Implement 'e' and 'f' modifiers for Neon inline asm. <rdar://problem/10551006>

These modifiers simply select either the low or high D subregister of a Neon
Q register.  I've also removed the unimplemented 'p' modifier, which turns out
to be a bit different than the comment here suggests and as far as I can tell
was only intended for internal use in Apple's version of gcc.

llvm-svn: 146417

llvm/lib/Target/ARM/ARMAsmPrinter.cpp

@@ -493,11 +493,21 @@ bool ARMAsmPrinter::PrintAsmOperand(const MachineInstr *MI, unsigned OpNum,
       return false;
     }
 
-    // These modifiers are not yet supported.
-    case 'p': // The high single-precision register of a VFP double-precision
-              // register.
     case 'e': // The low doubleword register of a NEON quad register.
-    case 'f': // The high doubleword register of a NEON quad register.
+    case 'f': { // The high doubleword register of a NEON quad register.
+      if (!MI->getOperand(OpNum).isReg())
+        return true;
+      unsigned Reg = MI->getOperand(OpNum).getReg();
+      if (!ARM::QPRRegClass.contains(Reg))
+        return true;
+      const TargetRegisterInfo *TRI = MF->getTarget().getRegisterInfo();
+      unsigned SubReg = TRI->getSubReg(Reg, ExtraCode[0] == 'e' ?
+                                       ARM::dsub_0 : ARM::dsub_1);
+      O << ARMInstPrinter::getRegisterName(SubReg);
+      return false;
+    }
+
+    // These modifiers are not yet supported.
     case 'h': // A range of VFP/NEON registers suitable for VLD1/VST1.
     case 'H': // The highest-numbered register of a pair.
       return true;

llvm/test/CodeGen/ARM/inlineasm3.ll

@@ -110,3 +110,13 @@ entry:
   call void asm "str $1, $0", "=*Q,r"(i8** %f.addr, i32 %g) nounwind
   ret void
 }
+
+; Radar 10551006
+
+define <4 x i32> @t11(i32* %p) nounwind {
+entry:
+; CHECK: t11
+; CHECK: vld1.s32 {d16[], d17[]}, [r0]
+  %0 = tail call <4 x i32> asm "vld1.s32 {${0:e}[], ${0:f}[]}, [$1]", "=w,r"(i32* %p) nounwind
+  ret <4 x i32> %0
+}