On Mon, 4 Feb 2019 at 05:27, Richard Henderson <richard.hender...@linaro.org> wrote: > > Two cleanups to surrounding code preceed the actual implementation. > > I did not find the ARM FPToFixedJS pseudo-function particularly helpful, > so I mostly cribed off of the Alpha cvttq implementation. But I think > it all makes sense -- convert, produce the mod 2^32 result with NaN > getting INT32_MAX, set ZF iff the conversion is exact. > > I've tested the aa64 version vs FVP with RISU. I need to adjust my > FVP configuration in order to test aa32.
I've reviewed patch 1; I'll leave 2 and 3 til you post v2. thanks -- PMM