- ABPS currently uses two kind of benchmarks for simulation purposes:
Stanford benchmarks (very helpful for didactic purposes - students);
SPEC 2000 benchmarks (more accurate, useful for a fine grained branch prediction study)
Detecting difficult to predict branches can be easily found with ABPS, in a highly configurable manner;
ABPS includes several detection schemes, based on:
-Local history (HrL);
-Global history (HrG);
-HrL + HrG;
-HrG + Path;
-HrL + HrG + Path.
Or, more exactly: a collection of predictors, fully configurable;
ABPS integrates two level predictors (GAg, GShare, PAg, PAp) and (state of the art) neural predictors (Simple perceptron, Fast-Path based percpetron)
The new version abps v1.0 beta is available for download. New features added: CBP2 successfully integrated; possibility to integrate your own benchmarks by editing a simple XML file; cascaded detection; results can be saved for futher reuse
We succeded to integrate Championship Branch Prediction (CBP2) benchmarks. The new version of ABPS is available for download.
The abpsv0.4 version is available for download. New cool things added. Check it out.
Project accepted at the 6 th EUROSIM Congress on Modelling and Simulation, september 9-13, 2007, Ljubljana, Slovenia.