More fixes.

Add more tests.  Correctly spill p12-p15.  Fix issue spilling an odd
number of predicates.

Author: efriedma-quic

llvm/lib/Target/AArch64/AArch64CallingConvention.td

@@ -622,7 +622,7 @@ def CSR_Darwin_AArch64_SVE_AAPCS : CalleeSavedRegs<(add (sequence "Z%u", 8, 23),
                                                         LR, FP, X19, X20, X21, X22,
                                                         X23, X24, X25, X26, X27, X28)>;
 
-def CSR_Win_AArch64_SVE_AAPCS : CalleeSavedRegs<(add (sequence "P%u", 4, 11),
+def CSR_Win_AArch64_SVE_AAPCS : CalleeSavedRegs<(add (sequence "P%u", 4, 15),
                                                      (sequence "Z%u", 8, 23),
                                                      X19, X20, X21, X22, X23, X24,
                                                      X25, X26, X27, X28, FP, LR)>;

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

@@ -3184,6 +3184,12 @@ static void computeCalleeSaveRegisterPairs(
         RPI.isPaired()) // RPI.FrameIdx must be the lower index of the pair
       RPI.FrameIdx = CSI[i + RegInc].getFrameIdx();
 
+    // Realign the scalable offset if necesary.  This is relevant when
+    // spilling predicates on Windows.
+    if (RPI.isScalable() && ScalableByteOffset % Scale != 0) {
+      ScalableByteOffset = alignTo(ScalableByteOffset, Scale);
+    }
+
     int OffsetPre = RPI.isScalable() ? ScalableByteOffset : ByteOffset;
     assert(OffsetPre % Scale == 0);