Many thanks to Segher and Richard for pointing out that my removal
of optimizations of ABS(ABS(x)) and ABS(FFS(x)) in the original version
of this patch was incorrect, and my assumption that these would be
subsumed by val_signbit_known_clear_p was mistaken. That the
tests for ABS and FFS looked
"Roger Sayle" writes:
> Many thanks to Segher and Richard for pointing out that my removal
> of optimizations of ABS(ABS(x)) and ABS(FFS(x)) in the original version
> of this patch was incorrect, and my assumption that these would be
> subsumed by val_signbit_known_clear_p was mistaken. That the
Richard Sandiford via Gcc-patches writes:
> "Roger Sayle" writes:
>> Many thanks to Segher and Richard for pointing out that my removal
>> of optimizations of ABS(ABS(x)) and ABS(FFS(x)) in the original version
>> of this patch was incorrect, and my assumption that these would be
>> subsumed by
