Code Monkey home page Code Monkey logo

Comments (3)

samarion avatar samarion commented on June 23, 2024

I mentioned this at some point (maybe to @gsps ?): we should be able to move the mergeCalls logic into the template generator which would make it much cleaner.

The idea is to change the IfExpr case in TemplateGenerator.mkExprClauses. We would change the following lines to call mkExprClauses recursively and then unify function calls which occur in both branches:

val trec = rec(newBool1, thenn, pol)
val erec = rec(newBool2, elze, pol)

would become

val (trec, tClauses) = mkExprClauses(newBool1, thenn, localSubst, pol)
val (erec, eClauses) = mkExprClauses(newBool2, elze, localSubst, pol)
builder ++= mergeCalls(pathVar, localSubst, tClauses, eClauses)

The new mergeCalls logic (inside TemplateGenerator.scala) would look something like this:

def mergeCalls(pathVar: Variable, condVar: Variable, substMap: Map[Variable, Encoded],
               thenClauses: TemplateClauses, elseClauses: TemplateClauses): TemplateClauses = {
  val builder = new Builder(pathVar, substMap)
  builder ++= thenClauses
  builder ++ elseClauses

  // Clear all guardedExprs in builder since we're going to transform them by merging calls.
  // The transformed guardedExprs will be added to builder at the end of the function.
  builder.guardedExprs = Map.empty

  def collectCalls(expr: Expr): Set[FunctionInvocation] =
    exprOps.collect { case fi: FunctionInvocation => Set(fi) case _ => Set.empty[FunctionInvocation] }(expr)
  def countCalls(expr: Expr): Int =
    exprOps.count { case fi: FunctionInvocation => 1 case _ => 0}(expr)
  def replaceCall(call: FunctionInvocation, newExpr: Expr)(e: Expr): Expr =
    exprOps.replace(Map(call, newExpr), e)

  def getCalls(guardedExprs: Map[Variable, Seq[Expr]]): Map[TypedFunDef, Seq[(FunctionInvocation, Set[Variable])]] =
    (for {(b, es) <- guardedExprs; e <- es; fi <- collectCalls(e)) yield (b -> fi))
    .groupBy(_._2)
    .mapValues(_.map(_._1).toSet)
    .toSeq
    .groupBy(_._1.tfd)
    .mapValues(_.toList.distinct.sortBy(p => countCalls(p._1)))  // place inner calls first
    .toMap

  var thenGuarded = thenClauses._4
  var elseGuarded = elseClauses._4

  val thenCalls = getCalls(thenGuarded)
  val elseCalls = getCalls(elseGuarded)

  // We sort common function calls in order to merge nested calls first.
  var toMerge = (thenCalls.keys() & elseCalls.keys())
    .flatMap(tfd => thenCalls(tfd) zip elseCalls(tfd))
    .toSeq
    .sortBy(p => countCalls(p._1._1) + countCalls(p._2._1))

  while (toMerge.nonEmpty) {
    val ((thenCall, thenBlockers), (elseCall, elseBlockers)) = toMerge.head
    toMerge = toMerge.tail

    val newExpr: Variable = Variable.fresh("call", thenCall.tfd.getType, true)
    builder.storeExpr(newExpr)

    val replaceThen = replaceCall(thenCall, newExpr) _
    val replaceElse = replaceCall(elseCall, newExpr) _

    thenGuarded = thenGuarded.mapValues(_.map(replaceThen)
    elseGuarded = elseGuarded.mapValues(_.map(replaceElse)
    toMerge = toMerge.map(p => (replaceThen(p._1), replaceElse(p._2))

    val newBlocker: Variable = Variable.fresh("bm", BooleanType(), true)
    // TODO: make this a method in Builder similar to storeCond
    builder.storeConds(thenBlockers ++ elseBlockers, newBlocker)
    builder.iff(orJoin((thenBlockers ++ elseBlockers).toSeq), newBlocker)

    val newArgs = (thenCall.args zip elseCall.args).map { case (thenArg, elseArg) =>
      val (newArg, argClauses) = mkExprClauses(newBlocker, IfExpr(condVar, thenArg, elseArg), builder.localSubst)
      builder ++= argClauses
      newArg
    }

    val newCall = thenCall.tfd.applied(newArgs)
    builder.storeGuarded(newBlocker, Equals(newExpr, newCall))
  }

  for ((b, es) <- thenGuarded; e <- es) builder.storeGuarded(b, e)
  for ((b, es) <- elseGuarded; e <- es) builder.storeGuarded(b, e)
  builder.result
}

from inox.

jad-hamza avatar jad-hamza commented on June 23, 2024

Should liftCalls remain?

from inox.

samarion avatar samarion commented on June 23, 2024

No, I think it shouldn't be necessary.

from inox.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.