Fix checked arithmetic for i8 on X86

[loladiro]

Summary

This is http://llvm.org/bugs/show_bug.cgi?id=19858

[andreadb]

Hi Keno,

Thanks for the patch.
Can you add a test for it? I think you can probably reuse the example you posted on the bugzilla to verify that we correctly test and branch.

lib/Target/X86/X86ISelLowering.cpp

11472–11475

You can simplify this into:

X86::CondCode X86Cond = Inverted ? X86::COND_E : X86::COND_NE;

[loladiro]

Added test. Simplified conditional.

[andreadb]

Hi Keno,

I only have a small request (see below). Otherwise, the patch looks good to me. Thanks!

test/CodeGen/X86/i8-umulo.ll

2–3

Please add a reference to the bugzilla.
Something like '; PR19858'

[loladiro]

Add reference to the bugzilla bug.

[loladiro]

Done. I don't have commit access though.

lib/Target/X86/X86ISelLowering.cpp

if (Cond.getOpcode() == ISD::AND && Cond.hasOneUse()) {
         CC = NewSetCC.getOperand(0);
         Cond = NewSetCC.getOperand(1);
         addTest = false;
       }
     }
   }
 
   if (addTest) {
-    CC = DAG.getConstant(X86::COND_NE, MVT::i8);
-    Cond = EmitTest(Cond, X86::COND_NE, dl, DAG);
+    X86::CondCode X86Cond = Inverted ? X86::COND_E : X86::COND_NE;
+    CC = DAG.getConstant(X86Cond, MVT::i8);
+    Cond = EmitTest(Cond, X86Cond, dl, DAG);
   }
   Cond = ConvertCmpIfNecessary(Cond, DAG);
   return DAG.getNode(X86ISD::BRCOND, dl, Op.getValueType(),
                      Chain, Dest, CC, Cond);
 }
 
 // Lower dynamic stack allocation to _alloca call for Cygwin/Mingw targets.
 // Calls to _alloca is needed to probe the stack when allocating more than 4k
 // bytes in one go. Touching the stack at 4K increments is necessary to ensure

test/CodeGen/X86/i8-umulo.ll

+; RUN: llc -mcpu=generic -march=x86 < %s | FileCheck %s
+; PR 19858
+
+declare {i8, i1} @llvm.umul.with.overflow.i8(i8 %a, i8 %b)
+define i8 @testumulo(i32 %argc) {
+; CHECK: imulw
+; CHECK: testb %{{.+}}, %{{.+}}
+; CHECK: je [[NOOVERFLOWLABEL:.+]]
+; CHECK: {{.*}}[[NOOVERFLOWLABEL]]:
+; CHECK-NEXT: movb
+; CHECK-NEXT: retl
+top:
+  %RHS = trunc i32 %argc to i8
+  %umul = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 25, i8 %RHS)
+  %ex = extractvalue { i8, i1 } %umul, 1
+  br i1 %ex, label %overflow, label %nooverlow
+
+overflow:
+  ret i8 %RHS
+
+nooverlow:
+  %umul.value = extractvalue { i8, i1 } %umul, 0
+  ret i8 %umul.value
+}