[LLVMdev] getting closer!

Terence Parr parrt at cs.usfca.edu
Tue Apr 22 17:55:39 PDT 2008

On Apr 22, 2008, at 5:43 PM, Gordon Henriksen wrote:
> This is not possible for instructions (which must be in SSA form). You
> can only use it for constants.
> http://llvm.org/docs/LangRef.html#constantexprs

ah!  ok, but really helpful for structs init. cool.

>> I'm always using the pure text input headline generating everything
>> from Java...
> Normally I'd say that's masochism, but you may have a handle on text
> templates. :)

Yep, generating text is my "speciality".  http:// 
www.stringtemplate.org :)  ANTLR+StringTemplates makes it pretty darn  
easy to generate the IR using any ANTLR target (C, C#, Python,  
ActionScript, Java, soon others).

Much better than having to code in C/C++ in my view [ducking]. ;)

>>> For example, something like this:
>> [snip]
>> Oooooooh! man!  that is exactly what I was looking for.  hooray!
>> Input/Output pairs are the best way to learn this stuff as far as I
>> can tell.
>> I believe I have enough from this to construct a collector using the
>> "shadow-stack" mechanism.  thanks so much!
> Excellent!

Yeah, probably too advanced for the intro DSL book I'm doing; i'll try  
to get a small writeup together sometime for the doc at your site to  
help others...


More information about the llvm-dev mailing list