问题
I was reading the following presentation: http://wingolog.org/pub/qc-2012-js-slides.pdf which talks about (4,10,19) inline ASM generation as a technique used in Javascript optimisation.
In the following paper: https://sites.google.com/site/juliangamble/Home/Compilers%20Tutorial%202006-09-16.pdf?attredirects=0&d=1 at page 30 and 31 they talk about using scheme to generate ASM that is subsequently linked and executed in a subsequent OS process.
What about the scenario where you want to generate the ASM and execute it inside your existing process? (ie no subsequent link and execute in separate steps).
I assume this comes down to (in C for example) generating the ASM, writing the bytes to an area of memory (code as data) and adding a function header and return (compatible with a C caller (similar to what we see on page 3 of the Goloum paper above). Then we take that data pointer and convert it to a function pointer and call it (code as code - from code as data).
My questions are:
(a) What is the terminology for this inline code generation and execution?
(b) Is there a simple 'hello world' C example of this available?
回答1:
There is quite a good example here. Again they use the terminology of self-modifying code.
Another one here.
Here they give it four categories (use cases):
- Metamorphism
- Trampolining
- JIT compilation
- Security implications (insecure coding/malware)
There is a discussion here about whether LISP is truly self-modifying - and they end up concluding that it is not.
来源：https://stackoverflow.com/questions/12168575/executing-generated-assembler-inline