TI MSP430 Microcontroller

Hardware Multiplier

Found in some versions such as the MSP430F1611

C compilers do not necessarily do a good job of using it. Below is an example using pure C vs the multiplier accumulator equivalent.

typedef unsigned short      UINT16;     ///< unsigned 16-bit
typedef unsigned long       UINT32;     ///< unsigned 32-bit


    UINT32 u32Answer;
    UINT16 uX,uY,uZ;
    uX = 10;
    uY = 2;
    uZ = 40000;
The line below translates to lots of code (even with the "use multiplier"
option checked in the compiller):
    u32Answer = (UINT32)uX*3 + (UINT32)uY*4 + (UINT32)uZ*5 + (UINT32)uX*6;

Efficientcy by using multiply accumulate:

    *(INT32 *)RESLO_ = 0L; // Clear accumulator
    MAC = 3;
    OP2 = uX;
    MAC = 4;
    OP2 = uY;
    MAC = 5;
    OP2 = uZ;
    MAC = 6;
    OP2 = uX;
    NOP();
    u32Answer = *(INT32 *)RESLO_;

Code in CPU looks like:
     94              *(INT32 *)RESLO_ = 0L;
   \   000072   82433a01     MOV.W   #0x0, &0x13a
   \   000076   82433c01     MOV.W   #0x0, &0x13c
     95              MAC = 3;
   \   00007A   B24003003401 MOV.W   #0x3, &0x134
     96              OP2 = uX;
   \   000080   824E3801     MOV.W   R14, &0x138
     97              MAC = 4;
   \   000084   A2423401     MOV.W   #0x4, &0x134
     98              OP2 = uY;
   \   000088   824C3801     MOV.W   R12, &0x138
     99              MAC = 5;
   \   00008C   B24005003401 MOV.W   #0x5, &0x134
    100              OP2 = uZ;
   \   000092   824F3801     MOV.W   R15, &0x138
    101              MAC = 6;
   \   000096   B24006003401 MOV.W   #0x6, &0x134
    102              OP2 = uX;
   \   00009C   824E3801     MOV.W   R14, &0x138
    103              NOP();
   \   0000A0   0343         NOP
    104              u32Answer = *(INT32 *)RESLO_;
   \   0000A2   92423A01.... MOV.W   &0x13a, &u32Answer
   \   0000A8   92423C01.... MOV.W   &0x13c, &u32Answer + 2