From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
To: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Subject: [PATCH 2/5] drm/rect: Handle rounding errors in drm_rect_clip_scaled, v2.
Date: Mon, 30 Apr 2018 15:46:08 +0200 [thread overview]
Message-ID: <20180430134611.50988-3-maarten.lankhorst@linux.intel.com> (raw)
In-Reply-To: <20180430134611.50988-1-maarten.lankhorst@linux.intel.com>
No matter how you perform the clip adjustments, a small
error may push the scaling factor to the other side of
0x10000. Solve this with a macro that will fixup the
scale to 0x10000 if we accidentally wrap to the other side.
Changes since v1:
- Adjust dst immediately, else drm_rect_width/height on dst gives bogus
results.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
---
drivers/gpu/drm/drm_rect.c | 71 ++++++++++++++++++++++++++++----------
1 file changed, 52 insertions(+), 19 deletions(-)
diff --git a/drivers/gpu/drm/drm_rect.c b/drivers/gpu/drm/drm_rect.c
index b179c7c73cc5..9d6194780a74 100644
--- a/drivers/gpu/drm/drm_rect.c
+++ b/drivers/gpu/drm/drm_rect.c
@@ -50,6 +50,24 @@ bool drm_rect_intersect(struct drm_rect *r1, const struct drm_rect *r2)
}
EXPORT_SYMBOL(drm_rect_intersect);
+static int drm_calc_scale(int src, int dst)
+{
+ int scale = 0;
+
+ if (WARN_ON(src < 0 || dst < 0))
+ return -EINVAL;
+
+ if (dst == 0)
+ return 0;
+
+ if (src > (dst << 16))
+ return DIV_ROUND_UP(src, dst);
+ else
+ scale = src / dst;
+
+ return scale;
+}
+
/**
* drm_rect_clip_scaled - perform a scaled clip operation
* @src: source window rectangle
@@ -71,49 +89,64 @@ bool drm_rect_clip_scaled(struct drm_rect *src, struct drm_rect *dst,
{
int diff;
+ /*
+ * When scale is near 1.0 rounding errors may cause the scaling
+ * factor to the other side. Some hardware may support
+ * upsampling, but not downsampling, and that would break when
+ * rounding.
+ */
+#define FIXUP(oldscale, fn, m, second) do { \
+ if (oldscale != 1 << 16) { \
+ int newscale = drm_calc_scale(fn(src), fn(dst)); \
+ \
+ if (newscale < 0) \
+ return false; \
+ \
+ if ((oldscale < 0x10000) != (newscale < 0x10000)) { \
+ if (!second) \
+ src->m##1 = src->m##2 - (fn(dst) << 16); \
+ else \
+ src->m##2 = src->m##1 + (fn(dst) << 16); \
+ } \
+ } \
+ } while (0)
+
diff = clip->x1 - dst->x1;
if (diff > 0) {
int64_t tmp = src->x1 + (int64_t) diff * hscale;
src->x1 = clamp_t(int64_t, tmp, INT_MIN, INT_MAX);
+ dst->x1 = clip->x1;
+ FIXUP(hscale, drm_rect_width, x, 0);
}
+
diff = clip->y1 - dst->y1;
if (diff > 0) {
int64_t tmp = src->y1 + (int64_t) diff * vscale;
src->y1 = clamp_t(int64_t, tmp, INT_MIN, INT_MAX);
+ dst->y1 = clip->y1;
+ FIXUP(vscale, drm_rect_height, y, 0);
}
+
diff = dst->x2 - clip->x2;
if (diff > 0) {
int64_t tmp = src->x2 - (int64_t) diff * hscale;
src->x2 = clamp_t(int64_t, tmp, INT_MIN, INT_MAX);
+ dst->x2 = clip->x2;
+ FIXUP(hscale, drm_rect_width, x, 1);
}
diff = dst->y2 - clip->y2;
if (diff > 0) {
int64_t tmp = src->y2 - (int64_t) diff * vscale;
src->y2 = clamp_t(int64_t, tmp, INT_MIN, INT_MAX);
+ dst->y2 = clip->y2;
+ FIXUP(vscale, drm_rect_height, y, 1);
}
+#undef FIXUP
- return drm_rect_intersect(dst, clip);
+ return drm_rect_visible(dst);
}
EXPORT_SYMBOL(drm_rect_clip_scaled);
-static int drm_calc_scale(int src, int dst)
-{
- int scale = 0;
-
- if (WARN_ON(src < 0 || dst < 0))
- return -EINVAL;
-
- if (dst == 0)
- return 0;
-
- if (src > (dst << 16))
- return DIV_ROUND_UP(src, dst);
- else
- scale = src / dst;
-
- return scale;
-}
-
/**
* drm_rect_calc_hscale - calculate the horizontal scaling factor
* @src: source window rectangle
--
2.17.0
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
next prev parent reply other threads:[~2018-04-30 13:46 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-30 13:46 [PATCH 0/5] drm: Fix rounding errors and use scaling in i915 Maarten Lankhorst
2018-04-30 13:46 ` [PATCH 1/5] drm/rect: Round above 1 << 16 upwards to correct scale calculation functions Maarten Lankhorst
2018-04-30 13:46 ` Maarten Lankhorst [this message]
2018-04-30 13:46 ` [PATCH 3/5] drm/i915: Do not adjust scale when out of bounds, v2 Maarten Lankhorst
2018-04-30 13:46 ` [PATCH 4/5] drm/selftests: Rename the Kconfig option to CONFIG_DRM_DEBUG_SELFTEST Maarten Lankhorst
2018-04-30 13:46 ` [PATCH 5/5] drm/selftests: Add drm helper selftest Maarten Lankhorst
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=20180430134611.50988-3-maarten.lankhorst@linux.intel.com \
--to=maarten.lankhorst@linux.intel.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=vidya.srinivas@intel.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox