| 09:43 | topi`_ | we have a need for communicating over RS485 ... what are my options if I want to still use the Hummingboard? |
| 09:52 | jnettlet | topi`_, you can do a breakout board that does uart to rs485 |
| 10:52 | topi`_ | jnettlet: I think the best bet is the Edge paired with a RS485 mikroUSB card |
| 10:52 | topi`_ | mikroBUS |
| 10:53 | topi`_ | how can you emulate rs485 with a simple uart? AFAIK the rs485 is mostly for the feature of connecting up to 32 devies on the same wire |
| 10:53 | jnettlet | topi`_, that will work. that most likely runs over the uart wired up to the mikrobus |
| 10:53 | topi`_ | jnettlet: what might be the cause of a (infrequently) occurring issue like this: after U-BOOT, I get just a black screen and nothing is being booted |
| 10:54 | topi`_ | however, the Red leght is on, and Green is blinking |
| 10:54 | topi`_ | I suppose U-BOOT initializes the Eth transceiver? |
| 10:54 | topi`_ | now I press the Reboot switch, and the kernel boots |
| 10:55 | jnettlet | topi`_, you have a display plugged in? |
| 10:55 | topi`_ | I think this only occurs after Ctrl-Alt-Del? |
| 10:55 | topi`_ | yeah :) |
| 10:55 | jnettlet | get the latest u-boot. |
| 10:55 | topi`_ | my uboot is probably from last April |
| 10:55 | jnettlet | yeah, I fixed that problem over the summer |
| 10:56 | topi`_ | is it some kind of race condition? since it doesn't occur on every reboot |
| 10:56 | jnettlet | yeah, they have a while loop in the display code that doesn't have a timeout. |
| 10:56 | jnettlet | u-boot code is sloppy as hell |
| 10:56 | topi`_ | hehe |
| 10:57 | jnettlet | screw booting the device which is our primary function...we are going to spin in a while loop until this display is ready damnit |
| 10:58 | topi`_ | I reckon the HDMI display detection stuff is quite hairy... maybe they just don't get it right every time? |
| 11:01 | jnettlet | it is u-boot! :) That is why I like coreboot. Google has the right idea. Don't build a second set of drivers, build just enough to boot a kernel. |
| 11:01 | jnettlet | then run a recovery initramfs if needed |
| 11:01 | jnettlet | write one set of drivers, debug one set of drivers for everything but the lowest level stuff. |
| 11:01 | topi`_ | there's also the LittleKernel (LK) but I think it's biased towards Snapdragon |
| 11:03 | jnettlet | if you strip down a linux kernel you can get to an initramfs prompt in around 1 second |
| 11:04 | topi`_ | what about using that kernel to boot another linux kernel? |
| 11:04 | jnettlet | yes it is supported |
| 11:04 | topi`_ | then your bootloader would just essentially be a linux kernel + initramfs |
| 11:05 | jnettlet | well your bootloader, would be what it is supposed to be a bootloader :) |
| 11:05 | jnettlet | then you are using linux to configure everything |
| 11:09 | topi`_ | I think the Linux kernel can be started just by setupping DRAM and setting the cpu to the right mode |
| 11:10 | topi`_ | of course, you need a way to read the kernel off of the media somehow, so configuring mmc is needed as well |
| 11:10 | topi`_ | unless your board has heaploads of NOR :) |
| 11:15 | jnettlet | that is what my SPL fastboot does. SPL loads into SRAM, setup dram, mmc, clocks, load zImage and dtb then jump to linux. |
| 11:15 | jnettlet | takes .3 seconds...which is basically limited by the SDHC read speeds |
| 11:21 | topi`_ | I have a Utilite2 here with Snapdragon 600, and booting it with the LK takes ages (around 7 seconds) |
| 11:22 | topi`_ | after 7 sec, the penguins start to appear on screen |
| 11:25 | vpeter | jnettlet: I just need to replace gpu-viv-bin-mx6q and gpu-viv-g2d with files from imx-gpu-viv-5.0.11.p4.5-hfp got it in yocto fido? |
| 11:27 | jnettlet | vpeter, yes |
| 11:28 | vpeter | From what I see I can also strip libraries. |
| 11:32 | jnettlet | I yes |
| 12:40 | jnettlet | mk01, vpeter, you will be interested to know that I have found the GPU actually runs non-optimally at the 396Mhz clocking. If you provide more voltage to the SOC then you get a modest performance bump, but setting the cpu min frequency to 798Mhz gives you a whole new level of performance. |
| 12:41 | jnettlet | I am going to add an option so we can choose to ramp up the min cpufrequency when using the GPU and dropping it back down then everything is idle |
| 13:31 | vpeter | jnettlet: You must be a wizard :) |
| 19:14 | Exaga | so what's the latest and greatest? |
| 19:14 | Exaga | rabeeh: are you around? |
| 19:17 | jnettlet | Exaga, what do you want do do? rabeeh is probably getting ready for the weekend. |
| 19:17 | Exaga | jnettlet: heya :) |
| 19:17 | jnettlet | evening |
| 19:17 | Exaga | jnettlet: just catching up really. I've noticed a few devices appearing that have similar features to the hummingboard |
| 19:18 | Exaga | the orange pi plus being one of them |
| 19:18 | Exaga | just wondering what solidrun are up to lately |
| 19:18 | jnettlet | Exaga, oh yeah, I saw that as well |
| 19:18 | jnettlet | Well the Hummingboard Edge was just announce |
| 19:18 | Exaga | jnettlet: i need to get my hands on one |
| 19:18 | jnettlet | announced |
| 19:19 | jnettlet | we are waiting for the iMX6+ chips from freescale, so we can start producing SOM's with them |
| 19:19 | Exaga | the edge eh? i'll look into that :D |
| 19:19 | Exaga | oh a cpu upgrade? |
| 19:20 | jnettlet | well it is the same CPU, but they have re-architected their mmdc/axi so it doesn't suck. It is basically removing the artificially bottleneck for the 64-bit mmdc components. |
| 19:20 | jnettlet | mainly GPU and VPU. |
| 19:20 | Exaga | gotcha |
| 19:21 | jnettlet | they claim 50% GPU improvement, which based on our previous calculations is very possible |
| 19:21 | Exaga | my apologies i thought the hummingboards were just iMX6 not iMX6+ |
| 19:22 | jnettlet | they are, but all our designs are carrier + SOM. So once we start selling iMX6+ soms you can upgrade your current board. |
| 19:23 | Exaga | jnettlet: that will never happen hahaha! I would rather buy the carrier & som complete |
| 19:23 | jnettlet | Exaga, well we won't complain then :) |
| 19:24 | jnettlet | but regardless I think I just unlocked the major bottleneck in the GPU driver that I have been hunting for so long |
| 19:24 | jnettlet | it is kind of my "white whale" |
| 19:24 | Exaga | lol |
| 19:25 | Exaga | as opposed to white elephant? |
| 19:25 | jnettlet | unfortunately it doesn't provide this huge performance bump....although it does remove a big bottleneck |
| 19:26 | Exaga | jnettlet: well performance is a major consideration of course |
| 19:26 | jnettlet | however with the iMX6+ chip coming out soon, this fix could be a HUGE performance bump |
| 19:26 | Exaga | i'm going to keep my eyes on the edge then |
| 19:27 | Exaga | see how it goes |
| 19:28 | jnettlet | well the edge has some nice features. Specifically the Mikrobus connector will be great for developers |
| 19:28 | Exaga | my hummingboards i1 and i2eX have been busy running Slackware ARM and impressively so |
| 19:28 | Exaga | i have zero complaints about the hardware |
| 19:28 | Exaga | if the edge is a continuation of that it will be awesome |
| 19:30 | jnettlet | well with the edge we are hoping to make prototyping process faster and easier. |
| 19:31 | jnettlet | Mikrobus will allow devs to buy and add various click boards for testing easily and at minimal cost. |
| 19:31 | jnettlet | They already have like 50 different boards. http://www.mikroe.com/click/ |
| 19:31 | Exaga | thanks for the link |
| 19:32 | Exaga | the edge has a built in RTC |
| 19:32 | jnettlet | other than that I have been working to cut down power usage and optimize where I can |
| 19:32 | Exaga | very wise |
| 19:32 | jnettlet | yeah, same rtc |
| 19:32 | Exaga | same as the i2eX ? |
| 19:33 | jnettlet | the rtc is the same. The new kernel I am working on has fixes that address any rtc problems we have found |
| 19:33 | Exaga | good. my RTC is quite inaccurate |
| 19:34 | Exaga | i actually don't use it and prefer to install a ds3231 controller that plugs into the gpio |
| 19:34 | Exaga | if you guys can find it within your remit to start using that controller you'll find its accuracy most impressive |
| 19:35 | jnettlet | well actually with the joy of OSS xobs from the Novena project found the problem |
| 19:36 | Exaga | well it's good that an error has been located |
| 19:37 | Exaga | some of these click modules are cool |
| 19:38 | jnettlet | with my auto-detect patches we assumed the chip was accurate enough to only settle on the proper reference source. |
| 19:38 | jnettlet | that turned out to not be true. We have added a patch so we can for using the proper clocking |
| 19:41 | Exaga | excellent |
| 19:43 | Exaga | http://www.mikroe.com/click/rtc3/ = GAH! they use a RTC with a crystal oscillator ... |
| 19:44 | Exaga | but it works the same as most other ds3107 RTC via the i2c bus |
| 19:44 | Exag | 19:44 * Exaga likes his RTCs |
| 19:45 | jnettlet | Yeah we use a crystal as well. It is just too lenient with its calibratation |
| 19:46 | Exaga | that's why i love the ds3231 |
| 19:46 | Exaga | it has a TCXO |
| 19:47 | Exaga | temperature-compensated crystal oscillator |
| 19:50 | jnettlet | Well i will be interested to hear if the new kernel and patches fix the problems you had with the pcf85233 |
| 19:51 | Exaga | yes it will. mine was terrible, gaining 6-12 seconds per day |
| 19:52 | Exaga | pcf85233 = pcf8523 ? |
| 19:59 | jnettlet | Yeah sorry i am on my phone now. |
| 19:59 | Exaga | geen probleem |