Acer Nitro 5 with nVidia 1050 on Linux (Ubuntu 18.04)

Acer Nitro 5 with nVidia 1050 GPU is an interesting beast.

- HDMI Output is wired to the NVIDIA chip

- Internal display wired to Intel GPU.

This is different than Optimus where both outputs are driven by the integrated GPU, and is actually more efficient since it doesn't spend system RAM bandwidth for display refresh or for copying the GPU  discrete GPU frame-buffer to the internal GPU frame-buffer when rendering using the.

So it's important that switching between using external monitor and internal laptop panel is handled gracefully.

Windows driver probably handles this automatically (though it's possible it has to be a special Acer build). It maybe even allows both GPUs to be active at the same time (Intel for internal, nVidia for external display)

But Linux nVidia driver typically works either in Optimus mode where both internal and external device is driven by Intel GPU or in traditional mode where monitor is only connected to the discrete GPU. Bumblebee project supports a more flexible setup, but is maybe a bit more difficult to configure

This is a short guide to enable switching between the either using external monitor with nVidia or Internal monitor with Intel GPU somewhat easily on Ubuntu 18.04:

First install nvidia proprietary driver :
 - apt install nvidia-driver-390 nvidia-prime
- edit /etc/default/grub so that GRUB_CMDLINE_LINUX_DEFAULT line has nomodeset option eg, GRUB_CMDLINE_LINUX_DEFAULT="quiet splash nomodeset" & run sudo update-grub
- restart machine while external HDMI monitor is connected



To switch to internal monitor:
- run sudo prime-select intel while on external monitor (or if you don't have the external monitor, Ctrl-Alt-F3 to text console and run sudo prime-select intel)
- reboot
- to enable acceleration for intel remove nomodeset from the kernel command line

To switch back to external monitor:
- run sudo prime-select nvidia
- reboot
- make sure there is nomodeset in the kernel command line

Making a shell script that does nomodeset grub switching and prime-select in a single step should be possible, and a situation where there is no nomodeset but nvidia is configured (GDM is stuck in neverending start loop) can be fixed by manually adding nomodeset in the grub menu

If you want to try a more flexible setup with Bumblebee, try these links


https://unix.stackexchange.com/questions/321151/do-not-manage-to-activate-hdmi-on-a-laptop-that-has-optimus-bumblebee

https://github.com/Bumblebee-Project/Bumblebee/wiki/Multi-monitor-setup

https://wiki.archlinux.org/index.php/bumblebee#Output_wired_to_the_NVIDIA_chip

How to have a higher chance of success when restoring a big MySQL database

Restoring a MySQL database is fast and easy when you just copy files in datadir when the server is shutdown, or if you use Percona xtrabackup.

But if you for some reason (AWS RDS) only have MySQL protocol available for backup, you usually can have a compressed mysqldump, that is quite slow to restore, not because of the compression or because the decompressed version is a text file that needs to be parsed, but because MySQL is slow to push it through it's disk pipeline, and because it needs to build data indexes while doing a restore.

I've spent multiple days babysitting the process of restoring a 7GB gzip compressed MySQL dump file, and these are results and tips that could help you save some time.


So, make sure that:
- you have enough IO available: For restoring a 66 GB datadir 315.6 GB was written to the drive (as measured with iostat), with a tuned MySQL configuration. For a DB of this size a mechanical drive doesn't cut it, and restore will take multiple days. Use a good SDD.

- your database TRIGGERS all have BEGIN/END statements (even though you can create them without and even thought the bug was supposed to be fixed https://bugs.mysql.com/bug.php?id=16878),  it fails on restore, with all versions of MySQL 5.7/5.6 i tried

- you start with a really empty database in your datadir - DB I worked with had inconsistent data types on a foreign key, when the dependent table with an inconsistent key already exists MySQL will report a foreign key error (MariaDB will be more informative), but if it doesn't it will happily restore the database

- your max_allowed_packet conf value is big enough, or you'll get a MySQL server has gone away message from your client while restoring.

- your innodb_log_file_size is big enough (https://dev.mysql.com/doc/refman/5.6/en/innodb-parameters.html#sysvar_innodb_log_file_size) - if you have large BLOB values in your DB, restore will fail if the value is lower than 10% of your blob field. This setting is important for quick restore too

- you have log-bin turned off in order to minimize your chance to run out of drive space and save IO (log-bin=Off doesn't mean that it's disabled just that the log bin files start with Off, the documentation can be confusing here :)  What worked for me is having all log-bin lines in the mysqld config section commented out


Finally, if you want it to finish quickly, use the fastest SSD you have available, and consider tuning MySQL configuration a bit. I'm also considering using a ramdisk, because it would help both with restore speed and when you need to do some DB transformations. MySQL defaults are not reasonable, especially for innodb_log_file_size, max_allowed_packet.

I used excellent pv for figuring out if the restore process will finish in reasonable time

pv db_dump.gz |gunzip -c |mysql -uroot database_name

Here's a full list of my MySQLd configuration variables that worked for me on my dev laptop

#my dev laptop is low on memory, for prod server you would use a lot more
innodb_buffer_pool_size=512M
innodb_additional_mem_pool_size=256M
innodb_log_buffer_size=256M
innodb_log_file_size=512M
max_allowed_packet=64M

#for saving disk IO, dont use on prod
innodb_flush_log_at_trx_commit = 2
innodb_flush_method=O_DIRECT_NO_FSYNC
skip-innodb_doublewrite

Some thoughts on (Modern) PHP

I have experience in both Java and PHP. Java mostly for traditional desktop apps, embedded UIs and PHP for websites.

Custom PHP frameworks I've built or helped build took into account the way PHP is executed: you are stateless and need to setup everything on every request (runtime is fast to start with FPM and opcode cache). Namespaces based cheap autoloader worked great. We used singletons for getting the configuration and connections to DBs. There was almost no setup code that needed to be run every time other than loading .ini based configuration and connecting to the DB. My webapps responded under 20ms (DB and other services like sphinx included), and I could get it to respond in 1 ms for things where we needed to be quick and didn't have to output HTML with Forms. It was really small and you could read the whole framework code in 1-2 hours. It worked with SQL in a reasonable way. You didn't have to write your SQL for simple CRUD, but for larger things involving joining multiple tables and more complex expressions we wrote native SQL. Caching was done thoughtfully, using  APC user cache (SHM with zero copy). It just felt nice.

I switched jobs recently, and started with Symfony 3. The thing felt like some Java framework, but poorly documented and harder to use than it should be. It had lots and lots of setup code done before handling every request. There's a whole DI framework with it's load of setup code for every component. And you have to do setup even though you don't use the component in that particular request. There are ways of doing setup lazily, but you still waste time to wire that up. Framework overhead can be 30-100ms. Other modern PHP frameworks often have similar overhead. I know that there's PHP-PM, to save some of that work that isn't really $_REQUEST specific, but it doesn't seem to be used much for production. And using Silex (deprecated by Symfony 4?) is really not that different, you still either reuse Symfony components, or rewrite them, but with similar "best practices" that are inspired by Java.

Regarding persistence, Doctrine and it's verbosity feels very ugly to me. I'd much rather use SQL syntax for defining relationships than bunch of PHP with special syntax comments or xml or yaml. And also use real SQL for complex queries.

Everybody is using type hints wherever they can, and it feels as verbose as Java, but without compile time type safety, and you can't really put type information everywhere (class members for example).

So you are almost using a type safe language, but can't get performance or compile time benefits, because inevitably, you'll have to use some dynamic typing or other dynamic language features.

Even though PHP runtime has made great progress with 7.x (it's probably the fastest interpreted language, and it's great it has reference counted garbage collection), it feels like language is struggling to find it's identity, with it taking a lot from Java and still coping with ugly legacy ($, having to use $this inside a class function, php.ini, features for supporting templating even though it's rarely used as a templating language in modern frameworks, https://3v4l.org/mcpi7).

Learning Python and Flask (as an example) was much more enjoyable than switching from a nimble custom PHP framework to Symfony. Using NodeJS and minimalistic components to build my own framework was also nice. I'd love to try GoLang, Swift or Rust in the backend too.

And there's the thing that most of the PHP frameworks try too hard to be full stack, when nowdays it's not rare you only do REST APIs on the backend. So there's a lot of infrastructure and assumptions in place for rendering HTML that you really don't need to use and that gets in the way when learning the framework and is wasteful when the code is executing.

I'd argue you can write fast, simple and maintainable PHP, by using state PHP runtime has setup for you ($_POST, $_GET, $_SERVER etc), namespaces and a namespace based autoloader, trying to use pure functions when you can (using static classes shouldn't be a sin - use it to split your code in sensible parts), and using general good practices for writing readable and maintainable code (avoid long functions, huge classes, too much block nesting, decoupling, naming things in a good way). With some coding conventions you can write a decent and productive framework quickly, but you could do that with a nicer language too, so what's the point?

(Thankfully, I'm not using Symfony on my new job, and Yii2 does suffer from some issues too, it at least feels better for now)

Kabelski internet i oversubscription

Ovo je post iz 5.11.2014. U međuvremenu sam promjenio kabelskog operatera 

Ako imate kabelski internet to znači da najvjerovatnije koristite jednu od sljedećih kabelskih tehnologija za prijenos digitalnih podataka:

- DOCSIS 1.0, 1.1, 2.0, 3.0 ili EuroDOCSIS standardi

- PacketCable 1.0, 1.5, 2.0 standardi koji na DOCSIS bazi grade razne usluge poput telefonije i digitalne televizije

Frekvencijski pojas svakog kabela podjeljen je na kanale. Širina kanala ovisi o standardu pa tako EuroDOCSIS koristi europsku širinu kanala od 8 MHz , a DOCSIS koristi američku od 6 MHz.

Podjela bandwidtha koaksijalnog kabela (Maksimalni downstream bandwidth koaksijalnog kabela je 4864 megabita prema primjeru niže)

Svi spomenuti DOCSIS transportni standardi imaju slične karakteristike oko toga koliku downstream propusnost podržavaju po jednom megahertzu, pa tako DOCSIS podržava 38 megabita po kanalu downloada, a EuroDOCSIS 50 megabita po kanalu downloada.

DOCSIS 1.1 je donio bolju standardizaciju i mogućnosti kontroliranja kvalitete usluge (QoS)

DOCSIS 2.0 je donio bolje upload brzine (27 megabita po kanalu u odnosu na DOCSIS 1.0 9 megabita po kanalu)

DOCSIS 3.0 je donio mogućnost da jedan korisnik istovremeno koristi više kanala tako povećavajući bandwidth.

DOCSIS 3.1 izdan u Listopadu 2013. je prva veća promjena u standardu jer donosi novu modulaciju 4096 QAM i odustaje od podjele kanala na 6 ili 8 MHz i umjesto toga koristi manje OFDM podkanale i u idealnim uvjetima podržava brzine do 10 gigabita downstream i 1 gigabit upstream. Još nije u primjeni.

E sad, sve je to divno i krano, ali zašto je uz takve ogromne brojke moj internet spor?

Koaksijalni kabel je medij koji dijelimo sa drugim korisnicima, za razliku od DSL-a gdje svaki modem ima vlastitu bakrenu paricu do centrale, kod kabelskih mreža dijelimo medij sa neodređenim i samo vašem ISP-u poznatim brojem korisnika. Obično operater nudi i uslugu kabelske televizije te je prostor za vaš internet sužen sa brojem kanala koji se koriste za TV uslugu.

Ajmo vidjeti jedan primjer u praksi na zagrebačkom području, za downstream:

Motorola SBV5121E

Koristi se modem Motorola SBV5121E (DOCSIS 2.0 i niže), što prema specifikaciji [2] znači da ima bandwidth za downstream od 88 do 860 MHz sa američkom širinom kanala od 6 MHz. Znači 772/6 = 128 kanala. Operater koji sam analizirao po mom saznanju 40 analognih TV kanala i 113 digitalnih. Recimo da se za ovih 113 digitalnih troši 30 6 MHz kanala u kabelu. Što znači da recimo srijedu uvečer, kad se ljudi vrate sa posla i škole, samo 58 različitih kućanstava (kanala) može istovremeno surfati punom brzinom od 38 megabita, svaki sljedeći korisnik koji krene surfati smanjuje brzinu ovim ostalima. 

Graf latencije (do prvog hop-a) na primjeru Zagrebačkog ISP-a dok korisnik osim za mjerenje ne koristi uslugu.

Operater kojeg sam analizirao nudi brzine od 8 megabita, što znači da bi teoretski trebao moći dati traženi bandwidth za (38/8) *58 = 275 korisnika, no pošto se tu vrijeme provedeno na kanalu po korisniku mora smanjiti kako bi se jedan kanal podjelio na više kućanstava, u tim slučajevima, čak i da surfa samo 275 kućanstava, njihova latencija (ICMP ping) sa odličnih 6-7 ms počinje rasti na (worst case, puna utilizacija na 418 korisnika) 4.75*7= 33 ms (molim ispravak ako je računica netočna, uzimam u obzir najmanju veličinu ICMP paketa tj. najmanju moguću diskretnu jedinicu u kojoj je moguće ostvariti komunikaciju). 

Dodatni problem je što DOCSIS 2.0 i niži ne omogućavaju brzo prebacivanje među kanalima, što znatno otežava dobru iskoristivost frekventnog spektra kabela (možda na drugim kanalima ima značajno više prostora za prijenos podataka).

U svakom slučaju, ako je previše korisnika koji dijele isti resurs (isti 6 MHz kanal, isti kabel) dolazi do drastičnog brzine pristupa pa tako kod ISP-a koji sam analizrao bandwidth pada na ispod 1 megabita, a ping ide i iznad 140 ms, uz česti packet loss.

Rijetko kada svih korisnici žele istovremeno i na period dulji od nekoliko minuta maksimalni bandwidth, pa je moguće (prema brojkama u primjeru) imati 10x više korisnika nego što je ukupnog kapaciteta (npr. 418 korisnika na 8 megabita na 88 kanala nego 4180 korisnika) a da sami korisnici ne primjete probleme u brzini pristupa, ali to uvelike ovisi o načinu korištenja Interneta. Moguće je da će više učenja na daljinu, skidanja igara preko Steam-a i sličnih servisa itd. značajno promjeniti navike korisnika u budućnosti.

Posao dijeljenja bandwidtha kada je više korisnika od broja slobodnih kanala rade zajedno modem kod korisnika i CMTS uređaj kod operatera. CMTS radi mnoge slične funkcije koje u DSL sustavima radi DSLAM, ali uzevši u obzir karakteristike dijeljenog koaksijalnog medija. CMTS omogućava da i do 1000 korisnika dijeli isti 6 MHz kanal. Koristi tehniku zvanu Statistical time division multiplexing. Nisam našao na podatak može li jedan CMTS uređaj stvarno i napuniti svih 128 downstream kanala i još 60 Mhz upstream bandwidtha. Svakako mu za to treba barem 10 gbit ethernet sučelje.

ISP može poboljšati infrastrukturu tako da smanji broj korisnika koji dijele jedan jedini kabel, ili poveća broj kanala koji se koriste za DOCSIS ukoliko medij ima slobodne kanale.

Također, ISP može početi koristiti digitalnu TV kako bi iskoristio mogućnost digitalne kompresije video i audio zapisa i time smanjio potreban bandwidth po TV kanalu za bar 4 puta (moguće i više sa kompresijom naprednijom od MPEG2), no ovo znači da operater mora svim korisnicima zamjeniti receivere za TV, što može biti značajna investicija.

Osnovana je i Facebook grupa gdje se korisnici mogu požaliti na svog operatera ili raspravljati o boljim operaterima i tehnlogijama poput recimo FTTH ili VDSL-a.

Pridružite nam se na:

https://www.facebook.com/groups/hocuboljiinternet/

Linkovi:

[1] http://en.wikipedia.org/wiki/DOCSIS

[2] http://www.wiretechsa.com.ar/PDF/equipamientointernet/SBV5121.pdf

[3] http://computer.howstuffworks.com/cable-modem.htm

[4] http://www.lightreading.com/cable-video/docsis/docsis-31-whats-next/d/d-id/708425
[5] http://www.cisco.com/c/dam/en/us/solutions/collateral/service-provider/cable-high-speed-data-hsd-solutions/gateway_to_connected_life_white_paper.pdf

Smart public transport with small automated, semi-automated or manually driven vehicles

Here's just and idea (feel free to use it in any way):

Imagine having a network of small (4-6 passengers) vehicles servicing a city for daily transportation needs. Users would enter a desired location and arrival time. The arrival time could be flexible (within an hour, if not then the price could be appropriately higher) and the user would announce any regularity (for example detailing a weekly commute) that could be used for future planning.

The centralized system would optimize the problem of getting all passengers to their respective  locations and suggest departure time and location (preferably within a few minutes of walking distance).

An interesting open source implementation would use OpenStreetMap data and have simulations and visualizations. A commercial entity could deal with deployments on various locations and provide a stable software as a service around the core open implementation. Autonomous vehicles would provide much more efficient operation of such a network and lower the costs significantly.

A tale of false alarm by ConfigServer, CPanel and a hosting provider.

I'm responsible for a couple of CPanel/WHM managed dedicated servers.

We  keep them updated, and try to do as little customization as possible outside of what cPanel knows about. We enabled mod_proxy_fcgi and PHP-FPM, so we can use Apache 2.4 MPM Event for our fairly high traffic web site. It's a unfortunate that CPanel doesn't have this configuration available out of the box, but that's for another blog post.

Today early in the morning we got a message from our lfd daemon (a service installed by a free ConfigServer Security & Firewall CPanel plugin installed by our hosting provider):

The following list of files have FAILED the md5sum comparison test. This means that the file has been changed in some way. This could be a result of an OS update or application upgrade. If the change is unexpected it should be investigated:
/usr/bin/ghostscript: FAILED
/usr/bin/gs: FAILED

The funny thing is, nothing upgraded any RPM files in this time window, our /var/log/yum.log didn't mention any upgrades to ghostscript package that provides the /usr/bin/gs binary (/usr/bin/ghostscript is a symlink to gs), we have disabled automatic updates that can be initiated by the cpanel upcp --cron sciprt, but the system us regulagrly kept up to date manually with yum update.

I've reinstalled the package with yum reinstall ghostscript (ghostscript-8.70-19.el6.x86_64 was reinstalled)

and the binary size and md5sum changed like this:

before:
size: 19152 bytes
md5sum: c64b5016d94450b476148c31cfef61ff

after reinstall:
size: 6760 bytes
md5sum: 73db43e258c4b191757b7ba75a883321

This is what actually happened: Our managed hosting provider had apparently changed our setup to upgrade our system packages automatically (probably with best intentions due to recent gnutls issue). And prelinking seems to be enabled on our system, so when upcp (CPanel automatic upgrade cron script that runs periodically) executed /usr/local/cpanel/scripts/rpmup to upgrade system packages, it also did the prelinking step, adding extra prelinking stuff to our /usr/bin/gs binary.

Similar issue described here:

http://linsec.ca/blog/2012/01/23/rpm-v-and-prelinked-binaries/

Dota 2 Wine optimization for Intel GPUs

Dota 2 for Linux implements it's 3D engine by using a Direct3D to OpenGL translation layer called ToGL. I assume that this layer can be used in different ways, but for Dota 2 it seems to be used in a less than ideal way as documented previously here. In short, Dota for Linux compiles 11000 shaders on startup, compared to just 220 the Wine version does. This causes much higher memory usage (1.2 GB vs 2.6 GB) and start-up time (35 seconds vs 1:15 min).

With Wine we actually do get the source of their Direct3D to OpenGL layer called wined3d, since Wine is open source. It's funny, the stack used to run Windows version of Dota 2 is actually more open.

Since Dota 2 for Windows when run on Wine actually outperforms native Linux version in some important aspects, and it's framerate is just slightly less, I decided to take a look on improving its performance.

I've used a tool called apitrace to record a trace of a Dota 2 session with wine so I can analyze the OpenGL calls and look at driver performance warnings (INTEL_DEBUG=perf) with qapitrace.

I optimized two things:

1. Reduce the number of vs and ps constants checked

There were many calls to check values of VS (vertex shader) and PS (pixel shader, also called fragment shaders in OpenGL) constants each frame like this:

532550 glGetUniformLocationARB(programObj = 152, name = "vs_c[4095]") = -1

This function is called from shader_glsl_init_vs_uniform_locations() in glsl_shader.c in
wined3d.

It uses GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB, defined to be 4096 in #define MAX_UNIFORMS in Mesa source.

Dota 2 doesn't need so many uniforms, most checks return -1, and wined3d checks all of values for both VS
and PS uniforms.

I reduced this number to 256, just enough for Dota 2. This saved thousands of calls per frame.

2. Use fast clear depth more often

Intel driver complains about not being able to use fast depth clears because of scissor being enabled. Turns out that device_clear_render_targets() in wined3d device.c doesn't really need to do glScissor for Dota 2, it's probably an optimization that maps better to Direct3D driver.


A small patch including both optimizations is here:
https://gist.github.com/vrodic/6437312

This patch is a hack, and glScissor part probably breaks other apps, so this is just for Dota 2. It maybe could be made in a better way so it could be merged in Wine, but I'm not wined3d expert.

So how faster is it? A solo mid hero on a setup described in the previous blog post used to get 41 FPS. Now it gets 46-49 FPS. Native version is similar to optimized Wine, but in some situations it gets worse than Wine optimized.

Ideas for improvement:

Dota 2 for Linux needs  ~7500 calls per frame. Wine version, even after my optimizations needs 37000 (EDIT: just as I was writing this post, there were some improvements, now its about 22000).

There is probably a way to optimize this even more, but it's outside of the scope of an afternoon project,  like this was. I'd like to keep on digging though.

Dota 2 performance: Linux/native vs Linux/Wine vs Windows 7 on Intel GPU

So how well does the mega popular game Dota 2 work on Linux? I've had some time to make detailed tests on my Intel IvyBridge GPU laptop (Lenovo ThinkPad X230). The graphics settings are the same on all versions.

Dota 2 Windows binary under Wine 1.6
Startup: 35 secs (over ntfs3g userspace fs driver that is not that fast)
Mem: 1.0GB
FPS: 37.5 FPS
Power usage LP mode (patched wine): 34W

Dota 2 Linux native
Startup: 1 min 14 secs (native ext4)
Mem: 2.6 GB
FPS: 40 FPS

Dota 2 Windows native - Windows 7
Startup: 25 secs
Mem: 1.2GB (measured by Windows Task manager)
FPS: 80 FPS
Power usage LP mode: 24W

Test setup:

CPU: Core i5 3320M

Resolution: 1366x768

GPU settings (same on all Dota 2 versions): shadows MEDIUM, textures HIGH, render quality: HIGHEST, all other: OFF, vsync: disabled

GPU settings LP mode (for power measurments above): Shadowd low, effects OFF, textures MED, render quality: LOWEST, fpx_max: 30

Mesa version: git-8b5b5fe (with rendering regression fix from here: https://bugs.freedesktop.org/show_bug.cgi?id=67887)

Linux distro: Ubuntu 13.10, kernel 3.11 drm-intel-nightly, running LXDE

FPS Benchmark method:
in "dota 2 beta/dota/cfg/autoexec.cfg"
cl_showfps 2
playdemo test

For FPS measurement number: look at the last 240 frames when the demo is ending

Memory measuring: RES column with `top`
Startup time measuring: stopwatch until the map is loaded

Analysis of the apitrace trace file:

I've made a trace of Dota 2 with apitrace, revealing possible performance issues.

Before the first frame of the game is drawn 11038 shaders are compiled. That is most likely why the load time is so slow and memory usage is so high. In addition a lot of the shaders being used seem to be recompiled by the Intel driver when rendering frames.

There are 162 frames in the trace I've analyzed, 193 shader recompiles, and 643 different shader programs (each program has 1 VS and 1 FS) used.

In contrast, Wine version of Dota 2 compiles only 220 shaders.

Performance feedback from the Intel driver:

glretrace of apitrace prints driver performance warnings. A sample of some that repeat every frame. These include shader recompile warnings.

575332: glDebugOutputCallback: Medium severity API performance issue 13, Clear color unsupported by fast color clear. Falling back to slow clear.
576094: glDebugOutputCallback: Medium severity API performance issue 14, Failed to fast clear depth due to scissor being enabled. Possible 5% performance win if avoided.
577739: glDebugOutputCallback: Medium severity API performance issue 4, Using a blit copy to avoid stalling on 480b glBufferSubData() to a busy buffer object.
577801: glDebugOutputCallback: Medium severity API performance issue 8, Recompiling vertex shader for program 7901
577801: glDebugOutputCallback: Medium severity API performance issue 9, Didn't find previous compile in the shader cache for debug
577801: glDebugOutputCallback: Medium severity API performance issue 1, Recompiling fragment shader for program 7901
577801: glDebugOutputCallback: Medium severity API performance issue 10, Didn't find previous compile in the shader cache for debug
577801: glDebugOutputCallback: Medium severity API performance issue 3, FS compile took 2.266 ms and stalled the GPU

Warnings in sequence in URL below:
https://gist.github.com/vrodic/6235313

It's interesting that most of this warnings were added by request of Valve back in 2012:
http://lists.freedesktop.org/archives/mesa-dev/2012-August/025288.html




Conclusion:

If you have a memory constrained machine and want to run under Linux, maybe using Wine is a better choice.

I hope Valve cares enough about Linux to fix what they can on their side and work with folks from Intel to fix their performance problems.

You will probably be luckier if you run it on nVidia GPU, since people in general are claiming performance very similar to Windows. Though probably still with slower startup times and higher mem usage.

Some info on how to compile Mesa 32 bit on 64bit Ubuntu:

Unfortunately this doesn't work with old Ubuntu versions, just with 13.10. For old version you must also remove 64 bit versions of the compiler, which was a bit too much of a requirement for me.

apt-get install gcc-multilib g++-multilib binutils-multiarch libx11-dev:i386 libdrm-dev:i386 

and as configure complains install others

apt-get build-dep mesa -ai386 

wants to install too much and remove some 64 bit stuff I need, and we don't actually need llvm i386 dev stuff for  compiling just the Intel driver.

I use this script to compile just the i965 driver for mesa:

sh autogen.sh
make clean
export CFLAGS="-m32 -O3 -mtune=native -march=native -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security"
export CXXFLAGS="-m32 -O3 -mtune=native -march=native -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security"
export PKG_CONFIG_PATH=/usr/lib/i386-linux-gnu/pkgconfig
./configure --disable-egl --enable-glx-tls --with-gallium-drivers= --with-dri-drivers=i965 --enable-32-bit
make -j4

Samsung Galaxy S2 (ARM Phone) vs Ubuntu PC performance

Introduction 

(this post has been updated in 2016)

It seems that many people assume that 1.2 GHz dual core mobile ARM CPU should be almost as fast as a PC CPU running on a similar frequency. They're wrong.

ARM cores are indeed more power efficient per square mm of surface on a same production process than Intel x86 and AMD64 architecture processors. Most of the efficiency comes from a simpler and more space efficient instruction set, but that advantage typically benefits only front-end of the CPU, which is not the biggest spender of those precious miliwatts.

The other reasons why modern dual or quad core mobile phones can run on a fraction of power that notebook or desktop (PC) CPUs need:

• less computation units on CPU die (less SIMD, ALU, etc units)
• smaller cache than PC CPUs
• power gating parts of CPU (but laptop and desktop CPUs also do this for a number of years)
• significantly slower DRAM interface  than PC CPUs, using slower DDR RAM (LPDDR2)

RAM speed significantly impacts many parts of phone performance. Executing complex JavaScript, image or video processing, Web page rendering are just some of the tasks that significantly benefit from having more RAM bandwidth. 

Your ARM device having significantly less of RAM bandwidth is also a big reason why you will probably avoid developing software on your new shiny ASUS Transformer Prime tablet/laptop (though I would certainly try:) )

So how much slower is your Android cell phone RAM than your PC RAM?

Unfortunately, I couldn't find any RAM bench-marking software that would run both on a Linux PC and on a un-rooted android device. There is a nice port of NBench, but NBench is a bigger benchmark and it needs some time before it prints out the one thing we need, the memory index. Also, it doesn't output MB/sec number, which is kind of unfortunate, since it's a really clear metric. 

So I found the really simplistic mbw (apt-get install mbw), made it even more simple (removed memcpy tests and left only the dumb array assignment part), and made Android NDK version of it.

RAMbandwidth

https://github.com/vrodic/RAMbandwidth/blob/master/app-debug.apk?raw=true

Source here. Be sure to close any apps before running it on a PC or your phone. Default array size being copied is 20 MB (the app needs 40 MB to perform the test) to better support low memory devices. 

Here are some results (20MB array size, 20 repetitions avg, run "mbw -t1 20 -n 20", default settings on RAMbandwidth, on some larger boxes 200MB size was used ):
~12500 MB/sec Intel Core i7-6700, (DDR4 x2 2133 MHz), dedicated GPU
~12300 MB/sec -Intel Core i7-9700 (DDR4 x2 2133 MHz), driving 2560x1440@60Hz display, Ubuntu 19.04, Asrock H310M-STX DeskMini 310
~9000 MB/sec - Intel Core i7-8550U (DDR3 x2 2133 Mhz, Asus UX430UNR)
~9000 MB/sec - Intel Core i7-5600U (DDR3 x2 1600 MHz)
~8200 MB/sec - Asus N56JR (Intel  i7-4700HQ, 2x DDR3 1600 Mhz memory)
~6800 MB/sec - Intel Xeon E5-1650 v2 4x DDR3 1600 MHz)
~5400 MB/sec - Intel Xeon X3430, DDR3 memory, under moderate MySQL load( 2009)
~6000 MB/sec - Thinkpad X230 Core i5 3320M (2x  DDR3 1600Mhz)

~3800 MB/sec - Core i3-2310M 2x DDR3 1333Mhz

~2200 MB/sec - Intel Core 2 E8200, PC 6400 DDR2 RAM, Desktop PC (2008).

~1100 MB/sec - Intel Core duo L2400, PC 5300 DDR2 RAM on a  Thinkpad X60S laptop (2006). 

and our mobile contenders

~6000 MB/sec - Xiaomi Pocophone F1 (Snapdragon 845 varies between 5700-7000)
~6000 MB/sec - LG G5 (Snapdragon 820 4 GB LPDDR4 2016, varies between 5800-6500)
~1500 MB/sec - LG G3 (3GB D855 - It varies from 800-1700)
~1200 MB/sec - Raspberry Pi 3
~690 MB/sec - Doogee Valencia2 Y100 Pro
~530 MB/sec- Raspberry Pi 2

~500 MB/sec - Samsung  Galaxy S2 (2011)

~250 MB/sec - HTC Desire (2010)
~120 MB/sec - Raspberry PI (2012, under X, fbdev 720p it falls to ~90 MB/sec) 

~55 MB/sec - HTC Magic (2009, had to use smaller 10MB array size because of limited RAM available) 

Samsung Galaxy S2 sometimes reports around 440 MB/sec, and sometimes 550 MB/sec. I guess it depends where kernel allocates the memory, maybe one of the memory banks shares the bus with the GPU, GSM CPU or some other greedy device. 

It should be easy to post some test results of your own hardware, so please share. 

EDIT: Check comments for some more results

Budget Surfer 0.0.1

Jučer sam htio istražiti koliko je našeg zajedničkog novca utrošeno na softverske licence za MS Windowse, Office i ostale proizvode više ili manje lako zamjenjive sa FOSS ekvivalentima. Jedine informacije koje sam zasad uspio dobiti su u dokumentu "Poseban dio Državnog proračuna Repulike Hrvatske za 2010. godinu i projekcije za 2011. i 2012." koji se nalazi ovdje http://mfin.hr/hr/drzavni-proracun-2010 .

Predpostavljam da je i Vizualizacija proračuna od projekta vjetrenjača napravljena iz istog podatkovnog izvora.

Kako bih produktivno gubio vrijeme odlučio sam napraviti mali program koji će importirati ovaj Excel u SQL bazu i omogućiti lakše "surfanje", filtriranje itd.

Ako se predpostavi da je softver u stavkama "INFORMATIZACIJA*", u podstavkama "Rashodi za nabavu neproizvedene imovine", onda ukupni iznos iznosi oko 24.5 milijuna kuna.

Informatizacija, Rashodi za nabavu neproizvedene imovine

No zanimljivo, misteriozne podstavke s naslovom "Rashodi za nabavu neproizvedene imovine" (koje nemaju nikad detaljnije podstavke) ne pojavljuju se samo u stavkama "INFORMATIZACIJA*", nego i npr. "RAČUNALNO KOMUNIKACIJSKA INFRASTRUKTURA U VISOKIM UČILIŠTIMA I JAVNIM INSTITUTIMA" od 14 milijuna kn i mnoge druge kategorije što nas dovodi do ukupnih oko 105 milijuna kuna.

Rashodi za nabavu neproizvedene imovine

Ono za što je predpostaviti da ima neku vrijednost u nabavkama softvera u Informatizaciji su stavke

"Nematerijalna proizvedena imovina" za koje je predpostaviti da se radi o softveru rađenom po narudžbi, tj koji se proizvodi za potrebe rada države. Ukupni iznos ovoga je oko 74 milijuna kuna.

Softver po narudžbi

Moj kupus, ŠBBKBB iliti TODO lista

Ako niste voljni čitati nadobudne i vjerovatno pretenciozne osobne rantove i brain dumpove slobodno preskočite ovaj post.

Pošto izgleda da je danas dan kada dajem otkaz još jednoj u nizu closed source (i closed mind dodao bih) firmi, možda je najbolje da se opustim uz nekoliko naivnih i praktičnih ideja za koje znam kako ih izvesti. Vjerovatno ću šarati od ideja za pojednostavljenje državne birokracije preko ideja za Android aplikacije, do bojnog plana za osvajanje javnog sektora od Linux desktopa u Hrvata.

No prvo o tome što mi je činiti u sljedećem razdoblju: Trebam naći poslodavca koji će shvaćati i prihvaćati da su mi briga za obitelj, prijatelje, zajednicu i moje sebične interese (fakultet recimo) na prvom mjestu. Vjerujem da takav već postoji, a ako netko ima ideju ili prijedlog slobodno se javite.

Ajmo polako s idejama:

Down to earth:

1. Proširiti postojeću Kostovu inicijativu za Android Market account tako da razvijemo i održavamo nekakav HULK-ov sistem donacija za domaće Free Software Android / Whatever aplikacije.

2. Android aplikacija koja skenira barcodove prehrambenih proizvoda i ima offline bazu sa crnom listom Ebrojeva u proizvodu. Samo uzmete proizvod u dućanu u ruke, skenirate mu barcod i dobijete crveno ili zeleno ovisno o tome ima li proizvod štetnih E dodataka (naravno postoji opcija za pregled detalja E brojeva a ako smo online i drugih detalja vezanih uz proizvode). Dodatno bi se ta radnja mogla iskoristiti da se proizvod zabilježi kao kupljen, te tako možemo imati i neki pregled / povijest potrošnje, a i jednostavniji način izbora namirnice kod brojanje kalorija ili neko drugo sofisticiranije praćenje prehrane. Naravno za organski proizvedene proizvode bez bar koda treba smisliti nešto drugo za jednostavno identificiranje (brzi izbor voće, povrće, boja, oblik, last recently used etc)


Blue sky:

Ideje u domeni odnosa države i građana:

I. Zdravstvena knjižica:

Kako relativno često mijenjam poslove, primjetio sam jednu glupu ostavštinu prošlih vremena: Posebnu zdravstvenu knjižicu koju iz nekog čudnog razloga treba mijenjati svaki put kada se promjeni radni status, a valjda i svaki put kada završite osnovnu, srednju ili fakultet. Ta zdravstvena knjižica ionako sadrži datum isteka koji ne znači baš ništa (jer osiguranje formalno može "isteći" i prije - o tome kako se uopće u socijalnoj državi može dogoditi da nemate osnovno zdravstveno osiguranje malo poslje). Dakle jedino što bi na njoj trebalo biti bitno je broj zdravstvenog osiguranja, odnosno OIB odnosno JMBG odnosno broj osobne izkaznice odnosno broj putovnice. Sve su to u biti ključevi za identifikaciju građana RH, i u teoriji bi bilo dovoljno imati recimo samo osobnu ili putovnicu (sa barcodom ili npr. qr codeom za strojno čitanje tog malog ključa). Status zdravtvenog osiguranja bi se transparentno u pozadini promjenio kada bi dobili prvi posao, dali otkaz, završili fakultet itd itd. Ionako ga uvijek zapravo imamo (neki imaju i bonus dopunsko ako su bez posla ili slično) i jedino što je bitno je da se to na nekog "knjiži" radi poreznih razloga, ali to stvarno ne znači da svaki put moramo mijenjati tu plastiku. Predpostavljam da fino zarađuje na toj plastici, a ako nije u pitanju neka korupcija, opet me fascinira glupost u javnom sektoru (nadam se da je samo glupost, jer tada je valjda lakše popraviti ovo).

A drugi problem/pizdarija oko zdravstvenog osiguranja je to što se može dogoditi da ostanete neosigurani jer ste recimo mislili da datum isteka koji piše na zdravstvenoj kartici zapravo znači datum isteka zdravstvenog osiguranja. Ne, to znači jedino i samo to koliko vrijedi taj komad plastike. U idealnom slučaju status zdravstvenog osiguranja bi se računalno provjerio prilikom svakog korištenja neke zdravstvene usluge, i gdje bi se korisnik samo upozorio da njegov status nije jasan (recimo zbog završetka studentskih prava gdje je uobičajeno da zdravstveno osiguranje sa roditelja/fakulteta "prelazi" na zavod za zapošljavanje ili budućeg poslodavca).

II. Još malo o zdravstvu:

Bilo bi lijepo kada bi laboratoriji za pretrage odnosno specijalisti automatizmom digitalno slali informacije o nalazima liječniku opće prakse, odnosno nadležnom specijalistu. Vjerujte, ovo bi uštedjelo kilometre i kilometre koji ljudi sami prelaze kako bi raznosili nalaze okolo liječnicima. U tom slučaju bi i liječnici mogli nazvati pacijente čim nalazi stignu, a ne da pacijenti moraju raditi upite svako malo. Dakle u računalnom žargonu, "digital event based a ne analog poll bazed event handling". Bilo bi lijepo i da pacijenti građani) imaju token za pristup centralnoj birokraciji gdje mogu sami pogledati svoj zdravstveni karton, rezultate pretraga, plan liječenja, dijetu, prisjetiti se ocjena u školi ili za one za nekom vrstom alzeihmerove bolesti, vidjeti koliko nekretnina imaju, tko su im roditelji, gdje su rođeni, itd itd :) I tako, ima tu puno etičkih i drugih pitanja koje treba razmotriti prije deploymenta. No predpostavljam da ne želimo da ovako nešto jednom za nas radi neki closed source sistem ili nedaj bože Facebook :)

U varijanti gdje JMBG ili OIB ili broj osobne ili broj putovnice stvarno postaju ono što jesu, a to su ključevi s kojim se identificira građanin-pojedinac, nije nam više potrebna niti radna knjižica niti pokaz niti Xica, a u idealnom slučaju (kada bi imali taj "token").

Od domaćih sustava, možda je upravo sustav iza Xice (ili studomata) baza za ovakvo nešto.


III. Stavimo Linux na desktope javnog i državnog sektora.

Stvarno, želimo li i danas, uz solidan i uoptrebljiv desktop na linuxu plaćati svake godine milijune za MS licence? Možda možemo otvoriti malu firmu koja bi se bavila supportom i specifičnim potrebama korisnika u državnom sektoru za novce utrošene na MS licence i time bar malo spriječiti odlijev novca odnosno nepovoljan omjer uvoza i izvoza?

IV. Pomognimo Marku Rakaru da http://proracun.pollitika.com/2010/korisnici.html ima više informacija, real time informacija, podataka o aktivnim natječajima u javnoj nabavi i slično ili nagovorimo državu/vladu/sabor da ovakave projekte kvalitetno odrađuju Narodne novine tj da ih financira sama država.

V. Pripremimo sve ove projekte tako da lako postanu dio nečeg većeg kada uđemo u EU ili se ujedinimo sa USA (kiberkomunist approved). Koristimo postojeća (Free software) riješenja gdje god možemo. Nemojmo se bojati ulagati u globalnu zajedničku infrastrukturu.

VI. Podržimo Zajedničko i zadružno stanovanje http://bit.ly/bEbcwK, Reciklirano imanje - Vukomerić http://bit.ly/arRK0U i slične projekte.

VII. Nacionalizirajmo ponovo Hrvatske Telekomunikacije ili barem:

Definirajmo jasnije javno pravo na DTK. Prisilimo HT na članstvo u CIX-u. Dajmo im dozvolu da postavljaju optiku, stavimo cijenu za najam niti na 100 kn po mjesecu po korisniku kako bi se brže vraćala investicija, ali svakako omogućimo da infrastrukturu od prvog dana mogu koristiti i alternativci. Regulirajmo maksimalne i minimalne cijene. Postavimo u zakon da svaka telefonska parica / veza može prenijeti neki normalni broadband a ne trenutnih 128K kako bi prisilili monopolistu da izgradi lokalnu infrastrukturu (ili stvorimo budžet za ovo dodatnim oporezivanjem operatera sa veliki tržišnim udjelom). Borimo se za simetrični broadband (sa korisnim upload brzinama). Borimo se za low latency broadband. Borimo se za 100megabitne, i gigabitne internet veze. Zadnje tri stavke omogućavaju i vrlo dobar peer to peer cloud - što nam treba za nekakav Diaspora-like social networking). Iskoristimo stvarno podatkovne kapacitete javnih poduzeća. Kad planiramo novu infrastrukturu za nova ili postojeća naselja, predvidimo i prostor sve bitno (da i optički internet pristup), kako nebi nepotrebno prekopavali kvart/naselje svako malo.

VIII. Dolazimo na razmjene vještina.

IX. Stvorimo računala koja možemo uzgajati kao što uzgajamo biljke i koja si stvaraju energiju sama kao što ih stvaraju biljke i koja imaju mrežu korjenja za povezivanje slično kao što i biljke imaju korijenje (prvi korak, pogledajte http://bbf.openwetware.org/).

X. Razno.

Diskusija o ovom blog postu nalazi se i na Friendfeed-u

Making Android development more enjoyable

Here's some things Google should look into:

0. Include android market in the emulator/test environment

That should be fast and easy to do.

1. Try finding some ways to speed up the compile/test cycle:

a) Avoid moving too much stuff arround

Currently every time an .apk file must be produced (compressed) on the developers machine, containing all the code and application resources. This apk that contains everything is then transferred to the phone or emu. It's then unpacked on the phone and the install procedure is run. It doesn't matter if you just changed one class file, the whole thing is moved around.

I understand that permissions are managed during the install process, but this could be solved in some other way like trusting the code for that app by default.

b) Make the test environment run in native code during the development process

I know that using an emulator was a nice and fast solution for you, but a lot of time and energy is lost on emulating ARM. Davlik runs on x86, and all other code also. Why don't we just make a jailed/chrooted native environment for testing available? I know that this is not a straightforward thing to implement on Windows, and there might be difficulties making this run on Mac OSX, but it's really worth it.

2. Improve the visual layout editor.

There's so many ways this could be improved, but making it faster/more intuitive is the general idea.

One quick suggestion: Try making it easier to jump to respective code implementations or have an option to generate event handlers etc. code if no code referencing that UI object is found.

Example of a great code editor/UI designer integration is Borland Delphi. Even really old versions have ease of use that Android developers could only dream of. Android API is more abstract and UIs have relative layouts, but most of the great concepts from Delphi still apply.

Item 1 b) could also create possibilities for the UI designer tool.

HTC Desire prvi dojmovi

EDIT: 26.5.2010: par novih detalja, + vesela informacija da je Samsung I9000 Galaxy S koliko vidim HTC Desire killer.

Kupih na pretplatu u VIP-u. Onu najmanju od 30kn, + 10 kn za radijsku frekvenciju + još

oko 10 kn za iskorištavanje "besplatnih" minuta + po želji oko 30 kn za 200 MB prometa. Dakle obvezao sam se na trošenje minimalno 50tak kuna mjesečno sljedeće dvije godine.

+ 2400 kn za sam telefon i nije toliko loš deal.

Kad sam kupovao HTC Desire znao sam da nije savršen, da ekran nije najbolje čitljiv na suncu, da uskoro dolaze novi uređaji temeljeni na bržem 45 nm Qualcomm Snapdragonu (moguće da će biti 1.5 GHz dual core!) i da će vjerovatno HTC kilaviti sa update-om na Android 2.2 za Desire , ali sam znao i da ima trenutno najbrži procesor, najviše memorije i odličnu kvalitetu zvuka, što mi je jako važno.

Opasno me zainteresirala Motorola Milestone (Droid u USA), koja ima tipkovnicu, nešto sporoji procesor, ali zato dosta brži GPU. No ona je pala na audio testu na gsmarena.com, što provjeram za svaki mobitel koji uzimam u obzir pošto sam se opekao sa Nokiom N79 gdje mi je glazba bila neslušljiva od distorzija. Ta Nokia je bila blizu savršenstva, tj. Nokia E52 bi bila još bliža, ali sam odlučio zamjeniti sada prilično zrelu Symbian platformu sa Android-om koji je ipak njubi...

• Još nema kvalitetne offline navigacije (Symbian Nokie imaju Garmin XT koji je odličan). iGo ima aplikaciju Motonav,(ima i potpuniji My way, ali neznam dali postoji verzija za Androide sa velikom rezolucijom), detalji ovdje. Navteq karte su bolje od Teleatlas karti izgleda. Garmin XT je još uvijek bolji, lijepše renderira karte i brži je.

• Izgleda da je potrebno koristiti nekakav Task manager za ubijanje aplikacija, koje u pravilu nemaju exit ili quit opciju i vole sjediti u backgroundu.

• Android issue 3708, Akcelerometar ne može dojavljivati podatke kada je ekran ugašen, tj. kad je mobitel u sleep mode-u. Ovime aplikacije koje recimo mjere broj koraka postaju skoro pa beskorisne jer puno brže popiju bateriju. Drugi primjer je navodno odlični Smart alarm. A i ja planiram napraviti aplikaciju koja inteligentno koristi akcelerometar u pozadini.

• Još nisam naletio na dobar DivX player. HTC je nešto obećao u kasnije SW updateu. Ali poznato je da se ne može vjerovati ovakvim obećanjima od proizvođača mobitela.

• Opera 5 Mini je još uvijek brzinom navigacije (back, forward, history) bolja od Android browsera. Komentar dolje spominje i Dolphin, ali to nisam probao. Android stock browser naravno bolje podržava JavaScript i HTML5 od Opere 5 mini.

• root-anje Androida (za stavljanje custom ROM-ova) još nije stabilno za HTC Desire, neki ljudi uspiju brickati telefone kod flashanja, neki ne (meni osobno ovo nije toliko bitno, iako me živcira što se neke meni nepotrebne HTC aplikacije, Stocks recimo, stalno vraćaju u listu aktivnih procesa, iako ih nisam dozvao)

• Još nema custom ROM-a koji je stabilan, tj implementira sve HW mogućnosti uređaja

• Hardver je dovoljno različit od Nexus One-a da se na njega ne može staviti Nexus One ROM, što će vjerovatno značiti da će HTC Desire korisnici dobiti Android 2.2 kasnije od Nexus One korisnika (nadajmo se da griješim)

• Aplikacije se po defaultu ne mogu instalirati na SD karticu, što ograničava prostor za instalaciju, HTC Desire ima samo 147 MB za instalaciju aplikacija. Ovo riješava Apps2SD (treba root), ili Android 2.2, ali oboje je trenutno nedostupno za Desire

• Poneki bug u browseru

• 3D je definitivno sporiji od iPhone-a 3GS, Motorole Droid, a vjerovatno i iPhone-a 3G

• FM radio softver je dosta neintuitivan, i zna se skršiti tako da je potreban reboot mobitela da bi se radoi ponovo osposobio, dok uredno nastavi trošiti 100% CPU-a, što je smrt za bateriju

• Aplikacije koje se plaćaju nisu dostupne kod nas?

Aplikacije koje još tražim:

• Music player koji ima file browser i playa po redu iz trenutnog foldera.

Bonus

Aplikacije koje sam instalirao, redom kojim sam ih instalirao (- neke igre):

Google Sky Map, Compass (Snaptic), ConnectBot, ASTRO File manager, Accelerometer Values, Android Sensor Data, Sensor List, Input Benchmark, Ultimate Stopwatch, Spare Parts, PowerTutor, fring (Skype, SIP, odlično), GPS Status, GPS Test, GPS Essentials, SID Player, Robo Defense FREE, Task Manager (Rhythm Software), android-vnc-viewer, Android System Info, NetCounter, Neocore (benchmark za grafiku), NenaMark1 (još jedan GPU bench), Tuner - gStrings, Robotic Guitarist Free, Splash Piano (prilično loš klavir, ali nema bolji), Buddy Runner, Endomondo Sports Tracker, SportPal OS1.5, Hi-Hiker Pro, Pedometer, Cardio Trainer (da, testiram personal GPS enabled aplikacije za trčanje/hodanje/bicikliranje, Cardio Trainer je simpa), Android SPL METER, KeepScreen, Fractoid, 2 Ballz Physics Toz, System Panel App / Task Manager, Free Advanced Task Manger, Spectral Audio Analyzer, FxCamera, PicSay, Ethereal Dialpad (synthezer), Camera Illusion, RTA Audio Analyzer, Barcode Scanner, Shazam, Google Goggles (malo nestabilno), ColorFlashlight Fun Flashlight, Call History (xiaoheng.liu), Mobile Metronome, Paint Joy, Raging Thunder LITE, Open GPS Tracker, WiFi Analyzer, Google Translete, OpenGL Demos, Sipdroid VoIP, Linpack for Android

Photo by Goran Zec