Lower inlining threshold when the caller has minsize attribute

[eraman]

Summary

When the caller has optsize attribute, we reduce the inlinining threshold to OptSizeThreshold (=75) if it is not already lower than that. We don't do the same for minsize and I suspect it was not intentional. This also addresses a FIXME regarding checking optsize attribute explicitly instead of using the right wrapper.

[echristo]

LGTM.

[eraman]

Hal, do you have any comments?

[hfinkel]

In D16493#338808, @eraman wrote:

Hal, do you have any comments?

LGTM too.

llvm/trunk/lib/Analysis/InlineCost.cpp

if (isAllocaDerivedArg(V))
     // alloca-derived value and null. Note that this fires regardless of
     // SROA firing.
     return true;
   
   return false;
 }
 
 void CallAnalyzer::updateThreshold(CallSite CS, Function &Callee) {
-  // If -inline-threshold is not given, listen to the optsize attribute when it
-  // would decrease the threshold.
+  // If -inline-threshold is not given, listen to the optsize and minsize
+  // attributes when they would decrease the threshold.
   Function *Caller = CS.getCaller();
 
-  // FIXME: Use Function::optForSize()
-  bool OptSize = Caller->hasFnAttribute(Attribute::OptimizeForSize);
-
-  if (!(DefaultInlineThreshold.getNumOccurrences() > 0) && OptSize &&
-      OptSizeThreshold < Threshold)
+  if (!(DefaultInlineThreshold.getNumOccurrences() > 0)) {
+    if (Caller->optForMinSize() && OptMinSizeThreshold < Threshold)
+      Threshold = OptMinSizeThreshold;
+    else if (Caller->optForSize() && OptSizeThreshold < Threshold)
       Threshold = OptSizeThreshold;
+  }
 
   // If profile information is available, use that to adjust threshold of hot
   // and cold functions.
   // FIXME: The heuristic used below for determining hotness and coldness are
   // based on preliminary SPEC tuning and may not be optimal. Replace this with
   // a well-tuned heuristic based on *callsite* hotness and not callee hotness.
   uint64_t FunctionCount = 0, MaxFunctionCount = 0;
   bool HasPGOCounts = false;

llvm/trunk/test/Transforms/Inline/inline-optsize.ll

 ; RUN: opt -S -Oz < %s | FileCheck %s -check-prefix=OZ
 ; RUN: opt -S -O2 < %s | FileCheck %s -check-prefix=O2
+; RUN: opt -S -Os < %s | FileCheck %s -check-prefix=OS
 
 ; The inline threshold for a function with the optsize attribute is currently
 ; the same as the global inline threshold for -Os. Check that the optsize
 ; function attribute doesn't alter the function-specific inline threshold if the
 ; global inline threshold is lower (as for -Oz).
 
 @a = global i32 4
 
 ; This function should be larger than the inline threshold for -Oz (25), but
 ; smaller than the inline threshold for optsize (75).
 define i32 @inner() {
   %a1 = load volatile i32, i32* @a
   %x1 = add i32 %a1,  %a1
   %a2 = load volatile i32, i32* @a
   %x2 = add i32 %x1, %a2
   %a3 = load volatile i32, i32* @a
   %x3 = add i32 %x2, %a3
   %a4 = load volatile i32, i32* @a
   %x4 = add i32 %x3, %a4
   %a5 = load volatile i32, i32* @a
   %x5 = add i32 %x3, %a5
   ret i32 %x5
 }
 
-; @inner() should be inlined for -O2 but not for -Oz.
+; @inner() should be inlined for -O2 and -Os but not for -Oz.
 ; OZ: call
 ; O2-NOT: call
+; OS-NOT: call
 define i32 @outer() optsize {
    %r = call i32 @inner()
    ret i32 %r
 }
+
+; @inner() should not be inlined for -O2, -Os and -Oz.
+; OZ: call
+; O2: call
+; OS: call
+define i32 @outer2() minsize {
+   %r = call i32 @inner()
+   ret i32 %r
+}