resolution independent px
Hi, attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling. Is there a need for something like this? Should it be enabled via config option or should we do it automatically if we detect a DPI value that differs a lot from the standard value? Cheers, Johannes [1] http://www.w3.org/TR/CSS2/syndata.html#x39
On Fri, Jan 16, 2015 at 08:50:13AM +0100, Johannes Hofmann wrote:
Hi,
attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling.
A long time ago, image scaling was relatively low overhead, but since dillo got correct gamma correction, it became much more expensive. I remember testing this and the difference was noticeable. It was left anyway because the amount of scaled images in the average web page is low.
Is there a need for something like this?
I don't know...
Should it be enabled via config option or should we do it automatically if we detect a DPI value that differs a lot from the standard value?
For me, it is hard to say without knowing the main use of it. What I do know is that our scaling is CPU intensive, so I tend to have it off unless necessary. Another way around is to optimize our gamma correction algorithm with integer arithmetic. An option that was not taken at the time because of the above stated reasons. HTH. -- Cheers Jorge.-
On Fr, Jan 16, 2015, Jorge Arellano Cid wrote:
On Fri, Jan 16, 2015 at 08:50:13AM +0100, Johannes Hofmann wrote:
Hi,
attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling.
A long time ago, image scaling was relatively low overhead, but since dillo got correct gamma correction, it became much more expensive.
It is not so much gamma correction, but some improvements before. BTW: if you look at dw/fltkimgbuf.cc, you see that the old fast code is still available, by setting scaleMode to SIMPLE. Perhaps this should be configurable (should be simple), or depending on estimated CPU performance. If you run dillo (3.1-dev, not 3.0.4.1) with the attached page, change the window width, and compare the results with Firefox, you'll notice that Firefox responds a bit faster. I'm not sure whether Firefox scales faster, but in other circumstances I've noticed that Firefox seems to apply two scaling algorithms: one simple but fast one, with higher priority, and a better but slower one, which is applied with lower priority. Sebastian
On Fri, Jan 16, 2015 at 12:14:19PM -0300, Jorge Arellano Cid wrote:
On Fri, Jan 16, 2015 at 08:50:13AM +0100, Johannes Hofmann wrote:
Hi,
attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling.
A long time ago, image scaling was relatively low overhead, but since dillo got correct gamma correction, it became much more expensive.
I remember testing this and the difference was noticeable. It was left anyway because the amount of scaled images in the average web page is low.
Is there a need for something like this?
I don't know...
Should it be enabled via config option or should we do it automatically if we detect a DPI value that differs a lot from the standard value?
For me, it is hard to say without knowing the main use of it. What I do know is that our scaling is CPU intensive, so I tend to have it off unless necessary.
It's only useful if you have a high resolution device where DPI is much higher than about 70. In that case images will be rendered very small. I currently don't have such a device, so personally I don't have any need for this now. I just wanted to get feedback from dillo-dev in case someone already ran into this issue. Cheers, Johannes
Hi Johannes, Reading the link, it looks like normally viewed displays with any resolution between about 50 DPI and 140 DPI will have one CSS pixel exactly equal to one device pixel; so resolutions outside that are the definition of 'differs a lot' in the standard. What's the overhead of asking the library to re-scale the image to the size that it already is? If it's significant, then maybe there needs to be a special case for 1:1. I start out with a bias against adding another run-time option, but it's based on my not-very-relevant experience of adding an option to dpid. Contradicting myself, the standard implies that there could be an option for viewing distance, in which case there could be a special case to that option for 'never scale'. Regards. James. On 1/18/15, Johannes Hofmann <Johannes.Hofmann at gmx.de> wrote:
On Fri, Jan 16, 2015 at 12:14:19PM -0300, Jorge Arellano Cid wrote:
On Fri, Jan 16, 2015 at 08:50:13AM +0100, Johannes Hofmann wrote:
Hi,
attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling.
A long time ago, image scaling was relatively low overhead, but since dillo got correct gamma correction, it became much more expensive.
I remember testing this and the difference was noticeable. It was left anyway because the amount of scaled images in the average web page is low.
Is there a need for something like this?
I don't know...
Should it be enabled via config option or should we do it automatically if we detect a DPI value that differs a lot from the standard value?
For me, it is hard to say without knowing the main use of it. What I do know is that our scaling is CPU intensive, so I tend to have it off unless necessary.
It's only useful if you have a high resolution device where DPI is much higher than about 70. In that case images will be rendered very small.
I currently don't have such a device, so personally I don't have any need for this now. I just wanted to get feedback from dillo-dev in case someone already ran into this issue.
Cheers, Johannes
_______________________________________________ Dillo-dev mailing list Dillo-dev at dillo.org http://lists.dillo.org/cgi-bin/mailman/listinfo/dillo-dev
Hi, On Sun, Jan 18, 2015 at 09:20:06PM +1300, James C wrote:
Hi Johannes,
Reading the link, it looks like normally viewed displays with any resolution between about 50 DPI and 140 DPI will have one CSS pixel exactly equal to one device pixel; so resolutions outside that are the definition of 'differs a lot' in the standard.
What's the overhead of asking the library to re-scale the image to the size that it already is? If it's significant, then maybe there needs to be a special case for 1:1.
the overhead is significant in case images would not be scaled otherwise, so we would want to special case it somehow.
I start out with a bias against adding another run-time option, but it's based on my not-very-relevant experience of adding an option to dpid.
Contradicting myself, the standard implies that there could be an option for viewing distance, in which case there could be a special case to that option for 'never scale'.
As nobody actually has a problem at the moment I think we can postpone this. It will come up again as high resolution devices are becoming more popular. Cheers, Johannes
Regards. James.
On 1/18/15, Johannes Hofmann <Johannes.Hofmann at gmx.de> wrote:
On Fri, Jan 16, 2015 at 12:14:19PM -0300, Jorge Arellano Cid wrote:
On Fri, Jan 16, 2015 at 08:50:13AM +0100, Johannes Hofmann wrote:
Hi,
attached is a patch that would enable resolution independent handling of px as described in [1]. It would result in roughly equally sized images on high and low resolution devices at the cost of a higher overhead for additional image scaling.
A long time ago, image scaling was relatively low overhead, but since dillo got correct gamma correction, it became much more expensive.
I remember testing this and the difference was noticeable. It was left anyway because the amount of scaled images in the average web page is low.
Is there a need for something like this?
I don't know...
Should it be enabled via config option or should we do it automatically if we detect a DPI value that differs a lot from the standard value?
For me, it is hard to say without knowing the main use of it. What I do know is that our scaling is CPU intensive, so I tend to have it off unless necessary.
It's only useful if you have a high resolution device where DPI is much higher than about 70. In that case images will be rendered very small.
I currently don't have such a device, so personally I don't have any need for this now. I just wanted to get feedback from dillo-dev in case someone already ran into this issue.
Cheers, Johannes
_______________________________________________ Dillo-dev mailing list Dillo-dev at dillo.org http://lists.dillo.org/cgi-bin/mailman/listinfo/dillo-dev
_______________________________________________ Dillo-dev mailing list Dillo-dev at dillo.org http://lists.dillo.org/cgi-bin/mailman/listinfo/dillo-dev
Johannes wrote:
As nobody actually has a problem at the moment I think we can postpone this. It will come up again as high resolution devices are becoming more popular.
Speaking of high resolution devices, does it look like there's much value these days in trying to get dillo to work better in the world of touch interfaces, or has dillo been utterly passed by there for one reason or another?
On So, Jan 18, 2015, Johannes Hofmann wrote:
As nobody actually has a problem at the moment I think we can postpone this. It will come up again as high resolution devices are becoming more popular.
Furthermore, if such devices become more common, web designers may become more aware of this problem; something like this ---------------------------------------------------------------------- Some text and an image: <img src="http://www.dillo.org/Icons/ProgramIcon32.png" style="min-height: 1em"> ---------------------------------------------------------------------- would solve the problem, too (test it with different font sizes in dillo 3.1-dev). Sebastian
participants (5)
-
eocene＠gmx.com -
james.from.wellington＠gmail.com -
jcid＠dillo.org -
Johannes.Hofmann＠gmx.de -
sgeerken＠dillo.org