[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. by trviv · Pull Request #7145 · pytorch/executorch

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

trviv · 2024-12-02T19:48:29Z

Stack from ghstack (oldest at bottom):

[ET-VK] Adding convenience functions in Compute graph to get PushConstantDataInfo for various attributes of a tensor. #7224
[ET-VK] Add PushConstantDataInfo and vector to hold push constants data in DispatchNode. #7223
[ET-VK] Adding UniformData struct in vTensor class to store uniform data, which will be stored using shared ptr and can be shared with push constants. #7222
[ET-VK] Adding function to set push constants in Command buffer. #7221
[ET-VK] Removing unnecessary and redundant members from StagingBuffer. #7220
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
->[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7145
[ET-VK] Removing unnecessary and redundant members from VulkanBuffer and ParamsBuffer. #7144

This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage.

Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

pytorch-bot · 2024-12-02T19:48:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit cee5e89 with merge base 63238ab ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

facebook-github-bot · 2024-12-02T19:48:51Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) ghstack-source-id: 256050917 Pull Request resolved: #7145

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-02T23:53:46Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-04T00:33:36Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-05T05:10:18Z

This pull request was exported from Phabricator. Differential Revision: D66655991

…ined tensor object, to reduce Value struct size from 448 to 80 bytes." This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) [ghstack-poisoned]

facebook-github-bot · 2024-12-06T16:53:30Z

This pull request was exported from Phabricator. Differential Revision: D66655991

… object, to reduce Value struct size from 448 to 80 bytes. Pull Request resolved: #7145 This diff aims to reduce the size of the Value struct in the Executorch Vulkan runtime by storing a unique pointer to the Tensor object instead of an inlined tensor object. This change reduces the size of the Value struct from 448 bytes to 80 bytes, which can improve performance and reduce memory usage. ghstack-source-id: 256911524 @exported-using-ghexport Differential Revision: [D66655991](https://our.internmc.facebook.com/intern/diff/D66655991/) --------- Co-authored-by: Vivek Trivedi <5340687+trivedivivek@users.noreply.github.com>

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Dec 2, 2024

facebook-github-bot added the fb-exported label Dec 2, 2024

junpi3 approved these changes Dec 2, 2024
View reviewed changes

trviv mentioned this pull request Dec 2, 2024
[ET-VK] Store unique ptr to StagingBuffer and SymInt in Value instead of inlined object, to reduce Value struct size from 80 to 32 bytes. #7152
Merged

trviv added the release notes: backends label Dec 2, 2024

facebook-github-bot merged commit f738e55 into gh/trivedivivek/17/baseDec 6, 2024

facebook-github-bot deleted the gh/trivedivivek/17/head branch December 6, 2024 19:06

facebook-github-bot temporarily deployed to cherry-pick-bot December 6, 2024 19:06 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Dec 6, 2024
[ET-VK] Store unique ptr to Tensor in Value instead of inlined tensor object, to reduce Value struct size from 448 to 80 bytes. #7227
Merged

trviv mentioned this pull request Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for view op #7229
Merged

This was referenced Dec 6, 2024
[ET-VK] Replace Uniform buffers with push constants for binary op #7230
Merged
[ET-VK] Replace Uniform buffers with push constants for permute op #7231
Merged

Uh oh!

Conversation

trviv commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-botBot commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

✅ No Failures

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 4, 2024

Uh oh!

facebook-github-bot commented Dec 5, 2024

Uh oh!

facebook-github-bot commented Dec 6, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

trviv commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-botBot commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

✅ No Failures

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 4, 2024

Uh oh!

facebook-github-bot commented Dec 5, 2024

Uh oh!

facebook-github-bot commented Dec 6, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

trviv commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-botBot commented Dec 2, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7145

✅ No Failures

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 2, 2024

Uh oh!

facebook-github-bot commented Dec 4, 2024

Uh oh!

facebook-github-bot commented Dec 5, 2024

Uh oh!

facebook-github-bot commented Dec 6, 2024

Uh oh!

Reviewers

trviv commented Dec 2, 2024•
edited
Loading

pytorch-botBot commented Dec 2, 2024•
edited
Loading

trviv commented Dec 2, 2024•
edited
Loading

pytorch-botBot commented Dec 2, 2024•
edited
Loading

trviv commented Dec 2, 2024•
edited
Loading

pytorch-botBot commented Dec 2, 2024•
edited
Loading

trviv commented Dec 2, 2024•
edited
Loading

pytorch-botBot commented Dec 2, 2024•
edited
Loading

trviv commented Dec 2, 2024•
edited
Loading

pytorch-botBot commented Dec 2, 2024•
edited
Loading