Recherche avancée

Sur d’autres sites (10571)

• My SBC Collection

  31 décembre 2023, par Multimedia Mike — General

  Like many computer nerds in the last decade, I have accumulated more than a few single-board computers, or “SBCs”, which are small computers based around a system-on-a-chip (SoC) that nearly always features an ARM CPU at its core. Surprisingly few of these units are Raspberry Pi units, though that brand has come to exemplify and dominate the product category.

  Also, as is the case for many computer nerds, most of these SBCs lay fallow for years at a time. Equipped with an inexpensive lightbox that I procured in the last year, I decided I could at least create glamour shots of various units and catalog them in a blog post.

  While Raspberry Pi still enjoys the most mindshare far and away, and while I do have a few Raspberry Pi units in my inventory, I have always been a bigger fan of the ODROID brand, which works with convenient importers around the world (in the USA, I can vouch for Ameridroid, to whom I’ve forked over a fair amount of cash for these computing toys).

  As mentioned, Raspberry Pi undisputedly has the most mindshare of all these SBC brands and I often wonder why… and then I immediately remind myself that it has the biggest ecosystem, and has a variety of turnkey projects and applications (such as Pi-hole and PiVPN) that promise a lower barrier to entry — as well as a slightly lower price point — than some of these other options. ODROID had a decent ecosystem for awhile, especially considering the monthly ODROID Magazine, though that ceased publication in July 2020. The Raspberry Pi and its variants were famously difficult to come by due to the global chip shortage from 2021-2023. Meanwhile, I had no trouble procuring these boards during the same timeframe.

  So let’s delve into the collection…
  

  Cubieboard
  The Raspberry Pi came out in 2012 and by 2013 I was somewhat coveting one to hack on. Finally ! An accessible ARM platform to play with. I had heard of the BeagleBoard for years but never tried to get my hands on one. I was thinking about taking the plunge on a new Raspberry Pi, but a colleague told me I should skip that and go with this new hotness called the Cubieboard, based on an Allwinner SoC. The big value-add that this board had vs. a Raspberry Pi was that it had a SATA adapter. Although now that it has been a decade, it only now occurs to me to quander whether it was true SATA or a USB-to-SATA bridge. Looking it up now, I’m led to believe that the SoC supported the functionality natively.

  Anyway, I did get it up and running but never did much with it, thus setting the tone for future SBC endeavors. No photos because I gave it to another tech enthusiast years ago, whose SBC collection dwarfs my own.

  ODROID-XU4
  I can’t recall exactly when or how I first encountered the ODROID brand. I probably read about it on some enthusiast page or another circa 2014 and decided to try one out. I eventually acquired a total of 3 of these ODROID-XU4 units, each with a different case, 1 with a fan and 2 passively-cooled :

  

  Collection of ODROID-XU4 SBCs

  This is based on the Samsung Exynos 5422 SoC, the same series as was used in their Note 3 phone released in 2013. It has been a fun chip to play with. The XU4 was also my first introduction to the eMMC storage solution that is commonly supported on the ODROID SBCs (alongside micro-SD). eMMC offers many benefits over SD in terms of read/write speed as well as well as longevity/write cycles. That’s getting less relevant these days, however, as more and more SBCs are being released with direct NVMe SSD support.

  I had initially wanted to make a retro-gaming device built on this platform (see the handheld section later for more meditations on that). In support of this common hobbyist goal, there is this nifty case XU4 case which apes the aesthetic of the Nintendo N64 :

  

  ODROID-XU4 N64-style case

  It even has a cool programmable LCD screen. Maybe one day I’ll find a use for it.

  For awhile, one of these XU4 units (likely the noisy, fan-cooled one) was contributing results to the FFmpeg FATE system.

  While it features gigabit ethernet and a USB3 port, I once tried to see if I could get 2 Gbps throughput with the unit using a USB3-gigabit dongle. I had curious results in that the total amount of traffic throughput could never exceed 1 Gbps across both interfaces. I.e., if 1 interface was dealing with 1 Gbps and the other interface tried to run at 1 Gbps, they would both only run at 500 Mbps. That remains a mystery to me since I don’t see that limitation with Intel chips.

  Still, the XU4 has been useful for a variety of projects and prototyping over the years.

  ODROID-HC2 NAS
  I find that a lot of my fellow nerds massively overengineer their homelab NAS setups. I’ll explore this in a future post. For my part, people tend to find my homelab NAS solution slightly underengineered. This is the ODROID-HC2 (the “HC” stands for “Home Cloud”) :

  

  ODROID-HC2 NAS

  It has the same guts as the ODROID-XU4 except no video output and the USB3 function is leveraged for a SATA bridge. This allows you to plug a SATA hard drive directly into the unit :

  

  ODROID-HC2 NAS uncovered

  Believe it or not, this has been my home NAS solution for something like 6 or 7 years now– I don’t clearly remember when I purchased it and put it into service.

  But isn’t this sort of irresponsible ? What about a failure of the main drive ? That’s why I have an external drive connected for backing up the most important data via rsync :

  

  ODROID-HC2 NAS backup enclosure

  The power consumption can’t be beat– Profiling for a few weeks of average usage worked out to 4.5 kWh for the ODROID-HC2… per month.

  ODROID-C2
  I was on a kick of ordering more SBCs at one point. This is the ODROID-C2, equipped with a 64-bit Amlogic SoC :

  

  ODROID-C2

  I had this on the FATE farm for awhile, performing 64-bit ARM builds (vs. the XU4’s 32-bit builds). As memory serves, it was unreliable and would occasionally freeze up.

  Here is a view of the eMMC storage through the bottom of the translucent case :

  

  Bottom of ODROID-C2 with view of eMMC storage

  ODROID-N2+
  Out of all my ODROID SBCs, this is the unit that I long to “get back to” the most– the ODROID-N2+ :

  

  ODROID-N2+

  Very capable unit that makes a great little desktop. I have some projects I want to develop using it so that it will force me to have a focused development environment.

  Raspberry Pi
  Eventually, I did break down and get a Raspberry Pi. I had a specific purpose in mind and, much to my surprise, I have stuck to it :

  

  Original Raspberry Pi

  I was using one of the ODROID-XU4 units as a VPN gateway. Eventually, I wanted to convert the XU4 to something else and I decided to run the VPN gateway as an appliance on the simplest device I could. So I procured this complete hand-me-down unit from eBay and went to work. This was also the first time I discovered the DietPi distribution and this box has been in service running Wireguard via PiVPN for many years.

  I also have a Raspberry Pi 3B+ kicking around somewhere. I used it as a Steam Link device for awhile.

  SOPINE + Baseboard
  Also procured when I was on this “let’s buy random SBCs” kick. The Pine64 SOPINE is actually a compute module that comes in the form factor of a memory module.

  

  Pine64 SOPINE Compute Module

  Back to using Allwinner SoCs. In order to make this thing useful, you need to place it in something. It’s possible to get a mini-ITX form factor board that can accommodate 7 of these modules. Before going to that extreme, there is this much simpler baseboard which can also use eMMC for storage.

  

  Baseboard with SOPINE, eMMC, and heat sinks

  I really need to find an appropriate case for this one as it currently performs its duty while sitting on an anti-static bag.

  NanoPi NEO3
  I enjoy running the DietPi distribution on many of these SBCs (as it’s developed not just for Raspberry Pi). I have also found their website to be a useful resource for discovering new SBCs. That’s how I found the NanoPi series and zeroed in on this NEO3 unit, sporting a Rockchip SoC, and photographed here with some American currency in order to illustrate its relative size :

  

  NanoPi NEO3

  I often forget about this computer because it’s off in another room, just quietly performing its assigned duty.

  MangoPi MQ-Pro
  So far, I’ve heard of these fruits prepending the Greek letter pi for naming small computing products :

  • Raspberry – the O.G.
  • Banana – seems to be popular for hobbyist router/switches
  • Orange
  • Atomic
  • Nano
  • Mango

  Okay, so the AtomicPi and NanoPi names don’t really make sense considering the fruit convention.

  Anyway, the newest entry is the MangoPi. These showed up on Ameridroid a few months ago. There are 2 variants : the MQ-Pro and the MQ-Quad. I picked one and rolled with it.

  

  MangoPi MQ-Pro pieces arrive

  When it arrived, I unpacked it, assembled the pieces, downloaded a distro, tossed that on a micro-SD card, connected a monitor and keyboard to it via its USB-C port, got the distro up and running, configured the wireless networking with a static IP address and installed sshd, and it was ready to go as a headless server for an edge application.

  

  MangoPi MQ-Pro components, ready for assembly

  The unit came with no instructions that I can recall. After I got it set up, I remember thinking, “What is wrong with me ? Why is it that I just know how to do all of this without any documentation ?”

  

  MangoPi MQ-Pro in first test

  Only after I got it up and running and poked around a bit did I realize that this SBC doesn’t have an ARM SoC– it’s a RISC-V SoC. It uses the Allwinner D1, so it looks like I came full circle back to Allwinner.

  

  MangoPi MQ-Pro with more US coinage for scale

  So I now have my first piece of RISC-V hobbyist kit, although I learned recently from Kostya that it’s not that great for multimedia.

  Handheld Gaming Units
  The folks at Hardkernel have also produced a series of handheld retro-gaming devices called ODROID-GO. The first one resembled the original Nintendo Game Boy, came as a kit to be assembled, and emulated 5 classic consoles. It also had some hackability to it. Quite a cool little device, and inexpensive too. I have since passed it along to another gaming enthusiast.

  Later came the ODROID-GO Advance, also a kit, but emulating more devices. I was extremely eager to get my hands on this since it could emulate SNES in addition to NES. It also features a headphone jack, unlike the earlier model. True to form, after I received mine, it took me about 13 months before I got around to assembling it. After that, the biggest challenge I had was trying to find an appropriate case for it.

  

  ODROID-GO Advance with case and headphones

  Even though it may try to copy the general aesthetic and form factor of the Game Boy Advance, cases for the GBA don’t fit this correctly.

  Further, Hardkernel have also released the ODROID-GO Super and Ultra models that do more and more. The Advance, Super, and Ultra models have powerful SoCs and feature much more hackability than the first ODROID-GO model.

  I know that the guts of the Advance have been used in other products as well. The same is likely true for the Super and Ultra.

  Ultimately, the ODROID-GO Advance was just another project I assembled and then set aside since I like the idea of playing old games much more than actually doing it. Plus, the fact has finally crystalized in my mind over the past few years that I have never enjoyed handheld gaming and likely will never enjoy handheld gaming, even after I started wearing glasses. Not that I’m averse to old Game Boy / Color / Advance games, but if I’m going to play them, I’d rather emulate them on a large display.

  The Future
  In some of my weaker moments, I consider ordering up certain Banana Pi products (like the Banana Pi BPI-R2) with a case and doing my own router tricks using some open source router/firewall solution. And then I remind myself that my existing prosumer-type home router is doing just fine. But maybe one day…

  The post My SBC Collection first appeared on Breaking Eggs And Making Omelettes.

• Custom Segmentation Guide : How it Works & Segments to Test

  13 novembre 2023, par Erin — Analytics Tips, Uncategorized

  Struggling to get the insights you’re looking for with premade reports and audience segments in your analytics ?

  Custom segmentation can help you better understand your customers, app users or website visitors, but only if you know what you’re doing.

  You can derive false insights with the wrong segments, leading your marketing campaigns or product development in the wrong direction.

  In this article, we’ll break down what custom segmentation is, useful custom segments to consider, how new privacy laws affect segmentation options and how to create these segments in an analytics platform.

  What is custom segmentation ?

  Custom segmentation is when you divide your audience (customers, users, website visitors) into bespoke segments of your own design, not premade segments designed by the analytics or marketing platform provider.

  To do this, you single out “custom segment input” — data points you will use to pinpoint certain users. For example, it could be everyone who has visited a certain page on your site.

  

  Segmentation isn’t just useful for targeting marketing campaigns and also for analysing your customer data. Creating segments is a great way to dive deeper into your data beyond surface-level insights.

  You can explore how various factors impact engagement, conversion rates, and customer lifetime value. These insights can help guide your higher-level strategy, not just campaigns.

  How custom segments can help your business

  As the global business world clamours to become more “data-driven,” even smaller companies collect all sorts of data on visitors, users, and customers.

  However, inexperienced organisations often become “data hoarders” without meaningful insights. They have in-house servers full of data or gigabytes stored by Google Analytics and other third-party providers.

  

  One way to leverage this data is with standard customer segmentation models. This can help you get insights into your most valuable customer groups and other standard segments.

  Custom segments, in turn, can help you dive deeper. They help you unlock insights into the “why” of certain behaviours. They can help you segment customers and your audience to figure out :

  • Why and how someone became a loyal customer
  • How high-order-value customers interact with your site before purchases
  • Which behaviours indicate audience members are likely to convert
  • Which traffic sources drive the most valuable customers

  This specific insight’s power led Gartner to predict that 70% of companies will shift focus from “big data” to “small and wide” by 2025. The lateral detail is what helps inform your marketing strategy. 

  You don’t need the same volume of data if you’re analysing and segmenting it effectively.

  Custom segment inputs : 6 data points you can use to create valuable custom segments 

  To help you get started, here are six useful data points you can use as a basis to create segments — AKA customer segment inputs :

  

  Visits to certain pages

  A basic data point that’s great for custom segments is visits to certain pages. Create segments for popular middle-of-funnel pages and compare their engagement and conversion rates. 

  For example, if a user visits a case study page, you can compare their likelihood to convert vs. other visitors.

  This is a type of behavioural segmentation, but it is the easiest custom segment to set up in terms of analysis and marketing efforts.

  Visitors who perform certain actions

  The other important type of behavioural segment is visitors or users who take certain actions. Think of things like downloading a file, clicking a link, playing a video or scrolling a certain amount.

  For instance, you can create a segment of all visitors who have downloaded a white paper. This can help you explore, for example, what drives someone to download a white paper. You can look at the typical user journey and make it easier for them to access the white paper — especially if your sales reps indicate many inbound leads mention it as a key driver of their interest.

  User devices

  Device-based segmentation lets you compare engagement and conversion rates on mobile, desktop and tablets. You can also get insights into their usage patterns and potential issues with certain mobile elements.

  

  This is one aspect of technographic segmentation, where you segment based on users’ hardware or software. You can also create segments based on browser software or even specific versions.

  Loyal or high-value customers

  The best way to get more loyal or high-value customers is to explore their journey in more detail. These types of segments can help you better understand your ideal customers and how they act on your site.

  You can then use this insight to alter your campaigns or how you communicate with your target audience.

  For example, you might notice that high-value customers tend to come from a certain source. You can then focus your marketing efforts on this source to reach more of your ideal customers.

  Visitor or customer source

  You need to track the results if you’re investing in marketing (like an influencer campaign or a sponsored post) outside platforms with their own analytics.

  

  Before you can create a reliable segment, you need to make sure that you use campaign tracking parameters to reliably track the source. You can use our free campaign tracking URL builder for that.

  Demographic segments — location (country, state) and more

  Web analytics tools, such as Matomo, use visitors’ IP addresses to pinpoint their location more accurately by cross-referencing with a database of known and estimated IP locations. In addition, these tools can detect a visitor’s location through the language settings in their browser. 

  This can help create segments based on location or language. By exploring these trends, you can identify patterns in behaviour, tailor your content to specific audiences, and adapt your overall strategy to better meet the preferences and needs of your diverse visitor base.

  How new privacy laws affect segmentation options

  Over the past few years, new legislation regarding privacy and customer data has been passed globally. The most notable privacy laws are the GDPR in the EU, the CCPA in California and the VCDPA in Virginia.

  

  For most companies, it can save a lot of work and future headaches to choose a GDPR-compliant web analytics solution not only streamlines operations, saving considerable effort and preventing future headaches, but also ensures peace of mind by guaranteeing the collection of compliant and accurate data. This approach allows companies to maintain compliance with privacy regulations while remaining firmly committed to a data-driven strategy.

  Create your very own custom segments in Matomo (while ensuring compliance and data accuracy)

  Crafting precise marketing messages and optimising ROI is crucial, but it becomes challenging without the right tools, especially when it comes to maintaining accurate data.

  That’s where Matomo comes in. Our privacy-friendly web analytics platform is GDPR-compliant and ensures accurate data, empowering you to effortlessly create and analyse precise custom segments.

  If you want to improve your marketing campaigns while remaining GDPR-compliant, start your 21-day free trial of Matomo. No credit card required.

• Revision e8b133c79c : Factor inverse transform functions into vpx_dsp This commit moves the module in

  31 juillet 2015, par Jingning Han

  Changed Paths :
   Modify /test/idct8x8_test.cc

  
   Delete /vp9/common/arm/neon/vp9_idct16x16_1_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct16x16_1_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct16x16_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct16x16_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct16x16_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct32x32_1_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct32x32_1_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct32x32_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct32x32_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct4x4_1_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct4x4_1_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct4x4_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct4x4_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct8x8_1_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct8x8_1_add_neon_asm.asm

  
   Delete /vp9/common/arm/neon/vp9_idct8x8_add_neon.c

  
   Delete /vp9/common/arm/neon/vp9_idct8x8_add_neon_asm.asm

  
   Modify /vp9/common/vp9_idct.c

  
   Modify /vp9/common/vp9_idct.h

  
   Modify /vp9/common/vp9_rtcd_defs.pl

  
   Modify /vp9/common/x86/vp9_idct_intrin_sse2.c

  
   Delete /vp9/common/x86/vp9_idct_intrin_sse2.h

  
   Delete /vp9/common/x86/vp9_idct_sse2.asm

  
   Delete /vp9/common/x86/vp9_idct_ssse3_x86_64.asm

  
   Modify /vp9/encoder/x86/vp9_dct_ssse3.c

  
   Modify /vp9/vp9_common.mk

  
   Add /vpx_dsp/arm/idct16x16_1_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct16x16_1_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct16x16_1_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct16x16_1_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct16x16_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct16x16_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct16x16_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct16x16_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct16x16_neon.c
  (from /vp9/common/arm/neon/vp9_idct16x16_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct32x32_1_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct32x32_1_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct32x32_1_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct32x32_1_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct32x32_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct32x32_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct32x32_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct32x32_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct4x4_1_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct4x4_1_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct4x4_1_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct4x4_1_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct4x4_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct4x4_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct4x4_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct4x4_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct8x8_1_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct8x8_1_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct8x8_1_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct8x8_1_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct8x8_add_neon.asm
  (from /vp9/common/arm/neon/vp9_idct8x8_add_neon_asm.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/arm/idct8x8_add_neon.c
  (from /vp9/common/arm/neon/vp9_idct8x8_add_neon.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/inv_txfm.c
  (from /vp9/common/vp9_idct.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/inv_txfm.h

  
   Modify /vpx_dsp/vpx_dsp.mk

  
   Modify /vpx_dsp/vpx_dsp_rtcd_defs.pl

  
   Add /vpx_dsp/x86/inv_txfm_sse2.asm
  (from /vp9/common/x86/vp9_idct_sse2.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/x86/inv_txfm_sse2.c
  (from /vp9/common/x86/vp9_idct_intrin_sse2.c
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/x86/inv_txfm_sse2.h
  (from /vp9/common/x86/vp9_idct_intrin_sse2.h
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
   Add /vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
  (from /vp9/common/x86/vp9_idct_ssse3_x86_64.asm
  :7f8dd35329fcec4f32eba6f56ada69224e30f238)
  
  Factor inverse transform functions into vpx_dsp

  This commit moves the module inverse transform functions from vp9
  to vpx_dsp folder. The hybrid transform wrapper functions stay in
  the vp9 folder, since it involves codec-specific data structures.

  Change-Id : Ib066367c953d3d024c73ba65157bbd70a95c9ef8