diff a/hat/docs/Build/hat.md b/hat/docs/Build/hat.md --- /dev/null +++ b/hat/docs/Build/hat.md @@ -0,0 +1,182 @@ +# Building HAT +[Back to Index ../](../index.md) + +We use `maven` and `cmake` to build HAT. + +Maven controls the build but delegates to cmake for native artifacts (such as ffi-backends), we also use cmake to provide jextract +with location information needed to reference various include/lib paths. (See extractions/pom.xml) + +Use `sudo apt` on Linux or `brew install`. + +```bash +sudo apt install cmake maven +``` + +```bash +brew install cmake +brew install maven +``` + +## Setting environment variables JAVA_HOME and PATH +To build HAT we need to ensure that `JAVA_HOME` points to our babylon jdk (the one we built [here](babylon.md)) + +We also need to ensure that `${JAVA_HOME}/bin` is in our PATH (before any other JAVA location). + +We also need jextract for some maven targets. + +We can download and install jextract from [here](https://jdk.java.net/jextract/) + +The `env.bash` shell script can be sourced (dot included) in your shell to set JAVA_HOME and PATH +```bash +cd hat +. ./env.bash +``` +This should detect the arch type (AARCH64 or X86_46) and select the correct relative parent dir and inject that dir in your PATH if you are working with +HAT in the original subdir of the babylon project. + +It will also check if jextract is in your PATH. If it is not, you will need to add it (if you decide to use any of the jextract artifacts). +```bash +cd hat +. ./env.bash +export PATH=${PATH}:/path/to/my/jextract/bin +echo ${JAVA_HOME} +/Users/ME/github/babylon/hat/../build/macosx-aarch64-server-release/jdk +echo ${PATH} +/Users/ME/github/babylon/hat/../build/macosx-aarch64-server-release/jdk/bin:/usr/local/bin:...... +``` +## Building +Now we should be able to use maven to build, if successful maven will place all jars and libs in a newly created `build` dir in your top level hat dir. +```bash +cd hat +. ./env.bash +mvn clean package +ls build +hat-core-1.0.jar hat-example-heal-1.0.jar +hat-backend-ffi-cuda-1.0.jar hat-example-mandel-1.0.jar +hat-backend-ffi-mock-1.0.jar hat-example-squares-1.0.jar +hat-backend-ffi-opencl-1.0.jar hat-example-view-1.0.jar +hat-example-violajones-1.0.jar hat-example-experiments-1.0.jar +libopencl_backend.dylib +``` + +## Running an example +To run we use some common java `opt` files. + +When executing with java we canplace common java params in files (say opts) then use `java @opts` to avoid typing all the opts each time. + +To declutter the hat dir, we prefix some useful `opt` files with `.` + +For example in the .ffi-opencl-example opt file we have +```bash +cat .ffi-opencl-example +--enable-preview --add-modules=jdk.incubator.code --enable-native-access=ALL-UNNAMED -Djava.library.path=build --class-path build/hat-ffi-opencl-examples-1.0.jar +``` + +So you can run an example (say nbody) using the opencl backend using +```bash +java @.ffi-opencl-example nbody.Main +``` + +Similarly we can run HAT's test suite. + +```bash +java @.ffi-opencl-test-suite +.... +``` + +To list available `@.` files +```bash +ls .* +``` + +---- +### HAT runtime environment variable. + +During ffi-backend development we added some useful flags to pass to native code to allow us to trace calls, inject logging, control options. + +These should be considered just development flags. + +At runtime the ffi-backends all communicate from the java side to the native side via a 32 bit 'config' int. + +The Java side class [hat.backend.ffi.Config](https://github.com/openjdk/babylon/blob/code-reflection/hat/backends/ffi/shared/src/main/java/hat/backend/ffi/Config.java) + +Is initialized from an environment variable `HAT` at runtime + +So for example when we launch +```bash +java @hat/run ffi-opencl heal +``` + +We can pass config info to `HAT` via either `-DHAT=xxx` or via the `HAT` ENV variable. + +So for example to get HAT to dump the text form of the kernel code models. + +``` +HAT=INFO,SHOW_KERNEL_MODEL java @hat/run ffi-opencl heal +``` + +Or to show generated opencl or cuda code +``` +HAT=INFO,SHOW_CODE java @hat/run ffi-opencl heal +``` + +This is particularly useful for selecting PLATFORM + DEVICE if you have multiple GPU devices. + +Here we select DEVICE 0 on PLATFORM 0 (actually the default) +``` +HAT=INFO,PLATFORM:0,DEVICE:0,SHOW_CODE ... +``` +or for DEVICE 1 on PLATFORM 1 +``` +HAT=INFO,PLATFORM:1,DEVICE:1,SHOW_CODE ... +``` + +Note the platform and device id's are 4 bits. This probably works for development, but we will need a more expressive way of capturing this via the accelerator selection. + +This just allows us to test without code changes. + +To keep he java code and the native code in sync. The `main` method at the end of + [hat.backend.ffi.Config](https://github.com/openjdk/babylon/blob/code-reflection/hat/backends/ffi/shared/src/main/java/hat/backend/ffi/Config.java) +is actually used to create the C99 header for Cuda and OpenCL ffi-backends. + +So whenever we change the Java config class we should run the main method to generate the header. This is not really robust, (but proved way better than trying to remember for all backends) but you need to know, if you add move config bits to the Java side. + +The Main method uses a variant of the code builders used for C99 style code (CUDA/OpenCL) to generate the config.h header. + +This is how we keep the ffi based backends in sync + +Some more useful `config bits` + +``` +HAT=PTX java @hat/run ffi-cuda .... +``` +Sends PTX generated by our prototype PTX generator to the backend (for CUDA) rather than C99 code. + +``` +HAT=INFO java @hat/run .... +``` + +Will dump all of the `config bits` from the native side. + +At present this yields +``` +native minimizeCopies 0 +native trace 0 +native profile 0 +native showCode 0 +native showKernelModel 0 +native showComputeModel 0 +native info 1 +native traceCopies 0 +native traceSkippedCopies 0 +native traceEnqueues 0 +native traceCalls 0 +native showWhy 0 +native showState 0 +native ptx 0 +native interpret 0 +``` +Generally the flags all represent single bits (except PLATFORM:n and DEVICE:n) and are set using comma separated uppercase+underscore string forms. + + +So to experiment with `minimizeCopies` and to `showCode` we set `HAT=MINIMIZE_COPIES,SHOW_CODE`