On 2018-12-03 15:11, Richard Henderson wrote:
> On 12/3/18 8:05 AM, Thomas Huth wrote:
>> +++ b/include/qemu/host-utils.h
>> @@ -28,7 +28,6 @@
>>
>> #include "qemu/bswap.h"
>>
>> -#ifdef CONFIG_INT128
>> static inline void mulu64(uint64_t *plow, uint64_t *phigh,
>> uint64_t a, uint64_t b)
>> {
>> @@ -76,35 +75,6 @@ static inline int divs128(int64_t *plow, int64_t *phigh,
>> int64_t divisor)
>> return result != *plow;
>> }
>> }
>> -#else
>> -void muls64(uint64_t *phigh, uint64_t *plow, int64_t a, int64_t b);
>> -void mulu64(uint64_t *phigh, uint64_t *plow, uint64_t a, uint64_t b);
>
> No, because __int128 support requires a 64-bit host.
> This will not compile for a 32-bit host.
Oops, my bad, you're right, of course. This would have been too easy, I
guess ... So it seems that the only thing that we could do here is to
get rid of the Clang version check in the configure script...
Thomas
