/default.etcd

# Also ignore protoc installed by hack/install-protoc.sh
/third_party/protoc*

# User cluster configs
.kubeconfig

.tags*

# Version file for dockerized build
.dockerized-kube-version-defs

# Web UI
/www/master/node_modules/
/www/master/npm-debug.log
/www/master/shared/config/development.json

# Karma output
/www/test_out

    tbl_outs = [
        (
            ["-gen-op-decls"],
            "tfrt_ops.h.inc",
        ),
        (
            ["-gen-op-defs"],
            "tfrt_ops.cc.inc",
        ),
    ],
    tblgen = "@llvm-project//mlir:mlir-tblgen",
    td_file = "tfrt_ops.td",
    deps = [
        ":tensorflow_tfrt_ops_td_files",
    ],
)

cc_library(
    name = "tensorflow_tfrt_ops",
    srcs = [

  for (Node* node : graph->nodes()) {
    if (CanCreateXlaKernel(node->def())) {
      const FunctionBody* fbody = nullptr;
      std::vector<int> constant_arg_indices;
      std::vector<int> resource_arg_indices;

      NameAttrList function;
      TF_RETURN_IF_ERROR(NameAndAttrsFromFunctionCall(node->def(), &function));

      // Force all constants to be on the host memory.

	// Type-check the package.
	// We create an empty map for each kind of input
	// we're interested in, and Check populates them.
	info := types.Info{
		Types: make(map[ast.Expr]types.TypeAndValue),
		Defs:  make(map[*ast.Ident]types.Object),
		Uses:  make(map[*ast.Ident]types.Object),
	}
	var conf types.Config
	pkg, err := conf.Check("fib", fset, []*ast.File{f}, &info)
	if err != nil {
		log.Fatal(err)
	}

load("@python_version_repo//:py_version.bzl", "TF_PYTHON_VERSION")

python_register_toolchains(
    name = "python",
    ignore_root_user_error = True,
    python_version = TF_PYTHON_VERSION,
)

load("@python//:defs.bzl", "interpreter")
load("@rules_python//python:pip.bzl", "package_annotation", "pip_parse")

NUMPY_ANNOTATIONS = {
    "numpy": package_annotation(
        additive_build_content = """\
filegroup(

			}
			ctxt.nonpkgdefs = append(ctxt.nonpkgdefs, s)
			nonpkgidx++
		} else {
			s.PkgIdx = goobj.PkgIdxSelf
			s.SymIdx = idx
			if idx != int32(len(ctxt.defs)) {
				panic("bad index")
			}
			ctxt.defs = append(ctxt.defs, s)
			idx++
		}
		s.Set(AttrIndexed, true)
	})

	ipkg := int32(1) // 0 is invalid index
	nonpkgdef := nonpkgidx

		if f != nil && fnDecls == nil {
			fnDecls = make(map[*types.Func]*ast.FuncDecl)
			for _, file := range files {
				for _, decl := range file.Decls {
					if fnDecl, ok := decl.(*ast.FuncDecl); ok {
						if fn, ok := info.Defs[fnDecl.Name].(*types.Func); ok {
							fnDecls[fn] = fnDecl
						}
					}
				}
			}
		}
		// TODO: set f = f.Origin() here.
		return fnDecls[f]
	}
}

				continue
			}
			if s.OnList() {
				panic("a symbol is added to defs multiple times")
			}
			s.PkgIdx = goobj.PkgIdxSelf
			s.SymIdx = symidx
			s.Set(AttrIndexed, true)
			s.Set(AttrOnList, true)
			symidx++
			infosyms = append(infosyms, s)
		}
	}
	ctxt.defs = append(ctxt.defs, infosyms...)
}

func writeAuxSymDebug(ctxt *Link, par *LSym, aux *LSym) {

  bool found = false;
  for (Node* node : graph->nodes()) {
    if (CanCreateXlaKernel(node->def())) {
      EXPECT_FALSE(found);
      found = true;
      std::vector<int32> hostmem_attr;
      EXPECT_TRUE(TryGetNodeAttr(node->def(), "_input_hostmem", &hostmem_attr));
      EXPECT_EQ(hostmem_attr.size(), 1);
      EXPECT_EQ(hostmem_attr[0], 1);
    }
  }

                   "defs argument",
                   i);
      return nullptr;
    }

    // Parse op def from character array.
    tensorflow::OpDef opdef;
    if (!tensorflow::protobuf::TextFormat::ParseFromString(tf_opdefs, &opdef)) {
      PyErr_Format(
          PyExc_ValueError,
          "Failed to parse opdefs at index %d of custom op defs argument: %s",
          i, tf_opdefs);