[Intel-gfx] [PATCH 00/13] Gen7 batch buffer command parser

[Volkin, Bradley D]

On Tue, Mar 25, 2014 at 06:15:36AM -0700, Daniel Vetter wrote:
> On Thu, Mar 20, 2014 at 04:43:05PM +0200, Jani Nikula wrote:
> > 
> > Hi Bradley -
> > 
> > Apologies for my procrastination with the review; I don't easily recall
> > as tedious a review as the command and register tables. And I sure have
> > reviewed a lot of miserable stuff in the past.
> > 
> > Most infuriatingly, I did not find a single real bug in the code!
> > 
> > I think we'll need to automate some things going forward, for example
> > listing the non-conforming length encoding with Damien's tools for cross
> > checking.
> > 
> > There are a few subtle points we need to discuss (separate mails
> > internally) but all in all this series is:
> > 
> > Reviewed-by: Jani Nikula <jani.nikula at intel.com>
> 
> Ok, pulled this one in, thanks a lot for the patches&review. I think it's
> time we start to move on with the next bits, the batch copy stuff seams
> like a suitable piece. There's still issues with launching the entire
> thing in the end, but we can start with the copy infrastructure.
> 
> Open issues I see still:
> 
> - The littel issue we're discussing internally. Like I've said that one is
>   blocking us and needs to be resolved before we can switch to enforcing
>   mode.
> 
> - Secure batch dispatch is still fubar.

I'm not sure that this will still impact us once we implement the batch copy
step. I was only using the secure dispatch stuff because it was a convenient
way to get the batch into GGTT. But with the copy step, we could just have
separate code to do that.

> 
> - CodingStyle says: Functions should be a) at most 3 indent levels b) at
>   most 3 ansi screens long (i.e. 75 lines). i915_parse_cmds violates both
>   metrics pretty deftly. I think a few refactoring patches to extract
>   helper functions and structure the flow a bit would be good.

:)

I'll start with a patch to move all of the actual checking logic into a
separate function, with maybe an extra helper for the bitmask checks. That
seems like it should cut the size down sufficiently.

Brad

> 
> Cheers, Daniel
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch