experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
STT-tensorflow/tensorflow/compiler/xla/service
History
Yunxing Dai 3b94e9cfdb Support 0-sized shape with dynamic reshapes. 
- Return finer granule control in CommonFactors when input and output are the same and one of dimensions is zero.
- Skip instruction with dynamic size in ZeroSizedHloElimination.

PiperOrigin-RevId: 352189050
Change-Id: Ieda2d8c89d20e66670b3cd85484f4e1e97bc4c93
2021-01-16 10:23:25 -08:00
..
cpu
[XLA:CPU] Move elemental conv emitter to generic code path so it can share the multiply-add logic with dot
2021-01-13 03:07:21 -08:00
g3doc
[XLA] Remove unsupported sparse layout
2020-01-02 18:04:52 -08:00
gpu
Merge pull request #46368 from ROCmSoftwarePlatform:google_upstream_rocm_csb_fix_210112
2021-01-15 05:07:56 -08:00
graphcycles
Update benchmarks to use newer API
2020-11-24 11:59:06 -08:00
interpreter
[XLA] Pack some parameters in xla::Compiler member functions into a struct. This is for later expanding for more parameters, like a thread pool for parallel compilation.
2020-12-08 01:54:36 -08:00
llvm_ir
[XLA/GPU] Fix issues brought by multi-threaded compilation:
2021-01-06 13:11:05 -08:00
mlir_gpu
Use mlir::OpState::operator->() to get to methods of mlir::Operation.
2020-12-23 01:58:41 -08:00
spmd
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
algebraic_simplifier_proof_distributive_property.py
…
algebraic_simplifier_test.cc
[XLA] Transform more cases of reshape(iota) to a mixed radix calculation with
2021-01-07 08:59:55 -08:00
algebraic_simplifier.cc
[XLA] Transform more cases of reshape(iota) to a mixed radix calculation with
2021-01-07 08:59:55 -08:00
algebraic_simplifier.h
[XLA] Convert Abs(a)*Abs(a) to a*a and add an option to allow for numerically unsafe algebraic simplifications
2020-09-06 12:51:40 -07:00
all_gather_decomposer_test.cc
[XLA] Fix some all-gather issues.
2020-05-19 17:55:08 -07:00
all_gather_decomposer.cc
[XLA] Fix some all-gather issues.
2020-05-19 17:55:08 -07:00
all_gather_decomposer.h
[XLA] Fix some all-gather issues.
2020-05-19 17:55:08 -07:00
all_reduce_combiner_test.cc
Verify replica count from AllReduce replica group config
2020-04-27 08:53:44 -07:00
all_reduce_combiner.cc
[XLA] Skip AllReduceCombiner when threshold is 0
2020-08-14 19:34:44 -07:00
all_reduce_combiner.h
[XLA:GPU] Add an AllReduceCombiner pass, that merges AllReduce operations.
2020-02-19 22:07:07 -08:00
all_reduce_simplifier_test.cc
[XLA] replace all-reduces with a singleton replica_group_size with the operand.
2020-09-11 08:58:29 -07:00
all_reduce_simplifier.cc
[XLA] replace all-reduces with a singleton replica_group_size with the operand.
2020-09-11 08:58:29 -07:00
all_reduce_simplifier.h
…
allocation_tracker.cc
Remove platform field from shaped buffer.
2020-10-30 14:26:55 -07:00
allocation_tracker.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
ar_crs_combiner_test.cc
Verify replica count from AllReduce replica group config
2020-04-27 08:53:44 -07:00
ar_crs_combiner.cc
[XLA] Identical check shouldn't skip operands for all-reduces.
2020-12-18 10:30:34 -08:00
ar_crs_combiner.h
…
backend.cc
…
backend.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
batch_dot_simplification_test.cc
…
batch_dot_simplification.cc
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
batch_dot_simplification.h
…
batchnorm_expander_test.cc
…
batchnorm_expander.cc
…
batchnorm_expander.h
…
bfloat16_conversion_folding_test.cc
Enable using global ids for replica groups in AllReduce
2020-02-21 19:01:52 -08:00
bfloat16_conversion_folding.cc
[XLA] Stop the bf16 conversion folding from converting unused tuple outputs.
2020-11-09 13:08:11 -08:00
bfloat16_conversion_folding.h
…
bfloat16_normalization_test.cc
Add optional layout constraint to AllToAll
2020-04-27 18:37:21 -07:00
bfloat16_normalization.cc
…
bfloat16_normalization.h
…
bfloat16_propagation_test.cc
[XLA] Bug fix for bfloat16 propagation where fixed-point repetition of resolving aliasing operands are needed.
2020-11-04 14:12:17 -08:00
bfloat16_propagation.cc
[XLA] Bug fix for bfloat16 propagation where fixed-point repetition of resolving aliasing operands are needed.
2020-11-04 14:12:17 -08:00
bfloat16_propagation.h
[XLA] Support conditional in bf16 propagation
2020-04-05 16:46:27 -07:00
bfloat16_support.cc
[XLA] Introducing AllGather HLO and a decomposition pass
2020-05-09 12:22:17 -07:00
bfloat16_support.h
…
buffer_assignment_test.cc
[XLA] Try committing DUS buffer sharing again with fixes.
2020-09-17 09:20:11 -07:00
buffer_assignment.cc
[XLA/GPU] Minor cleanup.
2020-10-13 12:37:44 -07:00
buffer_assignment.h
[XLA/GPU] Add an LMHLO -> Execution result test.
2021-01-08 15:39:11 -08:00
buffer_value_containers.h
…
buffer_value.cc
[XLA] BufferValue::Color now type aliases int64.
2020-05-27 09:35:14 -07:00
buffer_value.h
[XLA] Deprecate BufferValue::color.
2020-06-29 14:37:43 -07:00
BUILD
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
call_graph_test.cc
…
call_graph.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.
2020-04-06 17:55:23 -07:00
call_graph.h
…
call_inliner_test.cc
Convert absl string_view constants to const, rather than constexpr.
2020-02-05 11:15:53 -08:00
call_inliner.cc
[XLA] Fix VLOG interaction with CallInliner
2021-01-06 16:15:13 -08:00
call_inliner.h
[XLA:CPU] Restrict CallInliner to functions with a single call site
2020-01-23 09:19:45 -08:00
channel_tracker.cc
…
channel_tracker.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
cholesky_expander.cc
[XLA] Add complex number support to HLO cholesky decomposition implementation.
2020-09-18 07:08:30 -07:00
cholesky_expander.h
[XLA] Make the inner block kernel of CholeskyExpander override-able.
2020-08-31 09:33:25 -07:00
collective_ops_utils_test.cc
Add GetParticipatingDevices() function.
2020-12-04 03:11:26 -08:00
collective_ops_utils.cc
Add GetParticipatingDevices() function.
2020-12-04 03:11:26 -08:00
collective_ops_utils.h
[XLA-GPU] Fix undefined behaviour in NCCL utils.
2020-12-10 04:24:24 -08:00
comparison_expander.cc
Adding total-order comparison support in proto and HloInstruction.
2020-08-10 09:40:29 -07:00
comparison_expander.h
Adding total-order comparison support in proto and HloInstruction.
2020-08-10 09:40:29 -07:00
compilation_cache.cc
…
compilation_cache.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
compilation_stats.cc
…
compilation_stats.h
…
compile_only_service.cc
Enables per-host dummy args for TPUExecute (TF1) and adds XLA options.
2021-01-14 17:03:51 -08:00
compile_only_service.h
…
compiler.cc
[mlir] Add FusionOp to XLA HLO -> LMHLO
2020-10-13 14:20:52 -07:00
compiler.h
Enables per-host dummy args for TPUExecute (TF1) and adds XLA options.
2021-01-14 17:03:51 -08:00
computation_layout.cc
This CL adds the following changes:
2020-07-01 16:45:34 -07:00
computation_layout.h
This CL adds the following changes:
2020-07-01 16:45:34 -07:00
computation_placer.cc
Add DeviceAssignment::LogicalIdsForDevice method.
2020-11-27 00:11:51 -08:00
computation_placer.h
Add DeviceAssignment::LogicalIdsForDevice method.
2020-11-27 00:11:51 -08:00
conditional_canonicalizer_test.cc
Canonicalize dense array conditional into tuple conditional with one element.
2020-07-17 11:55:43 -07:00
conditional_canonicalizer.cc
Fix conditional canonicalizer to use clone instead of mutating existing conditional's shape.
2020-07-27 13:44:30 -07:00
conditional_canonicalizer.h
Canonicalize dense array conditional into tuple conditional with one element.
2020-07-17 11:55:43 -07:00
conditional_code_motion_test.cc
[XLA] Expose configuration space of conditional code motion to support auto-tuning of its cost model.
2021-01-14 13:06:50 -08:00
conditional_code_motion.cc
[XLA] Expose configuration space of conditional code motion to support auto-tuning of its cost model.
2021-01-14 13:06:50 -08:00
conditional_code_motion.h
[XLA] Expose configuration space of conditional code motion to support auto-tuning of its cost model.
2021-01-14 13:06:50 -08:00
conditional_simplifier_test.cc
[XLA] Add token support to conditional_simplifier
2020-06-02 11:00:01 -07:00
conditional_simplifier.cc
Condintional Simplifier: Return early before creating any instruction.
2020-11-20 18:39:27 -08:00
conditional_simplifier.h
…
conditional_to_select_test.cc
…
conditional_to_select.cc
…
conditional_to_select.h
…
convert_operand_folding_test.cc
[XLA] IntegralUpcaster->OperandUpcaster to allow auto upcasting floating point types.
2020-12-29 11:14:04 -08:00
convert_operand_folding.cc
[XLA] IntegralUpcaster->OperandUpcaster to allow auto upcasting floating point types.
2020-12-29 11:14:04 -08:00
convert_operand_folding.h
[XLA] Add a pass to fold hlo(convert(a),convert(b)) into hlo(a,b) to enable an alternative way of specifying wider accumulation type than the shape inference result for HLOs that support it.
2020-11-25 12:57:30 -08:00
convolution_4d_expander_test.cc
Convert 4D convolutions with trivial dimensions to lower-dimensional convolutions.
2020-01-28 02:40:48 -08:00
convolution_4d_expander.cc
Convert 4D convolutions with trivial dimensions to lower-dimensional convolutions.
2020-01-28 02:40:48 -08:00
convolution_4d_expander.h
Convert 4D convolutions with trivial dimensions to lower-dimensional convolutions.
2020-01-28 02:40:48 -08:00
convolution_group_converter_test.cc
[XLA] Fix the condition for rewriting batch group convolutions to include when the input batch is not equal to the batch group count.
2020-04-08 16:50:13 -07:00
convolution_group_converter.cc
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
convolution_group_converter.h
Fix confusing way of specifying filter_expansion parameter.
2020-01-08 00:43:03 -08:00
copy_insertion_test.cc
Internal change
2020-12-16 20:47:15 -08:00
copy_insertion.cc
Internal change
2020-12-16 20:47:15 -08:00
copy_insertion.h
…
custom_call_target_registry.cc
…
custom_call_target_registry.h
…
defuser_test.cc
…
defuser.cc
…
defuser.h
…
despecializer.cc
…
despecializer.h
…
dfs_hlo_visitor_with_default_test.cc
…
dfs_hlo_visitor_with_default.h
Introduce dynamic reshape op.
2020-08-14 15:48:19 -07:00
dfs_hlo_visitor.cc
…
dfs_hlo_visitor.h
Introduce dynamic reshape op.
2020-08-14 15:48:19 -07:00
dot_as_convolution_util.cc
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
dot_as_convolution_util.h
[XLA:SPMD] 1st step to refactor convolution_handler.
2020-09-11 19:33:47 -07:00
dot_decomposer_test.cc
Improve DotDecomposer to not add unnecessary non-contracting dimensions.
2020-05-26 05:49:19 -07:00
dot_decomposer.cc
Polish some comments in dot decomposer.
2020-06-15 14:41:41 -07:00
dot_decomposer.h
…
dump.cc
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
dump.h
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
dynamic_dimension_inference_test.cc
Fix a bug reshape dynamic dimension inference bug.
2020-11-18 16:51:43 -08:00
dynamic_dimension_inference.cc
Support dynamic input feature for convolutions.
2020-11-18 17:30:27 -08:00
dynamic_dimension_inference.h
Support dynamic spatial dimensions in convolution.
2020-10-27 10:26:50 -07:00
dynamic_index_splitter_test.cc
…
dynamic_index_splitter.cc
…
dynamic_index_splitter.h
…
dynamic_padder_test.cc
Support dynamic input feature for convolutions.
2020-11-18 17:30:27 -08:00
dynamic_padder.cc
[XLA] Add implementation support for variadic reduce window, including HLO, cost analysis, etc.
2020-12-15 19:59:45 -08:00
dynamic_padder.h
[XLA]Add HLO infrastructure to support dynamic op lowering.
2020-04-30 17:00:49 -07:00
dynamic_parameter_binding_test.cc
…
dynamic_parameter_binding.cc
…
dynamic_parameter_binding.h
…
dynamic_update_slice_test.cc
…
dynamic_window_utils.cc
Support dynamic spatial dimensions in convolution.
2020-10-27 10:26:50 -07:00
dynamic_window_utils.h
Support dynamic spatial dimensions in convolution.
2020-10-27 10:26:50 -07:00
elemental_ir_emitter_test.cc
…
elemental_ir_emitter.cc
[XLA:CPU] Move elemental conv emitter to generic code path so it can share the multiply-add logic with dot
2021-01-13 03:07:21 -08:00
elemental_ir_emitter.h
[XLA:CPU] Move elemental conv emitter to generic code path so it can share the multiply-add logic with dot
2021-01-13 03:07:21 -08:00
executable.cc
Remove platform field from shaped buffer.
2020-10-30 14:26:55 -07:00
executable.h
Do not creates MaybeOwningDeviceMemory and ExecutionInput object in TF/XLA. This lower XLA launch overhead.
2021-01-07 07:04:22 -08:00
execution_tracker.cc
…
execution_tracker.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
flatten_call_graph_test.cc
…
flatten_call_graph.cc
…
flatten_call_graph.h
…
fusion_node_indexing_evaluation_test.cc
Evaluate the maximum code duplication in fusion nodes instead of average.
2020-09-02 07:36:23 -07:00
fusion_node_indexing_evaluation.cc
Prevent creating fusions which could cause problems.
2020-10-29 09:12:46 -07:00
fusion_node_indexing_evaluation.h
Refactor the code to avoid duplication (NFC).
2020-10-01 05:16:36 -07:00
fusion_queue.h
Remove fusion config utils code from fusion_queue.
2020-12-26 22:42:26 -08:00
gather_expander_test.cc
Expand simple gathers into dynamic-slice.
2020-08-07 16:06:09 -07:00
gather_expander.cc
Expand simple gathers into dynamic-slice.
2020-08-07 16:06:09 -07:00
gather_expander.h
Expand simple gathers into dynamic-slice.
2020-08-07 16:06:09 -07:00
generic_transfer_manager.cc
[XLA] Don't pass on_host_shape to ShapedBuffer/ScopedShapedBuffer inside XLA.
2020-10-08 12:00:12 -07:00
generic_transfer_manager.h
…
global_device_id.cc
[NFC] Moved GlobalDeviceId into its own file as it is also used by the CPU runtime.
2020-11-24 04:21:38 -08:00
global_device_id.h
[NFC] Moved GlobalDeviceId into its own file as it is also used by the CPU runtime.
2020-11-24 04:21:38 -08:00
heap_simulator_test.cc
[XLA/GPU] Size-constrained buffer allocation.
2020-10-07 15:40:30 -07:00
heap_simulator.cc
Merge pull request #43850 from trentlo:size_constrained_buf_alloc_upstream
2020-10-28 04:19:38 -07:00
heap_simulator.h
[TF/XLA] Fix a compilation warning due to -Werror,-Wreorder-ctor.
2020-10-22 10:33:49 -07:00
hlo_alias_analysis_test.cc
[XLA] Try committing DUS buffer sharing again with fixes.
2020-09-17 09:20:11 -07:00
hlo_alias_analysis.cc
[XLA] Try committing DUS buffer sharing again with fixes.
2020-09-17 09:20:11 -07:00
hlo_alias_analysis.h
Internal change
2020-09-12 16:03:35 -07:00
hlo_buffer.cc
…
hlo_buffer.h
…
hlo_casting_utils_test.cc
…
hlo_casting_utils.h
…
hlo_clone_context.h
…
hlo_computation_test.cc
…
hlo_computation.cc
[XLA] Fix VLOG interaction with CallInliner
2021-01-06 16:15:13 -08:00
hlo_computation.h
Add add new option in HloComputation::AddInstruction to add instruction with different name.
2020-10-08 13:08:57 -07:00
hlo_constant_folding_test.cc
Replace absl::Span where we use initializer lists in tests with std::vector.
2020-02-28 17:37:42 -08:00
hlo_constant_folding.cc
…
hlo_constant_folding.h
…
hlo_cost_analysis_test.cc
[XLA] Add implementation support for variadic reduce window, including HLO, cost analysis, etc.
2020-12-15 19:59:45 -08:00
hlo_cost_analysis.cc
[XLA] Add implementation support for variadic reduce window, including HLO, cost analysis, etc.
2020-12-15 19:59:45 -08:00
hlo_cost_analysis.h
Update hlo cost function visibility
2020-12-17 10:52:08 -08:00
hlo_creation_utils_test.cc
…
hlo_creation_utils.cc
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
hlo_creation_utils.h
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
hlo_cse_test.cc
…
hlo_cse.cc
[XLA] Avoid hash collisions in CseHash.
2020-02-20 00:01:30 -08:00
hlo_cse.h
…
hlo_dataflow_analysis_test.cc
[XLA] Try committing DUS buffer sharing again with fixes.
2020-09-17 09:20:11 -07:00
hlo_dataflow_analysis.cc
Internal change
2020-09-24 10:22:25 -07:00
hlo_dataflow_analysis.h
Internal change
2020-09-24 10:22:25 -07:00
hlo_dce_test.cc
…
hlo_dce.cc
[XLA] Introducing AllGather HLO and a decomposition pass
2020-05-09 12:22:17 -07:00
hlo_dce.h
[XLA] Resubmit DCE change
2020-03-03 17:43:00 -08:00
hlo_domain_isolator.cc
…
hlo_domain_isolator.h
…
hlo_domain_map.cc
…
hlo_domain_map.h
…
hlo_domain_metadata.h
…
hlo_domain_remover.cc
…
hlo_domain_remover.h
…
hlo_domain_test.cc
…
hlo_domain_verifier.cc
…
hlo_domain_verifier.h
…
hlo_element_type_converter_test.cc
…
hlo_element_type_converter.cc
…
hlo_element_type_converter.h
…
hlo_evaluator_test.cc
[XLA] Add support for C64 to the sort evaluator
2020-11-28 10:29:36 -08:00
hlo_evaluator_typed_visitor_bfloat16.cc
…
hlo_evaluator_typed_visitor_bool.cc
…
hlo_evaluator_typed_visitor_complex64.cc
…
hlo_evaluator_typed_visitor_complex128.cc
…
hlo_evaluator_typed_visitor_double.cc
…
hlo_evaluator_typed_visitor_float.cc
…
hlo_evaluator_typed_visitor_half.cc
…
hlo_evaluator_typed_visitor_int8.cc
…
hlo_evaluator_typed_visitor_int16.cc
…
hlo_evaluator_typed_visitor_int32.cc
…
hlo_evaluator_typed_visitor_int64.cc
…
hlo_evaluator_typed_visitor_uint8.cc
…
hlo_evaluator_typed_visitor_uint16.cc
…
hlo_evaluator_typed_visitor_uint32.cc
…
hlo_evaluator_typed_visitor_uint64.cc
…
hlo_evaluator_typed_visitor.h
[XLA] Add S16/U16 support for Map in the HLO evaluator
2020-11-22 16:26:00 -08:00
hlo_evaluator.cc
[XLA] Add support for C64 to the sort evaluator
2020-11-28 10:29:36 -08:00
hlo_evaluator.h
[XLA] Rollforward of Rolledback CL after fixing mistakes.
2020-11-03 12:07:01 -08:00
hlo_execution_profile_data.proto
…
hlo_execution_profile_test.cc
Rewrite the pattern we want to match in hlo_execution_profile_test.
2020-01-27 05:22:03 -08:00
hlo_execution_profile.cc
[XLA/GPU] Make module parameter to GpuExecutable optional.
2020-12-29 14:33:16 -08:00
hlo_execution_profile.h
[XLA/GPU] Make module parameter to GpuExecutable optional.
2020-12-29 14:33:16 -08:00
hlo_get_dimension_size_rewriter.h
…
hlo_graph_dumper_test.cc
…
hlo_graph_dumper.cc
Some instructions have giant replica group fields, and make it hard to navigate the graph. Truncate the replica group field string to 128 chars.
2020-11-30 13:48:14 -08:00
hlo_graph_dumper.h
Roll forward "Add a show_fusion_subcomputations command to interactive_graphviz" with fix
2020-05-27 11:27:39 -07:00
hlo_input_output_alias_config_test.cc
[XLA] Unify aliasing types
2020-07-07 16:07:00 -07:00
hlo_input_output_alias_config.cc
[Resubmit] If an input-output pair is configured to be must-alias(off by default), they must be aliased at runtime.
2020-08-07 14:06:08 -07:00
hlo_input_output_alias_config.h
Fix priting of input/output must/may aliasing.
2020-08-11 13:48:12 -07:00
hlo_instruction_test.cc
…
hlo_instruction.cc
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_instruction.h
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_instructions.cc
[XLA] Adding printing and parsing of outfeed_shape attribute for outfeed HLO.
2021-01-11 17:36:13 -08:00
hlo_instructions.h
Support multiple computations in custom call.
2020-11-03 12:51:13 -08:00
hlo_lexer.cc
[XLA:SPMD] Define manual sharding, instead of using fake replication
2020-11-25 11:10:08 -08:00
hlo_lexer.h
[XLA:SPMD] Define manual sharding, instead of using fake replication
2020-11-25 11:10:08 -08:00
hlo_live_range_test.cc
[XLA] Unify aliasing types
2020-07-07 16:07:00 -07:00
hlo_live_range.cc
…
hlo_live_range.h
…
hlo_liveness_analysis_test.cc
…
hlo_liveness_analysis.cc
…
hlo_liveness_analysis.h
…
hlo_matchers_test.cc
[XLA] Add an optional bool is_cross_program_prefetch field to kCopyStart HLOs.
2020-08-21 15:10:38 -07:00
hlo_matchers.cc
…
hlo_matchers.h
[XLA] Add AllGather to the HLO matchers.
2020-05-21 22:29:00 -07:00
hlo_memory_scheduler_test.cc
[XLA] Use templates in heap simulator to allow opaque type to be different than HloValue (NFC)
2020-08-14 15:58:46 -07:00
hlo_memory_scheduler.cc
If an outfeed instruction has a config, schedule it early. If an infeed instruction has a config, schedule it late.
2020-06-03 22:15:37 -07:00
hlo_memory_scheduler.h
…
hlo_module_config.cc
Rename GenerateCubinForTfCode to GenerateGpuBinaryForTfCode and add ROCm support for it
2020-08-26 13:50:01 +00:00
hlo_module_config.h
Enables per-host dummy args for TPUExecute (TF1) and adds XLA options.
2021-01-14 17:03:51 -08:00
hlo_module_dce_test.cc
Update HloModuleDCE to call DCE and WhileLoopSimplifier after removing dead tuple elements in while loops.
2020-10-27 07:58:09 -07:00
hlo_module_dce.cc
Update HloModuleDCE to call DCE and WhileLoopSimplifier after removing dead tuple elements in while loops.
2020-10-27 07:58:09 -07:00
hlo_module_dce.h
…
hlo_module_group_metadata.cc
…
hlo_module_group_metadata.h
…
hlo_module_group_test.cc
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
hlo_module_group_util.cc
…
hlo_module_group_util.h
…
hlo_module_group.cc
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
hlo_module_group.h
[XLA] Add option to print out hlo pass hash.
2020-02-27 14:54:06 -08:00
hlo_module_metadata_test.cc
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
hlo_module_metadata.cc
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
hlo_module_metadata.h
[XLA] Add creation_pass_id and dummy op_names to OpMetadata.
2020-12-11 17:48:13 -08:00
hlo_module_test.cc
Add method to retrieve a filtered list of computations in post-order.
2020-12-16 20:26:32 -08:00
hlo_module_util.cc
Enables per-host dummy args for TPUExecute (TF1) and adds XLA options.
2021-01-14 17:03:51 -08:00
hlo_module_util.h
Move CreateModuleConfig to a new hlo_module_util header.
2020-12-14 15:14:29 -08:00
hlo_module.cc
Enables per-host dummy args for TPUExecute (TF1) and adds XLA options.
2021-01-14 17:03:51 -08:00
hlo_module.h
Add method to retrieve a filtered list of computations in post-order.
2020-12-16 20:26:32 -08:00
hlo_op_metadata.cc
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_op_metadata.h
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_opcode_test.cc
[XLA] Add xla builder support for variadic reduce window op. This is the first CL leading to full support of variaduc reduce window.
2020-10-22 00:22:54 -07:00
hlo_opcode.cc
…
hlo_opcode.h
[XLA] Add xla builder support for variadic reduce window op. This is the first CL leading to full support of variaduc reduce window.
2020-10-22 00:22:54 -07:00
hlo_ordering_test.cc
…
hlo_ordering.cc
Internal change
2020-12-16 20:47:15 -08:00
hlo_ordering.h
Internal change
2020-12-16 20:47:15 -08:00
hlo_parser_test.cc
[XLA] Adding printing and parsing of outfeed_shape attribute for outfeed HLO.
2021-01-11 17:36:13 -08:00
hlo_parser.cc
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_parser.h
…
hlo_pass_fix.h
Better warning. With that we can just to replay_computation directly for further analysis.
2020-10-06 11:01:41 -07:00
hlo_pass_interface.h
…
hlo_pass_pipeline_test.cc
[XLA] Fix timestamps in HLO pipeline test.
2020-11-16 11:48:23 -08:00
hlo_pass_pipeline.cc
[XLA] Add creation_pass_id and dummy op_names to OpMetadata.
2020-12-11 17:48:13 -08:00
hlo_pass_pipeline.h
[XLA] Add metadata for HLO modules and passes.
2020-11-12 11:34:07 -08:00
hlo_phi_graph_test.cc
Properly support nest phi reduction in reverse order.
2020-06-17 18:47:22 -07:00
hlo_phi_graph.cc
Properly support nest phi reduction in reverse order.
2020-06-17 18:47:22 -07:00
hlo_phi_graph.h
a bit more
2020-10-29 18:16:19 +03:00
hlo_profile_printer_data.proto
…
hlo_profile_printer.cc
…
hlo_profile_printer.h
…
hlo_proto_util_test.cc
…
hlo_proto_util.cc
With other internal changes, extend the TF XLA interface to support running post-optimizations HLO modules. Also fix a few bugs in XLA related to saving and loading HLO from protobuf.
2020-11-04 12:25:32 -08:00
hlo_proto_util.h
With other internal changes, extend the TF XLA interface to support running post-optimizations HLO modules. Also fix a few bugs in XLA related to saving and loading HLO from protobuf.
2020-11-04 12:25:32 -08:00
hlo_query.cc
Add some XLA frontend attribute names.
2020-01-16 19:01:38 -08:00
hlo_query.h
Add some XLA frontend attribute names.
2020-01-16 19:01:38 -08:00
hlo_reachability_test.cc
…
hlo_reachability.cc
Allow split-phase operations for HloReachabilityMap where we can
2020-06-14 14:46:08 -07:00
hlo_reachability.h
Allow split-phase operations for HloReachabilityMap where we can
2020-06-14 14:46:08 -07:00
hlo_rematerialization_test_utils_test.cc
…
hlo_rematerialization_test_utils.h
…
hlo_rematerialization_test.cc
Rollback of [XLA] Allow hlo_rematerialization pass to look through bitcasts when doing rematerialization.
2020-12-03 23:04:46 -08:00
hlo_rematerialization.cc
Rollback of [XLA] Allow hlo_rematerialization pass to look through bitcasts when doing rematerialization.
2020-12-03 23:04:46 -08:00
hlo_rematerialization.h
Express remat: Only consider remat with nodes that create big buffers.
2020-09-23 16:29:09 -07:00
hlo_replication_analysis_test.cc
Support x64<->x32 instructions in hlo replication analysis.
2020-10-01 13:25:16 -07:00
hlo_replication_analysis.cc
Support x64<->x32 instructions in hlo replication analysis.
2020-10-01 13:25:16 -07:00
hlo_replication_analysis.h
…
hlo_runner_interface.cc
Added support to visualise HloModule Protobufs
2021-01-08 02:06:03 -08:00
hlo_runner_interface.h
Added support to visualise HloModule Protobufs
2021-01-08 02:06:03 -08:00
hlo_runner.cc
Internal build rule change.
2020-12-01 17:46:15 -08:00
hlo_runner.h
Internal build rule change.
2020-12-01 17:46:15 -08:00
hlo_schedule_test.cc
…
hlo_schedule.cc
…
hlo_schedule.h
…
hlo_sharding_metadata.cc
…
hlo_sharding_metadata.h
…
hlo_sharding_test.cc
Add OpMetadata list to OpShardingProto/HloSharding.
2021-01-12 12:56:24 -08:00
hlo_sharding_util_test.cc
[XLA:SPMD] Improve resharding
2020-07-22 23:18:17 -07:00
hlo_sharding_util.cc
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
hlo_sharding_util.h
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
hlo_sharding.cc
[XLA:SPMD] Adding support for detecting potential parallel dimensions in gather.
2021-01-13 09:34:18 -08:00
hlo_sharding.h
[XLA:SPMD] Adding support for detecting potential parallel dimensions in gather.
2021-01-13 09:34:18 -08:00
hlo_subcomputation_unification_test.cc
…
hlo_subcomputation_unification.cc
…
hlo_subcomputation_unification.h
…
hlo_value.cc
[XLA] BufferValue::Color now type aliases int64.
2020-05-27 09:35:14 -07:00
hlo_value.h
[XLA] Add hash function to HloUse/HloPosition.
2020-07-29 23:27:09 -07:00
hlo_verifier_test.cc
[XLA] Verify the comparison type for comparisons
2020-11-20 15:41:22 -08:00
hlo_verifier.cc
[XLA] Add implementation support for variadic reduce window, including HLO, cost analysis, etc.
2020-12-15 19:59:45 -08:00
hlo_verifier.h
Introduce dynamic reshape op.
2020-08-14 15:48:19 -07:00
hlo.proto
[XLA] Fix timestamps in HLO pipeline test.
2020-11-16 11:48:23 -08:00
human_readable_profile_builder.cc
…
human_readable_profile_builder.h
…
indexed_array_analysis_test.cc
Remove 1 + "string" cases and undo -Wno-string-plus-int.
2020-02-05 16:58:37 -08:00
indexed_array_analysis.cc
…
indexed_array_analysis.h
…
instruction_fusion_test.cc
rng loop fusion
2020-06-05 10:34:16 -07:00
instruction_fusion.cc
Remove fusion config utils code from fusion_queue.
2020-12-26 22:42:26 -08:00
instruction_fusion.h
Add caching for reused parameters.
2020-09-01 06:49:41 -07:00
layout_assignment_test.cc
Make linear layout more explicit.
2020-06-18 06:55:02 -07:00
layout_assignment.cc
Internal change
2020-08-21 13:30:52 -07:00
layout_assignment.h
Internal change
2020-08-21 13:30:52 -07:00
llvm_compiler.cc
[XLA] Pack some parameters in xla::Compiler member functions into a struct. This is for later expanding for more parameters, like a thread pool for parallel compilation.
2020-12-08 01:54:36 -08:00
llvm_compiler.h
[XLA] Pack some parameters in xla::Compiler member functions into a struct. This is for later expanding for more parameters, like a thread pool for parallel compilation.
2020-12-08 01:54:36 -08:00
local_service.cc
Move CreateModuleConfig to a new hlo_module_util header.
2020-12-14 15:14:29 -08:00
local_service.h
Change LocalClient::Compile to support returning multiple executables (one per partition).
2020-01-29 01:03:57 -08:00
logical_buffer_analysis.cc
[XLA] Avoid structured bindings.
2021-01-12 16:13:55 -08:00
logical_buffer_analysis.h
[XLA] Add missing custom-call aliasing-related changes in analysis passes
2021-01-12 13:57:06 -08:00
logical_buffer.cc
[XLA] BufferValue::Color now type aliases int64.
2020-05-27 09:35:14 -07:00
logical_buffer.h
…
logistic_expander_test.cc
[XLA] Add expander pass for kLogistic.
2020-06-24 22:32:37 -07:00
logistic_expander.cc
[XLA] Add expander pass for kLogistic.
2020-06-24 22:32:37 -07:00
logistic_expander.h
[XLA] Add expander pass for kLogistic.
2020-06-24 22:32:37 -07:00
loop_schedule_linearizer_test.cc
[XLA] Insert control edges from write to read instructions for same buffers inside loops
2020-10-26 17:30:57 -07:00
loop_schedule_linearizer.cc
[XLA] Insert control edges from write to read instructions for same buffers inside loops
2020-10-26 17:30:57 -07:00
loop_schedule_linearizer.h
[XLA] Insert control edges from write to read instructions for same buffers inside loops
2020-10-26 17:30:57 -07:00
map_inliner_test.cc
…
map_inliner.cc
…
map_inliner.h
…
maybe_owning_device_memory.cc
Make XRT CPU/GPU use MaybeOwning buffer interface, so the new copy protection CL won't break aliasing.
2020-06-22 12:32:43 -07:00
maybe_owning_device_memory.h
Make XRT CPU/GPU use MaybeOwning buffer interface, so the new copy protection CL won't break aliasing.
2020-06-22 12:32:43 -07:00
memory_space_assignment_best_fit_repacker_test.cc
[XLA] Implement a repacker that wraps heap simulator's best fit algorithm.
2020-08-18 09:21:37 -07:00
memory_space_assignment_best_fit_repacker.cc
[XLA] Implement a repacker that wraps heap simulator's best fit algorithm.
2020-08-18 09:21:37 -07:00
memory_space_assignment_best_fit_repacker.h
[XLA] Implement a repacker that wraps heap simulator's best fit algorithm.
2020-08-18 09:21:37 -07:00
memory_space_assignment_repacking.h
[XLA] Use templates in heap simulator to allow opaque type to be different than HloValue (NFC)
2020-08-14 15:58:46 -07:00
memory_space_assignment_test.cc
[XLA] Remove an overly pessimistic assertion from memory space assignment.
2020-12-30 13:02:28 -08:00
memory_space_assignment_utils.cc
Internal change
2020-09-24 10:22:25 -07:00
memory_space_assignment_utils.h
[XLA] Exposes the IsValueAllowedInAlternateMemory() for better reuse.
2020-08-18 14:26:53 -07:00
memory_space_assignment.cc
[XLA] Remove an overly pessimistic assertion from memory space assignment.
2020-12-30 13:02:28 -08:00
memory_space_assignment.h
[XLA] Add mechanism to disable alternative mem allocation at uses
2020-12-11 13:12:08 -08:00
memory_space_propagation_test.cc
[XLA] Propagate memory spaces recursively inside nested fusions.
2020-06-18 13:49:23 -07:00
memory_space_propagation.cc
[XLA] Propagate memory spaces recursively inside nested fusions.
2020-06-18 13:49:23 -07:00
memory_space_propagation.h
[XLA] Propagate memory spaces recursively inside nested fusions.
2020-06-18 13:49:23 -07:00
multi_output_fusion.cc
[XLA] Try committing DUS buffer sharing again with fixes.
2020-09-17 09:20:11 -07:00
multi_output_fusion.h
[Resubmit] Fix several issues of multi output fusion.
2020-02-18 20:50:32 -08:00
name_uniquer_test.cc
…
name_uniquer.cc
[-Wsign-compare] warning fixes batch 4
2020-06-24 21:58:00 +00:00
name_uniquer.h
…
op_expander_pass.cc
…
op_expander_pass.h
…
operand_upcaster_test.cc
[XLA] IntegralUpcaster->OperandUpcaster to allow auto upcasting floating point types.
2020-12-29 11:14:04 -08:00
operand_upcaster.cc
[XLA] IntegralUpcaster->OperandUpcaster to allow auto upcasting floating point types.
2020-12-29 11:14:04 -08:00
operand_upcaster.h
[XLA] IntegralUpcaster->OperandUpcaster to allow auto upcasting floating point types.
2020-12-29 11:14:04 -08:00
optimize_input_output_buffer_alias_test.cc
Automated rollback of commit afd980b3ce068c30bd3a43d54929e2590be8bd29
2020-02-12 14:31:05 -08:00
optimize_input_output_buffer_alias.cc
[XLA] Unify aliasing types
2020-07-07 16:07:00 -07:00
optimize_input_output_buffer_alias.h
Automated rollback of commit afd980b3ce068c30bd3a43d54929e2590be8bd29
2020-02-12 14:31:05 -08:00
pattern_matcher_gmock_test.cc
[XLA] Remove unsupported sparse layout
2020-01-02 18:04:52 -08:00
pattern_matcher_gmock.h
…
pattern_matcher_test.cc
[XLA] Remove unsupported sparse layout
2020-01-02 18:04:52 -08:00
pattern_matcher.h
[XLA] Add select and scatter to pattern matchers.
2020-10-20 10:35:38 -07:00
platform_util.cc
…
platform_util.h
…
qr_expander.cc
[XLA] Add support for complex numbers to Qr decomposition expander.
2020-09-22 19:36:07 -07:00
qr_expander.h
[XLA] Change the QR decomposition implementation to lower to a CustomCall operator that is later expanded by a new QrExpander pass. This allows backends to lower the Qr decomposition differently if they choose.
2020-09-21 17:35:34 -07:00
reshape_mover_test.cc
…
reshape_mover.cc
…
reshape_mover.h
…
rng_bit_generator_expander.cc
Rewrite output state to match the input.
2020-04-29 15:51:55 -07:00
rng_bit_generator_expander.h
Introduce new RngBitGenerator HLO
2020-02-06 05:06:44 -08:00
rng_expander.cc
Introduce new RngBitGenerator HLO
2020-02-06 05:06:44 -08:00
rng_expander.h
[XLA] Cache expanded RNG computations
2020-01-23 10:39:48 -08:00
root_instruction_sinker_test.cc
[XLA] Add a RootInstructionSinker pass.
2020-04-22 14:46:15 -07:00
root_instruction_sinker.cc
[XLA] Add a RootInstructionSinker pass.
2020-04-22 14:46:15 -07:00
root_instruction_sinker.h
[XLA] Add a RootInstructionSinker pass.
2020-04-22 14:46:15 -07:00
scatter_expander_test.cc
[XLA] Expand simple scatter operations into dynamic-update-slice.
2020-08-18 13:21:28 -07:00
scatter_expander.cc
[XLA] Expand simple scatter operations into dynamic-update-slice.
2020-08-18 13:21:28 -07:00
scatter_expander.h
[XLA] Expand simple scatter operations into dynamic-update-slice.
2020-08-18 13:21:28 -07:00
service_executable_run_options.h
…
service.cc
Dump HLO snapshots even when execution failed
2020-12-15 13:44:37 -08:00
service.h
[XLA/GPU] Add parallel compilation for LLVM and ptxas.
2020-12-08 09:22:49 -08:00
shape_inference_test.cc
[XLA] Be more consistent about inferring mixed operand shapes
2021-01-12 13:29:04 -08:00
shape_inference.cc
[XLA] Be more consistent about inferring mixed operand shapes
2021-01-12 13:29:04 -08:00
shape_inference.h
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
shaped_buffer_test.cc
Internal tests cleanup.
2020-10-28 07:40:36 -07:00
shaped_buffer.cc
Remove platform field from shaped buffer.
2020-10-30 14:26:55 -07:00
shaped_buffer.h
Remove platform field from shaped buffer.
2020-10-30 14:26:55 -07:00
sharding_propagation_test.cc
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
sharding_propagation.cc
[XLA] Add sharding propagation support for Gather sharding on parallel dimensions.
2021-01-16 01:56:57 -08:00
sharding_propagation.h
[XLA] Move sharding propagation to third party
2020-05-22 18:18:30 -07:00
slice_sinker_test.cc
…
slice_sinker.cc
…
slice_sinker.h
…
slow_operation_alarm.cc
Print HLO module name on slow compile
2020-12-10 23:19:52 -08:00
slow_operation_alarm.h
Print HLO module name on slow compile
2020-12-10 23:19:52 -08:00
sort_simplifier_test.cc
…
sort_simplifier.cc
…
sort_simplifier.h
…
source_map_util.h
…
space_to_batch_converter_test.cc
Allow space-to-batch to occur on base dilated convolutions
2020-11-23 12:49:19 -08:00
space_to_batch_converter.cc
[XLA] Enable RHS space-to-batch on backprop filter convolutions.
2021-01-08 18:49:11 -08:00
space_to_batch_converter.h
Extend space to batch to apply to larger batch sizes
2020-10-06 09:56:50 -07:00
stable_sort_expander_test.cc
…
stable_sort_expander.cc
…
stable_sort_expander.h
…
stream_pool_test.cc
…
stream_pool.cc
…
stream_pool.h
Prefixing TensorFlow thread annotation macros with TF_.
2020-03-05 08:42:01 -08:00
topk_rewriter_test.cc
[XLA] Rewrite 1d sort to TopK
2020-08-05 02:40:46 -07:00
topk_rewriter.cc
[XLA] Rewrite 1d sort to TopK
2020-08-05 02:40:46 -07:00
topk_rewriter.h
[XLA] Add TopK rewriter pass
2020-08-05 01:44:08 -07:00
transfer_manager.cc
Remove the use of host shape in TransferManager::ReadDynamicShapes.
2020-11-17 13:46:04 -08:00
transfer_manager.h
Remove the use of host shape in TransferManager::ReadDynamicShapes.
2020-11-17 13:46:04 -08:00
transpose_folding_test.cc
[XLA] Add an optional preferred_element_type argument to Dot/Conv builder methods to enable generating HLOs that has wider accumulation type than default shape inference result.
2020-11-20 13:16:56 -08:00
transpose_folding.cc
[XLA] Teach TransposeFolding to fold dots with batch dimensions
2020-10-08 09:13:33 -07:00
transpose_folding.h
[XLA] Add the ability to disable the Reduce(Reshape(X)) to Reduce(X) rewrite and have TransposeFolding default to folding all operands of Convolution and Dot
2020-03-07 00:36:17 -08:00
tree_reduction_rewriter.cc
…
tree_reduction_rewriter.h
…
triangular_solve_expander_test.cc
Merge pull request #39690 from whoozle:triangular-solve-expander-block-size
2020-07-15 06:57:41 -07:00
triangular_solve_expander.cc
[XLA] Fix bug in triangular solve expander.
2020-11-04 10:32:06 -08:00
triangular_solve_expander.h
[XLA] Add an HLO implementation of a direct triangular solve, and use it for small matrices (n<32) with large batches (>8).
2020-11-02 17:14:35 -08:00
tuple_points_to_analysis_test.cc
[XLA] Add missing custom-call aliasing-related changes in analysis passes
2021-01-12 13:57:06 -08:00
tuple_points_to_analysis.cc
[XLA] Avoid structured bindings.
2021-01-12 16:13:55 -08:00
tuple_points_to_analysis.h
[XLA] Add missing custom-call aliasing-related changes in analysis passes
2021-01-12 13:57:06 -08:00
tuple_simplifier_test.cc
…
tuple_simplifier.cc
[XLA:HLO] Small refactoring and more comments in tuple_simplifier.
2020-06-08 11:52:07 -07:00
tuple_simplifier.h
[XLA:HLO] Small refactoring and more comments in tuple_simplifier.
2020-06-08 11:52:07 -07:00
tuple_util_test.cc
…
tuple_util.cc
…
tuple_util.h
[XLA] Add a RootInstructionSinker pass.
2020-04-22 14:46:15 -07:00
while_loop_all_reduce_code_motion_test.cc
Rolling forward after fixing C++ version problems.
2020-12-30 13:24:09 -08:00
while_loop_all_reduce_code_motion.cc
Rolling forward after fixing C++ version problems.
2020-12-30 13:24:09 -08:00
while_loop_all_reduce_code_motion.h
Rolling forward after fixing C++ version problems.
2020-12-30 13:24:09 -08:00
while_loop_analysis_test.cc
…
while_loop_analysis.cc
…
while_loop_analysis.h
…
while_loop_constant_sinking_test.cc
…
while_loop_constant_sinking.cc
…
while_loop_constant_sinking.h
…
while_loop_invariant_code_motion_test.cc
…
while_loop_invariant_code_motion.cc
[XLA] Correct WhileLoopInvariantCodeMotion log messages.
2020-05-07 13:35:38 -07:00
while_loop_invariant_code_motion.h
…
while_loop_simplifier_test.cc
While loop simplification to remove repeated parameters.
2020-08-28 09:18:08 -07:00
while_loop_simplifier.cc
While loop simplification to remove repeated parameters.
2020-08-28 09:18:08 -07:00
while_loop_simplifier.h
…
while_loop_trip_count_annotator_test.cc
…
while_loop_trip_count_annotator.cc
…
while_loop_trip_count_annotator.h
…
while_util_test.cc
…
while_util.cc
[XLA][TF2XLA] Support tensor list with dynamic dimension.
2020-04-22 14:37:18 -07:00
while_util.h
[XLA][TF2XLA] Support tensor list with dynamic dimension.
2020-04-22 14:37:18 -07:00
zero_sized_hlo_elimination_test.cc
…
zero_sized_hlo_elimination.cc
Support 0-sized shape with dynamic reshapes.
2021-01-16 10:23:25 -08:00
zero_sized_hlo_elimination.h
…