[erlang-questions] HiPE compiler
Sun Jan 29 21:35:23 CET 2012
On 01/29/12 19:01, Matthew Evans wrote:
> Hi group,
> My understanding is that HiPE takes Erlang byte code beam files as the
> input to produce modified combined HiPE + byte code beam files.
> I'm thinking of using HiPE on a project. Unfortunately the build system
> is mixed, and I'd rather not go down the cross-compiler route. My
> question is - is there a way to create "HiPE targets" just from beam
> files on the target platform? The source will, obviously, not be
> available on the target platform.
I am not sure I understand your question well, but I'll give it a shot
in answering it anyway.
Unlike BEAM byte code, the native code that the HiPE compiler generates
has a very strong dependence on the specifics of the Erlang runtime
system which is running the code. This means that you cannot native
compile the code using Erlang/OTP R-X and expect it to run on R-Y where
X and Y are different. This holds even for minor revisions of the same
major release (i.e. code compiled for R14B03 cannot run on R14B03.)
Also obviously there is a strong dependency on the target architecture:
i.e. you cannot compile code for an x86 and expect it to run on an ARM.
But you can compile your code natively and run it on another target if
the machine that you compile in and the target machine are running
exactly the same Erlang/OTP release and have the same underlying
architecture. In this case you can use:
erlc +native FILE.erl -o FILE.beam
and transfer these FILE.beam(s) on the target machine. I am pretty sure
But what you can also do the following: compile the files to BEAM byte
code and transfer these files to the target machine and then native
compile these files on the fly (as part of the start of your
application) by using something like the following:
lists:foreach(fun (F) -> hipe:c(F) end, BEAM_FILES)
This will compile the BEAM files in memory, generate native code for
them and load them in the running system. (Obviously, you can choose to
compile only a subset of your files, perhaps only those that are time
critical.) There is obviously a small start up cost in doing this, but
in most applications this should not be an issue.
Let me know how it goes. We can continue this perhaps offline.
More information about the erlang-questions