Merged
Conversation
3e8cd49 to
c8ee0a1
Compare
apaszke
reviewed
Mar 25, 2017
Contributor
apaszke
left a comment
There was a problem hiding this comment.
Looks great! I only have some suggestions about the naming, but that's it. I only want to take a look about the generated files before merging it.
tools/cwrap/cwrap.py
Outdated
| call_arg = ', '.join(call_arg) | ||
| for plugin in self.plugins: | ||
| arg_unpack = plugin.process_all_unpacks(arg_unpack, option) | ||
| call_arg = plugin.process_all_unpacks(call_arg, option) |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
tools/cwrap/cwrap.py
Outdated
| result.append(res) | ||
| return result | ||
|
|
||
| def build_arg_assign(self, arguments, arg_unpack): |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
tools/cwrap/cwrap.py
Outdated
| assignement = [] | ||
| call_arg = [] | ||
| # If type names needs to be changed | ||
| arguments = self.get_formal_args(arguments) |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
Contributor
|
The generated code looks good too! |
apaszke
approved these changes
Mar 26, 2017
Contributor
|
@pytorchbot add to whitelist |
Contributor
|
Thank you! |
jjsjann123
pushed a commit
to jjsjann123/pytorch
that referenced
this pull request
Nov 5, 2021
…ions (pytorch#1131) Fixes pytorch#1102 This PR implements the second approach mentioned in pytorch#1102 For example, indexing and predicates are changed from: ``` = T0[(((((nvfuser_index_t)blockIdx.x) * ((nvfuser_index_t)blockDim.y)) + ((nvfuser_index_t)threadIdx.y)) * T0.stride[0])] ``` to: ``` = T0[(((((nvfuser_index_t)blockIdx.x) * 4) + ((nvfuser_index_t)threadIdx.y)) * T0.stride[0])] ``` The use of `blockDim.y` is replaced by the extent of the second axis of `T0`, which is `4` in this case. This change only matters when a parallel type is not exact (in this case `TIDy`). The indexing change only needed to change `getExtent` in index_compute.cpp. However, we also need to predicate `threadIdx` and `blockIdx` to be smaller than IterDomain extents. That's implemented as `ParallelizedDomainPredicate` in predicate_compute.h.
syed-ahmed
pushed a commit
to syed-ahmed/pytorch
that referenced
this pull request
Sep 22, 2022
* Validate MacOS conda packages and wheels * Call all matrices using reusable workflow
akashveramd
pushed a commit
to ROCm/pytorch
that referenced
this pull request
Jun 13, 2025
Implement iRoPE for llama4 Needs the fix from PyTorch: pytorch#151270
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
My first
cwrapPR, thorough review needed !Extend cwrap stucture to assign the arguments to variables named
arg_*before the function call.This simplifies the
before_callthat can now access directly the arguments (the previous behaviour with${arg0}is still allowed right now for backward compatibility).Also allows the
before_callto modify the arguments!!Also fixes #146 as the unpacking is done before.
Sample before:
Sample after:
THTensor* arg_self = ((THPTensor*)self)->cdata; PyThreadState *_save = NULL; try { long ndim = arg_self->nDimension; THPUtils_assert(ndim == 2, "t_() expects a 2D tensor, but self is %ldD", ndim); Py_UNBLOCK_THREADS; THTensor_(transpose)(LIBRARY_STATE arg_self, arg_self, 0, 1); Py_BLOCK_THREADS; Py_INCREF(self); return (PyObject*)self; } catch (...) { if (_save) { Py_BLOCK_THREADS; } throw; }