from numba.core.typing.templates import ConcreteTemplate from numba.core import types, typing, funcdesc, config, compiler, sigutils from numba.core.compiler import (sanitize_compile_result_entries, CompilerBase, DefaultPassBuilder, Flags, Option, CompileResult) from numba.core.compiler_lock import global_compiler_lock from numba.core.compiler_machinery import (LoweringPass, AnalysisPass, PassManager, register_pass) from numba.core.errors import NumbaInvalidConfigWarning, TypingError from numba.core.typed_passes import (IRLegalization, NativeLowering, AnnotateTypes) from warnings import warn from numba.cuda.api import get_current_device def _nvvm_options_type(x): if x is None: return None else: assert isinstance(x, dict) return x class CUDAFlags(Flags): nvvm_options = Option( type=_nvvm_options_type, default=None, doc="NVVM options", ) compute_capability = Option( type=tuple, default=None, doc="Compute Capability", ) # The CUDACompileResult (CCR) has a specially-defined entry point equal to its # id. This is because the entry point is used as a key into a dict of # overloads by the base dispatcher. The id of the CCR is the only small and # unique property of a CompileResult in the CUDA target (cf. the CPU target, # which uses its entry_point, which is a pointer value). # # This does feel a little hackish, and there are two ways in which this could # be improved: # # 1. We could change the core of Numba so that each CompileResult has its own # unique ID that can be used as a key - e.g. a count, similar to the way in # which types have unique counts. # 2. At some future time when kernel launch uses a compiled function, the entry # point will no longer need to be a synthetic value, but will instead be a # pointer to the compiled function as in the CPU target. class CUDACompileResult(CompileResult): @property def entry_point(self): return id(self) def cuda_compile_result(**entries): entries = sanitize_compile_result_entries(entries) return CUDACompileResult(**entries) @register_pass(mutates_CFG=True, analysis_only=False) class CUDABackend(LoweringPass): _name = "cuda_backend" def __init__(self): LoweringPass.__init__(self) def run_pass(self, state): """ Back-end: Packages lowering output in a compile result """ lowered = state['cr'] signature = typing.signature(state.return_type, *state.args) state.cr = cuda_compile_result( typing_context=state.typingctx, target_context=state.targetctx, typing_error=state.status.fail_reason, type_annotation=state.type_annotation, library=state.library, call_helper=lowered.call_helper, signature=signature, fndesc=lowered.fndesc, ) return True @register_pass(mutates_CFG=False, analysis_only=False) class CreateLibrary(LoweringPass): """ Create a CUDACodeLibrary for the NativeLowering pass to populate. The NativeLowering pass will create a code library if none exists, but we need to set it up with nvvm_options from the flags if they are present. """ _name = "create_library" def __init__(self): LoweringPass.__init__(self) def run_pass(self, state): codegen = state.targetctx.codegen() name = state.func_id.func_qualname nvvm_options = state.flags.nvvm_options state.library = codegen.create_library(name, nvvm_options=nvvm_options) # Enable object caching upfront so that the library can be serialized. state.library.enable_object_caching() return True @register_pass(mutates_CFG=False, analysis_only=True) class CUDALegalization(AnalysisPass): _name = "cuda_legalization" def __init__(self): AnalysisPass.__init__(self) def run_pass(self, state): # Early return if NVVM 7 from numba.cuda.cudadrv.nvvm import NVVM if NVVM().is_nvvm70: return False # NVVM < 7, need to check for charseq typmap = state.typemap def check_dtype(dtype): if isinstance(dtype, (types.UnicodeCharSeq, types.CharSeq)): msg = (f"{k} is a char sequence type. This type is not " "supported with CUDA toolkit versions < 11.2. To " "use this type, you need to update your CUDA " "toolkit - try 'conda install cudatoolkit=11' if " "you are using conda to manage your environment.") raise TypingError(msg) elif isinstance(dtype, types.Record): for subdtype in dtype.fields.items(): # subdtype is a (name, _RecordField) pair check_dtype(subdtype[1].type) for k, v in typmap.items(): if isinstance(v, types.Array): check_dtype(v.dtype) return False class CUDACompiler(CompilerBase): def define_pipelines(self): dpb = DefaultPassBuilder pm = PassManager('cuda') untyped_passes = dpb.define_untyped_pipeline(self.state) pm.passes.extend(untyped_passes.passes) typed_passes = dpb.define_typed_pipeline(self.state) pm.passes.extend(typed_passes.passes) pm.add_pass(CUDALegalization, "CUDA legalization") lowering_passes = self.define_cuda_lowering_pipeline(self.state) pm.passes.extend(lowering_passes.passes) pm.finalize() return [pm] def define_cuda_lowering_pipeline(self, state): pm = PassManager('cuda_lowering') # legalise pm.add_pass(IRLegalization, "ensure IR is legal prior to lowering") pm.add_pass(AnnotateTypes, "annotate types") # lower pm.add_pass(CreateLibrary, "create library") pm.add_pass(NativeLowering, "native lowering") pm.add_pass(CUDABackend, "cuda backend") pm.finalize() return pm @global_compiler_lock def compile_cuda(pyfunc, return_type, args, debug=False, lineinfo=False, inline=False, fastmath=False, nvvm_options=None, cc=None): if cc is None: raise ValueError('Compute Capability must be supplied') from .descriptor import cuda_target typingctx = cuda_target.typing_context targetctx = cuda_target.target_context flags = CUDAFlags() # Do not compile (generate native code), just lower (to LLVM) flags.no_compile = True flags.no_cpython_wrapper = True flags.no_cfunc_wrapper = True # Both debug and lineinfo turn on debug information in the compiled code, # but we keep them separate arguments in case we later want to overload # some other behavior on the debug flag. In particular, -opt=3 is not # supported with debug enabled, and enabling only lineinfo should not # affect the error model. if debug or lineinfo: flags.debuginfo = True if lineinfo: flags.dbg_directives_only = True if debug: flags.error_model = 'python' else: flags.error_model = 'numpy' if inline: flags.forceinline = True if fastmath: flags.fastmath = True if nvvm_options: flags.nvvm_options = nvvm_options flags.compute_capability = cc # Run compilation pipeline from numba.core.target_extension import target_override with target_override('cuda'): cres = compiler.compile_extra(typingctx=typingctx, targetctx=targetctx, func=pyfunc, args=args, return_type=return_type, flags=flags, locals={}, pipeline_class=CUDACompiler) library = cres.library library.finalize() return cres @global_compiler_lock def compile_ptx(pyfunc, sig, debug=False, lineinfo=False, device=False, fastmath=False, cc=None, opt=True): """Compile a Python function to PTX for a given set of argument types. :param pyfunc: The Python function to compile. :param sig: The signature representing the function's input and output types. :param debug: Whether to include debug info in the generated PTX. :type debug: bool :param lineinfo: Whether to include a line mapping from the generated PTX to the source code. Usually this is used with optimized code (since debug mode would automatically include this), so we want debug info in the LLVM but only the line mapping in the final PTX. :type lineinfo: bool :param device: Whether to compile a device function. Defaults to ``False``, to compile global kernel functions. :type device: bool :param fastmath: Whether to enable fast math flags (ftz=1, prec_sqrt=0, prec_div=, and fma=1) :type fastmath: bool :param cc: Compute capability to compile for, as a tuple ``(MAJOR, MINOR)``. Defaults to ``(5, 0)``. :type cc: tuple :param opt: Enable optimizations. Defaults to ``True``. :type opt: bool :return: (ptx, resty): The PTX code and inferred return type :rtype: tuple """ if debug and opt: msg = ("debug=True with opt=True (the default) " "is not supported by CUDA. This may result in a crash" " - set debug=False or opt=False.") warn(NumbaInvalidConfigWarning(msg)) nvvm_options = { 'fastmath': fastmath, 'opt': 3 if opt else 0 } args, return_type = sigutils.normalize_signature(sig) cc = cc or config.CUDA_DEFAULT_PTX_CC cres = compile_cuda(pyfunc, return_type, args, debug=debug, lineinfo=lineinfo, fastmath=fastmath, nvvm_options=nvvm_options, cc=cc) resty = cres.signature.return_type if resty and not device and resty != types.void: raise TypeError("CUDA kernel must have void return type.") if device: lib = cres.library else: tgt = cres.target_context code = pyfunc.__code__ filename = code.co_filename linenum = code.co_firstlineno lib, kernel = tgt.prepare_cuda_kernel(cres.library, cres.fndesc, debug, lineinfo, nvvm_options, filename, linenum) ptx = lib.get_asm_str(cc=cc) return ptx, resty def compile_ptx_for_current_device(pyfunc, sig, debug=False, lineinfo=False, device=False, fastmath=False, opt=True): """Compile a Python function to PTX for a given set of argument types for the current device's compute capabilility. This calls :func:`compile_ptx` with an appropriate ``cc`` value for the current device.""" cc = get_current_device().compute_capability return compile_ptx(pyfunc, sig, debug=debug, lineinfo=lineinfo, device=device, fastmath=fastmath, cc=cc, opt=True) def declare_device_function(name, restype, argtypes): return declare_device_function_template(name, restype, argtypes).key def declare_device_function_template(name, restype, argtypes): from .descriptor import cuda_target typingctx = cuda_target.typing_context targetctx = cuda_target.target_context sig = typing.signature(restype, *argtypes) extfn = ExternFunction(name, sig) class device_function_template(ConcreteTemplate): key = extfn cases = [sig] fndesc = funcdesc.ExternalFunctionDescriptor( name=name, restype=restype, argtypes=argtypes) typingctx.insert_user_function(extfn, device_function_template) targetctx.insert_user_function(extfn, fndesc) return device_function_template class ExternFunction(object): def __init__(self, name, sig): self.name = name self.sig = sig