[PATCH] D146171: InstCombine: Introduce new is.fpclass from logic of fcmp

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: jcranmer-intel, sepavloff, andrew.w.kaylor, kpn, cameron.mcinally.
Herald added a subscriber: hiraditya.
Herald added a project: All.
arsenm requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

Fixes regressions from patch to turn more classes into fcmp.


https://reviews.llvm.org/D146171

Files:
  llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
  llvm/test/Transforms/InstCombine/is_fpclass.ll


Index: llvm/test/Transforms/InstCombine/is_fpclass.ll
===================================================================
--- llvm/test/Transforms/InstCombine/is_fpclass.ll
+++ llvm/test/Transforms/InstCombine/is_fpclass.ll
@@ -1322,9 +1322,7 @@
 
 define i1 @test_fold_xor3_class_f32_0(float %a) {
 ; CHECK-LABEL: @test_fold_xor3_class_f32_0(
-; CHECK-NEXT:    [[CLASS0:%.*]] = fcmp uno float [[A:%.*]], 0.000000e+00
-; CHECK-NEXT:    [[CLASS2:%.*]] = fcmp oeq float [[A]], 0xFFF0000000000000
-; CHECK-NEXT:    [[XOR_1:%.*]] = xor i1 [[CLASS0]], [[CLASS2]]
+; CHECK-NEXT:    [[XOR_1:%.*]] = fcmp ueq float [[A:%.*]], 0xFFF0000000000000
 ; CHECK-NEXT:    ret i1 [[XOR_1]]
 ;
   %class0 = call i1 @llvm.is.fpclass.f32(float %a, i32 1)
@@ -1337,24 +1335,7 @@
 
 define i1 @test_fold_xor_all_tests_class_f32_0(float %a) {
 ; CHECK-LABEL: @test_fold_xor_all_tests_class_f32_0(
-; CHECK-NEXT:    [[CLASS0:%.*]] = fcmp uno float [[A:%.*]], 0.000000e+00
-; CHECK-NEXT:    [[CLASS2:%.*]] = fcmp oeq float [[A]], 0xFFF0000000000000
-; CHECK-NEXT:    [[CLASS3:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 8)
-; CHECK-NEXT:    [[CLASS4:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 16)
-; CHECK-NEXT:    [[CLASS5:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 32)
-; CHECK-NEXT:    [[CLASS6:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 64)
-; CHECK-NEXT:    [[CLASS7:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 128)
-; CHECK-NEXT:    [[CLASS8:%.*]] = call i1 @llvm.is.fpclass.f32(float [[A]], i32 256)
-; CHECK-NEXT:    [[CLASS9:%.*]] = fcmp oeq float [[A]], 0x7FF0000000000000
-; CHECK-NEXT:    [[XOR_1:%.*]] = xor i1 [[CLASS0]], [[CLASS2]]
-; CHECK-NEXT:    [[XOR_2:%.*]] = xor i1 [[XOR_1]], [[CLASS3]]
-; CHECK-NEXT:    [[XOR_3:%.*]] = xor i1 [[XOR_2]], [[CLASS4]]
-; CHECK-NEXT:    [[XOR_4:%.*]] = xor i1 [[XOR_3]], [[CLASS5]]
-; CHECK-NEXT:    [[XOR_5:%.*]] = xor i1 [[XOR_4]], [[CLASS6]]
-; CHECK-NEXT:    [[XOR_6:%.*]] = xor i1 [[XOR_5]], [[CLASS7]]
-; CHECK-NEXT:    [[XOR_7:%.*]] = xor i1 [[XOR_6]], [[CLASS8]]
-; CHECK-NEXT:    [[XOR_8:%.*]] = xor i1 [[XOR_7]], [[CLASS9]]
-; CHECK-NEXT:    ret i1 [[XOR_8]]
+; CHECK-NEXT:    ret i1 true
 ;
   %class0 = call i1 @llvm.is.fpclass.f32(float %a, i32 1)
   %class1 = call i1 @llvm.is.fpclass.f32(float %a, i32 2)
@@ -1421,10 +1402,7 @@
 
 define i1 @test_fold_xor_class_f32_2(float %a) {
 ; CHECK-LABEL: @test_fold_xor_class_f32_2(
-; CHECK-NEXT:    [[CLASS0:%.*]] = fcmp ueq float [[A:%.*]], 0xFFF0000000000000
-; CHECK-NEXT:    [[CLASS1:%.*]] = fcmp ueq float [[A]], 0xFFF0000000000000
-; CHECK-NEXT:    [[XOR:%.*]] = xor i1 [[CLASS0]], [[CLASS1]]
-; CHECK-NEXT:    ret i1 [[XOR]]
+; CHECK-NEXT:    ret i1 false
 ;
   %class0 = call i1 @llvm.is.fpclass.f32(float %a, i32 7)
   %class1 = call i1 @llvm.is.fpclass.f32(float %a, i32 7)
Index: llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
===================================================================
--- llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -1577,9 +1577,8 @@
   bool IsRHSClass =
       match(Op1, m_OneUse(m_Intrinsic<Intrinsic::is_fpclass>(
                      m_Value(ClassVal1), m_ConstantInt(ClassMask1))));
-  if (((IsLHSClass && IsRHSClass) ||
-       ((!IsLHSClass && matchIsFPClassLikeFCmp(Op0, ClassVal0, ClassMask0)) ||
-        (!IsRHSClass && matchIsFPClassLikeFCmp(Op1, ClassVal1, ClassMask1)))) &&
+  if ((((IsLHSClass || matchIsFPClassLikeFCmp(Op0, ClassVal0, ClassMask0)) &&
+        (IsRHSClass || matchIsFPClassLikeFCmp(Op1, ClassVal1, ClassMask1)))) &&
       ClassVal0 == ClassVal1) {
     unsigned NewClassMask;
     switch (BO.getOpcode()) {
@@ -1596,10 +1595,24 @@
       llvm_unreachable("not a binary logic operator");
     }
 
-    auto *II = IsLHSClass ? cast<IntrinsicInst>(Op0) : cast<IntrinsicInst>(Op1);
-    II->setArgOperand(
-        1, ConstantInt::get(II->getArgOperand(1)->getType(), NewClassMask));
-    return replaceInstUsesWith(BO, II);
+    if (IsLHSClass) {
+      auto *II = cast<IntrinsicInst>(Op0);
+      II->setArgOperand(
+          1, ConstantInt::get(II->getArgOperand(1)->getType(), NewClassMask));
+      return replaceInstUsesWith(BO, II);
+    }
+
+    if (IsRHSClass) {
+      auto *II = cast<IntrinsicInst>(Op1);
+      II->setArgOperand(
+          1, ConstantInt::get(II->getArgOperand(1)->getType(), NewClassMask));
+      return replaceInstUsesWith(BO, II);
+    }
+
+    CallInst *NewClass =
+        Builder.CreateIntrinsic(Intrinsic::is_fpclass, {ClassVal0->getType()},
+                                {ClassVal0, Builder.getInt32(NewClassMask)});
+    return replaceInstUsesWith(BO, NewClass);
   }
 
   return nullptr;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D146171.505634.patch
Type: text/x-patch
Size: 4726 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230315/d2d12760/attachment.bin>