Difference between revisions of "Work Package 2"
From Superoptimization
(→Work package summary: Added start and end dates) |
(→Work package summary: Added status) |
||
| (One intermediate revision by the same user not shown) | |||
| Line 7: | Line 7: | ||
| ''End date:'' || 25 July 2014 | | ''End date:'' || 25 July 2014 | ||
|} | |} | ||
| + | |||
| + | Revised, since [[Work Package 1|work package 1]] has established that peephold optimization is not immediately relevant and has also indicated prototyping of key technologies is feasible. | ||
=== Description === | === Description === | ||
| − | + | Implement prototypes for the superoptimization techniques identified in [[Work Package 1|work package 1]]. | |
=== Deliverables === | === Deliverables === | ||
| − | + | Prototype implementations of superoptimization techniques. | |
=== Dependencies === | === Dependencies === | ||
Cannot finish before [[Work Package 1|work package 1]] is finished. | Cannot finish before [[Work Package 1|work package 1]] is finished. | ||
| + | |||
| + | === Status === | ||
| + | |||
| + | Complete. | ||
== Approach == | == Approach == | ||
Latest revision as of 10:09, 8 August 2014
Contents
Work package summary
| Start date: | 30 June 2014 |
| End date: | 25 July 2014 |
Revised, since work package 1 has established that peephold optimization is not immediately relevant and has also indicated prototyping of key technologies is feasible.
Description
Implement prototypes for the superoptimization techniques identified in work package 1.
Deliverables
Prototype implementations of superoptimization techniques.
Dependencies
Cannot finish before work package 1 is finished.
Status
Complete.
Approach
- Design a peephole optimization to integrate a superoptimizer into GCC and/or LLVM.
Notes
A peephole optimizer is no longer relevant - not enough potential for gains.
Optimizing simultaneous sequences.
- Peephole superoptimizer study could bruteforce multiple sequences simultenously, by hashing the test vectors
- Unclear to see how this can be done for SMT based methods
- Naively attempting to find multiple solutions at the same time will limit the 'branch pruning' of the solver – it will be unable to rule out large areas of the search space due to multiple divergant solutions
- Could be the thing that limits the effectiveness of the SMT method
- Possible solutions
- Attempt to find multiple solutions to 'similar' sequences. Based on the assumption that the solutions will be similar. How to define a similarity metric?
Deliverables
The above notes, and notes on the Meetings page are the output of WP2, to be formulated into the project report (WP4)
