From jon at restlesslemon.co.uk Mon Jan 28 16:11:56 2008 From: jon at restlesslemon.co.uk (Jon Senior) Date: Tue Jan 29 09:26:17 2008 Subject: [Savage40] xv Overlay on crtc2 on SuperSavage (T23) / Argyll display calibration (Resend) Message-ID: <20080129011156.ece1148a.jon@restlesslemon.co.uk> Resent because I wasn't a member of the list last time so couldn't be sure if it was blocked (checking the archives didn't reveal anything coming up). ------------ I think the title says it all. ;-) Looking through the archives I notice that this got a mention as a known issue, but web searches only seem to produce hits dating from 2004/2005. Being as we are now in 2008 and this a device likely to lose rather than gain coders, I was wondering what the problem actually was. The problem in question is that (as best as I can tell) it's not been possible to enable the overlay in crtc2. There is also a related issue in that in crtc1 with working overlay, video with a horizontal resolution greater than 800 pixels (IIRC) shows pink gradient bars from the 801st pixel onwards. This is hard to demonstrate as screen grabs using xine / player show the correct video. A final query: I was using the Argyll colour management suite to calibrate the internal display (It's not very good, but it's now a calibrated not-very-good!). Having added a second display (external CRT) and set up for xinerama calibration no longer works. I can run the software and generate profiles and using dispwin to load the profiles doesn't return any errors, but nothing actually changes on either display. Does xinerama on SuperSavage do something strange to the video LUTs? I ask because I got a similar effect when connecting an external display and using Fn-F7 to switch. The first press changed nothing, but unloaded the calibration settings. Thanks in advance. Edit (For resend): I had problems with the scrolling hang in xinerama, but adding Option "ShadowStatus" "true" didn't seem to solve it. Interestingly, adding that option to both Device sections killed xinerama and left me with only the external CRT which has perfect xv. I assumed the crtc's would be hardwired. If this is not the case, is there a way to specify which one gets used where? Is it just screen ordering? I could handle only having xv on the CRT when both are enabled. :-) -- Jon Senior From timr at probo.com Tue Jan 29 10:23:05 2008 From: timr at probo.com (Tim Roberts) Date: Tue Jan 29 10:23:09 2008 Subject: [Savage40] xv Overlay on crtc2 on SuperSavage (T23) / Argyll display calibration (Resend) In-Reply-To: <20080129011156.ece1148a.jon@restlesslemon.co.uk> References: <20080129011156.ece1148a.jon@restlesslemon.co.uk> Message-ID: <479F6F09.5000407@probo.com> Jon Senior wrote: > Resent because I wasn't a member of the list last time so couldn't be sure if it was blocked (checking the archives didn't reveal anything coming up). > Yes, I long ago configured the list to unceremoniously and silently discard all messages from non-members. Spam management was just too painful. > I think the title says it all. ;-) Looking through the archives I notice that this got a mention as a known issue, but web searches only seem to produce hits dating from 2004/2005. Being as we are now in 2008 and this a device likely to lose rather than gain coders, I was wondering what the problem actually was. > I'm not sure there are *any* active coders left. This mailing list had more than 1,000 subscribers at its peak, but it's down below 450 now. The crtc2 support ("duo view") was added as a labor of love by Alex Deucher. If he's listening, he may be able to address this. If not, you may be out of luck. The Savage graphics engine is celebrating the end of its first decade this year. In human terms, that makes it 150 years old... > There is also a related issue in that in crtc1 with working overlay, video with a horizontal resolution greater than 800 pixels (IIRC) shows pink gradient bars from the 801st pixel onwards. This is hard to demonstrate as screen grabs using xine / player show the correct video. > Does this depend on resolution or refresh rate? There are a billion weird memory timing parameters in the Savage relating to memory access for the streams engine, but I thought all the common cases had been ironed out long ago. -- Tim Roberts, timr@probo.com Providenza & Boekelheide, Inc. From agd5f at yahoo.com Tue Jan 29 11:42:29 2008 From: agd5f at yahoo.com (Alex Deucher) Date: Tue Jan 29 11:42:32 2008 Subject: [Savage40] xv Overlay on crtc2 on SuperSavage (T23) / Argyll display calibration (Resend) In-Reply-To: <20080129011156.ece1148a.jon@restlesslemon.co.uk> Message-ID: <323779.36185.qm@web39611.mail.mud.yahoo.com> --- Jon Senior wrote: > Resent because I wasn't a member of the list last time so couldn't be > sure if it was blocked (checking the archives didn't reveal anything > coming up). > > ------------ > > I think the title says it all. ;-) Looking through the archives I > notice that this got a mention as a known issue, but web searches > only seem to produce hits dating from 2004/2005. Being as we are now > in 2008 and this a device likely to lose rather than gain coders, I > was wondering what the problem actually was. > > The problem in question is that (as best as I can tell) it's not been > possible to enable the overlay in crtc2. > The mobile savage chips have 2 overlays, one per crtc, and you can even source both of them to crtc 0. However, I was never able to get the second overlay working. IIRC, the code for the second overlay is in the driver, I was just never able to get it to work so it's disabled. I'm not sure what magic incantation was needed. > There is also a related issue in that in crtc1 with working overlay, > video with a horizontal resolution greater than 800 pixels (IIRC) > shows pink gradient bars from the 801st pixel onwards. This is hard > to demonstrate as screen grabs using xine / player show the correct > video. I recall similar bugs. I'm not sure what the problem is. The savage overlay was one of the most hateful blocks I've ever dealt with. Whether or not it would work was often luck. I suspect there are quite a few hw errata at work. > > A final query: I was using the Argyll colour management suite to > calibrate the internal display (It's not very good, but it's now a > calibrated not-very-good!). Having added a second display (external > CRT) and set up for xinerama calibration no longer works. I can run > the software and generate profiles and using dispwin to load the > profiles doesn't return any errors, but nothing actually changes on > either display. Does xinerama on SuperSavage do something strange to > the video LUTs? I ask because I got a similar effect when connecting > an external display and using Fn-F7 to switch. The first press > changed nothing, but unloaded the calibration settings. > Due to the way the xserver does gamma correction, it only works in depth 24, not depth 16. This was a software limitation, and could probably have been fixed in the driver, but no one ever got to it. > Thanks in advance. > > Edit (For resend): I had problems with the scrolling hang in > xinerama, but adding Option "ShadowStatus" "true" didn't seem to > solve it. Interestingly, adding that option to both Device sections > killed xinerama and left me with only the external CRT which has > perfect xv. I assumed the crtc's would be hardwired. If this is not > the case, is there a way to specify which one gets used where? Is it > just screen ordering? I could handle only having xv on the CRT when > both are enabled. :-) the crtcs can be routed to any output (fp, dac1, dac2), however the vbe calls we use to program the crtcs hardcode them (crtc0->lcd, crtc1->crt). I had planned to rewrite the driver to program the crtcs and outputs directly, but never had the time. when you only use one head, it's always driven by crtc0. Alex ____________________________________________________________________________________ Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping