On 03/18/2014 12:23 PM, Peter Maydell wrote:
> +#define USATACC(bits, shift) \
> + do { \
> + va = (int##bits##_t)((a >> shift) & ((1 << bits) - 1)); \
> + vb = (uint##bits##_t)((b >> shift) & ((1 << bits) - 1)); \
The masking seems redundant with the cast.
Perhaps better to use extract32/sextract32?
> + r |= (uint32_t) (vr & ((1 << bits) - 1)) << shift; \
Perhaps better with deposit32? That implies an unnecessary mask of R, but
perhaps the compiler can clean that up...
Otherwise I don't actually see anything wrong,
Reviewed-by: Richard Henderson <rth@xxxxxxxxxxx>
r~
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
