|Course on code generation email@example.com (Yaakov Yaari) (2001-03-26)|
|Re: Course on code generation firstname.lastname@example.org (Mike Spivey) (2001-03-27)|
|Re: Course on code generation email@example.com (2001-03-27)|
|From:||firstname.lastname@example.org (Ian L. Kaplan)|
|Date:||27 Mar 2001 23:46:11 -0500|
|Posted-Date:||27 Mar 2001 23:46:10 EST|
Yaakov Yaari <email@example.com> wrote:
>I am planning a course on compiler code generation. The idea is to use GCC
>as the basic infrastructure for projects.
>Does anyone have some suggestions for the design of such course? Any
>experience with a similar course? My main concern is the complexity of GCC's
I've found that there is a significant learning curve for GCC.
There is a lot of code that is only executed in specific cases of
instruction set families. The structure of the code is also pretty
grungy. Although GCC was advanced for the time it was written, I
really don't think that anyone would design a backend like this
today. If I were teaching such a course I'd use lcc which is much
easier to understand and modify. Another possibility is Zypher
(http://www.cs.virginia.edu/zephyr/index.html), but this looks like
ti has a big learning curve too. On the other hand Zypher is much
more in the mainstream of compiler research that is GCC (which I
think people only use for pragmatic reasons).
Return to the
Search the comp.compilers archives again.