drm/i915/pch: Fix integer math bugs in panel fitting
Subsystems (2)
PATCH 1/1 drm/i915/pch: Fix integer math bugs in panel fitting
Consider a 1600x900 panel, upscaling a 1360x768 mode, full-aspect. The old math would give you:
scaled_width = 1600 * 768; /* 1228800 */
scaled_height = 1360 * 900; /* 1224000 */
if (scaled_width > scaled_height) { /* pillarbox, and true */
width = 1224000 / 768; /* int(1593.75) = 1593 */
x = (1600 - 1593 + 1) / 2; /* 4 */
y = 0;
height = 768;
} /* ... */This is broken. The total width of scanout would then be 1593 + 4 + 4, or 1601, which is wider than the panel itself. The hardware very dutifully implements this, and you end up with a black 45° diagonal from the top-left corner to the bottom edge of the screen. It's a cool effect and all, but not what you wanted. Similar things happen for the letterbox case.
The problem is that you have an integer number of pixels, which means it's usually impossible to upscale equally on both axes. 1360/768 is 1.7708, 1600/900 is 1.7777. Since we're constrained on the one axis, the other one wants to come out as an even number of pixels (the panel is almost certainly even on both axes, and the x/y offsets will be applied on both sides). In the math above, if 'width' comes out even, rounding down is correct; if it's odd, you'd rather round up. So just increment width/height in those cases.
Tested on a Lenovo T500 (Ironlake).
--- drivers/gpu/drm/i915/intel_panel.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
drivers/gpu/drm/i915/intel_panel.c +4 -0
@@ -83,11 +83,15 @@ intel_pch_panel_fitting(struct drm_device *dev,u32 scaled_height = mode->hdisplay * adjusted_mode->vdisplay;if (scaled_width > scaled_height) {/* pillar */ width = scaled_height / mode->vdisplay; +if (width &1 ) + width++; x = (adjusted_mode->hdisplay - width +1 ) /2 ; y =0 ; height = adjusted_mode->vdisplay; }else if (scaled_width < scaled_height) {/* letter */ height = scaled_width / mode->hdisplay; +if (height &1 ) + height++; y = (adjusted_mode->vdisplay - height +1 ) /2 ; x =0 ; width = adjusted_mode->hdisplay;
Signature · 6 lines
1.7.6 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
drivers/gpu/drm/i915/intel_panel.c intel_pch_panel_fitting(struct drm_device *dev, · line 94
Quoted 6 lines · Adam Jackson
drivers/gpu/drm/i915/intel_panel.c +4
-0+if (width &1 ) + width++; +if (height &1 ) + height++;
You'll want to stick a comment in here describing why this is here. Otherwise, someone will just break it in the future.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 45 lines · Adam Jackson
Consider a 1600x900 panel, upscaling a 1360x768 mode, full-aspect. The old math would give you:
scaled_width = 1600 * 768; /* 1228800 */ scaled_height = 1360 * 900; /* 1224000 */ if (scaled_width > scaled_height) { /* pillarbox, and true */ width = 1224000 / 768; /* int(1593.75) = 1593 */ x = (1600 - 1593 + 1) / 2; /* 4 */ y = 0; height = 768; } /* ... */This is broken. The total width of scanout would then be 1593 + 4 + 4, or 1601, which is wider than the panel itself. The hardware very dutifully implements this, and you end up with a black 45° diagonal from the top-left corner to the bottom edge of the screen. It's a cool effect and all, but not what you wanted. Similar things happen for the letterbox case.
The problem is that you have an integer number of pixels, which means it's usually impossible to upscale equally on both axes. 1360/768 is 1.7708, 1600/900 is 1.7777. Since we're constrained on the one axis, the other one wants to come out as an even number of pixels (the panel is almost certainly even on both axes, and the x/y offsets will be applied on both sides). In the math above, if 'width' comes out even, rounding down is correct; if it's odd, you'd rather round up. So just increment width/height in those cases.
Tested on a Lenovo T500 (Ironlake).
Signed-off-by Adam Jackson <[email protected]>--- drivers/gpu/drm/i915/intel_panel.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)drivers/gpu/drm/i915/intel_panel.c +2
-0@@ -83,11 +83,15 @@ intel_pch_panel_fitting(struct drm_device *dev,u32 scaled_height = mode->hdisplay * adjusted_mode->vdisplay;if (scaled_width > scaled_height) {/* pillar */ width = scaled_height / mode->vdisplay; +if (width &1 ) + width++;
drivers/gpu/drm/i915/intel_panel.c intel_pch_panel_fitting(struct drm_device *dev, · line 97
Quoted 10 lines
drivers/gpu/drm/i915/intel_panel.c +2
-0x = (adjusted_mode->hdisplay - width +1 ) /2 ; y =0 ; height = adjusted_mode->vdisplay; }else if (scaled_width < scaled_height) {/* letter */ height = scaled_width / mode->hdisplay; +if (height &1 ) + height++; y = (adjusted_mode->vdisplay - height +1 ) /2 ; x =0 ; width = adjusted_mode->hdisplay;
We've seen this in Ubuntu in the form of
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/753994 so we can add a
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 12 lines · Adam Jackson
Consider a 1600x900 panel, upscaling a 1360x768 mode, full-aspect. The old math would give you:
scaled_width = 1600 * 768; /* 1228800 */ scaled_height = 1360 * 900; /* 1224000 */ if (scaled_width > scaled_height) { /* pillarbox, and true */ width = 1224000 / 768; /* int(1593.75) = 1593 */ x = (1600 - 1593 + 1) / 2; /* 4 */ y = 0; height = 768; } /* ... */
Quoted 6 lines
This is broken. The total width of scanout would then be 1593 + 4 + 4, or 1601, which is wider than the panel itself. The hardware very dutifully implements this, and you end up with a black 45° diagonal from the top-left corner to the bottom edge of the screen. It's a cool effect and all, but not what you wanted. Similar things happen for the letterbox case.
I'm going to show my ignorance but why did the hw apply the offset to both sides of the output and generate a new mode line? That's probably inherent in the nature of the hblanks, I guess? The entirety of the description I have is that the position is simply the origin of the top-left pixel of the fitted window.
So a comment here is necessary, but otherwise:
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=38851
And a big thank you.
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx