[Mesa-dev] [PATCH 00/13] i965: Make LOAD_PAYLOAD sane again

Kenneth Graunke kenneth at whitecape.org
Tue May 5 12:18:30 PDT 2015

On Wednesday, April 01, 2015 06:19:11 PM Jason Ekstrand wrote:
> This patch series comes with two apologies.  First, is for the fact that
> most of it has to be squashed into a single super-commit.  Unfortunately,
> this is one of those things that can't really be done incrementally.
> Second, is for the absolute disaster that I left LOAD_PAYLOAD in at the end
> of my send-from-GRF series.  That was the first time I had ever touched the
> compiler and I didn't know what I was doing.
> This series makes LOAD_PAYLOAD far more sane.  For details on how the new
> LOAD_PAYLOAD works, see patch 09/13.  Also, It's less code. :-)
> Jason Ekstrand (13):
>   i965/fs_cse: Factor out code to create copy instructions
>   i965: Change header_present to header_size in backend_instruction
>   i965/register_coalesce: Do register size checks before is_copy_payload

The first three are:

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

>   i965/fs_inst: Add an is_copy_payload helper

I requested changes here.

>   i965/fs: Make emit_single_fb_write take an explicit exec_size
>   i965/fs: Make LOAD_PAYLOAD take a header size
>   i965/fs: Rework fs_visitor::LOAD_PAYLOAD
>   SQUASH: i965/fs: Make destinations of load_payload have the
>     appropreate width
>   SQUASH: i965/fs: Rework fs_visitor::lower_load_payload
>   SQUASH: i965/fs_cse: Support the new-style LOAD_PAYLOAD

These five are:

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

(this seems much better than what we have before - we may as well do
this for now.  we can always change it later)

>   SQUASH: i965/fs_inst::is_copy_payload: Support the new-style

Not sure about this one.

>   SQUASH: i965/fs: Set up fb-write payloads directly
>   i965/fs_inst: Get rid of the effective_width field

These two are:
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

I'm not really sure why the FB write patch would need to be squashed.
It seems like a cleanup that follows from your LOAD_PAYLOAD series, but
shouldn't be strictly required...if you don't have to squash it, please
let it stand on its own.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20150505/76ac5bea/attachment.sig>

More information about the mesa-dev mailing list