On 2008-04-28, David Brown <[email protected]> wrote:
> Grant Edwards wrote:
>>>> Cast one or both operands to 32-bit and C will give you a 32
>>>> bit result. Or copy one to a 32-bit and use that as the
>>>> result.
>>>
>>> but, if you promote operands to 32-bit, the compiler will use
>>> 32x32->32 multiply, which is not implemented in hardware on
>>> msp430 (presumably it'll be a library call). The relevant
>>> question is how to trigger the code generator to emit the MPY
>>> instruction that takes two 16-bit operands and produces a
>>> 32-bit result.
>>
>> That question has come up a couple times in the past, and IIRC,
>> the answer is that there's no practical way to get GCC to do
>> that.
I must have mis-remembered the previous threads, because
casting either or both of the operands does exactly what the OP
wants:
#include <stdint.h>
uint32_t foo(uint16_t a, uint16_t b)
{
return (uint32_t)a * b;
}
foo:
push r2
dint
nop
mov r15, &__MPY
mov r14, &__OP2
mov &__RESLO, r14
mov &__RESHI, r15
pop r2
ret
I get the same generated code if I cast both operands, so I
don't really understand what the OP is complaining about...
--
Grant Edwards grante Yow! DIDI ... is that a
at MARTIAN name, or, are we
visi.com in ISRAEL?
