[Polly] Exclude Ignored Intrinsics from explicit instruction list

[nandini12396]

This comment was removed by nandini12396.

[nandini12396]

This comment was removed by nandini12396.

[nandini12396]

Ok Sir. I got it finally! This is wrong. I will update this diff. Thank you.

[Meinersbur]

test/ScopInfo/intrinsics_misc.ll

2

Remove -S. It is ignored when you also specify -analyze.

The result of this command is:

Printing analysis 'Polly - Create polyhedral description of Scops' for region: 'for.cond1 => for.end' in function '_Z2jdv':
Invalid Scop!
Printing analysis 'Polly - Create polyhedral description of Scops' for region: 'for.cond => for.end12' in function '_Z2jdv':
Invalid Scop!
Printing analysis 'Polly - Create polyhedral description of Scops' for region: 'entry => <Function Return>' in function '_Z2jdv':
Invalid Scop!

that is, there is no valid scop detection in this function.

Test cases should fail without the change you are intending to make.

[nandini12396]

yes, sorry forgot to remove that.

[Meinersbur]

The test case still does not fail when I undo the change in ScopBuilder.cpp

Instead of CHECK-NOT, try a positive match (CHECK-NEXT) of all the instruction list that are supposed to be in there.

[Meinersbur]

You test case is relatively large for such a simple check. Can you find a smaller one? The one in D33163 was a lot smaller. Only a loop and a store in its body is required for the smallest scop. I prefer manually crafted IR over clang-generated IR. Avoid unnamed values (e.g. %0) which can cause a problems with finding unique names in LLVM. The pass -instnamer gives ever value a name.

[nandini12396]

Addressed Michael Sir's comments.

lib/Analysis/ScopBuilder.cpp

void ScopBuilder::buildStmts(Region &SR) {
 
   for (auto I = SR.element_begin(), E = SR.element_end(); I != E; ++I)
     if (I->isSubRegion())
       buildStmts(*I->getNodeAs<Region>());
     else {
       std::vector<Instruction *> Instructions;
       for (Instruction &Inst : *I->getNodeAs<BasicBlock>()) {
         Loop *L = LI.getLoopFor(Inst.getParent());
-        if (!isa<TerminatorInst>(&Inst) && !canSynthesize(&Inst, *scop, &SE, L))
+        if (!isa<TerminatorInst>(&Inst) && !isIgnoredIntrinsic(&Inst) &&
+            !canSynthesize(&Inst, *scop, &SE, L))
           Instructions.push_back(&Inst);
       }
       Loop *SurroundingLoop = LI.getLoopFor(I->getNodeAs<BasicBlock>());
       scop->addScopStmt(I->getNodeAs<BasicBlock>(), SurroundingLoop,
                         Instructions);
     }
 }
 

test/ScopInfo/intrinsics_misc.ll

+; RUN: opt %loadPolly -polly-scops -analyze -polly-print-instructions < %s | FileCheck %s
+;
+; Verify that we remove the ignored intrinsics from the instruction list.
+;
+; CHECK-NOT: call void @llvm.donothing
+;
+;    int A[1024], B[1024];
+;    void jd() {
+;      for (int i = 0; i < 1024; i++) {
+;        for (int j = i; j < 1024; j++)
+;          B[i] += B[j];
+;        A[i] = B[i];
+;      }
+;    }
+;
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+@A = global [1024 x i32] zeroinitializer, align 16
+@B = global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: noinline nounwind uwtable
+define void @_Z2jdv() #0 {
+entry:
+  %i = alloca i32, align 4
+  %j = alloca i32, align 4
+  store i32 0, i32* %i, align 4
+  br label %for.cond
+
+  for.cond:                                         ; preds = %for.inc10, %entry
+  %0 = load i32, i32* %i, align 4
+  %cmp = icmp slt i32 %0, 1024
+  br i1 %cmp, label %for.body, label %for.end12
+
+  for.body:                                         ; preds = %for.cond
+  %1 = load i32, i32* %i, align 4
+  store i32 %1, i32* %j, align 4
+  br label %for.cond1
+
+  for.cond1:                                        ; preds = %for.inc, %for.body
+  %2 = load i32, i32* %j, align 4
+  %cmp2 = icmp slt i32 %2, 1024
+  br i1 %cmp2, label %for.body3, label %for.end
+
+  for.body3:                                        ; preds = %for.cond1
+  %3 = load i32, i32* %j, align 4
+  %idxprom = sext i32 %3 to i64
+  %arrayidx = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %idxprom
+  %4 = load i32, i32* %arrayidx, align 4
+  %5 = load i32, i32* %i, align 4
+  %idxprom4 = sext i32 %5 to i64
+  %arrayidx5 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %idxprom4
+  call void @llvm.donothing()
+  %6 = load i32, i32* %arrayidx5, align 4
+  %add = add nsw i32 %6, %4
+  store i32 %add, i32* %arrayidx5, align 4
+  br label %for.inc
+
+  for.inc:                                          ; preds = %for.body3
+  %7 = load i32, i32* %j, align 4
+  %inc = add nsw i32 %7, 1
+  store i32 %inc, i32* %j, align 4
+  br label %for.cond1
+
+  for.end:                                          ; preds = %for.cond1
+  %8 = load i32, i32* %i, align 4
+  %idxprom6 = sext i32 %8 to i64
+  %arrayidx7 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %idxprom6
+  %9 = load i32, i32* %arrayidx7, align 4
+  %10 = load i32, i32* %i, align 4
+  %idxprom8 = sext i32 %10 to i64
+  %arrayidx9 = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, i64 %idxprom8
+  store i32 %9, i32* %arrayidx9, align 4
+  br label %for.inc10
+
+  for.inc10:                                        ; preds = %for.end
+  %11 = load i32, i32* %i, align 4
+  %inc11 = add nsw i32 %11, 1
+  store i32 %inc11, i32* %i, align 4
+  br label %for.cond
+
+  for.end12:                                        ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind
+declare void @llvm.donothing() #1
+
+attributes #0 = { noinline nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone }
+