Re: [rvm-research] Inserting Assembly into Jikes

For an example of adding a Magic instruction, do

hg diff -r8829:8830

which shows the addition of the Intel PAUSE instruction, accessed by
Magic.pause().

regards,
Robin

On 25/10/13 06:49, Claire Murphy wrote:
> Thank you for your fast response; I will try the syscall route,
> because I should be able to do what I want by inserting C.
>
> I have one question about your points though, in case another "magic"
> ends up being what I want, (and also just so I understand things):
> If Jikes RVM offers no "assembly inserts", what are the "asm.emit"
> functions used
> in /jikesrvm/rvm/src/org/jikesrvm/compilers/baseline/ppc/BaselineCompilerImpl.java
> doing? Am I correct to think that I would make new emit functions in
> the Assembler and then modify the compilers as well to populate the
> "magic" functions by calling those Assembler functions?
>
> Thank you very much for your time,
> Claire
>
>
> On Thu, Oct 24, 2013 at 8:54 AM, Eliot Moss <[email protected]
> <mailto:[email protected]>> wrote:
>
>     On 10/24/2013 7:46 AM, David P Grove wrote:
>
>         Claire Murphy <[email protected]
>         <mailto:[email protected]>> wrote on 10/23/2013
>         09:35:25 PM:
>
>
>         If a fast path from Java to C is what you are looking for,
>         then the quickest way to get it working
>         is using VM.sysCall.  (Injects a call to a C function that you
>         write in the bootimage).
>
>         See org.jikesrvm.runtime.SysCall.java and follow the pattern
>         that is there for the other sysCalls.
>
>
>     Dave will know more about this, but here are some things I can
>     tell you:
>
>     1) Jikes RVM offers no "assembly inserts".
>     2) It might be possible to extend the set of "magics" to do what
>     you want.
>        Note that a magic should (in principle) produce some kind of
>     working, or
>        at least not totally broken, code *for each platform*.  (Jikes
>     RVM is not
>        just for x86.)  I suppose in a research project you could cut
>     that corner.
>     3) Adding a magic involves changing TWO compilers, BaseLine and
>     Opt.  BaseLine
>        is fairly straightforward (once you modify the assembler (see
>     below)); Opt
>        perhaps less so, but Dave would know that better.  I've not
>     dealt with adding
>        magics to Opt.
>     4) You need to extend the assembler(s), disassembler(s), etc.
>
>     You also need to understand the properties of the instructions --
>     do they affect
>     non-interruptibility of code sequences?  Do they involve pointers
>     or pointer
>     accesses, in which case you might need to understand interaction
>     with GC?  Etc.
>
>     A managed language environment is rather more complex that C/C++
>     in these respects!
>
>     Regards -- Eliot Moss
>
>
>
>
> ------------------------------------------------------------------------------
> October Webinars: Code for Performance
> Free Intel webinars can help you accelerate application performance.
> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from 
> the latest Intel processors and coprocessors. See abstracts and register >
> http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk
>
>
> _______________________________________________
> Jikesrvm-researchers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers


------------------------------------------------------------------------------
Android is increasing in popularity, but the open development platform that
developers love is also attractive to malware creators. Download this white
paper to learn more about secure code signing practices that can help keep
Android apps secure.
http://pubads.g.doubleclick.net/gampad/clk?id=65839951&iu=/4140/ostg.clktrk