RegCount NITER calculation #4159
Closed
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/4159
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 8, 2024
Summary: Pull Request resolved: pytorch#4159 Differential Revision: D59405012
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 8, 2024
Summary: Pull Request resolved: pytorch#4159 Differential Revision: D59405012
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 8, 2024
Summary: Pull Request resolved: pytorch#4159 Differential Revision: D59405012
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
SS-JIA
approved these changes
Jul 9, 2024
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 10, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
Esteb37
force-pushed
the
export-D59405012
branch
from
fdaace9 to
da22ab2
Compare
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Reviewed By: SS-JIA Differential Revision: D59405012
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Reviewed By: SS-JIA Differential Revision: D59405012
Esteb37
force-pushed
the
export-D59405012
branch
from
da22ab2 to
d3998ea
Compare
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Reviewed By: SS-JIA Differential Revision: D59405012
|
This pull request was exported from Phabricator. Differential Revision: D59405012 |
Esteb37
force-pushed
the
export-D59405012
branch
from
d3998ea to
23949a0
Compare
Esteb37
pushed a commit
to Esteb37/executorch
that referenced
this pull request
Jul 11, 2024
Summary: Pull Request resolved: pytorch#4159 This adds an internal implementation of https://github.com/microsoft/ArchProbe. This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See [this paper ](https://www.microsoft.com/en-us/research/uploads/prod/2022/02/mobigpu_mobicom22_camera.pdf), page 4, for more information. This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests. The kernel looks like the following for any K number of registers: float reg_data0 = float(niter) + 0; float reg_data1 = float(niter) + 1; ... float reg_dataK = float(niter) + K; int i = 0; for (; i < niter; ++i) { reg_data0 *= reg_dataK; reg_data1 *= reg_data0; reg_data2 *= reg_data1; ... reg_dataK *= reg_data(K-1); } i = i >> 31; buffer_out.data[0 * i] = reg_data0; buffer_out.data[1 * i] = reg_data1; ... buffer_out.data[K * i] = reg_dataK; Differential Revision: D59405012 Reviewed By: SS-JIA
|
This pull request has been merged in ac1c7d0. |
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.
This adds an internal implementation of https://github.com/microsoft/ArchProbe.
This stack introduces a kernel that can be used to get the number of available registers on a mobile GPU by gradually increasing the number of accessed elements and detecting dramatic drops in performance. See this paper , page 4, for more information.
This first diff gets the number of iterations (NITER) that can run in 1000us, to be used in the following tests.
The kernel looks like the following for any K number of registers:
Differential Revision: D59405012