All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Imre Deak <imre.deak@intel.com>
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH 18/23] drm/i915: Shrink the size of intel_remapped_plane_info struct
Date: Fri, 12 Mar 2021 20:09:16 +0200	[thread overview]
Message-ID: <YEuuTBaP2IUCWPp8@intel.com> (raw)
In-Reply-To: <20210311221942.GA3107177@ideak-desk.fi.intel.com>

On Fri, Mar 12, 2021 at 12:19:42AM +0200, Imre Deak wrote:
> On Thu, Mar 11, 2021 at 09:45:14PM +0200, Ville Syrjälä wrote:
> > On Thu, Mar 11, 2021 at 12:17:31AM +0200, Imre Deak wrote:
> > > Save some place in the GTT VMAs by using a u16 instead of unsigned int
> > > to store the view dimensions. The maximum FB stride is 256kB which is
> > > 4096 tiles in the worst case (yf-tiles), the maximum FB height is 16k
> > > pixels, which is 2048 tiles in the worst case (x-tiles).
> > 
> > Actually I think the worst case for height is remapping linear fbs
> > since we more or less treat it as 4kx1 tiles. But 16k is still< 64k
> > so should be all good.
> 
> Yes, thanks for catching that. Will fix the commit message.
> 
> > Integer promotion stuff/etc. is what worried me the most here, but
> > looks like rotate_pages()/remap_pages() at least gets everything
> > passed in as unsigned int, so we're not in danger of sign bit
> > shenanigans there at least.
> 
> Yes. Fwiw I can think only of the following kind of sign-extension
> problem scenario with u16:
> 
> u16 v1=-1;
> int i=v1;
> 
> So if u16 stored a negative result, we'd miss the expected
> sign-extension, but I can't see a way we wanted to store a negative
> value to these fields. So for instance in remap_pages() the
> 
> 	offset += src_stride - width;
> 
> would be still correct even if src_stride/width would be u16 and
> src_stride was less than width.

I suppose most issues would be along the lines of expecting
u16-u16 to produce a mod u16 result. But since things get
promoted to signed int that is no longer true when the result
is negative. Also if you then cast that result to some bigger
type you can get even more sign extension happening.

-- 
Ville Syrjälä
Intel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  reply	other threads:[~2021-03-12 18:09 UTC|newest]

Thread overview: 65+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-10 22:17 [Intel-gfx] [PATCH 00/23] drm/i915: Add support for FBs requiring a POT stride padding Imre Deak
2021-03-10 22:17 ` [Intel-gfx] [PATCH 01/23] drm/i915: Fix rotation setup during plane HW readout Imre Deak
2021-03-11 16:04   ` Ville Syrjälä
2021-03-11 16:52     ` Imre Deak
2021-03-11 17:25       ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 02/23] drm/i915/selftest: Fix error handling in igt_vma_remapped_gtt() Imre Deak
2021-03-11 16:05   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 03/23] drm/i915/selftest: Fix debug message " Imre Deak
2021-03-11 16:06   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 04/23] drm/i915: Make sure i915_ggtt_view is inited when creating an FB Imre Deak
2021-03-11 16:07   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 05/23] drm/i915/selftest: Make sure to init i915_ggtt_view in igt_vma_rotate_remap() Imre Deak
2021-03-11 16:11   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 06/23] drm/i915: Remove duplicate intel_surf_alignment() declaration Imre Deak
2021-03-11 16:12   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 07/23] drm/i915/intel_fb: Pull FB plane functions from intel_display_types.h Imre Deak
2021-03-11 16:15   ` Ville Syrjälä
2021-03-11 16:31     ` Imre Deak
2021-03-10 22:17 ` [Intel-gfx] [PATCH 08/23] drm/i915/intel_fb: Pull FB plane functions from skl_universal_plane.c Imre Deak
2021-03-11 16:18   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 09/23] drm/i915/intel_fb: Pull is_surface_linear() from intel_display.c/skl_universal_plane.c Imre Deak
2021-03-11 16:19   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 10/23] drm/i915/intel_fb: Pull FB plane functions from intel_sprite.c Imre Deak
2021-03-11 16:20   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 11/23] drm/i915/intel_fb: Pull FB plane functions from intel_display.c Imre Deak
2021-03-11 16:23   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 12/23] drm/i915/intel_fb: Unexport intel_fb_check_stride() Imre Deak
2021-03-11 16:23   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 13/23] drm/i915/intel_fb: s/dev_priv/i915/ Imre Deak
2021-03-11 16:23   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 14/23] drm/i915/intel_fb: Factor out convert_plane_offset_to_xy() Imre Deak
2021-03-11 16:32   ` Ville Syrjälä
2021-03-11 16:37     ` Ville Syrjälä
2021-03-11 16:57     ` Imre Deak
2021-03-10 22:17 ` [Intel-gfx] [PATCH 15/23] drm/i915/intel_fb: Factor out calc_plane_aligned_offset() Imre Deak
2021-03-11 16:39   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 16/23] drm/i915/intel_fb: Factor out calc_plane_normal_size() Imre Deak
2021-03-11 16:52   ` Ville Syrjälä
2021-03-11 17:02     ` Imre Deak
2021-03-11 17:26       ` Ville Syrjälä
2021-03-11 17:47         ` Imre Deak
2021-03-11 17:58           ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 17/23] drm/i915/intel_fb: Factor out plane_calc_remap_info() Imre Deak
2021-03-11 17:21   ` Ville Syrjälä
2021-03-11 19:04     ` Imre Deak
2021-03-11 19:35       ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 18/23] drm/i915: Shrink the size of intel_remapped_plane_info struct Imre Deak
2021-03-11 19:45   ` Ville Syrjälä
2021-03-11 22:19     ` Imre Deak
2021-03-12 18:09       ` Ville Syrjälä [this message]
2021-03-10 22:17 ` [Intel-gfx] [PATCH 19/23] drm/i915/selftest: Unify use of intel_remapped_plane_info in igt_vma_rotate_remap() Imre Deak
2021-03-11 21:17   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 20/23] drm/i915: s/stride/src_stride/ in the intel_remapped_plane_info struct Imre Deak
2021-03-12 17:51   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 21/23] drm/i915: Add support for FBs requiring a POT stride alignment Imre Deak
2021-03-12 18:02   ` Ville Syrjälä
2021-03-13 14:36     ` Imre Deak
2021-03-15 14:44       ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 22/23] drm/i915/selftest: Add remap/rotate vma subtests when dst_stride!=width/height Imre Deak
2021-03-12 18:03   ` Ville Syrjälä
2021-03-10 22:17 ` [Intel-gfx] [PATCH 23/23] drm/i915: For-CI: Force remapping the FB with a POT aligned stride Imre Deak
2021-03-10 23:53 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Add support for FBs requiring a POT stride padding Patchwork
2021-03-10 23:54 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-03-11  0:22 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork
2021-03-11  2:19 [Intel-gfx] [PATCH 18/23] drm/i915: Shrink the size of intel_remapped_plane_info struct kernel test robot

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YEuuTBaP2IUCWPp8@intel.com \
    --to=ville.syrjala@linux.intel.com \
    --cc=imre.deak@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.