PiperOrigin-RevId: 280227040
Change-Id: I39854dfbdf998b08c7d276a50cfb5ee4406b861b

This refactors the implementation of block signature(type) conversion to not insert fake cast operations to perform the type conversion, but to instead create a new block containing the proper signature. This has the benefit of enabling the use of pre-computed analyses that rely on mapping values. It also leads to a much cleaner implementation overall. The major user facing change is that applySignatureConversion will now replace the entry block of the region, meaning that blocks generally shouldn't be cached over calls to applySignatureConversion.

PiperOrigin-RevId: 280226936
Change-Id: Ibb8cb815c59bb1f805377d11bb66c6ff9808ec82

Add 'quantization_parameters' field to python interpreter to allow getting the full per-axis quantization parameters via python.

PiperOrigin-RevId: 280220657
Change-Id: I3fe27b9c88a717f9c85e59164cd12449679006e3

This creates a central place in the documentation where MLIR-specific terminology is defined. See discussion on the MLIR forum (https://groups.google.com/a/tensorflow.org/g/mlir/c/5YXDSdu76Hk).

PiperOrigin-RevId: 280220365
Change-Id: I370d5bd7bc329b81212d9cca37f107dd0d8243be

PiperOrigin-RevId: 280220301
Change-Id: Ib36618937995f92d65660d806ecf90e9d608ab9a

PiperOrigin-RevId: 280218429
Change-Id: I51024998ef3f85e1f99fface6f6f36a1443ef031

The current implementation silently fails if the '@' identifier isn't present, making it similar to the 'optional' parse methods. This change renames the current implementation to 'Optional' and adds a new 'parseSymbolName' that emits an error.

PiperOrigin-RevId: 280214610
Change-Id: I44da6f57992b814f6e503f06c220cd8973d3d013

Previously, `get_concrete_function` did not call _garbage_collector.release()
so the graph was deleted when the function was deleted.

Also, introduce _get_concrete_function_garbage_collected for places
that we want to keep the existing behavior.

PiperOrigin-RevId: 280213338
Change-Id: Ica498342b61c129132d254adf61e30256b24efe3

(cf #29871)

PiperOrigin-RevId: 280208993
Change-Id: Ia604e0ce832c7fb2adb5c2eff631a7cd1e0c2629

For some models, resolving constant expressions in the bridge leads to
non-linear slowdown.

PiperOrigin-RevId: 280205828
Change-Id: Icaa0394d8045c31544b66123857da7cb1b1701b9

PiperOrigin-RevId: 280204674
Change-Id: Idb551efdde6aa51c2fcfd6fe6e82d175bdcc9bb0

PiperOrigin-RevId: 280200721
Change-Id: Icfaffb567b970da140e9b0d3a6c2093452893f01

PiperOrigin-RevId: 280198370
Change-Id: I67e4117575ff3ac00cbda0ac6bfc971c2fbeaf74

PiperOrigin-RevId: 280185051
Change-Id: I5be656a7a4fa4c6d434b79e59a532dfd8a6dd455

PiperOrigin-RevId: 280184071
Change-Id: I6a0f22809a7d11c5bad7254c20e79525bf3ab88a

It will just be ignored by the starlark rule.

PiperOrigin-RevId: 280183097
Change-Id: I456d8a1ef071c5b3a82701cdc4de70d434811cba

Currently both works because of an alias rule, but we want to remove the alias
rule.

PiperOrigin-RevId: 280163952
Change-Id: Icd6e1938aacf9a76186049558f82f358c7783ea4

xla_proto and hlo_proto are now tf_proto_library_cc targets.

PiperOrigin-RevId: 280157517
Change-Id: I51a5c3720856908eb3be4e1b8ea6283d03e7fed6

Move the memory stats output back to BenchmarkModel::Run so that it's always there regardless which benchmark listener is used.

PiperOrigin-RevId: 280149185
Change-Id: Ic9932baea72083117e37c75062c2bd9e9c652d9e

PiperOrigin-RevId: 280146999
Change-Id: I6ca6f95e0fc6725bc23823934be1730b19f8478a

This was forgotten when migrating from xla_proto_library to
tf_proto_library_cc.

PiperOrigin-RevId: 280146847
Change-Id: Ieb837fe06a7af77f559bb287ff22cb9956964876

PiperOrigin-RevId: 280146467
Change-Id: Ibb580f0f13d9afd4f7c373bbb0eab7022b1a95ce

This was forgotten when migrating from xla_proto_library to
tf_proto_library_cc.

PiperOrigin-RevId: 280143885
Change-Id: Ibd23bda2239f819dc9ba96bd16e17c42a53ff55a

failure with TRT 6.0.

PiperOrigin-RevId: 280129198
Change-Id: Ife5ee7958c8cdc1438c11e059dd18c834304c9c5

Since VariableOp is serialized during processBlock, we add two more fields,
`functionHeader` and `functionBody`, to collect instructions for a function.
After all the blocks have been processed, we append them to the `functions`.

Also, fix a bug in processGlobalVariableOp. The global variables should be
encoded into `typesGlobalValues`.

PiperOrigin-RevId: 280105366
Change-Id: I5389451312d55869a1f6740d87597444013421b2

PiperOrigin-RevId: 280102761
Change-Id: I057c8074e3e25ee04a44e4ce77fd4981babd17f4

PiperOrigin-RevId: 280097036
Change-Id: I4578cacb876f3d27bb860c3794b685f2bc26da03

PiperOrigin-RevId: 280096904
Change-Id: I0d5e9233da52afd2f7c13094574a59c0d8403403

PiperOrigin-RevId: 280096487
Change-Id: I506633f77f4042fa19f81c078f307f843a9e0c1f

PiperOrigin-RevId: 280090706
Change-Id: I628bde953f31500ce3588127c7e42414f002f8b4

PiperOrigin-RevId: 280085436
Change-Id: If02ccdebcddeead5b80f85790a7d7e8a05d30c4d

PiperOrigin-RevId: 280080043
Change-Id: Id152c198ebf20256fc14b2ea1e16b8c5db71844c

For that we need to generate multiple device events (one per device) AND one driver callback events.

Add some trace me for nccl_manager.cc

PiperOrigin-RevId: 280077382
Change-Id: Ice0ebe4b98f3a6a222e210d266666b0e7fbf8171

Add 'quantization_parameters' field to python interpreter to allow getting the full per-axis quantization parameters via python.

PiperOrigin-RevId: 280077041
Change-Id: I32e7c8053e7550b47af74600a9d5a0d3b01b54bb

PiperOrigin-RevId: 280072888
Change-Id: I7293fe0632fea5a29f3c8fa644f1437adc37cdd0

OwnedIterators and MultiDeviceIterators get created using their components at least twice during function tracing. We don't want to run the deleter in those cases as they are just referring to the original resource and not creating a new one. So we just create the deleter the first time.

PiperOrigin-RevId: 280071920
Change-Id: I266ac500246354e6e4d6145835c4f90926ec5715

PiperOrigin-RevId: 280070797
Change-Id: I3c8d6ba04e51c05d0ffb515b87b8f686f74c31a0

PiperOrigin-RevId: 280067957
Change-Id: I0ff98b7754b3c4356db1ee32eede9582ab3c7a72

PiperOrigin-RevId: 280067091
Change-Id: I83939e8da82e9e488685afa36903ec9980dc5b50

PiperOrigin-RevId: 280065803
Change-Id: I9079f18e1064bf2cef8523d9e3c5f060eeda0a8e