OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 12:21:27 +01:00
Code Issues Actions 37 Packages Projects Releases Wiki Activity
72,046 Commits 8,304 Branches 150 Tags 1.5 GiB
fd90991790
Commit Graph

5 Commits

Author SHA1 Message Date
Jack Khuu
e9674d146c [Specialized Kernel] Propagate Specialized Kernel Support through ComputeCodegenUnboxedKernels (#103113) 
Updating ComputeCodegenUnboxedKernels to accept and write out kernel information to RegisterCodegenUnboxedKernels.cpp

Differential Revision: [D46486195](https://our.internmc.facebook.com/intern/diff/D46486195/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/103113
Approved by: https://github.com/larryliu0820, https://github.com/kirklandsign
 2023-06-14 10:18:16 +00:00
Jack Khuu
d0c0e13b69 [Specialized Kernel] Translate Kernel Assignment Logic from function.yaml to native_functions.yaml (#102576) 
Updating `gen_executorch.translate_native_yaml()` to translate kernel assignments when converting `functions.yaml` to `native_functions.yaml`
---
Functions.yaml format:
```
- func: add.out
	type_alias:
		T0: [<Type>, <Type>]
		T1: [<Type>]
	dim_order_alias:
		D0: [0, 1, 2, 3]
		D1: [0, 3, 2, 1]
	kernels:
		- arg_meta: null
		  kernel_name: default_impl
		- arg_meta:
			self: [T0, D0]
			other:[T0, D0]
			out: [T0, D0]
		  kernel_name: test_impl
```

native_functions.yaml format
```
func: add.out(Tensor self, Tensor other, *, Scalar alpha=1, Tensor(a!) out) -> Tensor(a!)
  kernel:
    default: default_impl
    v<Version>/<TYPE Enum>;<DIM Order>|<TYPE Enum>;<DIM Order>|<TYPE Enum>;<DIM Order>: test_impl
```
Example: **'v1/6;0,1,2,3|3;0,1,2,3|6;0,1,2,3' : 'test_impl'**

## Note:
- If a "kernels" field is not present in functions.yaml (as it currently is), the output is unaffected
---
Design Doc: https://docs.google.com/document/d/1gq4Wz2R6verKJ2EFseLyPdAF0wqomnCrVDDJpRkYsRw/edit?kh_source=GDOCS#

Differential Revision: [D45971107](https://our.internmc.facebook.com/intern/diff/D45971107/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102576
Approved by: https://github.com/larryliu0820
 2023-06-08 23:42:24 +00:00
Mengwei Liu
eebe0ee141 [Executorch][codegen] Add ETKernelIndex for aggregating all kernels for kernel (#102874) 
Summary:
keys and change codegen to take ETKernelIndex

We are adding support for dtype and dim order specialized kernel registration. This requires us to reorganize `BackendIndex` (which is a `Dict[DispatchKey, Dict[OperatorName, BackendMetadata]]`) to be `Dict[OperatorName, Dict[ETKernelKey, BackendMetadata]]`. This PR adds new data structures in order to support this change:

* `ETKernelKey` to retrieve a certain kernel from the registry.
* `ETKernelIndex`, the dictionary from operator name to kernel key to kernel mapping.

Note that the codegen logic is not changed yet, we need subsequent diffs to actually generate code for different kernel keys.

Test Plan: Added tests

Reviewed By: Jack-Khuu

Differential Revision: D46407096

Pull Request resolved: https://github.com/pytorch/pytorch/pull/102874
Approved by: https://github.com/Jack-Khuu, https://github.com/kirklandsign
 2023-06-03 17:23:42 +00:00
Nikita Shulga
fb0729054b Revert "[Executorch][codegen] Add ETKernelIndex for aggregating all kernels for kernel (#102565)" 
This reverts commit 019c38624c /
https://github.com/pytorch/pytorch/pull/102565 as it breaks
ExecutorchBuilds.
 2023-06-01 12:35:23 -07:00
Larry Liu
019c38624c [Executorch][codegen] Add ETKernelIndex for aggregating all kernels for kernel (#102565) 
keys and change codegen to take ETKernelIndex

We are adding support for dtype and dim order specialized kernel registration. This requires us to reorganize `BackendIndex` (which is a `Dict[DispatchKey, Dict[OperatorName, BackendMetadata]]`) to be `Dict[OperatorName, Dict[ETKernelKey, BackendMetadata]]`. This PR adds new data structures in order to support this change:

* `ETKernelKey` to retrieve a certain kernel from the registry.
* `ETKernelIndex`, the dictionary from operator name to kernel key to kernel mapping.

Note that the codegen logic is not changed yet, we need subsequent diffs to actually generate code for different kernel keys.

Differential Revision: [D46206339](https://our.internmc.facebook.com/intern/diff/D46206339/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102565
Approved by: https://github.com/Jack-Khuu
 2023-05-31 09:41:36 +00:00