armemu: Implement USAD8 and USADA8

src/core/arm/interpreter/armemu.cpp

@@ -5888,8 +5888,37 @@ L_stm_s_takeabort:
         case 0x75:
             printf ("Unhandled v6 insn: smmla/smmls/smmul\n");
             break;
-        case 0x78:
+        case 0x78: // USAD8 and USADA8
-            printf ("Unhandled v6 insn: usad/usada8\n");
+            {
+                const u8 rm_idx = BITS(8, 11);
+                const u8 rn_idx = BITS(0, 3);
+                const u8 rd_idx = BITS(16, 19);
+
+                const u32 rm_val = state->Reg[rm_idx];
+                const u32 rn_val = state->Reg[rn_idx];
+
+                // Equivalent of https://graphics.stanford.edu/~seander/bithacks.html#IntegerAbs for 8-bit.
+                auto branchlessAbs = [&](u8 shift) {
+                    const int diff = ((rn_val >> shift) & 0xFF) - ((rm_val >> shift) & 0xFF);
+                    const int mask = diff >> (CHAR_BIT - 1);
+
+                    return (diff + mask) ^ mask;
+                };
+
+                u32 finalDif = 0;
+                finalDif += branchlessAbs(0);
+                finalDif += branchlessAbs(8);
+                finalDif += branchlessAbs(16);
+                finalDif += branchlessAbs(24);
+
+                // Op is USADA8 if true.
+                const u8 ra_idx = BITS(12, 15);
+                if (ra_idx != 15)
+                    finalDif += state->Reg[ra_idx];
+
+                state->Reg[rd_idx] = finalDif;
+                return 1;
+            }
             break;
 #if 0
         case 0x7a: