4 Porting MLton to a new target platform (architecture or OS) involves
7 1. Make the necessary changes to the scripts, runtime system,
8 <:BasisLibrary: Basis Library> implementation, and compiler.
10 2. Get the regressions working using a cross compiler.
12 3. <:CrossCompiling: Cross compile> MLton and bootstrap on the target.
14 MLton has a native code generator only for AMD64 and X86, so, if you
15 are porting to another architecture, you must use the C code
16 generator. These notes do not cover building a new native code
19 Some of the following steps will not be necessary if MLton already
20 supports the architecture or operating system you are porting to.
23 == What code to change ==
28 * In `bin/platform`, add new cases to define `$HOST_OS` and `$HOST_ARCH`.
34 The goal of this step is to be able to successfully run `make` in the
35 `runtime` directory on the target machine.
37 * In `platform.h`, add a new case to include `platform/<arch>.h` and `platform/<os>.h`.
39 * In `platform/<arch>.h`:
40 ** define `MLton_Platform_Arch_host`.
42 * In `platform/<os>.h`:
43 ** include platform-specific includes.
44 ** define `MLton_Platform_OS_host`.
45 ** define all of the `HAS_*` macros.
47 * In `platform/<os>.c` implement any platform-dependent functions that the runtime needs.
49 * Add rounding mode control to `basis/Real/IEEEReal.c` for the new arch (if not `HAS_FEROUND`)
51 * Compile and install the <:GnuMP:>. This varies from platform to platform. In `platform/<os>.h`, you need to include the appropriate `gmp.h`.
54 * Basis Library implementation (`basis-library/*`)
57 * In `primitive/prim-mlton.sml`:
58 ** Add a new variant to the `MLton.Platform.Arch.t` datatype.
59 ** modify the constants that define `MLton.Platform.Arch.host` to match with `MLton_Platform_Arch_host`, as set in `runtime/platform/<arch>.h`.
60 ** Add a new variant to the `MLton.Platform.OS.t` datatype.
61 ** modify the constants that define `MLton.Platform.OS.host` to match with `MLton_Platform_OS_host`, as set in `runtime/platform/<os>.h`.
63 * In `mlton/platform.{sig,sml}` add a new variant.
65 * In `sml-nj/sml-nj.sml`, modify `getOSKind`.
67 * Look at all the uses of `MLton.Platform` in the Basis Library implementation and see if you need to do anything special. You might use the following command to see where to look.
70 find basis-library -type f | xargs grep 'MLton\.Platform'
73 If in doubt, leave the code alone and wait to see what happens when you run the regression tests.
79 * In `lib/stubs/mlton-stubs/platform.sig` add any new variants, as was done in the Basis Library.
81 * In `lib/stubs/mlton-stubs/mlton.sml` add any new variants in `MLton.Platform`, as was done in the Basis Library.
84 The string used to identify a particular architecture or operating
85 system must be the same (except for possibly case of letters) in the
86 scripts, runtime, Basis Library implementation, and compiler (stubs).
87 In `mlton/main/main.fun`, MLton itself uses the conversions to and
90 MLton.Platform.{Arch,OS}.{from,to}String
93 If the there is a mismatch, you may see the error message
94 `strange arch` or `strange os`.
97 == Running the regressions with a cross compiler ==
99 When porting to a new platform, it is always best to get all (or as
100 many as possible) of the regressions working before moving to a self
101 compile. It is easiest to do this by modifying and rebuilding the
102 compiler on a working machine and then running the regressions with a
103 cross compiler. It is not easy to build a gcc cross compiler, so we
104 recommend generating the C and assembly on a working machine (using
105 MLton's `-target` and `-stop g` flags, copying the generated files to
106 the target machine, then compiling and linking there.
108 1. Remake the compiler on a working machine.
110 2. Use `bin/add-cross` to add support for the new target. In particular, this should create `build/lib/mlton/targets/<target>/` with the platform-specific necessary cross-compilation information.
112 3. Run the regression tests with the cross-compiler. To cross-compile all the tests, do
115 bin/regression -cross <target>
118 This will create all the executables. Then, copy `bin/regression` and
119 the `regression` directory to the target machine, and do
122 bin/regression -run-only <target>
125 This should run all the tests.
127 Repeat this step, interleaved with appropriate compiler modifications,
128 until all the regressions pass.
133 Once you've got all the regressions working, you can build MLton for
134 the new target. As with the regressions, the idea for bootstrapping
135 is to generate the C and assembly on a working machine, copy it to the
136 target machine, and then compile and link there. Here's the sequence
139 1. On a working machine, with the newly rebuilt compiler, in the `mlton` directory, do:
142 mlton -stop g -target <target> mlton.mlb
145 2. Copy to the target machine.
147 3. On the target machine, move the libraries to the right place. That is, in `build/lib/mlton/targets`, do:
154 Also make sure you have all the header files in build/lib/mlton/include. You can copy them from a host machine that has run `make runtime`.
156 4. On the target machine, compile and link MLton. That is, in the mlton directory, do something like:
159 gcc -c -Ibuild/lib/mlton/include -Ibuild/lib/mlton/targets/self/include -O1 -w mlton/mlton.*.[cs]
160 gcc -o build/lib/mlton/mlton-compile \
161 -Lbuild/lib/mlton/targets/self \
164 -lmlton -lgmp -lgdtoa -lm
167 5. At this point, MLton should be working and you can finish the rest of a usual make on the target machine.
170 make basis-no-check script mlbpathmap constants libraries tools
173 6. Making the last tool, mlyacc, will fail, because mlyacc cannot bootstrap its own yacc.grm.* files. On the host machine, run `make -C mlyacc src/yacc.grm.sml`. Then copy both files to the target machine, and compile mlyacc, making sure to supply the path to your newly compile mllex: `make -C mlyacc MLLEX=mllex/mllex`.
175 There are other details to get right, like making sure that the tools
176 directories were clean so that the tools are rebuilt on the new
177 platform, but hopefully this structure works. Once you've got a
178 compiler on the target machine, you should test it by running all the
179 regressions normally (i.e. without the `-cross` flag) and by running a
180 couple rounds of self compiles.
185 The above description is based on the following emails sent to the
188 * http://www.mlton.org/pipermail/mlton/2002-October/013110.html
189 * http://www.mlton.org/pipermail/mlton/2004-July/016029.html