A multiplication by utilizing the MAC unit leads to following assembler listing
c = a*b; 4: 02 12 push r2 ; 6: 32 c2 dint 8: 03 43 nop a: 82 4c 32 01 mov r15, &0x0132 ; e: 82 4f 38 01 mov r14, &0x0138 ; 12: 1f 42 3a 01 mov &0x013a,r15 ;0x013a 16: 32 41 pop r2 ; 18: 82 4f 00 00 mov r15, &0x0000 ; Wouldn't it make sense to have a measure to avoid the interrupt enable/disable sequence if one is sure that his interrupts don't use multiplications? Say, with a command line parameter -mnointmac? If you need very tight code (e. g. fft), this would be very helpful. Optimal code would be: mov r15, &0x0132 mov r14, &0x0138 mov &0x013a,&0x0000 An even more sophisticated solution would be if an intermediate 32bit result were held in the MAC result register, and a "e = a*b + c*d" ended up in the appropriate multiply/accumulate sequence. Regards, Ralf
