numba_dpex.core.passes.passes#

Overview#

Classes#
`NoPythonBackend`	Base class for function passes
`DumpParforDiagnostics`	Base class for analysis passes (no modification made to state)
`QualNameDisambiguationLowering`	Qualified name disambiguation lowering pass

Classes#

class NoPythonBackend#

Bases: numba.core.compiler_machinery.FunctionPass

Base class for function passes

Overview

Methods#
`run_pass`(state)	Back-end: Generate LLVM IR from Numba IR, compile to machine code

Members

run_pass(state)#: Back-end: Generate LLVM IR from Numba IR, compile to machine code

class DumpParforDiagnostics#

Bases: numba.core.compiler_machinery.AnalysisPass

Base class for analysis passes (no modification made to state)

Overview

Methods#
`run_pass`(state)	Runs the pass itself. Must return True/False depending on whether

Members

run_pass(state)#: Runs the pass itself. Must return True/False depending on whether statement level modification took place.

class QualNameDisambiguationLowering#

Bases: numba.core.typed_passes.NativeLowering

Qualified name disambiguation lowering pass

If there are multiple @func decorated functions exist inside another @func decorated block, the numba compiler machinery creates same qualified names for different compiled function. Therefore, we utilize unique_name to resolve the ambiguity.

Parameters:

NativeLowering (CompilerPass) – Superclass from which this
inherited. (class has been) –

Returns:

True if run_pass() of the superclass is successful.

Return type:

bool

Overview

Methods#
`run_pass`(state)	Runs the pass itself. Must return True/False depending on whether

Members

run_pass(state)#: Runs the pass itself. Must return True/False depending on whether statement level modification took place.