Re: Requirements for Just-in-time Compilation

On 29 Apr 2001 02:11:17 -0400, jason petrone <jp@demonseed.net>
wrote:


>Thanks to all for your responses. They were very helpful.
>
>Ralph Corderoy <ralph@inputplus.demon.co.uk> wrote:
>> Yes, an external assembler pass would be too slow. Is anyone using the
>> New Jersey Machine-Code Toolkit as part of a JIT or dynarec?
>> http://www.eecs.harvard.edu/~nr/toolkit/
>
>I really knew it would be too slow, I just wanted to make sure, since I don't
>really want to write an assembler. I am aware of the NJ-MC-TK, but I'm
>thinking about trying GNU Lightning first(www.gnu.org/software/lightning).
>
>>> It seems to me that supporting multiple architectures would require making
>>> extra passes, and would slow things down.
>> Why extra passes? A different backend would be used per architecture,
>> no?
>
>I'm thinking extra passes since I would need to have an architecture
>independent intermediate representation, like RTL. I couldn't use something
>like the single pass compiler that is the first example in the dragon book.


Your symbol table and syntax tree gives you an intermediate.
You could generate code at the statement or function level.
No need to build the whole tree, convert to intermediate, then
generate code.
Your JIT compiler would not then be IT IMHO.


Thanks. Take care, Brian Inglis Calgary, Alberta, Canada
--
(Brian dot Inglis at SystematicSw dot ab dot ca)

Return to the comp.compilers page.
Search the comp.compilers archives again.