LLVM 19.0.0git
Functions
interactive_host Namespace Reference

Functions

def send (io.BufferedWriter f, Union[int, float] value, log_reader.TensorSpec spec)
 
def run_interactive (str temp_rootname, Callable[[List[log_reader.TensorValue]], Union[int, float]] make_response, List[str] process_and_args)
 

Detailed Description

Utility for testing InteractiveModelRunner.

Use it from pass-specific tests by providing a main .py which calls this library's
`run_interactive` with an appropriate callback to provide advice.

From .ll tests, just call the above-mentioned main as a prefix to the opt/llc
invocation (with the appropriate flags enabling the interactive mode)

Examples:
test/Transforms/Inline/ML/interactive-mode.ll
test/CodeGen/MLRegAlloc/interactive-mode.ll

Function Documentation

◆ run_interactive()

def interactive_host.run_interactive ( str  temp_rootname,
Callable[[List[log_reader.TensorValue]], Union[int, float]]  make_response,
List[str]  process_and_args 
)
Host the compiler.
Args:
  temp_rootname: the base file name from which to construct the 2 pipes for
  communicating with the compiler.
  make_response: a function that, given the current tensor values, provides a
  response.
  process_and_args: the full commandline for the compiler. It it assumed it
  contains a flag poiting to `temp_rootname` so that the InteractiveModeRunner
  would attempt communication on the same pair as this function opens.

This function sets up the communication with the compiler - via 2 files named
`temp_rootname`.in and `temp_rootname`.out - prints out the received features,
and sends back to the compiler an advice (which it gets from `make_response`).
It's used for testing, and also to showcase how to set up communication in an
interactive ML ("gym") environment.

Definition at line 35 of file interactive_host.py.

References log_reader.pretty_print_tensor_value(), print(), log_reader.read_header(), log_reader.read_one_observation(), and send().

◆ send()

def interactive_host.send ( io.BufferedWriter  f,
Union[int, float]  value,
log_reader.TensorSpec  spec 
)
Send the `value` - currently just a scalar - formatted as per `spec`.

Definition at line 23 of file interactive_host.py.

Referenced by run_interactive().