On Fri, Jan 13, 2017 at 09:04:46AM +0000, Tvrtko Ursulin wrote:
> On 12/01/2017 21:35, Chris Wilson wrote:
> >Save a lot of characters by making the union anonymous, with the
> >side-effect of ignoring unset bits when comparing views.
> Side-effect is not happening in this patch. Always tricky what to do
> with commit messages for split patches. :) Maybe just rewrite the
> commit message.
> >v2: Roll up the memcmps back into one.
> >v3: And split again as Ville points out we can't trust the compiler.
> Not sure what is split, memcmps? But there is only one. Needs v4? Or
> I am missing something?
The changelog just needs rewriting so that the preconditions for this
patch (wanting to able to use a single memcmp and so no uninitialised
bits within the memcmp) are clear. In a sense, it's why I liked having
this in one patch, since the patches are not really standalone and only
make sense together :| Revert one, we have to revert them all.
Chris Wilson, Intel Open Source Technology Centre
Intel-gfx mailing list