Re: Record activity performance
Just so everyone knows, there have recently been major speedups in the Theora encoder (libtheora), which is a bottleneck for Record performance. These speedups are available in version 1.2-alpha1, which (despite the name) is the recommended version for all but the most conservative applications. It has zero known bugs and is ABI-compatible with all preceding versions. http://downloads.xiph.org/releases/theora/libtheora-1.2.0alpha1.tar.gz I don't know how to get new versions of things into OLPC builds, but if people are interested in Record performance this is probably worthwhile. Version 1.2alpha1 at speed-level 2 should be faster and better than what is currently in use. --Ben signature.asc Description: OpenPGP digital signature ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: suppressing mesh
after testing the patch below i've opened d.l.o #10579 to track this. a revised patch is attached to that ticket. i can commit/push this change if it's wanted. i don't know when it would get into a release, but the kernel rpm would be available shortly after. the revised patch adds a libertas_disablemesh parameter, which, if set, permanently disables until the module is reloaded without it. paul paul wrote: i wrote: martin wrote: On Fri, Dec 17, 2010 at 6:15 PM, Martin Abente martin.abente.lah...@gmail.com wrote: I wrote this script that is a little bit better than just waiting N seconds, and it seems to work fine: ... great! But I like Martin's idea better, just not sure how to make it work atm. embed exactly that chunk of code you wrote at the end of the function I linked to earlier. I'm 99.9% sure it'll just work. If/when it does, we make it conditional on a config option and pester pgf to include it an release. this actually sounds like a perfect use for the resume script hook to me, since they're already spawned into the background. is mesh off a general problem? so it seems that being able to permanently suppress mesh on the XO-1 is becoming a desired feature, since ad-hoc is the way forward if mixing XO-1/1.5 in under a tree sharing. but powerd isn't the right place for fixing this. could others who've been inside libertas take a look at this small patch? it adds a module parameter that would allow keeping mesh disabled on card discovery, but also allow enabling it later on if desired. (uncompiled/untested) paul diff --git a/drivers/net/wireless/libertas/main.c b/drivers/net/wireless/libertas/main.c index 3f81289..f355b6a 100644 --- a/drivers/net/wireless/libertas/main.c +++ b/drivers/net/wireless/libertas/main.c @@ -39,6 +39,10 @@ unsigned int lbs_debug; EXPORT_SYMBOL_GPL(lbs_debug); module_param_named(libertas_debug, lbs_debug, int, 0644); +unsigned int lbs_startmesh = 1; +EXPORT_SYMBOL_GPL(lbs_startmesh); +module_param_named(libertas_startmesh, lbs_startmesh, int, 0644); + /* This global structure is used to send the confirm_sleep command as * fast as possible down to the firmware. */ @@ -1347,7 +1351,10 @@ int lbs_start_card(struct lbs_private *priv) /* Check mesh FW version and appropriately send the mesh start * command */ -if (priv-mesh_fw_ver == MESH_FW_OLD) { +if (!lbs_startmesh) { +priv-mesh_tlv = 0; + +} if (priv-mesh_fw_ver == MESH_FW_OLD) { /* Enable mesh, if supported, and work out which TLV it uses. 0x100 + 291 is an unofficial value used in 5.110.20.pXX 0x100 + 37 is the official value used in 5.110.21.pXX =- paul fox, p...@laptop.org ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel =- paul fox, p...@laptop.org ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Record activity performance
On 23 December 2010 17:10, James Cameron qu...@laptop.org wrote: On Thu, Dec 23, 2010 at 04:53:26PM +1100, Sridhar Dhanapalan wrote: Firstly, I tried Record after having used the XO with various activities (Musicpainter, Image Thumbnail, etc.) and closed them (leaving nothing else running). Recording in Low for 2 min appeared to work, but playback gave no sound, and it paused after about 10 seconds. That's interesting. I've not heard about that. What version of those activities? Where do you get Musicpainter from? I can't see it at activities.sugarlabs.org. Nor can I see anything called Thumbnail. If these activities change the mixer controls at all then this may lead to the no sound symptom you describe. Musicpainter and Image Thumbnail are new activities created by us, which we haven't yet fully released (I haven't submitted them to ASLO yet). I'm now suspecting that the Musicpainter is the culprit, so that's our problem to investigate. The project pages for those activities are: https://dev.laptop.org.au/projects/musicpainter https://dev.laptop.org.au/projects/image-thumbnail Restarting the XO, I ran Record again. Playback of the video was fine. Then I recorded a 2 min video in High quality. Fast movement, etc. seemed okay until the final ~30 seconds. At this point, the screen updated very slowly. I reduced the image complexity so that it was able to catch up by the end. Playing back the High quality video gave me what I had seen on the screen when recording, i.e. it was fine until about 30 secs remaining. Then the frame rate slowed to a crawl. Audio continued to be smooth despite the video. The video was fine by the final few seconds, in sync with the audio. We're tracking this here: https://dev.laptop.org.au/issues/86 Does anyone have any ideas on what can be done to address this? Everything else you said above, apart from the no sound symptom, matches what we placed in the release notes for 10.1.2. http://wiki.laptop.org/go/Release_notes/10.1.2#Record_activity_video_recording_may_pause Reading that again, yes it does match. I didn't think the scene was very complex, but I think the last sentence adequately explains the behaviour: In this release the frame rate is kept as fast as possible until the pause is unavoidable. What class SD card are you using? Class 2. I've been informed that Quanta do not offer anything faster, as none meet durability requirements. Since the version of Record has not changed between 10.1.2 and 10.1.3, it is expected that the same problem will affect 10.1.3. It sounds like High quality may need removing temporarily if this problem isn't fixed. I was thinking that. Unfortunately this bug makes Record look 'broken' to a lot of people, and it doesn't create a good perception of the XOs. I was thinking of releasing our own version with High disabled in hw.py (just as Best is), but thought it better to discuss here first. Cheers, Sridhar ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Installing 10.1.3 to SD Card/USB Stick
On Thu, Dec 23, 2010 at 06:09:19PM -0500, Chris Marshall wrote: I would like to try installing 10.1.3 to an external SD card (the slot) or a USB pen drive. The directions from the wiki: http://wiki.laptop.org/go/Installing_to_SD_Card say to use the ext3 image format file but there don't appear to be any of those being generated with the candidate osXXX.img files. Yes, the 10.1.x series no longer supports this method. So I've marked the page deprecated. Are the .usb files a new name for the ext3 image files? No, the .usb files are for olpc-update using USB drive. If you would like to try installing to an external SD card on an XO-1, the method would involve: 1. obtaining the osXXX.tree.tar.lzma file, 2. creating a filesystem, 3. unpacking the tree file, 4. preparing an olpc.fth for booting from external SD, 5. boot. Please update the Wiki page if you get this working. It isn't a focus of the development effort at the moment. -- James Cameron http://quozl.linux.org.au/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Record activity performance
On Fri, Dec 24, 2010 at 10:36:56AM +1100, Sridhar Dhanapalan wrote: The project pages for those activities are: https://dev.laptop.org.au/projects/musicpainter https://dev.laptop.org.au/projects/image-thumbnail Any chance of supporting http:// as well as https:// ? The extra few packets required by the SSL negotiation combined with rural and remote network latency makes visiting these URLs painful. I was thinking that. Unfortunately this bug makes Record look 'broken' to a lot of people, and it doesn't create a good perception of the XOs. I was thinking of releasing our own version with High disabled in hw.py (just as Best is), but thought it better to discuss here first. Sounds like a reasonable idea, unless you see some development happening. -- James Cameron http://quozl.linux.org.au/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Installing 10.1.3 to SD Card/USB Stick
On Thu, Dec 23, 2010 at 4:07 PM, James Cameron qu...@laptop.org wrote: On Thu, Dec 23, 2010 at 06:09:19PM -0500, Chris Marshall wrote: I would like to try installing 10.1.3 to an external SD card (the slot) or a USB pen drive. The directions from the wiki: http://wiki.laptop.org/go/Installing_to_SD_Card say to use the ext3 image format file but there don't appear to be any of those being generated with the candidate osXXX.img files. *snip* If you would like to try installing to an external SD card on an XO-1, the method would involve: My quick and dirty notes I have from the last time I did this are as follow. Note this needs to be done on a machine with enough free RAM to create an appropriately sized ramdisk 1) get the jffs2 image 2) sudo modprobe -v mtd 3) sudo modprobe -v jffs2 4) sudo modprobe -v mtdram total_size=70 erase_size=25 (note this just needs to be as large as the image. I have 8GB's of RAM so I don't skimp) 5) sudo modprobe -v mtdchar 6) sudo modprobe -v mtdblock ( 5 and 6 might not be needed. I have them *'d but didn't write myself a note why :-( ) 7) sudo dd if=imagename.jffs2 of=/dev/mtd0 8) sudo mount -t jffs2 /dev/mtdblock0 /where/the/image/shouldbe/mounted 9) partition your SD card 10) tar over the file files from the ramdisk image to the sd card. After this there may be some configuration changes however I did not document them. If you run into problems let me know and I can put together more complete instructions. -Jon ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: Record activity performance
On 24 December 2010 11:12, James Cameron qu...@laptop.org wrote: On Fri, Dec 24, 2010 at 10:36:56AM +1100, Sridhar Dhanapalan wrote: The project pages for those activities are: https://dev.laptop.org.au/projects/musicpainter https://dev.laptop.org.au/projects/image-thumbnail Any chance of supporting http:// as well as https:// ? The extra few packets required by the SSL negotiation combined with rural and remote network latency makes visiting these URLs painful. Interesting, I didn't consider that. We actually forced https by default because we thought users would appreciate some enforced security. I'll see what we can do. Sridhar ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel