commit b7d65cab4f3e9a6f66a496e723e53ed736c4d2e7
Author: Mark Wielaard <mark@klomp.org>
Date:   Sun Dec 9 00:55:42 2018 +0100

    Implement ppc64 lxvd2x as 128-bit load with double word swap for ppc64le.
    
    This makes it possible for memcheck to know which part of the 128bit
    vector is defined, even if the load is partly beyond an addressable block.
    
    Partially resolves bug 386945.

diff --git a/VEX/priv/guest_ppc_toIR.c b/VEX/priv/guest_ppc_toIR.c
index a81dace..7af4973 100644
--- a/VEX/priv/guest_ppc_toIR.c
+++ b/VEX/priv/guest_ppc_toIR.c
@@ -20590,16 +20590,22 @@ dis_vx_load ( UInt theInstr )
    }
    case 0x34C: // lxvd2x
    {
-      IROp addOp = ty == Ity_I64 ? Iop_Add64 : Iop_Add32;
-      IRExpr * high, *low;
-      ULong ea_off = 8;
-      IRExpr* high_addr;
+      IRExpr *t128;
       DIP("lxvd2x %d,r%u,r%u\n", XT, rA_addr, rB_addr);
-      high = load( Ity_I64, mkexpr( EA ) );
-      high_addr = binop( addOp, mkexpr( EA ), ty == Ity_I64 ? mkU64( ea_off )
-            : mkU32( ea_off ) );
-      low = load( Ity_I64, high_addr );
-      putVSReg( XT, binop( Iop_64HLtoV128, high, low ) );
+      t128 = load( Ity_V128, mkexpr( EA ) );
+
+      /* The data in the vec register should be in big endian order.
+         So if we just did a little endian load then swap around the
+         high and low double words. */
+      if (host_endness == VexEndnessLE) {
+         IRTemp high = newTemp(Ity_I64);
+         IRTemp low = newTemp(Ity_I64);
+         assign( high, unop(Iop_V128HIto64, t128) );
+         assign( low, unop(Iop_V128to64, t128) );
+         t128 = binop( Iop_64HLtoV128, mkexpr (low), mkexpr (high) );
+      }
+
+      putVSReg( XT, t128 );
       break;
    }
    case 0x14C: // lxvdsx