R600/SI: Use SGPR_64 as reister class for S_LOAD_DWORDX2

[arsenm]

Summary

There's no reason to ever load into the special registers
in the superclass SReg_64. This helps avoid the verifier
error "Expected a SGPR_64 register, but got a SReg_64 register"
because SGPR_64 is not a superclass of SReg_64 when trying
to use a loaded kernel argument register in some cases.

[arsenm]

Actually I think this will just hide a more general problem since lots of places use SReg_64 destinations

lib/Target/R600/SIInstructions.td

 // SMRD instructions, because the SGPR_32 register class does not include M0
 // and writing to M0 from an SMRD instruction will hang the GPU.
 defm S_LOAD_DWORD : SMRD_Helper <0x00, "S_LOAD_DWORD", SReg_64, SGPR_32>;
-defm S_LOAD_DWORDX2 : SMRD_Helper <0x01, "S_LOAD_DWORDX2", SReg_64, SReg_64>;
+defm S_LOAD_DWORDX2 : SMRD_Helper <0x01, "S_LOAD_DWORDX2", SReg_64, SGPR_64>;
 defm S_LOAD_DWORDX4 : SMRD_Helper <0x02, "S_LOAD_DWORDX4", SReg_64, SReg_128>;
 defm S_LOAD_DWORDX8 : SMRD_Helper <0x03, "S_LOAD_DWORDX8", SReg_64, SReg_256>;
 defm S_LOAD_DWORDX16 : SMRD_Helper <0x04, "S_LOAD_DWORDX16", SReg_64, SReg_512>;