RE: [RFC PATCH v3 26/34] Hexagon (target/hexagon) macros referenced in instruction semantics

[Taylor Simpson <tsimpson@quicinc.com>]

> -----Original Message-----
> From: Richard Henderson <richard.henderson@linaro.org>
> Sent: Friday, October 9, 2020 6:59 AM
>
> On 10/8/20 3:54 PM, Taylor Simpson wrote:
> > I'll reach out to the architects and ask if there is a case where doing the
> store first could be a problem - and also about the priority of exceptions.
>
> Of course there are cases where doing the store first will cause problems.
>
> > This may or may not be possible.  We already have to reorder to put .new
> consumers after the producers, and we have to keep the change-of-flow
> instructions in the original order.  So, I'm reluctant to go down this path
> worrying that there will be a long tail of corner case bugs to engineer out.
>
> Well, do as you like, I suppose, but I think the merging thing that you were
> trying to do is just as fraught with peril.

Thanks for your patience on this.  After speaking to the architects and putting more comments in the decoder (per your feedback),  I'm convinced that doing the store first won't lead to problems.  So, I'll move forward with your suggestion.

In other news, this change and the switch to qemu softfloat are the only blockers from your review.  Once these are done, I'll submit v5 of the series.  Thanks for your feedback, it's made our code significantly better.


Thanks
Taylor