Eliminate direct access to size/strides of THTensor; replace them with std::vector #9561
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
ezyang
requested review from
apaszke,
colesbury,
gchanan,
soumith and
zdevito
as code owners
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
Author
|
@pytorchbot retest this please |
1 similar comment
Contributor
Author
|
@pytorchbot retest this please |
Contributor
Author
|
@pytorchbot retest this please |
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Contributor
|
Not sure if std::vector is a good longterm choice since size and stride is so short. It also can't be compiled into very efficient code. Ideally we'd template on that size and branch somewhere, at least that led to the best code within CPUApplyUtils.h. Otherwise SmartVector could be a good choice for now. |
Contributor
Author
|
@cpuhrsch Yeah, some sort of inline size/stride is in the roadmap, but it's not blocking for the merge so I haven't done it. When we add a change like this some benchmarking will be needed. |
gchanan
approved these changes
Jul 19, 2018
Contributor
gchanan
left a comment
gchanan
left a comment
There was a problem hiding this comment.
I assume you are just trying to figure out where you need the to put the no-strict-overflow? Also, why do you need it?
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
aten/src/TH/THTensor.hpp
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
aten/src/TH/generic/THTensor.cpp
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
cmake/public/utils.cmake
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
setup.py
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
torch/CMakeLists.txt
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
CMakeLists.txt
Outdated
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
Contributor
facebook-github-bot
left a comment
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
This patch was very carefully constructed to avoid having to modify too many files; there are some obvious follow ups which I will be hitting later. - I didn't do stride. But the change for stride should look very similar. - I did NOT rename the field in question, so that direct accesses of the form foo->size[n] keep working. I intend to do a codemod to fix all of these shortly. - Anywhere a "public" API function made use of a int64_t* of sizes, I opted to just finagle it out of the tensor using THTensor_getSizePtr rather than try to rewrite all of these sites to use ArrayRef. They should use ArrayRef eventually, but not yet. - _THSizeDesc got an overload that understands ArrayRef (which a vector size is convertible to). Eventually we should get rid of all of these functions (because ArrayRef is printable via the AT_ERROR macros), but not today. - I ran into something very subtle in the implementation of sizes() for TensorDerived: I MUST use the dim as per Tensor::dim() (which correctly is zero for scalars), otherwise I'll give a nonsense sizes(). We can fix this eventually once Scalar is turned on internally. - I added two new functions THTensor_resizeSize and THTensor_setSize. Maybe these are eventually worth deifying as methods in the Tensor class, but for now I'm keeping them out-of-line just in case. Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
Signed-off-by: Edward Z. Yang <[email protected]>
zdevito
pushed a commit
to zdevito/ATen
that referenced
this pull request
Jul 19, 2018
…h std::vector (#9561) Summary: * THTensor now stores `sizes_` and `strides_` which is a `std::vector<int64_t>` * Anywhere a "public" API function made use of a int64_t* of sizes, I opted to just finagle it out of the tensor using THTensor_getSizePtr rather than try to rewrite all of these sites to use ArrayRef. They should use ArrayRef eventually, but not yet. * There are new utility functions for resizing sizes/strides in one go (THTensor_resizeDim), or replacing sizes and strides with completely new values (THTensor_setSizesAndStrides) * Anywhere you said `t->size[n] = 0`, we now say `THTensor_setSizeAt(t, n, 0)`, ditto for strides * Anywhere you said `t->size[n]`, we now say `t->size(n)` (coming soon: ditto for strides) Previous review of just the `std::vector` change in #9518, but I'm planning to merge this all in one go. Note for gchanan: review from commit "ci" and after Pull Request resolved: pytorch/pytorch#9561 Reviewed By: cpuhrsch Differential Revision: D8901926 Pulled By: ezyang fbshipit-source-id: 483cf275060ab0a13845cba1ece39dd127142510
ezyang
added a commit
to ezyang/pytorch
that referenced
this pull request
Jul 22, 2018
Summary: This pops off `refcount_`, `storage_`, `storage_offset_`; there are now no more direct accesses to these fields and we can make them private (with appropriate friending). Stacked on pytorch#9561 Pull Request resolved: pytorch#9591 Reviewed By: gchanan Differential Revision: D8922246 Pulled By: gchanan fbshipit-source-id: 63dd3eb9787c460373c3cc507886d3b0a30e7542
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 22, 2018
Summary: Pull Request resolved: #9683 This pops off `refcount_`, `storage_`, `storage_offset_`; there are now no more direct accesses to these fields and we can make them private (with appropriate friending). Stacked on #9561 Pull Request resolved: #9591 Reviewed By: SsnL Differential Revision: D8922246 Pulled By: ezyang fbshipit-source-id: dfae023d790e29ce652e2eab9a1628bbe97b318d
jramseyer
pushed a commit
to jramseyer/pytorch
that referenced
this pull request
Jul 30, 2018
…h std::vector (pytorch#9561) Summary: * THTensor now stores `sizes_` and `strides_` which is a `std::vector<int64_t>` * Anywhere a "public" API function made use of a int64_t* of sizes, I opted to just finagle it out of the tensor using THTensor_getSizePtr rather than try to rewrite all of these sites to use ArrayRef. They should use ArrayRef eventually, but not yet. * There are new utility functions for resizing sizes/strides in one go (THTensor_resizeDim), or replacing sizes and strides with completely new values (THTensor_setSizesAndStrides) * Anywhere you said `t->size[n] = 0`, we now say `THTensor_setSizeAt(t, n, 0)`, ditto for strides * Anywhere you said `t->size[n]`, we now say `t->size(n)` (coming soon: ditto for strides) Previous review of just the `std::vector` change in pytorch#9518, but I'm planning to merge this all in one go. Note for gchanan: review from commit "ci" and after Pull Request resolved: pytorch#9561 Reviewed By: cpuhrsch Differential Revision: D8901926 Pulled By: ezyang fbshipit-source-id: 483cf275060ab0a13845cba1ece39dd127142510
jramseyer
pushed a commit
to jramseyer/pytorch
that referenced
this pull request
Jul 30, 2018
Summary: Pull Request resolved: pytorch#9683 This pops off `refcount_`, `storage_`, `storage_offset_`; there are now no more direct accesses to these fields and we can make them private (with appropriate friending). Stacked on pytorch#9561 Pull Request resolved: pytorch#9591 Reviewed By: SsnL Differential Revision: D8922246 Pulled By: ezyang fbshipit-source-id: dfae023d790e29ce652e2eab9a1628bbe97b318d
goodlux
pushed a commit
to goodlux/pytorch
that referenced
this pull request
Aug 15, 2018
…h std::vector (pytorch#9561) Summary: * THTensor now stores `sizes_` and `strides_` which is a `std::vector<int64_t>` * Anywhere a "public" API function made use of a int64_t* of sizes, I opted to just finagle it out of the tensor using THTensor_getSizePtr rather than try to rewrite all of these sites to use ArrayRef. They should use ArrayRef eventually, but not yet. * There are new utility functions for resizing sizes/strides in one go (THTensor_resizeDim), or replacing sizes and strides with completely new values (THTensor_setSizesAndStrides) * Anywhere you said `t->size[n] = 0`, we now say `THTensor_setSizeAt(t, n, 0)`, ditto for strides * Anywhere you said `t->size[n]`, we now say `t->size(n)` (coming soon: ditto for strides) Previous review of just the `std::vector` change in pytorch#9518, but I'm planning to merge this all in one go. Note for gchanan: review from commit "ci" and after Pull Request resolved: pytorch#9561 Reviewed By: cpuhrsch Differential Revision: D8901926 Pulled By: ezyang fbshipit-source-id: 483cf275060ab0a13845cba1ece39dd127142510
goodlux
pushed a commit
to goodlux/pytorch
that referenced
this pull request
Aug 15, 2018
Summary: Pull Request resolved: pytorch#9683 This pops off `refcount_`, `storage_`, `storage_offset_`; there are now no more direct accesses to these fields and we can make them private (with appropriate friending). Stacked on pytorch#9561 Pull Request resolved: pytorch#9591 Reviewed By: SsnL Differential Revision: D8922246 Pulled By: ezyang fbshipit-source-id: dfae023d790e29ce652e2eab9a1628bbe97b318d
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
sizes_andstrides_which is astd::vector<int64_t>t->size[n] = 0, we now sayTHTensor_setSizeAt(t, n, 0), ditto for stridest->size[n], we now sayt->size(n)(coming soon: ditto for strides)Previous review of just the
std::vectorchange in #9518, but I'm planning to merge this all in one go.Note for @gchanan: review from commit "ci" and after