Week |
Activities |
Status |
1 |
(a) find the Itanium research compiler and documentation (b) formulate scheduling on the Itanium as a particular kind of job-shop problem |
done (Oct 31) 50%: it doesn't quite fit into standard frameworks, so we need to formulate it on our own. |
2 |
(c) find a PTAS or formulate a good anytime search
algorithm for that problem (d) design an interface for describing the job-shop problem (one that deals only in terms of jobs, machines, and prerequisites) |
20%: no PTAS exists; we must find our own algorithm. 50%: first draft done. |
3 |
(e) write code in the compiler architecture to transfer
from their IR to a job-shop problem (f) write code to implement the job-shop algorithm, conforming to the interface we designed |
|
4 |
(g) write a method to read a file of function names
and epsilon values so the compiler will use those values to determine how
much to optimize (h) run the compiler on some toy test cases (i) implement a set of scripts to quickly run and check test cases |
|
5 |
(j) implement gathering profile information (k) implement automatically using profile information to drive the choice of epsilon |
|
6 |
(l) run experiments: SPEC or some other relatively
standard benchmark. (m) Compare with the Itanium compiler's default instruction scheduler |