| .. |
|
graphcycles
|
During contraction of an edge from "a" to "b" there is a choice which node should be used to represent the union of the nodes. Using the node with the largest degree minimizes the number of operation that should be performed.
|
2020-03-16 18:10:55 -07:00 |
|
kernels
|
[TF2XLA] [NFC] Refactor datastructures for resource variables to not require snapshotting for compilation
|
2020-06-22 12:55:49 -07:00 |
|
ops
|
minor spelling tweaks
|
2019-12-09 18:21:12 +09:00 |
|
tests
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
BUILD
|
Moving TPU ops components as TPU kernels library.
|
2020-05-20 16:36:27 -07:00 |
|
build_xla_ops_pass_test.cc
|
Merge pull request #33063 from bas-aarts:xla-merge
|
2019-10-23 09:14:40 -07:00 |
|
build_xla_ops_pass.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
build_xla_ops_pass.h
|
[NFC] Clarify documentation of JIT passes
|
2019-12-19 11:20:28 -08:00 |
|
clone_constants_for_better_clustering_test.cc
|
Make constant cloning more aggressive (i.e. make it clone more often)
|
2019-10-07 12:23:37 -07:00 |
|
clone_constants_for_better_clustering.cc
|
Make constant cloning more aggressive (i.e. make it clone more often)
|
2019-10-07 12:23:37 -07:00 |
|
clone_constants_for_better_clustering.h
|
Add an optimization pass that clones Constant nodes to make larger clusters
|
2019-03-04 15:25:22 -08:00 |
|
cluster_scoping_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
cluster_scoping_pass.cc
|
Merge pull request #31594 from trentlo:github-cluster-scoping
|
2019-08-21 20:15:13 -07:00 |
|
cluster_scoping_pass.h
|
Merge pull request #31594 from trentlo:github-cluster-scoping
|
2019-08-21 20:15:13 -07:00 |
|
compilability_check_util_test.cc
|
[Build cleanup] Split "core_cpu_impl" into fine-grained targets (4/n).
|
2020-04-28 09:53:59 -07:00 |
|
compilability_check_util.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
compilability_check_util.h
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
compilation_passes_test_main.cc
|
Change heuristics around creating small clusters
|
2019-03-13 17:31:28 -07:00 |
|
deadness_analysis_internal.h
|
Address review comments.
|
2019-05-21 17:04:21 -07:00 |
|
deadness_analysis_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
deadness_analysis.cc
|
Code review changes
|
2020-01-14 23:58:14 -05:00 |
|
deadness_analysis.h
|
Print out the deadness predicate on a mismatch
|
2019-05-16 15:37:24 -07:00 |
|
defs.cc
|
Move must-compilation from experimental_compile=True to a separate attribute
|
2019-12-20 20:31:28 -08:00 |
|
defs.h
|
Move must-compilation from experimental_compile=True to a separate attribute
|
2019-12-20 20:31:28 -08:00 |
|
device_util_test.cc
|
[XLA] Allow multiple compatible devices in device_util
|
2019-07-12 18:47:39 -07:00 |
|
device_util.cc
|
[XLA] Allow multiple compatible devices in device_util
|
2019-07-12 18:47:39 -07:00 |
|
device_util.h
|
Add forward declarations of friend'ed classes.
|
2019-09-02 23:59:27 -07:00 |
|
encapsulate_subgraphs_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
encapsulate_subgraphs_pass.cc
|
[TF/XLA] Minor clean up
|
2020-02-09 17:23:44 -08:00 |
|
encapsulate_subgraphs_pass.h
|
[NFC] Clarify documentation of JIT passes
|
2019-12-19 11:20:28 -08:00 |
|
encapsulate_util_test.cc
|
Remove unused inclusions and add missing inclusions
|
2019-02-05 14:06:22 -08:00 |
|
encapsulate_util.cc
|
Move error_codes.proto part 2.
|
2019-09-24 10:07:35 -07:00 |
|
encapsulate_util.h
|
minor spelling tweaks
|
2019-12-09 18:21:12 +09:00 |
|
encapsulate_xla_computations_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
encapsulate_xla_computations_pass.cc
|
Split out node_def_util, op_def_builder, op_def_util and attr_value_util targets in tensorflow/core/framework/BUILD. Split out
|
2020-01-06 11:46:01 -08:00 |
|
encapsulate_xla_computations_pass.h
|
[NFC] Clarify documentation of JIT passes
|
2019-12-19 11:20:28 -08:00 |
|
extract_outside_compilation_pass_test.cc
|
minor spelling tweaks
|
2019-12-09 18:21:12 +09:00 |
|
extract_outside_compilation_pass.cc
|
This CL optimizes C++11 range-based for loops where the variable is copied in each iteration but it would suffice to obtain it by const reference. This is only applied to loop variables of types that are expensive to copy which means they are not trivially copyable or have a non-trivial copy constructor or destructor.
|
2020-04-07 14:38:02 -07:00 |
|
extract_outside_compilation_pass.h
|
Support summary in while loop.
|
2019-06-28 12:56:25 -07:00 |
|
flags.cc
|
Provides an environment variable via TF_XLA_FLAGS to turn on the MLIR bridge.
|
2020-06-12 09:26:57 -07:00 |
|
flags.h
|
Provides an environment variable via TF_XLA_FLAGS to turn on the MLIR bridge.
|
2020-06-12 09:26:57 -07:00 |
|
force_xla_constants_on_host_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
force_xla_constants_on_host_pass.cc
|
[TF/XLA] Force all tensors which need to be constant during the XLA compilation to be located on the host
|
2020-04-01 11:34:32 -07:00 |
|
force_xla_constants_on_host_pass.h
|
[TF/XLA] Force all tensors which need to be constant during the XLA compilation to be located on the host
|
2020-04-01 11:34:32 -07:00 |
|
increase_dynamism_for_auto_jit_pass_test.cc
|
Don't rewrite slices that have a constant begin input
|
2019-02-06 14:44:47 -08:00 |
|
increase_dynamism_for_auto_jit_pass.cc
|
Replace tf2xla/dump_graph.h with core/util/dump_graph.h.
|
2019-03-05 05:27:56 -08:00 |
|
increase_dynamism_for_auto_jit_pass.h
|
Introduce a pass to increase the amount of dynamism supported by an XLA cluster
|
2018-10-26 14:01:34 -07:00 |
|
introduce_floating_point_jitter_pass_internal.h
|
Add a debug-only pass that introduces a small error to a designated TF node
|
2019-03-28 12:25:47 -07:00 |
|
introduce_floating_point_jitter_pass_test.cc
|
Add a debug-only pass that introduces a small error to a designated TF node
|
2019-03-28 12:25:47 -07:00 |
|
introduce_floating_point_jitter_pass.cc
|
Add a debug-only pass that introduces a small error to a designated TF node
|
2019-03-28 12:25:47 -07:00 |
|
introduce_floating_point_jitter_pass.h
|
Add a debug-only pass that introduces a small error to a designated TF node
|
2019-03-28 12:25:47 -07:00 |
|
jit_compilation_pass_registration.cc
|
[TF/XLA] Force all tensors which need to be constant during the XLA compilation to be located on the host
|
2020-04-01 11:34:32 -07:00 |
|
mark_for_compilation_pass_test_helper.cc
|
Add a new unittest in mark_for_compilation_pass_test.
|
2019-08-16 13:37:19 -07:00 |
|
mark_for_compilation_pass_test_helper.h
|
Add a new unittest in mark_for_compilation_pass_test.
|
2019-08-16 13:37:19 -07:00 |
|
mark_for_compilation_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
mark_for_compilation_pass.cc
|
Add DeviceIndex xla op.
|
2020-06-22 15:06:38 -07:00 |
|
mark_for_compilation_pass.h
|
[NFC] Clarify documentation of JIT passes
|
2019-12-19 11:20:28 -08:00 |
|
node_matchers_test.cc
|
minor spelling tweaks
|
2019-12-09 18:21:12 +09:00 |
|
node_matchers.cc
|
This CL optimizes C++11 range-based for loops where the variable is copied in each iteration but it would suffice to obtain it by const reference. This is only applied to loop variables of types that are expensive to copy which means they are not trivially copyable or have a non-trivial copy constructor or destructor.
|
2020-04-07 14:38:02 -07:00 |
|
node_matchers.h
|
minor spelling tweaks
|
2019-12-09 18:21:12 +09:00 |
|
partially_decluster_pass_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
partially_decluster_pass.cc
|
Split out node_def_util, op_def_builder, op_def_util and attr_value_util targets in tensorflow/core/framework/BUILD. Split out
|
2020-01-06 11:46:01 -08:00 |
|
partially_decluster_pass.h
|
Decluster some must-be-constant ops to reduce XLA recompilations
|
2018-09-07 18:53:18 -07:00 |
|
rearrange_function_argument_pass_test.cc
|
Update RearrangeFunctionArgumentForFunctionTest to use correct index when creating Retval nodes.
|
2019-09-27 14:25:28 -07:00 |
|
report_clustering_info_pass.cc
|
Roll forward "Add an XLA "activity listener" mechanism."
|
2019-06-14 17:14:19 -07:00 |
|
report_clustering_info_pass.h
|
Roll forward "Add an XLA "activity listener" mechanism."
|
2019-06-14 17:14:19 -07:00 |
|
resource_operation_safety_analysis_test.cc
|
[Build cleanup] Split "core_cpu_impl" into fine-grained targets (4/n).
|
2020-04-28 09:53:59 -07:00 |
|
resource_operation_safety_analysis.cc
|
Storing kRead kinds doesn't have any effect, but might take a lot of RAM.
|
2020-03-17 09:54:28 -07:00 |
|
resource_operation_safety_analysis.h
|
[TF:XLA] Update resource_operation_safety_analysis file comment.
|
2019-07-16 18:03:24 -07:00 |
|
shape_inference_helpers.cc
|
Fix shape inference for outside_compilation clusters that include cycles.
|
2018-04-12 11:06:51 -07:00 |
|
shape_inference_helpers.h
|
Fix shape inference for outside_compilation clusters that include cycles.
|
2018-04-12 11:06:51 -07:00 |
|
shape_inference_test.cc
|
Handle shape inference for loop invariant Merge node.
|
2019-06-10 16:16:06 -07:00 |
|
shape_inference.cc
|
Fixes for outside compilation.
|
2020-01-03 14:27:55 -08:00 |
|
shape_inference.h
|
Move shape inference utility to open source.
|
2018-10-15 21:23:29 -07:00 |
|
test_util.cc
|
Use absl instead of deprecated str_util
|
2019-05-21 08:10:10 -07:00 |
|
test_util.h
|
[TF/XLA Bridge] [NFC] Reduce the amount of boilerplate required to create GraphOptimizationPassOptions
|
2019-09-30 20:48:18 -07:00 |
|
union_find.h
|
[TF:XLA] Split union-find implementation in mark_for_compilation_pass.cc into a separate library, make it more generic.
|
2017-06-02 11:08:26 -07:00 |
|
xla_activity_listener_test.cc
|
[TF:XLA] Logging of known cases of performance deficiencies using XLA with Tensorflow.
|
2019-08-07 04:04:01 -07:00 |
|
xla_activity_listener.cc
|
Prefixing TensorFlow thread annotation macros with TF_.
|
2020-03-05 08:42:01 -08:00 |
|
xla_activity_listener.h
|
[TF:XLA] Logging of known cases of performance deficiencies using XLA with Tensorflow.
|
2019-08-07 04:04:01 -07:00 |
|
xla_activity_logging_listener.cc
|
[TF:XLA] Logging of known cases of performance deficiencies using XLA with Tensorflow.
|
2019-08-07 04:04:01 -07:00 |
|
xla_activity.proto
|
[TF:XLA] Logging of known cases of performance deficiencies using XLA with Tensorflow.
|
2019-08-07 04:04:01 -07:00 |
|
xla_cluster_util_test.cc
|
[Build cleanup] Update #includes of moved header "graph/graph_constructor.h".
|
2020-04-29 09:20:48 -07:00 |
|
xla_cluster_util.cc
|
Utility function to figure out whether a cluster has reference variables
|
2019-09-19 17:07:52 -07:00 |
|
xla_cluster_util.h
|
Utility function to figure out whether a cluster has reference variables
|
2019-09-19 17:07:52 -07:00 |
|
xla_compilation_cache_test.cc
|
Performance improvements to speed up invocation of XLA code, by making
|
2019-10-23 09:39:27 -07:00 |
|
xla_compilation_cache.cc
|
Support XlaCompiler arguments of type kConstant in XlaCompilationCache
|
2020-05-05 16:12:29 -07:00 |
|
xla_compilation_cache.h
|
Roll forward change to run use MLIR based TensorFlow compiler in XLA on demand compiler
|
2020-03-23 14:50:24 -07:00 |
|
xla_compile_on_demand_op.cc
|
[TF2XLA] [NFC] Refactor datastructures for resource variables to not require snapshotting for compilation
|
2020-06-22 12:55:49 -07:00 |
|
xla_compile_on_demand_op.h
|
[TF2XLA] [NFC] Refactor datastructures for resource variables to not require snapshotting for compilation
|
2020-06-22 12:55:49 -07:00 |
|
xla_cpu_device.cc
|
[TF/XLA] Only enable XLA_ devices if TF_XLA_FLAGS=--tf_xla_enable_xla_devices is set. For now, set the flag to "true" by default.
|
2020-01-09 11:46:30 -08:00 |
|
xla_device_context.cc
|
Export memory fragmentation in memstats.
|
2020-04-14 22:18:56 -07:00 |
|
xla_device_context.h
|
Prefixing TensorFlow thread annotation macros with TF_.
|
2020-03-05 08:42:01 -08:00 |
|
xla_device_ops.cc
|
Use RefCountPtr in LookupResource to avoid leaks
|
2019-05-28 22:55:44 -07:00 |
|
xla_device_ops.h
|
Resubmitting the change to make iterator deleter not on host memory
|
2020-05-08 11:11:31 -07:00 |
|
xla_device.cc
|
Use INFO logging for device deprecation, for easier filtering.
|
2020-05-27 10:30:49 -07:00 |
|
xla_device.h
|
Prefixing TensorFlow thread annotation macros with TF_.
|
2020-03-05 08:42:01 -08:00 |
|
xla_gpu_device.cc
|
[ROCm] Fix to enable XLA_GPU device registration for ROCm platform
|
2020-01-15 14:50:10 +00:00 |
|
xla_interpreter_device.cc
|
address comments on commit e4c76ebc7790368227f2aa4695c177cff364e463
|
2019-10-08 08:38:05 -07:00 |
|
xla_kernel_creator_test.cc
|
Replace NodeDef with std::shared_ptr<NodeProperties> in the kernel creation code paths and try to avoid as many copies of NodeDefs as possible. This will in most cases allow sharing the NodeDef between the OpKernel and the graph Node from which it is created.
|
2020-02-18 13:20:06 -08:00 |
|
xla_kernel_creator_util.cc
|
[TF/XLA] On compilation failure, do not overflow the max size of the bad status by a huge list of function inputs
|
2020-05-27 11:59:06 -07:00 |
|
xla_kernel_creator_util.h
|
[TF/XLA] Force all tensors which need to be constant during the XLA compilation to be located on the host
|
2020-04-01 11:34:32 -07:00 |
|
xla_kernel_creator.cc
|
[TF/XLA] Force all tensors which need to be constant during the XLA compilation to be located on the host
|
2020-04-01 11:34:32 -07:00 |
|
xla_kernel_creator.h
|
Replace NodeDef with std::shared_ptr<NodeProperties> in the kernel creation code paths and try to avoid as many copies of NodeDefs as possible. This will in most cases allow sharing the NodeDef between the OpKernel and the graph Node from which it is created.
|
2020-02-18 13:20:06 -08:00 |
|
xla_launch_util.cc
|
Support dynamic outputs for XLA on demand ops.
|
2020-07-01 11:11:47 -07:00 |
|
xla_launch_util.h
|
[TF2XLA] [NFC] Refactor datastructures for resource variables to not require snapshotting for compilation
|
2020-06-22 12:55:49 -07:00 |
|
xla_tensor.cc
|
[TF/XLA] Remove the wrapping function XlaTensor::RefCountIsOne
|
2020-06-15 11:44:06 -07:00 |
|
xla_tensor.h
|
[XLA] Do not needlessly store wrapped Tensor and ScopedShapedBuffer inside XlaTensor on a heap
|
2020-06-15 12:52:08 -07:00 |
