Add the cuda.core.experiemental.Linker class

cuda_core/cuda/core/experimental/_linker.py

@@ -0,0 +1,282 @@
+from cuda.core.experimental._module import ObjectCode
+from cuda.core.experimental._utils import check_or_create_options
+from dataclasses import dataclass
+from typing import Optional, List
+from cuda.bindings import nvjitlink
+
+
+@dataclass
+class LinkerOptions:
+    """Customizable :obj:`LinkerOptions` for nvJitLink.
+
+    Attributes
+    ----------
+    arch : str
+        Pass SM architecture value. Can use compute_<N> value instead if only generating PTX.
+        This is a required option.
+        Acceptable value type: str
+        Maps to: -arch=sm_<N>
+    max_register_count : int, optional
+        Maximum register count.
+        Default: None
+        Acceptable value type: int
+        Maps to: -maxrregcount=<N>
+    time : bool, optional
+        Print timing information to InfoLog.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -time
+    verbose : bool, optional
+        Print verbose messages to InfoLog.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -verbose
+    link_time_optimization : bool, optional
+        Perform link time optimization.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -lto
+    ptx : bool, optional
+        Emit PTX after linking instead of CUBIN; only supported with -lto.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -ptx
+    optimization_level : int, optional
+        Set optimization level. Only 0 and 3 are accepted.
+        Default: None
+        Acceptable value type: int
+        Maps to: -O<N>
+    debug : bool, optional
+        Generate debug information.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -g
+    lineinfo : bool, optional
+        Generate line information.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -lineinfo
+    ftz : bool, optional
+        Flush denormal values to zero.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -ftz=<n>
+    prec_div : bool, optional
+        Use precise division.
+        Default: True
+        Acceptable value type: bool
+        Maps to: -prec-div=<n>
+    prec_sqrt : bool, optional
+        Use precise square root.
+        Default: True
+        Acceptable value type: bool
+        Maps to: -prec-sqrt=<n>
+    fma : bool, optional
+        Use fast multiply-add.
+        Default: True
+        Acceptable value type: bool
+        Maps to: -fma=<n>
+    kernels_used : List[str], optional
+        Pass list of kernels that are used; any not in the list can be removed. This option can be specified multiple times.
+        Default: None
+        Acceptable value type: list of str
+        Maps to: -kernels-used=<name>
+    variables_used : List[str], optional
+        Pass list of variables that are used; any not in the list can be removed. This option can be specified multiple times.
+        Default: None
+        Acceptable value type: list of str
+        Maps to: -variables-used=<name>
+    optimize_unused_variables : bool, optional
+        Assume that if a variable is not referenced in device code, it can be removed.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -optimize-unused-variables
+    xptxas : List[str], optional
+        Pass options to PTXAS. This option can be called multiple times.
+        Default: None
+        Acceptable value type: list of str
+        Maps to: -Xptxas=<opt>
+    split_compile : int, optional
+        Split compilation maximum thread count. Use 0 to use all available processors. Value of 1 disables split compilation (default).
+        Default: 1
+        Acceptable value type: int
+        Maps to: -split-compile=<N>
+    split_compile_extended : int, optional
+        A more aggressive form of split compilation available in LTO mode only. Accepts a maximum thread count value. Use 0 to use all available processors. Value of 1 disables extended split compilation (default). Note: This option can potentially impact performance of the compiled binary.
+        Default: 1
+        Acceptable value type: int
+        Maps to: -split-compile-extended=<N>
+    jump_table_density : int, optional
+        When doing LTO, specify the case density percentage in switch statements, and use it as a minimal threshold to determine whether jump table (brx.idx instruction) will be used to implement a switch statement. Default value is 101. The percentage ranges from 0 to 101 inclusively.
+        Default: 101
+        Acceptable value type: int
+        Maps to: -jump-table-density=<N>
+    no_cache : bool, optional
+        Do not cache the intermediate steps of nvJitLink.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -no-cache
+    device_stack_protector : bool, optional
+        Enable stack canaries in device code. Stack canaries make it more difficult to exploit certain types of memory safety bugs involving stack-local variables. The compiler uses heuristics to assess the risk of such a bug in each function. Only those functions which are deemed high-risk make use of a stack canary.
+        Default: False
+        Acceptable value type: bool
+        Maps to: -device-stack-protector
+    """
+    arch: str
+    max_register_count: Optional[int] = None
+    time: Optional[bool] = None
+    verbose: Optional[bool] = None
+    link_time_optimization: Optional[bool] = None
+    ptx: Optional[bool] = None
+    optimization_level: Optional[int] = None
+    debug: Optional[bool] = None
+    lineinfo: Optional[bool] = None
+    ftz: Optional[bool] = None
+    prec_div: Optional[bool] = None
+    prec_sqrt: Optional[bool] = None
+    fma: Optional[bool] = None
+    kernels_used: Optional[List[str]] = None
+    variables_used: Optional[List[str]] = None
+    optimize_unused_variables: Optional[bool] = None
+    xptxas: Optional[List[str]] = None
+    split_compile: Optional[int] = None
+    split_compile_extended: Optional[int] = None
+    jump_table_density: Optional[int] = None
+    no_cache: Optional[bool] = None
+    device_stack_protector: Optional[bool] = None
+
+    def __post_init__(self):
+        self.formatted_options = []
+        if self.arch is not None:
+            self.formatted_options.append(f"-arch={self.arch}")
+        if self.max_register_count is not None:
+            self.formatted_options.append(f"-maxrregcount={self.max_register_count}")
+        if self.time is not None:
+            self.formatted_options.append("-time")
+        if self.verbose is not None:
+            self.formatted_options.append("-verbose")
+        if self.link_time_optimization is not None:
+            self.formatted_options.append("-lto")
+        if self.ptx is not None:
+            self.formatted_options.append("-ptx")
+        if self.optimization_level is not None:
+            self.formatted_options.append(f"-O{self.optimization_level}")
+        if self.debug is not None:
+            self.formatted_options.append("-g")
+        if self.lineinfo is not None:
+            self.formatted_options.append("-lineinfo")
+        if self.ftz is not None:
+            self.formatted_options.append(f"-ftz={'true' if self.ftz else 'false'}")
+        if self.prec_div is not None:
+            self.formatted_options.append(f"-prec-div={'true' if self.prec_div else 'false'}")
+        if self.prec_sqrt is not None:
+            self.formatted_options.append(f"-prec-sqrt={'true' if self.prec_sqrt else 'false'}")
+        if self.fma is not None:
+            self.formatted_options.append(f"-fma={'true' if self.fma else 'false'}")
+        if self.kernels_used is not None:
+            for kernel in self.kernels_used:
+                self.formatted_options.append(f"-kernels-used={kernel}")
+        if self.variables_used is not None:
+            for variable in self.variables_used:
+                self.formatted_options.append(f"-variables-used={variable}")
+        if self.optimize_unused_variables is not None:
+            self.formatted_options.append("-optimize-unused-variables")
+        if self.xptxas is not None:
+            for opt in self.xptxas:
+                self.formatted_options.append(f"-Xptxas={opt}")
+        if self.split_compile is not None:
+            self.formatted_options.append(f"-split-compile={self.split_compile}")
+        if self.split_compile_extended is not None:
+            self.formatted_options.append(f"-split-compile-extended={self.split_compile_extended}")
+        if self.jump_table_density is not None:
+            self.formatted_options.append(f"-jump-table-density={self.jump_table_density}")
+        if self.no_cache is not None:
+            self.formatted_options.append("-no-cache")
+        if self.device_stack_protector is not None:
+            self.formatted_options.append("-device-stack-protector")
+
+
+class Linker:
+
+    __slots__ = "_handle"
+
+    def __init__(self, *object_codes : ObjectCode, options: LinkerOptions = None):
+        self._handle = None
+        options = check_or_create_options(LinkerOptions, options, "Linker options")
+        self._handle = nvjitlink.create(len(options.formatted_options), options.formatted_options)
+
+        if object_codes is not None:
+            for code in object_codes:
+                assert isinstance(code, ObjectCode)
+                self._add_code_object(code)
+
+
+    def _add_code_object(self, object_code: ObjectCode):
+        data = object_code._module
+        assert isinstance(data, bytes)
+        nvjitlink.add_data(
+            self._handle,
+            self._input_type_from_code_type(object_code._code_type),
+            data,
+            len(data),
+            f"{object_code._handle}_{object_code._code_type}",
+        )
+
+
+    def link(self, target_type) -> ObjectCode:
+        nvjitlink.complete(self._handle)
+        if target_type not in ["cubin", "ptx"]:
+            raise ValueError(f"Unsupported target type: {target_type}")
+        code = None
+        if target_type == "cubin":
+            cubin_size = nvjitlink.get_linked_cubin_size(self._handle)
+            code = bytearray(cubin_size)
+            nvjitlink.get_linked_cubin(self._handle, code)
+        else:
+            ptx_size = nvjitlink.get_linked_ptx_size(self._handle)
+            code = bytearray(ptx_size)
+            nvjitlink.get_linked_ptx(self._handle, code)
+
+        return ObjectCode(bytes(code), target_type)
+
+
+    def get_error_log(self) -> str:
+        log_size = nvjitlink.get_error_log_size(self._handle)
+        log = bytearray(log_size)
+        nvjitlink.get_error_log(self._handle, log)
+        return log.decode()
+
+
+    def get_info_log(self) -> str:
+        log_size = nvjitlink.get_info_log_size(self._handle)
+        log = bytearray(log_size)
+        nvjitlink.get_info_log(self._handle, log)
+        return log.decode()
+
+
+    def _input_type_from_code_type(self, code_type: str) -> nvjitlink.InputType:
+        # this list is based on the supported values for code_type in the ObjectCode class definition. nvjitlink supports other options for input type
+        if code_type == "ptx":
+            return nvjitlink.InputType.PTX
+        elif code_type == "cubin":
+            return nvjitlink.InputType.CUBIN
+        elif code_type == "fatbin":
+            return nvjitlink.InputType.FATBIN
+        elif code_type == "ltoir":
+            return nvjitlink.InputType.LTOIR
+        elif code_type == "object":
+            return nvjitlink.InputType.OBJECT
+        else:
+            raise ValueError(
+                f"Unknown code_type associated with ObjectCode: {code_type}"
+            )
+
+
+    @property
+    def handle(self) -> int:
+        return self._handle
+
+    def __del__(self):
+        if self._handle is not None:
+            nvjitlink.destroy(self._handle)
+            self._handle = None

cuda_core/tests/test_linker.py

@@ -0,0 +1,98 @@
+import pytest
+from cuda.core.experimental._linker import Linker, LinkerOptions
+from cuda.core.experimental._module import ObjectCode
+from cuda.core.experimental._program import Program
+
+ARCH = "sm_80" # use sm_80 for testing the oop nvJitLink wrapper
+empty_entrypoint_kernel = "__global__ void A() {}"
+empty_kernel = "__device__ void B() {}"
+addition_kernel = "__device__ int C(int a, int b) { return a + b; }"
+
+
+@pytest.fixture(scope="module")
+def compile_ptx_functions(init_cuda):
+
+    object_code_a_ptx = Program(empty_entrypoint_kernel, "c++").compile("ptx")
+    object_code_b_ptx = Program(empty_kernel, "c++").compile("ptx")
+    object_code_c_ptx = Program(addition_kernel, "c++").compile("ptx")
+
+    return object_code_a_ptx, object_code_b_ptx, object_code_c_ptx
+
+
+@pytest.fixture(scope="module")
+def compile_ltoir_functions(init_cuda):
+    object_code_a_ltoir = Program(empty_entrypoint_kernel, "c++").compile("ltoir", options=("-dlto",))
+    object_code_b_ltoir = Program(empty_kernel, "c++").compile("ltoir", options=("-dlto",))
+    object_code_c_ltoir = Program(addition_kernel, "c++").compile("ltoir", options=("-dlto",))
+
+    return object_code_a_ltoir, object_code_b_ltoir, object_code_c_ltoir
+
+
+@pytest.mark.parametrize("options", [
+    LinkerOptions(arch=ARCH),
+    LinkerOptions(arch=ARCH, max_register_count=32),
+    LinkerOptions(arch=ARCH, time=True),
+    LinkerOptions(arch=ARCH, verbose=True),
+    LinkerOptions(arch=ARCH, optimization_level=3),
+    LinkerOptions(arch=ARCH, debug=True),
+    LinkerOptions(arch=ARCH, lineinfo=True),
+    LinkerOptions(arch=ARCH, ftz=True),
+    LinkerOptions(arch=ARCH, prec_div=True),
+    LinkerOptions(arch=ARCH, prec_sqrt=True),
+    LinkerOptions(arch=ARCH, fma=True),
+    LinkerOptions(arch=ARCH, kernels_used=["kernel1"]),
+    LinkerOptions(arch=ARCH, variables_used=["var1"]),
+    LinkerOptions(arch=ARCH, optimize_unused_variables=True),
+    LinkerOptions(arch=ARCH, xptxas=["-v"]),
+    LinkerOptions(arch=ARCH, split_compile=0),
+    LinkerOptions(arch=ARCH, split_compile_extended=1),
+    LinkerOptions(arch=ARCH, jump_table_density=100),
+    LinkerOptions(arch=ARCH, no_cache=True)
+])
+def test_linker_init(compile_ptx_functions, options):
+    linker = Linker(*compile_ptx_functions, options=options)
+    object_code = linker.link("cubin")
+    assert isinstance(object_code, ObjectCode)
+
+
+def test_linker_init_invalid_arch():
+    options = LinkerOptions(arch=None)
+    with pytest.raises(TypeError):
+        Linker(options)
+
+
+def test_linker_link_ptx(compile_ltoir_functions):
+    options = LinkerOptions(arch=ARCH, link_time_optimization=True, ptx=True)
+    linker = Linker(*compile_ltoir_functions, options = options)
+    linked_code = linker.link("ptx")
+    assert isinstance(linked_code, ObjectCode)
+
+
+def test_linker_link_cubin(compile_ptx_functions):
+    options = LinkerOptions(arch=ARCH)
+    linker = Linker(*compile_ptx_functions, options=options)
+    linked_code = linker.link("cubin")
+    assert isinstance(linked_code, ObjectCode)
+
+
+def test_linker_link_invalid_target_type(compile_ptx_functions):
+    options = LinkerOptions(arch=ARCH)
+    linker = Linker(*compile_ptx_functions, options=options)
+    with pytest.raises(ValueError):
+        linker.link("invalid_target")
+
+
+def test_linker_get_error_log(compile_ptx_functions):
+    options = LinkerOptions(arch=ARCH)
+    linker = Linker(*compile_ptx_functions, options=options)
+    linker.link("cubin")
+    log = linker.get_error_log()
+    assert isinstance(log, str)
+
+
+def test_linker_get_info_log(compile_ptx_functions):
+    options = LinkerOptions(arch=ARCH)
+    linker = Linker(*compile_ptx_functions, options=options)
+    linker.link("cubin")
+    log = linker.get_info_log()
+    assert isinstance(log, str)