./bazel-bin/xla/tools/hlo-opt --help

hlo-opt-help.txt

This tool lets you run a given HloModule from a file (or stdin) and convert it
to expanded HLO, fully optimized HLO, or a binary depending on options.

HLO passes are always run, unless the HLO module is already scheduled (has
is_scheduled=True).

You can also pass in debug option flags for the HloModule.

Usage:

  bazel run opt -- --platform=[gpu|cpu|...] path/to/hlo_module


usage: ./bazel-bin/xla/tools/hlo-opt
Flags:
	--o="-"                          	string	Output filename, or '-' for stdout (default).
	--platform="gpu"                 	string	The platform for which we perform the translation
	--format=""                      	string	The format of the input file. By default inferred from the filename. Valid values:
			  hlo : HLO textual format
			  pb : xla::HloProto in binary proto format
			  pbtxt : xla::HloProto in text proto format
	--stage="hlo"                    	string	Output stage to dump. Valid values depend on the platform, for GPUs:
			 * hlo : HLO after all optimizations
			 * llvm : LLVM IR
			 * ptx : PTX dump
			 * buffer-assignment: Buffer Assignment
			 * hlo-backend: HLO after backend passes

	--list-stages=false              	bool	Print all supported stages for a given platform and exit
	--split-input-file=false         	bool	Splits the input file in pieces based on '// -----' substring, and processes each chunk independently
	--xla_cpu_enable_fast_math=false 	bool	Enable unsafe fast-math optimizations in the CPU compiler; this may produce faster code at the expense of some accuracy.
	--xla_cpu_fast_math_honor_nans=true	bool	When xla_cpu_enable_fast_math is true then this controls whether we allow operations to produce NaNs. Ignored when xla_cpu_enable_fast_math is false.
	--xla_cpu_fast_math_honor_infs=true	bool	When xla_cpu_enable_fast_math is true then this controls whether we allow operations to produce infinites. Ignored when xla_cpu_enable_fast_math is false.
	--xla_cpu_fast_math_honor_division=true	bool	When xla_cpu_enable_fast_math is true then this controls whether we forbid to use multiplication by the reciprocal instead of division. Ignored when xla_cpu_enable_fast_math is false.
	--xla_cpu_fast_math_honor_functions=true	bool	When xla_cpu_enable_fast_math is true then this controls whether we forbid to approximate calculations for functions. Ignored when xla_cpu_enable_fast_math is false.
	--xla_cpu_enable_fast_min_max=true	bool	Enable fast floating point min/max lowering that always propagates NaNs.
	--xla_gpu_enable_fast_min_max=false	bool	Enable fast floating point min/max lowering that does not propagate NaNs.
	--xla_llvm_enable_alias_scope_metadata=true	bool	In LLVM-based backends, enable the emission of !alias.scope metadata in the generated IR.
	--xla_llvm_enable_noalias_metadata=true	bool	In LLVM-based backends, enable the emission of !noalias metadata in the generated IR.
	--xla_llvm_enable_invariant_load_metadata=true	bool	In LLVM-based backends, enable the emission of !invariant.load metadata in the generated IR.
	--xla_llvm_disable_expensive_passes=false	bool	In LLVM-based backends, disable a custom set of expensive optimization passes.
	--xla_backend_optimization_level=3	int32	Numerical optimization level for the XLA compiler backend.
	--xla_disable_hlo_passes=""      	string	Comma-separated list of hlo passes to be disabled. These names must exactly match the passes' names; no whitespace around commas.
	--xla_enable_hlo_passes_only=""  	string	Comma-separated list of hlo passes to be enabled. These names must exactly match the passes' names; no whitespace around commas. The unspecified passes are all disabled.
	--xla_disable_all_hlo_passes=false	bool	Disables all HLO passes. Notes that some passes are necessary for correctness and the invariants that must be satisfied by 'fully optimized' HLO are different for different devices and may change over time. The only 'guarantee', such as it is, is that if you compile XLA and dump the optimized HLO for some graph, you should be able to run it again on the same device with the same build of XLA.
	--xla_embed_ir_in_executable=false	bool	Embed the compiler IR as a string in the executable.
	--xla_eliminate_hlo_implicit_broadcast=true	bool	Eliminate implicit broadcasts when lowering user computations to HLO instructions; use explicit broadcast instead.
	--xla_cpu_multi_thread_eigen=true	bool	When generating calls to Eigen in the CPU backend, use multi-threaded Eigen mode.
	--xla_gpu_cuda_data_dir="./cuda_sdk_lib"	string	If non-empty, specifies a local directory containing ptxas and nvvm libdevice files; otherwise we use those from runfile directories.
	--xla_gpu_ftz=false              	bool	If true, flush-to-zero semantics are enabled in the code generated for GPUs.
	--xla_gpu_ptx_file=""            	string	If non-empty, specifies a file containing ptx to use. The filename prefix must have the same pattern as PTX dumped by XLA. This allows to match one specific module. General workflow. Get the generated module ptx from XLA, modify it, then pass it back via this option.
	--xla_gpu_llvm_ir_file=""        	string	If non-empty, specifies a file containing textual LLVM IR to use. The filename prefix must have the same pattern as LLVM dumped by XLA (i.e. module_0001.ir-no-opt.ll -> module_0001.MY_NEW_FILE.ll). This allows to match one specific module. General workflow. Get the not optimized LLVM IR from XLA, modify it, then pass it back via this option.
	--xla_test_all_output_layouts=false	bool	Let ClientLibraryTestBase::ComputeAndCompare* test all permutations of output layouts. For example, with a 3D shape, all permutations of the set {0, 1, 2} are tried.
	--xla_test_all_input_layouts=false	bool	Let ClientLibraryTestBase::ComputeAndCompare* test all permutations of *input* layouts. For example, for 2 input arguments with 2D shape and 4D shape, the computation will run 2! * 4! times for every possible layouts
	--xla_hlo_profile=false          	bool	Instrument the computation to collect per-HLO cycle counts
	--xla_backend_extra_options=""   	string	Extra options to pass to a backend; comma-separated list of 'key=val' strings (=val may be omitted); no whitespace around commas.
	--xla_cpu_use_mkl_dnn=false      	bool	Generate calls to MKL-DNN in the CPU backend.
	--xla_cpu_use_acl=false          	bool	Generate calls to ACL (Arm Compute Library) in the CPU backend.
	--xla_cpu_use_xla_runtime=false  	bool	Enable XLA Runtime in the CPU backend.
	--xla_cpu_sparse_cuda_threads=0  	int32	Sets number fo CUDA threads for sparse GPU acceleration in the CPU backend (0 = off).
	--xla_gpu_crash_on_verification_failures=false	bool	Crashes the program on extra verification failures, e.g. cuDNN cross checking failures
	--xla_gpu_strict_conv_algorithm_picker=true	bool	Upgrades warnings to failures when all algorithms fail conv autotuning.
	--xla_gpu_autotune_level=4       	int32	Set GEMM and Convolution auto-tuning level. 0 = off; 1 = on; 2 = on+init; 3 = on+init+reinit; 4 = on+init+reinit+check.
	--xla_force_host_platform_device_count=1	int32	Force the host platform to pretend that there are these many host "devices". All of these host devices are backed by the same threadpool. Setting this to anything other than 1 can increase overhead from context switching but we let the user override this behavior to help run tests on the host that run models in parallel across multiple devices.
	--xla_gpu_disable_gpuasm_optimizations=false	bool	In XLA:GPU run ptxas in -O0 (default is -O3).
	--xla_gpu_asm_extra_flags=""     	string	Pass extra parameters to the GPU assembler tool (i.e., ptxas for CUDA). If multiple parameters, separate them by comma.
	--xla_fuel=""                    	string	Sets compiler fuel, useful for bisecting bugs in passes. Format --xla_fuel=PASS1=NUM1,PASS2=NUM2,...
	--xla_dump_to=""                 	string	Directory into which debugging data is written. If not specified but another dumping flag is passed, data will be written to stdout. To explicitly write to stdout, set this to "-". The values "sponge" and "test_undeclared_outputs_dir" have a special meaning: They cause us to dump into the directory specified by the environment variable TEST_UNDECLARED_OUTPUTS_DIR.
	--xla_dump_hlo_as_text=false     	bool	Dumps HLO modules as text before and after optimizations. debug_options are written to the --xla_dump_to dir, or, if no dir is specified, to stdout.
	--xla_dump_hlo_as_long_text=false	bool	Dumps HLO modules as long text before and after optimizations. debug_options are written to the --xla_dump_to dir, or, if no dir is specified, to stdout. Ignored unless xla_dump_hlo_as_text is true.
	--xla_dump_hlo_as_proto=false    	bool	Dumps HLO modules as HloProtos to the directory specified by --xla_dump_to.
	--xla_dump_hlo_as_dot=false      	bool	Dumps HLO modules rendered as dot files to the directory specified by --xla_dump_to.
	--xla_dump_hlo_as_html=false     	bool	Dumps HLO modules rendered as HTML files to the directory specified by --xla_dump_to.
	--xla_dump_hlo_as_url=false      	bool	Tries to dump HLO modules rendered as URLs to stdout (and also to the directory specified by --xla_dump_to). This is not implemented by default; you need to add a plugin which calls RegisterGraphToURLRenderer().
	--xla_dump_fusion_visualization=false	bool	Tries to generate HLO fusion visualization as an HTML page to the directory specified by --xla_dump_to). This is not implemented by default; you need to add a plugin which calls RegisterGraphToURLRenderer(). Generates a file per computation. Currently only implemented for the GPU backend.
	--xla_dump_hlo_snapshots=false   	bool	Every time an HLO module is run, dumps an HloSnapshot to the directory specified by --xla_dump_to.
	--xla_dump_hlo_module_re=""      	string	Limits dumping only to modules which match this regular expression. Default is to dump all modules.
	--xla_dump_hlo_pass_re=""        	string	If specified, dumps HLO before and after optimization passes which match this regular expression, in addition to dumping at the very beginning and end of compilation.
	--xla_dump_include_timestamp=false	bool	If specified, includes a timestamp in the dumped filenames.
	--xla_dump_max_hlo_modules=-1    	int32	Max number of hlo module dumps in a directory. Set to < 0 for unbounded.
	--xla_dump_module_metadata=false 	bool	Dumps HloModuleMetadata as text protos to the directory specified by --xla_dump_to.
	--xla_dump_compress_protos=false 	bool	Gzip-compress protos dumped by --xla_dump_hlo_as_proto.
	--xla_hlo_graph_addresses=false  	bool	When rendering graphs (--xla_dump_hlo_as_{dot,html,url}), displays the address in memory of each HloInstruction object.
	--xla_hlo_graph_sharding_color=false	bool	Assign colors based on sharding assignments when generating the HLO graphs.
	--xla_allow_excess_precision=true	bool	Allow xla to increase the output precision of an instruction.
	--xla_gpu_force_conv_nchw=false  	bool	For cuDNN convolutions, always use NCHW layouts.
	--xla_gpu_force_conv_nhwc=false  	bool	For cuDNN convolutions, always use NHWC layouts.
	--xla_gpu_algorithm_denylist_path=""	string	An AlgorithmDenylist text proto file as a denylist of convolutions to avoid to use.
	--xla_gpu_use_runtime_fusion=false	bool	For using cuDNN runtime compiled fusion kernels.
	--xla_tpu_detect_nan=false       	bool	Trigger error on execution on TPU if a NAN value is detected
	--xla_tpu_detect_inf=false       	bool	Trigger error on execution on TPU if a INF value is detected
	--xla_cpu_enable_xprof_traceme=false	bool	If true, XLA CPU generates code to call TraceMe::Activity{Start|End} around HLO operations.
	--xla_gpu_unsafe_fallback_to_driver_on_ptxas_not_found=false	bool	If true, XLA GPU falls back to the driver if ptxas is not found. Note that falling back to the driver can have drawbacks like using more memory and/or other bugs during compilation, so we recommend setting this flag to false.
	--xla_multiheap_size_constraint_per_heap=-1	int32	Generates multiple heaps (i.e., temp buffers) with a size constraint on each heap to avoid Out-of-Memory due to memory fragmentation. The constraint is soft, so it works with tensors larger than the given constraint size. -1 corresponds to no constraints.
	--xla_gpu_force_compilation_parallelism=0	int32	Overrides normal multi-threaded compilation setting to use this many threads. Setting to 0 (the default value) means no enforcement.
	--xla_gpu_enable_llvm_module_compilation_parallelism=false	bool	Decides whether we can do LLVM module compilation in a parallelised way. If set to false, then it will be single threaded, otherwise the number of threads depends on the --xla_gpu_force_compilation_parallelism flag and the thread pool supplied to GpuCompiler.
	--xla_gpu_deterministic_ops=false	bool	Guarantees run-to-run determinism on GPU.
	--xla_gpu_enable_async_collectives=true	bool	Converts synchronous collective ops into asynchronous.
	--xla_gpu_enable_async_all_reduce=true	bool	Converts synchronous all-reduce ops into asynchronous.
	--xla_gpu_enable_async_collective_permute=false	bool	Converts synchronous collective-permute ops into asynchronous.
	--xla_gpu_enable_async_all_gather=false	bool	Converts synchronous all-gather ops into asynchronous.
	--xla_gpu_enable_async_reduce_scatter=false	bool	Converts synchronous reduce-scatter ops into asynchronous.
	--xla_gpu_enable_async_all_to_all=false	bool	Converts synchronous all-to-all ops into asynchronous.
	--xla_gpu_all_reduce_combine_threshold_bytes=31457280	int64	Size threshold (in bytes) for the GPU all-reduce combiner.
	--xla_gpu_all_gather_combine_threshold_bytes=31457280	int64	Size threshold (in bytes) for the GPU all-gather combiner.
	--xla_gpu_reduce_scatter_combine_threshold_bytes=31457280	int64	Size threshold (in bytes) for the GPU reduce-scatter combiner.
	--xla_gpu_enable_all_gather_combine_by_dim=true	bool	Combine all-gather ops with the same gather dimension or irrespective of their dimension.
	--xla_gpu_enable_reduce_scatter_combine_by_dim=true	bool	Combine reduce-scatter ops with the same dimension or irrespective of their dimension.
	--xla_gpu_all_reduce_contiguous=false	bool	Combine all-reduces into a single operation over a contiguous buffer.
	--xla_gpu_all_reduce_blueconnect_num_devices_per_host=0	int32	Number of devices per host for first stage of BlueConnect decomposition pass. The pass will attempt to decompose all-reduces ops into a ReduceScatter-AllReduce-AllGather sequence, with the initial ReduceScatter being performed over all of the devices in the same host. Set to < 1 to disable all-reduce decomposition.
	--xla_gpu_enable_while_loop_reduce_scatter_code_motion=false	bool	Enable hoisting of reduce-scatter outside while loops.
	--xla_gpu_collective_inflation_factor=1	int32	Inflation factor for collectives. If set to > 1, each XLA/GPU collective will execute multiple times (will yield incorrect results)
	--xla_gpu_enable_reassociation_for_converted_a