #5934 Changes following from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246658454 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
…ture" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246672275 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
…ture" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 7, 2024
Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246725187 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246850704 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
…ture" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
…ture" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246869292 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
…ture" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 8, 2024
Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246872869 @exported-using-ghexport Differential Revision: [D63918659](https://our.internmc.facebook.com/intern/diff/D63918659/)
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 8, 2024
Summary: Pull Request resolved: #5934 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. ghstack-source-id: 246872869 exported-using-ghexport bypass-github-export-checks bypass-github-pytorch-ci-checks bypass-github-executorch-ci-checks Reviewed By: SS-JIA Differential Revision: D63918659 fbshipit-source-id: d4095228fac24be118af8bb0bf42b453ca0a03a2
jorgep31415
added a commit
that referenced
this pull request
Oct 9, 2024
…ge, take 2" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support texture2d and support uint8, respectively. Differential Revision: [D64076249](https://our.internmc.facebook.com/intern/diff/D64076249/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 9, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support texture2d and support uint8, respectively. Differential Revision: [D64076249](https://our.internmc.facebook.com/intern/diff/D64076249/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 9, 2024
…ge, take 2" "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support texture2d and support uint8, respectively. Differential Revision: [D64076249](https://our.internmc.facebook.com/intern/diff/D64076249/) [ghstack-poisoned]
jorgep31415
added a commit
that referenced
this pull request
Oct 9, 2024
"bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support texture2d and support uint8, respectively. Differential Revision: [D64076249](https://our.internmc.facebook.com/intern/diff/D64076249/) [ghstack-poisoned]
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 9, 2024
Summary: Pull Request resolved: #6028 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from #4485 to support `texture2d` and support `uint8`, respectively. (Redo of #5934) ghstack-source-id: 247078906 Reviewed By: SS-JIA Differential Revision: D64076249 fbshipit-source-id: 541cfddf92c55ebd4c6e39c6bd041fb5aa78b3b9
kedarnath03
pushed a commit
to kedarnath03/executorch
that referenced
this pull request
Jun 25, 2025
Pull Request resolved: pytorch/executorch#6028 "bitw8" = bit width 8, which is equivalent to 8-bit. We use "bitw8" as the name since shader compilation disallows names starting with a digit. Changes follow from pytorch/executorch#4485 to support `texture2d` and support `uint8`, respectively. (Redo of pytorch/executorch#5934) ghstack-source-id: 247078906 Differential Revision: [D64076249](https://our.internmc.facebook.com/intern/diff/D64076249/)
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.
Summary:
Context
By default, storage buffers in Vulkan must contain 32 bit data types; using 8 bit and 16 bit data types in buffers can be enabled optionally by supporting the VK_KHR_8bit_storage extension or the VK_KHR_16bit_storage extension respectively.
Previously, 8-bit and 16-bit tensors were enabled by using those extensions; however, this meant that 8-bit and 16-bit tensors could not be used if the Vulkan driver does not support the corresponding extension.
This diff adds support for 8-bit texture-backed tensors without the need for the VK_KHR_8bit_storage extension. This is done by introducing shaders that manually pack and repack 4 8-bit integers into a single int32 value. Once the tensor data has been transferred to an image texture (which will use the
VK_FORMAT_R8G8B8A8_SINTimage format) the extension will no longer be required.Differential Revision: D60536832