Skip to content
Branch: master
Find file History
yunxing and tensorflower-gardener [XLA] Use module scheduler to reduce compile time.
Previously, we only have computation scheduler, which runs heap
simulator once per computation. For models with large number of
computation, this creates extremly slow compilation time.

This cl introduces module scheduler, that only runs heap simulator
after the whole module is scheduled. It also contains a helper
function that automatically converts a computation scheduler to module
scheduler.

PiperOrigin-RevId: 258436352
Latest commit a07965c Jul 16, 2019
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
tests
BUILD
buffer_info_util.cc
buffer_info_util.h
build_defs.bzl
compiler_functor.cc
compiler_functor.h
conv_canonicalization.cc
conv_canonicalization.h
conv_canonicalization_test.cc
cpu_compiler.cc
cpu_compiler.h
cpu_eigen_tensor_alignment_test.cc
cpu_executable.cc
cpu_executable.h
cpu_hlo_support_checker.cc
cpu_hlo_support_checker.h
cpu_hlo_support_checker_test.cc
cpu_instruction_fusion.cc
cpu_instruction_fusion.h
cpu_instruction_fusion_test.cc
cpu_layout_assignment.cc
cpu_layout_assignment.h
cpu_layout_assignment_test.cc
cpu_options.cc
cpu_options.h
cpu_runtime.cc
cpu_runtime.h
cpu_runtime_test.cc
cpu_transfer_manager.cc
cpu_transfer_manager.h
disassembler.cc
disassembler.h
dot_op_emitter.cc
dot_op_emitter.h
dot_op_emitter_internal.h
elemental_ir_emitter.cc
elemental_ir_emitter.h
ir_emission_utils.cc
ir_emission_utils.h
ir_emission_utils_test.cc
ir_emitter.cc
ir_emitter.h
ir_function.cc
ir_function.h
llvm_ir_runtime.cc
llvm_ir_runtime.h
orc_jit_memory_mapper.cc
orc_jit_memory_mapper.h
parallel_loop_emitter.cc
parallel_loop_emitter.h
parallel_task_assignment.cc
parallel_task_assignment.h
parallel_task_assignment_test.cc
runtime_conv2d.cc
runtime_conv2d.h
runtime_conv2d_impl.h
runtime_conv2d_mkl.cc
runtime_conv2d_mkl.h
runtime_fft.cc
runtime_fft.h
runtime_fft_impl.h
runtime_fork_join.cc
runtime_fork_join.h
runtime_fp16.cc
runtime_fp16.h
runtime_key_value_sort.cc
runtime_key_value_sort.h
runtime_lightweight_check.h
runtime_matmul.cc
runtime_matmul.h
runtime_matmul_mkl.cc
runtime_matmul_mkl.h
runtime_matvec.h
runtime_single_threaded_conv2d.cc
runtime_single_threaded_conv2d.h
runtime_single_threaded_fft.cc
runtime_single_threaded_fft.h
runtime_single_threaded_matmul.cc
runtime_single_threaded_matmul.h
sample_harness.cc
shape_partition.cc
shape_partition.h
shape_partition_test.cc
simple_orc_jit.cc
simple_orc_jit.h
target_machine_features.cc
target_machine_features.h
target_machine_features_fake.h
tiled_dot_emitter.cc
tiled_dot_emitter.h
vector_support_library.cc
vector_support_library.h
windows_compatibility.cc
windows_compatibility.h
xfeed_manager.cc
xfeed_manager.h
xfeed_manager_test.cc
You can’t perform that action at this time.