[PATCH] D139140: [ModuleInliner] Add MLPriority

[Kazu Hirata via Phabricator via llvm-commits]

kazu created this revision.
Herald added a subscriber: hiraditya.
Herald added a project: All.
kazu requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

This patch adds MLPriority as the first step toward the ML-based
function inlining with the module inliner.

For now, MLPriority is completely identical to CostPriority.

Once this patch lands, I'm planning to:

- integrate NoInferenceModelRunner,

- memoize the priority computation so that the priority remains the same for given values of metrics even with the noise injected during training, and

- port/take more features into account.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D139140

Files:
  llvm/lib/Analysis/InlineOrder.cpp


Index: llvm/lib/Analysis/InlineOrder.cpp
===================================================================
--- llvm/lib/Analysis/InlineOrder.cpp
+++ llvm/lib/Analysis/InlineOrder.cpp
@@ -22,7 +22,7 @@
 
 #define DEBUG_TYPE "inline-order"
 
-enum class InlinePriorityMode : int { Size, Cost, CostBenefit };
+enum class InlinePriorityMode : int { Size, Cost, CostBenefit, ML };
 
 static cl::opt<InlinePriorityMode> UseInlinePriority(
     "inline-priority-mode", cl::init(InlinePriorityMode::Size), cl::Hidden,
@@ -32,7 +32,9 @@
                clEnumValN(InlinePriorityMode::Cost, "cost",
                           "Use inline cost priority."),
                clEnumValN(InlinePriorityMode::CostBenefit, "cost-benefit",
-                          "Use cost-benefit ratio.")));
+                          "Use cost-benefit ratio."),
+               clEnumValN(InlinePriorityMode::ML, "ml",
+                          "Use ML.")));
 
 static cl::opt<int> ModuleInlinerTopPriorityThreshold(
     "moudle-inliner-top-priority-threshold", cl::Hidden, cl::init(0),
@@ -175,6 +177,26 @@
   Optional<CostBenefitPair> CostBenefit;
 };
 
+class MLPriority {
+public:
+  MLPriority() = default;
+  MLPriority(const CallBase *CB, FunctionAnalysisManager &FAM,
+             const InlineParams &Params) {
+    auto IC = getInlineCostWrapper(const_cast<CallBase &>(*CB), FAM, Params);
+    if (IC.isVariable())
+      Cost = IC.getCost();
+    else
+      Cost = IC.isNever() ? INT_MAX : INT_MIN;
+  }
+
+  static bool isMoreDesirable(const MLPriority &P1, const MLPriority &P2) {
+    return P1.Cost < P2.Cost;
+  }
+
+private:
+  int Cost;
+};
+
 template <typename PriorityT>
 class PriorityInlineOrder : public InlineOrder<std::pair<CallBase *, int>> {
   using T = std::pair<CallBase *, int>;
@@ -274,6 +296,10 @@
     LLVM_DEBUG(
         dbgs() << "    Current used priority: cost-benefit priority ---- \n");
     return std::make_unique<PriorityInlineOrder<CostBenefitPriority>>(FAM, Params);
+  case InlinePriorityMode::ML:
+    LLVM_DEBUG(
+        dbgs() << "    Current used priority: ML priority ---- \n");
+    return std::make_unique<PriorityInlineOrder<MLPriority>>(FAM, Params);
   }
   return nullptr;
 }


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D139140.479401.patch
Type: text/x-patch
Size: 2213 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221201/3c36e699/attachment.bin>