diff --git a/hpvm/docs/developerdocs/components/transformdse.rst b/hpvm/docs/developerdocs/components/transformdse.rst index 8739ee8eadb9385f24eb6cddb5a478647787757d..02ad687d28d0c89662e96282c37f5adfdcf653ef 100644 --- a/hpvm/docs/developerdocs/components/transformdse.rst +++ b/hpvm/docs/developerdocs/components/transformdse.rst @@ -122,7 +122,8 @@ Design Space Exploration Our optimization framework sets up the parameter space for DSE by performing a static analysis of the HPVM IR and extracting the corresponding parameters for every possible optimization. Then, it -interacts with HyperMapper (HM), where in every iteration HM sends a +interacts with `HyperMapper (HM) <https://github.com/luinardi/hypermapper>`, +where in every iteration HM sends a chosen design sample (i.e. set of values for the parameters) and waits for a response. For each sample, we apply the optimizations with the provided parameter values, and then estimate the execution time. diff --git a/hpvm/docs/how_to_guides/using-dse.rst b/hpvm/docs/how_to_guides/using-dse.rst index c9a680c617956166b60c2d63d8b4e6eaa1775d21..1fc46f90aa66f2108d0a69eb5e464d85ab188e3f 100644 --- a/hpvm/docs/how_to_guides/using-dse.rst +++ b/hpvm/docs/how_to_guides/using-dse.rst @@ -3,7 +3,8 @@ Using ``hpvm-dse`` ``hpvm-dse`` provides a DSE workflow allowing the optimization of HPVM programs, supporting any language that can target the IR (currently HPVM-C and HeteroC++). -As an autotuning engine, the it uses HyperMapper, a bayesian optimization tool. +The framework uses `HyperMapper (HM) <https://github.com/luinardi/hypermapper>` +to explore the design space. Tool Flow ---------