tf_1.8_xla_doc
|
namespace for xla More...
Namespaces | |
anonymous_namespace{user_computation.cc} | |
cpu | |
Functions | |
StatusOr< std::vector< const HloInstruction * > > | DefaultMemoryScheduler (const HloComputation &computation, const TuplePointsToAnalysis &points_to_analysis, const LogicalBuffer::SizeFunction &size_function) |
StatusOr< SequentialHloOrdering::HloModuleSequence > | CreateMemoryMinimizingSequence (const HloModule &module, const LogicalBuffer::SizeFunction &size_function, const MemorySchedulerAlgorithm &algorithm) |
namespace for xla
namespace of xla
StatusOr< SequentialHloOrdering::HloModuleSequence > xla::CreateMemoryMinimizingSequence | ( | const HloModule & | module, |
const LogicalBuffer::SizeFunction & | size_function, | ||
const MemorySchedulerAlgorithm & | algorithm | ||
) |
xla::TuplePointsToAnalysis::Run
xla::HloModule::MakeNonfusionComputations
xla::anonymous_namespace{hlo_scheduling.cc}::CreateMemoryMinimizingSequence
Google Docs:
Returns an HloModuleSequence which seeks to minimize the memory required for the computation. size_function is the function returning the number of bytes required for a LogicalBuffer.
StatusOr< std::vector< const HloInstruction * > > xla::DefaultMemoryScheduler | ( | const HloComputation & | computation, |
const TuplePointsToAnalysis & | points_to_analysis, | ||
const LogicalBuffer::SizeFunction & | size_function | ||
) |
Try list-scheduler based ordering and DFS based ordering choose the one requires less memory and ignore framentation.