X86: Implement the versions of PUNPCKLBW that use MMX registers.

src/arch/x86/isa/decoder/two_byte_opcodes.isa

@@ -551,7 +551,7 @@
             0x0C: decode LEGACY_DECODEVAL {
                 // no prefix
                 0x0: decode OPCODE_OP_BOTTOM3 {
-                    0x0: punpcklbw_Pq_Qd();
+                    0x0: Inst::PUNPCKLBW(Pq,Qd);
                     0x1: punpcklwd_Pq_Qd();
                     0x2: punpckldq_Pq_Qd();
                     0x3: packsswb_Pq_Qq();

src/arch/x86/isa/insts/simd64/integer/data_reordering/unpack_and_interleave.py

@@ -54,10 +54,23 @@
 # Authors: Gabe Black
 
 microcode = '''
+def macroop PUNPCKLBW_MMX_MMX {
+    unpack mmx, mmx, mmxm, sel=0, size=1
+};
+
+def macroop PUNPCKLBW_MMX_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    unpack mmx, mmx, ufp1, sel=0, size=1
+};
+
+def macroop PUNPCKLBW_MMX_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    unpack mmx, mmx, ufp1, sel=0, size=1
+};
+'''
 # PUNPCKHBW
 # PUNPCKHWD
 # PUNPCKHDQ
-# PUNPCKLBW
 # PUNPCKLWD
 # PUNPCKLDQ
-'''