[PATCH] D57352: [InstCombine] canonicalize cmp/select form of uadd saturate with constant

[Sanjay Patel via Phabricator via llvm-commits]

spatel created this revision.
spatel added reviewers: nikic, RKSimon, lebedev.ri.
Herald added a subscriber: mcrosier.

I'm not sure where we are in the process of optimizing the saturating/overflowing ops, so I'm circling back around to a loose end from D51929 <https://reviews.llvm.org/D51929>.

The backend (either CGP or DAG) doesn't recognize this pattern, so we end up with different asm for these IR variants.

Regardless of anything else, I think we want to get raw IR variations into the minimal number of raw IR forms. If/when we can canonicalize to intrinsics, that will make that step easier.


https://reviews.llvm.org/D57352

Files:
  lib/Transforms/InstCombine/InstCombineSelect.cpp
  test/Transforms/InstCombine/saturating-add-sub.ll


Index: test/Transforms/InstCombine/saturating-add-sub.ll
===================================================================
--- test/Transforms/InstCombine/saturating-add-sub.ll
+++ test/Transforms/InstCombine/saturating-add-sub.ll
@@ -655,9 +655,9 @@
 define i32 @uadd_sat_constant_commute(i32 %x) {
 ; CHECK-LABEL: @uadd_sat_constant_commute(
 ; CHECK-NEXT:    [[A:%.*]] = add i32 [[X:%.*]], 42
-; CHECK-NEXT:    [[C:%.*]] = icmp ult i32 [[X]], -43
-; CHECK-NEXT:    [[R:%.*]] = select i1 [[C]], i32 [[A]], i32 -1
-; CHECK-NEXT:    ret i32 [[R]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ugt i32 [[X]], -43
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 [[A]]
+; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
   %a = add i32 %x, 42
   %c = icmp ult i32 %x, -43
@@ -681,9 +681,9 @@
 define <4 x i32> @uadd_sat_constant_vec_commute(<4 x i32> %x) {
 ; CHECK-LABEL: @uadd_sat_constant_vec_commute(
 ; CHECK-NEXT:    [[A:%.*]] = add <4 x i32> [[X:%.*]], <i32 42, i32 42, i32 42, i32 42>
-; CHECK-NEXT:    [[C:%.*]] = icmp ult <4 x i32> [[X]], <i32 -43, i32 -43, i32 -43, i32 -43>
-; CHECK-NEXT:    [[R:%.*]] = select <4 x i1> [[C]], <4 x i32> [[A]], <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>
-; CHECK-NEXT:    ret <4 x i32> [[R]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ugt <4 x i32> [[X]], <i32 -43, i32 -43, i32 -43, i32 -43>
+; CHECK-NEXT:    [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, <4 x i32> [[A]]
+; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
   %a = add <4 x i32> %x, <i32 42, i32 42, i32 42, i32 42>
   %c = icmp ult <4 x i32> %x, <i32 -43, i32 -43, i32 -43, i32 -43>
Index: lib/Transforms/InstCombine/InstCombineSelect.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -675,6 +675,23 @@
   return IsNegative ? Builder.CreateSub(B, Max) : Builder.CreateSub(Max, B);
 }
 
+static Value *canonicalizeSaturatedAdd(ICmpInst *Cmp, Value *TVal, Value *FVal,
+                                       InstCombiner::BuilderTy &Builder) {
+  // Match an unsigned saturated add with constant.
+  Value *X = Cmp->getOperand(0);
+  const APInt *CmpC, *AddC;
+  if (!Cmp->hasOneUse() || Cmp->getPredicate() != ICmpInst::ICMP_ULT ||
+      !match(Cmp->getOperand(1), m_APInt(CmpC)) || !match(FVal, m_AllOnes()) ||
+      !match(TVal, m_Add(m_Specific(X), m_APInt(AddC))) || ~(*AddC) != *CmpC)
+    return nullptr;
+
+  // Commute compare and select operands:
+  // select (icmp ult X, C), (add X, ~C), -1 -->
+  // select (icmp ugt X, C), -1, (add X, ~C)
+  Value *NewCmp = Builder.CreateICmp(ICmpInst::ICMP_UGT, X, Cmp->getOperand(1));
+  return Builder.CreateSelect(NewCmp, FVal, TVal);
+}
+
 /// Attempt to fold a cttz/ctlz followed by a icmp plus select into a single
 /// call to cttz/ctlz with flag 'is_zero_undef' cleared.
 ///
@@ -1048,6 +1065,9 @@
   if (Value *V = canonicalizeSaturatedSubtract(ICI, TrueVal, FalseVal, Builder))
     return replaceInstUsesWith(SI, V);
 
+  if (Value *V = canonicalizeSaturatedAdd(ICI, TrueVal, FalseVal, Builder))
+    return replaceInstUsesWith(SI, V);
+
   return Changed ? &SI : nullptr;
 }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D57352.183940.patch
Type: text/x-patch
Size: 3205 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190128/bece81f4/attachment.bin>