Consider the process of compilation.
Once your source files have been compiled, there should be no need to recompile them unless something has changed that affects the output, such as the modification of a source file or the removal of an output file.
And compilation can take a significant amount of time, so skipping the step when it’s not needed saves a lot of time.

* Support for several source languages.
* Support for building different variants of the same software, for different architectures, operating systems, or for any purpose.
* Incremental parallel compilation, precompiled headers.
* Dependency management between native software components.
* Unit test execution.
* Generate Visual studio solution and project files.

      True, ('Would you like to override eigen strong inline for some C++ '
             'compilation to reduce the compilation time?'),
      'Eigen strong inline overridden.', 'Not overriding eigen strong inline, '
      'some compilations could take more than 20 mins.'):
    # Due to a known MSVC compiler issue
    # https://github.com/tensorflow/tensorflow/issues/10521

Similarly, tasks from the Groovy and Scala plugins also rely on toolchains to determine on which JVM they are executed.

==== Scala compilation target

With the toolchain changes described above, Scala compilation tasks are now always provided with a `target` or `release` parameter.
The exact parameter and value depend on toolchain usage, or not, and Scala version.

Note that there is a known issue with Kotlin compilation avoidance that can cause `OutOfMemory` exceptions in `compileKotlin` tasks if the compilation classpath contains very large JAR files.
This applies to builds applying the Kotlin plugin v1.8.20 or the `kotlin-dsl` plugin.

You can work around it by disabling Kotlin compilation avoidance in your `gradle.properties` file:

[source,properties]
----

The main reason for this is the slower script compilation for Kotlin DSL.
* In IntelliJ IDEA, you must link:https://www.jetbrains.com/help/idea/gradle.html#gradle_import[import your project from the Gradle model] in order to get content assist and refactoring support for your Kotlin DSL build scripts.
* Kotlin DSL script compilation avoidance has known issues.

# not on sm_90 GPUs. compute_80 kernels though can also run on sm_90 GPUs.
build:cuda_clang --repo_env=TF_CUDA_COMPUTE_CAPABILITIES="sm_60,sm_70,sm_80,sm_89,compute_90"

# Set up compilation CUDA version and paths and use the CUDA Clang toolchain.
build:cuda_clang_official --config=cuda_clang
build:cuda_clang_official --action_env=TF_CUDA_VERSION="12"
build:cuda_clang_official --action_env=TF_CUDNN_VERSION="8"

      if (compile_op == nullptr) {
        compile_op = op;
      } else {
        ops_to_erase.insert(op);
      }
    }
    // If there's no outside compilation, we can exit early because this isn't
    // a TPU function.
    if (compile_op == nullptr) {
      return LogicalResult::success();
    }

    TF::TPUReplicateMetadataOp metadata_op = nullptr;

	    fuzz tests matched by -fuzz.

	-cover
	    Enable coverage analysis.
	    Note that because coverage works by annotating the source
	    code before compilation, compilation and test failures with
	    coverage enabled may report line numbers that don't correspond
	    to the original sources.

	-covermode set,count,atomic
	    Set the mode for coverage analysis for the package[s]

                    list.add(f.getPath());
                }
            }
        }
        return list;
    }

    /**
     * Returns the elements placed on the classpath for compilation.
     * This method can be invoked when the caller does not support module-path.
     *
     * @throws DependencyResolutionRequiredException if an artifact file is used, but has not been resolved
     *