90f0d7f935
Add support for lowering mhlo.conv to Linalg on tensors.
...
This pattern only works for normal convolutions. It does not work for depthwise
convolutions. The Linalg conv ops are defined with static rank, so it only
supports 1d/2d/3d cases, which are the most typical cases.
This also refactors out the same check in lmhlo.conv lowering.
PiperOrigin-RevId: 359503527
2021-02-25 05:59:08 -08:00
c5f5d13930
[MLIR] Add dependency from "lhlo_ops_structs_inc_gen" to "lhlo" target.
...
PiperOrigin-RevId: 359488020
2021-02-25 03:51:36 -08:00
45a1249fe2
Add support for lowering mhlo.pad to linalg.pad_tensor
...
The change upstreams the pattern from IREE repo to MHLO repo.
PiperOrigin-RevId: 359481543
2021-02-25 03:00:39 -08:00
459362b206
Integrate LLVM at llvm/llvm-project@d8a8e5d624
...
Updates LLVM usage to match
[d8a8e5d6240a](https://github.com/llvm/llvm-project/commit/d8a8e5d6240a )
PiperOrigin-RevId: 359475769
2021-02-25 02:20:12 -08:00
7dc1543bb2
Integrate LLVM at llvm/llvm-project@96a3dfeb93
...
Updates LLVM usage to match
[96a3dfeb9303](https://github.com/llvm/llvm-project/commit/96a3dfeb9303 )
PiperOrigin-RevId: 359394503
2021-02-24 16:17:23 -08:00
d9d6ab105e
Integrate LLVM at llvm/llvm-project@5c74c6be3c
...
Updates LLVM usage to match
[5c74c6be3c29](https://github.com/llvm/llvm-project/commit/5c74c6be3c29 )
PiperOrigin-RevId: 359354498
2021-02-24 13:19:53 -08:00
89f7f2bd65
Lower integer matmuls to linalg
...
PiperOrigin-RevId: 359306495
2021-02-24 09:45:07 -08:00
475b4a06a5
Add support for lowering mhlo.slice to subtensor.
...
PiperOrigin-RevId: 359297978
2021-02-24 09:06:09 -08:00
b478bdf00e
Integrate LLVM at llvm/llvm-project@de40423c85
...
Updates LLVM usage to match
[de40423c8512](https://github.com/llvm/llvm-project/commit/de40423c8512 )
PiperOrigin-RevId: 359280337
2021-02-24 07:32:25 -08:00
3e2ff6e253
Integrate LLVM at llvm/llvm-project@861dbe1a02
...
Updates LLVM usage to match
[861dbe1a021e](https://github.com/llvm/llvm-project/commit/861dbe1a021e )
PiperOrigin-RevId: 359091800
2021-02-23 11:28:36 -08:00
ac0552f127
[MLIR][HLO] Remove duplicate `PopulateTransformUnrankedHloPatterns`
...
PiperOrigin-RevId: 359046173
2021-02-23 07:50:47 -08:00
2df1bb3d6b
Integrate LLVM at llvm/llvm-project@6c9541d4dd
...
Updates LLVM usage to match
[6c9541d4ddfd](https://github.com/llvm/llvm-project/commit/6c9541d4ddfd )
PiperOrigin-RevId: 359020255
2021-02-23 04:38:26 -08:00
855cd484fc
Integrate LLVM at llvm/llvm-project@557d2ade01
...
Updates LLVM usage to match
[557d2ade016f](https://github.com/llvm/llvm-project/commit/557d2ade016f )
PiperOrigin-RevId: 358925409
2021-02-22 16:05:25 -08:00
b45eaf8d33
Integrate LLVM at llvm/llvm-project@12edddafac
...
Updates LLVM usage to match
[12edddafac45](https://github.com/llvm/llvm-project/commit/12edddafac45 )
PiperOrigin-RevId: 358889182
2021-02-22 13:21:11 -08:00
0da7ea2545
[MLIR][HLO] Cleanup CMakeLists and comments.
...
- Cleanup CMakeLists file to remove unused argument and use a new function for
setting up lmhlo and lmhlo_gpu dialect targets.
- Fix inconsistently formatted copyright comment and fix header include guards.
PiperOrigin-RevId: 358865838
2021-02-22 11:36:29 -08:00
5adb7c6e12
[MLIR:LHLO] Add optional call target arg mapping to LMHLO CustomCall operations.
...
- XLA:HLO -> LMHLO conversion drops all token arguments and return values, however
custom calls that users write still expect to get buffer pointers for these token types.
- To be able to support this, add an optional call target argument mapping attribute to
LMHLO custom calls. When this attribute is present, it indicates the number of
arguments and returns that the custom call expects and also indicates which LMHLO
arg() or output() maps to which arg or result number of the custom call.
PiperOrigin-RevId: 358826664
2021-02-22 08:43:00 -08:00
a9cc1dcfa0
[mlir][hlo] Add basic rank-specialization for select
...
This just blows up everything to ranked (up to 6) and is probably quite slow.
This is sufficient to make kernelgen compile SelectV2.
PiperOrigin-RevId: 358777728
2021-02-22 02:41:12 -08:00
909574e393
Pass result element type to XlaBuilder for `mhlo.dot_general` and `mhlo.convolution` ops.
...
`mhlo.dot_general` and `mhlo.convolution` result element type might be different from operand element type. See `preferred_element_type` attribute that allows i8xi8 to i32 dot computation. `mhlo` to HLO exporter should pass the result element type to Xla builder to override the shape inference of XLA.
PiperOrigin-RevId: 358580718
2021-02-20 07:07:17 -08:00
f63c93399a
Integrate LLVM at llvm/llvm-project@7e54d7304b
...
Updates LLVM usage to match
[7e54d7304be2](https://github.com/llvm/llvm-project/commit/7e54d7304be2 )
PiperOrigin-RevId: 358516349
2021-02-19 17:49:00 -08:00
5cc9fa9d9d
Integrate LLVM at llvm/llvm-project@cd8cc00b9e
...
Updates LLVM usage to match
[cd8cc00b9e2b](https://github.com/llvm/llvm-project/commit/cd8cc00b9e2b )
PiperOrigin-RevId: 358329421
2021-02-18 21:09:36 -08:00
5400287d69
Integrate LLVM at llvm/llvm-project@b7e05c874b
...
Updates LLVM usage to match
[b7e05c874b5b](https://github.com/llvm/llvm-project/commit/b7e05c874b5b )
PiperOrigin-RevId: 358289004
2021-02-18 16:19:04 -08:00
27a8b18526
Integrate LLVM at llvm/llvm-project@f70cdc5b5c
...
Updates LLVM usage to match
[f70cdc5b5c7c](https://github.com/llvm/llvm-project/commit/f70cdc5b5c7c )
PiperOrigin-RevId: 358240325
2021-02-18 12:40:08 -08:00
72d3eee118
Integrate LLVM at llvm/llvm-project@5b094bfeb3
...
Updates LLVM usage to match
[5b094bfeb3cc](https://github.com/llvm/llvm-project/commit/5b094bfeb3cc )
PiperOrigin-RevId: 358214975
2021-02-18 10:52:25 -08:00
b42def4612
[mlir][hlo] Refactor rank specialization to allow an arbitrary number of inputs
...
This actually simplifies the code a bit.
PiperOrigin-RevId: 358201038
2021-02-18 09:53:03 -08:00
ca4034b56e
[mlir][hlo] Make select ready for dynamic shapes (ranked only for now)
...
Move tf.SelectV2 broadcast lowering to a chlo.broadcast_select op, and lower it
to broadcasts on mhlo from there.
PiperOrigin-RevId: 358179975
2021-02-18 08:08:40 -08:00
a6f03aecfb
Integrate LLVM at llvm/llvm-project@624fccba87
...
Updates LLVM usage to match
[624fccba87cc](https://github.com/llvm/llvm-project/commit/624fccba87cc )
PiperOrigin-RevId: 358167564
2021-02-18 06:50:29 -08:00
37e31f8b26
Lower Expm1 kernel to math.ExpM1.
...
PiperOrigin-RevId: 358152908
2021-02-18 04:54:23 -08:00
90b222cf6e
Integrate LLVM at llvm/llvm-project@55756f32f7
...
Updates LLVM usage to match
[55756f32f735](https://github.com/llvm/llvm-project/commit/55756f32f735 )
PiperOrigin-RevId: 358093073
2021-02-17 20:33:14 -08:00
b579bd5d9e
Support dynamic-shaped operand in verification of BroadcastInDim.
...
Verification of HLO_BroadcastInDimOp was previously failing or crashing if the
operand had a dynamic shape or was unranked. Update the verification code to
allow the operand to be unranked or have dynamic shape.
PiperOrigin-RevId: 358056793
2021-02-17 16:18:09 -08:00
dd237d4267
Integrate LLVM at llvm/llvm-project@892d2822b6
...
Updates LLVM usage to match
[892d2822b62e](https://github.com/llvm/llvm-project/commit/892d2822b62e )
PiperOrigin-RevId: 358038509
2021-02-17 14:52:43 -08:00
220deb3709
[MLIR][CHLO] Add legalization for `chlo.polygamma` to MHLO
...
PiperOrigin-RevId: 357954624
2021-02-17 08:33:01 -08:00
c06de24f6c
[MLIR][CHLO] Generalize lowering with upcast to n-ary operation
...
Allows reuse for zeta lowering now and for the polygamma lowering soon.
PiperOrigin-RevId: 357739910
2021-02-16 09:47:24 -08:00
81abaf364d
[MLIR][MHLO] Add polygamma op to the CHLO dialect
...
PiperOrigin-RevId: 357724465
2021-02-16 08:32:33 -08:00
02dab94054
Integrate LLVM at llvm/llvm-project@9f581815ae
...
Updates LLVM usage to match
[9f581815ae4d](https://github.com/llvm/llvm-project/commit/9f581815ae4d )
PiperOrigin-RevId: 357683119
2021-02-16 03:33:12 -08:00
9447374676
Integrate LLVM at llvm/llvm-project@715dc556b7
...
Updates LLVM usage to match
[715dc556b782](https://github.com/llvm/llvm-project/commit/715dc556b782 )
PiperOrigin-RevId: 357607282
2021-02-15 13:35:33 -08:00
af5343ff86
Integrate LLVM at llvm/llvm-project@2dbe88db58
...
Updates LLVM usage to match
[2dbe88db5804](https://github.com/llvm/llvm-project/commit/2dbe88db5804 )
PiperOrigin-RevId: 357587981
2021-02-15 09:58:25 -08:00
b594254c79
[mhlo] Lower int->bool to a comparison with zero
...
This matches what TF (and C++) do in this case.
PiperOrigin-RevId: 357566262
2021-02-15 06:38:09 -08:00
240a44de82
[mhlo] Lower int->int cast to sign extension instead of zero extension
...
Signless does not mean unsigned here. Currently mhlo only has signed types.
PiperOrigin-RevId: 357561712
2021-02-15 05:58:47 -08:00
8672735e9a
[mhlo] Lower float->bool to a comparison with zero
...
This matches what TF (and C++) do in this case.
PiperOrigin-RevId: 357553098
2021-02-15 04:36:36 -08:00
2fe0c33083
Integrate LLVM at llvm/llvm-project@16428a8d91
...
Updates LLVM usage to match
[16428a8d91a9](https://github.com/llvm/llvm-project/commit/16428a8d91a9 )
PiperOrigin-RevId: 357550807
2021-02-15 04:17:58 -08:00
89d81adf6d
[mhlo] Lower float->bool to a comparison with zero
...
This matches what TF (and C++) do in this case.
PiperOrigin-RevId: 357541594
2021-02-15 03:11:56 -08:00
3e80d91e73
[mhlo] Lower float->bool to a comparison with zero
...
This matches what TF (and C++) do in this case.
PiperOrigin-RevId: 357534118
2021-02-15 02:17:19 -08:00
824bc9c425
Improve broadcast transformation to treat dynamic shapes with 1 element as scalar.
...
A shape that contains exactly one element is effectively a scalar. This leads
to a speedup in cases where we have a binary op with one operand that is
effectively a scalar, because we can use the fast path.
PiperOrigin-RevId: 357515552
2021-02-14 23:25:41 -08:00
4060a86fe2
Integrate LLVM at llvm/llvm-project@2bfe27da17
...
Updates LLVM usage to match
[2bfe27da171e](https://github.com/llvm/llvm-project/commit/2bfe27da171e )
PiperOrigin-RevId: 357196336
2021-02-12 08:32:03 -08:00
e993082b97
Integrate LLVM at llvm/llvm-project@6f04addc8b
...
Updates LLVM usage to match
[6f04addc8b2e](https://github.com/llvm/llvm-project/commit/6f04addc8b2e )
PiperOrigin-RevId: 357170209
2021-02-12 05:21:12 -08:00
f7ac575dd9
Integrate LLVM at llvm/llvm-project@0c118831a3
...
Updates LLVM usage to match
[0c118831a37a](https://github.com/llvm/llvm-project/commit/0c118831a37a )
PiperOrigin-RevId: 357158589
2021-02-12 03:32:55 -08:00
a7365aadbe
Integrate LLVM at llvm/llvm-project@8151c1b442
...
Updates LLVM usage to match
[8151c1b44211](https://github.com/llvm/llvm-project/commit/8151c1b44211 )
PiperOrigin-RevId: 357080940
2021-02-11 16:17:46 -08:00
077df4f8cc
Integrate LLVM at llvm/llvm-project@418c218efa
...
Updates LLVM usage to match
[418c218efa95](https://github.com/llvm/llvm-project/commit/418c218efa95 )
PiperOrigin-RevId: 357010454
2021-02-11 10:49:27 -08:00
6fa6974e8d
[XLA/GPU] Plumb through Bitcast op for LMHLO.
...
Also remove BitcastOp. XLA bitcast requires the input buffer to alias the output buffer, which makes bitcast always a no-op.
PiperOrigin-RevId: 356884383
2021-02-10 19:45:40 -08:00
4e202135d1
Integrate LLVM at llvm/llvm-project@9db6e97a86
...
Updates LLVM usage to match
[9db6e97a8605](https://github.com/llvm/llvm-project/commit/9db6e97a8605 )
PiperOrigin-RevId: 356753027
2021-02-10 09:32:14 -08:00
Hanhan Wang
Rahul Joshi
A. Unique TensorFlower
Geoffrey Martin-Noble
Benjamin Kramer
Prakalp Srivastava
Adrian Kuegel
Richard Uhler
Tim Shen