[llvm-commits] [llvm] r135126 - in /llvm/trunk: lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp lib/Target/README.txt test/CodeGen/X86/switch-bt.ll

[Benjamin Kramer]

Author: d0k
Date: Wed Jul 13 20:38:42 2011
New Revision: 135126

URL: http://llvm.org/viewvc/llvm-project?rev=135126&view=rev
Log:
Don't emit a bit test if there is only one case the test can yield false. A simple SETNE is sufficient.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
    llvm/trunk/lib/Target/README.txt
    llvm/trunk/test/CodeGen/X86/switch-bt.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=135126&r1=135125&r2=135126&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Wed Jul 13 20:38:42 2011
@@ -1727,7 +1727,8 @@
   SDValue ShiftOp = DAG.getCopyFromReg(getControlRoot(), getCurDebugLoc(),
                                        Reg, VT);
   SDValue Cmp;
-  if (CountPopulation_64(B.Mask) == 1) {
+  unsigned PopCount = CountPopulation_64(B.Mask);
+  if (PopCount == 1) {
     // Testing for a single bit; just compare the shift count with what it
     // would need to be to shift a 1 bit in that position.
     Cmp = DAG.getSetCC(getCurDebugLoc(),
@@ -1735,6 +1736,13 @@
                        ShiftOp,
                        DAG.getConstant(CountTrailingZeros_64(B.Mask), VT),
                        ISD::SETEQ);
+  } else if (PopCount == BB.Range) {
+    // There is only one zero bit in the range, test for it directly.
+    Cmp = DAG.getSetCC(getCurDebugLoc(),
+                       TLI.getSetCCResultType(VT),
+                       ShiftOp,
+                       DAG.getConstant(CountTrailingOnes_64(B.Mask), VT),
+                       ISD::SETNE);
   } else {
     // Make desired shift
     SDValue SwitchVal = DAG.getNode(ISD::SHL, getCurDebugLoc(), VT,

Modified: llvm/trunk/lib/Target/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/README.txt?rev=135126&r1=135125&r2=135126&view=diff
==============================================================================
--- llvm/trunk/lib/Target/README.txt (original)
+++ llvm/trunk/lib/Target/README.txt Wed Jul 13 20:38:42 2011
@@ -1762,7 +1762,6 @@
 
 //===---------------------------------------------------------------------===//
 
-Switch lowering generates less than ideal code for the following switch:
 define void @a(i32 %x) nounwind {
 entry:
   switch i32 %x, label %if.end [
@@ -1783,19 +1782,15 @@
 Generated code on x86-64 (other platforms give similar results):
 a:
 	cmpl	$5, %edi
-	ja	.LBB0_2
-	movl	%edi, %eax
-	movl	$47, %ecx
-	btq	%rax, %rcx
-	jb	.LBB0_3
+	ja	LBB2_2
+	cmpl	$4, %edi
+	jne	LBB2_3
 .LBB0_2:
 	ret
 .LBB0_3:
 	jmp	foo  # TAILCALL
 
-The movl+movl+btq+jb could be simplified to a cmpl+jne.
-
-Or, if we wanted to be really clever, we could simplify the whole thing to
+If we wanted to be really clever, we could simplify the whole thing to
 something like the following, which eliminates a branch:
 	xorl    $1, %edi
 	cmpl	$4, %edi

Modified: llvm/trunk/test/CodeGen/X86/switch-bt.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/switch-bt.ll?rev=135126&r1=135125&r2=135126&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/switch-bt.ll (original)
+++ llvm/trunk/test/CodeGen/X86/switch-bt.ll Wed Jul 13 20:38:42 2011
@@ -79,3 +79,23 @@
 }
 
 declare void @bar()
+
+define void @test3(i32 %x) nounwind {
+; CHECK: test3:
+; CHECK: cmpl $5
+; CHECK: ja
+; CHECK: cmpl $4
+; CHECK: jne
+  switch i32 %x, label %if.end [
+    i32 0, label %if.then
+    i32 1, label %if.then
+    i32 2, label %if.then
+    i32 3, label %if.then
+    i32 5, label %if.then
+  ]
+if.then:
+  tail call void @bar() nounwind
+  ret void
+if.end:
+  ret void
+}