Add a little dag combine to compile this:

int %AreSecondAndThirdElementsBothNegative(<4 x float>* %in) {
entry:
        %tmp1 = load <4 x float>* %in           ; <<4 x float>> [#uses=1]
        %tmp = tail call int %llvm.ppc.altivec.vcmpgefp.p( int 1, <4 x float> < float 0x7FF8000000000000, float 0.000000e+00, float 0.000000e+00, float 0x7FF8000000000000 >, <4 x float> %tmp1 )           ; <int> [#uses=1]
        %tmp = seteq int %tmp, 0                ; <bool> [#uses=1]
        %tmp3 = cast bool %tmp to int           ; <int> [#uses=1]
        ret int %tmp3
}

into this:

_AreSecondAndThirdElementsBothNegative:
        mfspr r2, 256
        oris r4, r2, 49152
        mtspr 256, r4
        li r4, lo16(LCPI1_0)
        lis r5, ha16(LCPI1_0)
        lvx v0, 0, r3
        lvx v1, r5, r4
        vcmpgefp. v0, v1, v0
        mfcr r3, 2
        rlwinm r3, r3, 27, 31, 31
        mtspr 256, r2
        blr

instead of this:

_AreSecondAndThirdElementsBothNegative:
        mfspr r2, 256
        oris r4, r2, 49152
        mtspr 256, r4
        li r4, lo16(LCPI1_0)
        lis r5, ha16(LCPI1_0)
        lvx v0, 0, r3
        lvx v1, r5, r4
        vcmpgefp. v0, v1, v0
        mfcr r3, 2
        rlwinm r3, r3, 27, 31, 31
        xori r3, r3, 1
        cntlzw r3, r3
        srwi r3, r3, 5
        mtspr 256, r2
        blr

llvm-svn: 27356

1 files changed, 33 insertions, 0 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index cf585a82dd9..f0003f0d6cd 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -1601,6 +1601,39 @@ SDOperand DAGCombiner::visitSRL(SDNode *N) {
     return DAG.getNode(ISD::SRL, VT, N0.getOperand(0), 
                        DAG.getConstant(c1 + c2, N1.getValueType()));
   }
+  
+  // fold (srl (ctlz x), "5") -> x  iff x has one bit set (the low bit).
+  if (N1C && N0.getOpcode() == ISD::CTLZ && 
+      N1C->getValue() == Log2_32(MVT::getSizeInBits(VT))) {
+    uint64_t KnownZero, KnownOne, Mask = MVT::getIntVTBitMask(VT);
+    TLI.ComputeMaskedBits(N0.getOperand(0), Mask, KnownZero, KnownOne);
+    
+    // If any of the input bits are KnownOne, then the input couldn't be all
+    // zeros, thus the result of the srl will always be zero.
+    if (KnownOne) return DAG.getConstant(0, VT);
+    
+    // If all of the bits input the to ctlz node are known to be zero, then
+    // the result of the ctlz is "32" and the result of the shift is one.
+    uint64_t UnknownBits = ~KnownZero & Mask;
+    if (UnknownBits == 0) return DAG.getConstant(1, VT);
+    
+    // Otherwise, check to see if there is exactly one bit input to the ctlz.
+    if ((UnknownBits & (UnknownBits-1)) == 0) {
+      // Okay, we know that only that the single bit specified by UnknownBits
+      // could be set on input to the CTLZ node.  If this bit is set, the SRL
+      // will return 0, if it is clear, it returns 1.  Change the CTLZ/SRL pair
+      // to an SRL,XOR pair, which is likely to simplify more.
+      unsigned ShAmt = CountTrailingZeros_64(UnknownBits);
+      SDOperand Op = N0.getOperand(0);
+      if (ShAmt) {
+        Op = DAG.getNode(ISD::SRL, VT, Op,
+                         DAG.getConstant(ShAmt, TLI.getShiftAmountTy()));
+        AddToWorkList(Op.Val);
+      }
+      return DAG.getNode(ISD::XOR, VT, Op, DAG.getConstant(1, VT));
+    }
+  }
+  
   return SDOperand();
 }