[IndVars] Eliminate redundant type cast between unsigned integer and float

[Allen]

Summary

Extend for unsigned integer according the comment of D129191.

[Allen]

ping ?

[nikic]

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

697

For fptoui, shouldn't we also check that it's non-negative? Wouldn't your current code optimize an IV that does -100 to 100 for example?

[Allen]

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

697

thanks @nikic very much, and I verified that the gcc also transform with negative value, see detail in https://godbolt.org/z/1hxebdYfr

[nikic]

LGTM

The next step here could be to also support uitofp.

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

697

Ah yes, it is fine because FPToUI with negative value results in poison anyway. It would probably still be valuable to add a test with negative IV.

[Allen]

Add a new test case including negative value according comment

[Allen]

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

697

it doesn't have scvtf, also see the range (-100, 100] in https://godbolt.org/z/jMhxfnbs9

697

Done, thanks.

[Allen]

update test case

[Allen]

In D129358#3647986, @nikic wrote:

LGTM

The next step here could be to also support uitofp.

Thanks, addressed here, https://reviews.llvm.org/D129756.

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

if (UseInst->getOpcode() != CastInst::SIToFP)
     return false;
 
   Value *IVOperand = UseInst->getOperand(0);
   // Get the symbolic expression for this instruction.
   ConstantRange IVRange = SE->getSignedRange(SE->getSCEV(IVOperand));
   unsigned DestNumSigBits = UseInst->getType()->getFPMantissaWidth();
   if (IVRange.getActiveBits() <= DestNumSigBits) {
     for (User *U : UseInst->users()) {
-      // Match for fptosi of sitofp and with same type.
-      auto *CI = dyn_cast<FPToSIInst>(U);
+      // Match for fptosi/fptoui of sitofp and with same type.
+      auto *CI = dyn_cast<CastInst>(U);
       if (!CI || IVOperand->getType() != CI->getType())
         continue;
 
+      CastInst::CastOps Opcode = CI->getOpcode();
+      if (Opcode != CastInst::FPToSI && Opcode != CastInst::FPToUI)
+        continue;
+
       CI->replaceAllUsesWith(IVOperand);
       DeadInsts.push_back(CI);
       LLVM_DEBUG(dbgs() << "INDVARS: Replace IV user: " << *CI
                         << " with: " << *IVOperand << '\n');
 
       ++NumFoldedUser;
       Changed = true;
     }

llvm/test/Transforms/IndVarSimplify/floating-point-small-iv.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -indvars -S | FileCheck %s
 
 @array = dso_local global [16777219 x i32] zeroinitializer, align 4
 
-define void @small_const_bound(i32 %index) {
+define void @small_const_bound() {
 ; CHECK-LABEL: @small_const_bound(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV_INT:%.*]] = phi i32 [ 100, [[ENTRY:%.*]] ], [ [[DEC_INT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[IV_INT]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [16777219 x i32], [16777219 x i32]* @array, i64 0, i64 [[IDXPROM]]
 ; CHECK-NEXT:    store i32 [[IV_INT]], i32* [[ARRAYIDX]], align 4
for.body:                                         ; preds = %for.body, %entry
   %cmp = icmp ugt i32 %dec.int, 0
   br i1 %cmp, label %for.body, label %cleanup
 
 cleanup:                                          ; preds = %for.body
   ret void
 }
 
 ; Negative test: The transform is *not* valid because there are too many significant bits
-define void @overflow_masked_const_bound(i32 %index) {
+define void @overflow_masked_const_bound() {
 ; CHECK-LABEL: @overflow_masked_const_bound(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV_INT:%.*]] = phi i32 [ 16777218, [[ENTRY:%.*]] ], [ [[DEC_INT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[INDVAR_CONV:%.*]] = sitofp i32 [[IV_INT]] to float
 ; CHECK-NEXT:    [[CONV:%.*]] = fptosi float [[INDVAR_CONV]] to i32
 ; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CONV]] to i64
for.body:                                         ; preds = %for.body, %entry
   %cmp = icmp ugt i32 %dec.int, 0
   br i1 %cmp, label %for.body, label %cleanup
 
 cleanup:                                          ; preds = %for.body
   ret void
 }
 
 ; Negative test: Type mismatch between the integer IV and the fptosi result
-define void @mismatch_type_const(i32 %index) {
+define void @mismatch_type_const() {
 ;
 ; CHECK-LABEL: @mismatch_type_const(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV_INT:%.*]] = phi i32 [ 100, [[ENTRY:%.*]] ], [ [[DEC_INT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[INDVAR_CONV:%.*]] = sitofp i32 [[IV_INT]] to float
 ; CHECK-NEXT:    [[CONV:%.*]] = fptosi float [[INDVAR_CONV]] to i16
for.body:                                         ; preds = %for.body, %entry
   store i32 %idxprom32, i32* %arrayidx, align 4
   %dec.int = add nsw i32 %iv.int, -1
   %cmp = icmp ugt i32 %dec.int, 0
   br i1 %cmp, label %for.body, label %cleanup
 
 cleanup:                                          ; preds = %for.body
   ret void
 }
+
+define void @unsigned_const_bound() {
+; CHECK-LABEL: @unsigned_const_bound(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV_INT:%.*]] = phi i32 [ 100, [[ENTRY:%.*]] ], [ [[DEC_INT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = zext i32 [[IV_INT]] to i64
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [16777219 x i32], [16777219 x i32]* @array, i64 0, i64 [[IDXPROM]]
+; CHECK-NEXT:    store i32 [[IV_INT]], i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[DEC_INT]] = add nsw i32 [[IV_INT]], -1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[DEC_INT]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[CLEANUP:%.*]]
+; CHECK:       cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %iv.int = phi i32 [ 100, %entry ], [ %dec.int, %for.body ]
+  %indvar.conv = sitofp i32 %iv.int to float
+  %conv = fptoui float %indvar.conv to i32
+  %idxprom = zext i32 %conv to i64
+  %arrayidx = getelementptr inbounds [16777219 x i32], [16777219 x i32]* @array, i64 0, i64 %idxprom
+  store i32 %conv, i32* %arrayidx, align 4
+  %dec.int = add nsw i32 %iv.int, -1
+  %cmp = icmp sgt i32 %dec.int, 0
+  br i1 %cmp, label %for.body, label %cleanup
+
+cleanup:                                          ; preds = %for.body
+  ret void
+}
+
+; TODO: Range including negative value.
+define void @unsigned_const_bound_with_negative () {
+; CHECK-LABEL: @unsigned_const_bound_with_negative(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV_INT:%.*]] = phi i32 [ 100, [[ENTRY:%.*]] ], [ [[DEC_INT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[INDVAR_CONV:%.*]] = sitofp i32 [[IV_INT]] to float
+; CHECK-NEXT:    [[CONV:%.*]] = fptoui float [[INDVAR_CONV]] to i32
+; CHECK-NEXT:    [[IDXPROM:%.*]] = zext i32 [[CONV]] to i64
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [16777219 x i32], [16777219 x i32]* @array, i64 0, i64 [[IDXPROM]]
+; CHECK-NEXT:    store i32 [[CONV]], i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[DEC_INT]] = add nsw i32 [[IV_INT]], -1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[DEC_INT]], -100
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[CLEANUP:%.*]]
+; CHECK:       cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %iv.int = phi i32 [ 100, %entry ], [ %dec.int, %for.body ]
+  %indvar.conv = sitofp i32 %iv.int to float
+  %conv = fptoui float %indvar.conv to i32
+  %idxprom = zext i32 %conv to i64
+  %arrayidx = getelementptr inbounds [16777219 x i32], [16777219 x i32]* @array, i64 0, i64 %idxprom
+  store i32 %conv, i32* %arrayidx, align 4
+  %dec.int = add nsw i32 %iv.int, -1
+  %cmp = icmp sgt i32 %dec.int, -100
+  br i1 %cmp, label %for.body, label %cleanup
+
+cleanup:                                          ; preds = %for.body
+  ret void
+}