maybe only twice as fast as 32 bit floating point
#code-talk
fickle aurora
https://pastebin.com/AiqEAidL is a multithreaded test code that I wrote a while ago and it just works by passing the second value as your number of threads, and it is quite interesting how you can get different speeds depending on the number of threads - they aren't linear speeds by any chance
these are the test results I got for speed, though I think what affected results the most was dividing the unsigned long int by the number of threads on line 17
dividing by 3 gave the worst performance from everything .. though it is good to note, this was done on ARM7v71 with 1200Mhz and 4c/4t, it is interesting that it had 9 seconds of run time for both 4 threads and 8 threads and I have not been able to replicate that at any other number of threads
devout ivy
Their FPU is integral to the CPU core in armv6hf onwards. Compulsory in all A profile ARMv7 implementations (optional in M profile, I'll admit I dont know about R) and all ARMv8 cores regardless of profile
Very little functional difference between how ARM and x86 handle floating point or any other math. Only major difference is the more widespread adoption of SIMD operations in x86, 128 bit wide SIMD is present in ARMv8 but modern intel and AMD both have 256 bit wide and intel also have some cores with 512 bit wide
pure sierra
would it be hard to make a fpu bitcoin miner ?
fickle aurora
well floating point math can be done on the general purpose registers, it's just that it's more precise and faster to calculate in the FPU which is what I am trying to get at
for example, AVR has no fancy FPU at all, and it can certainly calculate floating point
pure sierra
i like arduino
fickle aurora
yeah arduino is like python, it runs horribly slow
pure sierra
dont tell me that
fickle aurora
you need to use Atmel Studio to get the most of AVR
pure sierra
damn u puritans
fickle aurora
well, the studio isn't what makes it, I mean, you could write your own compiler for AVR but it would be a waste of time
pure sierra
are you saying the compiling is slow or the result ?
fickle aurora
well arduino is a framework, like a runtime for your programs
pure sierra
i thought it was just an editor & bootloader
fickle aurora
whether you use libraries in it or not, anything you compile in it will rely on the arduino framework when it runs
yeah the bootloader is essentially like the runtime
fickle aurora
when you program AVR in avr-asm and upload it with an Atmel programmer, it will not use a bootloader, it will run natively
you can use arduino bootloader in atmel studio as well
when I started using avr-asm I stopped using arduino for it entirely and focused on doing everything in assembly to further expand my capability
some of the chips don't even have a multiply or divide instruction so you have to make a function to do that
pure sierra
but here's the thing, time, for me the time saved using arduino and all its available libraries is worth way more than any performance difference on a 8Mhz chip...
fickle aurora
well with something like the attiny10, you can only fit tiny programs on it, and you probably may never need to divide numbers on it
it has add, subtract but no multiply or divide instruction
pure sierra
sure, but think about all the modules that make arduino so good, having to write your own 'drivers' for them
fickle aurora
yeah it is a pain to deal with i2c modules in assembly because you have to write all the functions to handle the serial communication
but once you make what you need for the basics of data transfer, you can do way more than arduino can do since you can choose when to send a bit and whatnot
pure sierra
i think sure, say you had a big project that you needed every little bit of performance and it was going to sell a million units, go atmel/assembly, but generally im just tinkering for myself so quicker the better as hhalf my projects never get finished
fickle aurora
yeah the performance is better but also the thing is that if you wanted to send custom messages via serial, it is much harder to do in arduino compared to at the assembly level
pure sierra
how custom, ? i can send bits or bytes via serial
fickle aurora
timing is a big deal for serial comms as well
pure sierra
baud rate ?
fickle aurora
so if arduino's code happens to run too slow for something that requires precise timing, that can cause issues
separately from baud rate, there are things like buffer registers for example
when data comes in or goes out, you have to read or write it from a buffer register, then deal with that data in a timely manner before another message is sent or received
pure sierra
ive wondered about this, can a serial input fill up if you dont read from it ?
fickle aurora
I have to read up on it, but I think it said in the protocol that it blocks until the buffer is cleared
pure sierra
so , the sender would get some kind of fail event ?
fickle aurora
the same thing happens with computers when too many connections start but then never acknowledge, filling up socket queue/buffers and then this causes denial of service, since all the queue slots are filled
so to fix a basic denial of service with a socket queue, you have to basically make a timeout function that measures the time a connection to the socket started, and if the 3 way handshake is not completed in a timely fashion, you should close() the connection to free up a spot in the queue
in serial communication, you can have the same issue theoretically because of the data buffer
well in i2c, parity is very simple, so it just wouldn't send an acknowledge back to the sender if the buffer blocked
long raft
would it be hard to make a fpu bitcoin miner ?
not profitable because of electricity costs. need specialized hardware
fickle aurora
well ideally for bitcoin, there should be dedicated math coprocessors, and those used to be a thing at one time in history
long raft
its more like just the circuitry for calculating the hash
generally modern processors are beasts with floating point calculations, but always memory access is a throttle. same on the GPUs
pure sierra
but they do it with asic
long raft
this is why everybody tries to turn everything into matrix math because they know how to optimize those calculations for good memory access
so the compute power is there, but its not worth the cost in energy
fickle aurora
it is kind of sad that most software will never utilize all of the FPU
long raft
its really hard
fickle aurora
like for me, I don't have AVX-512 support nor AVX2, but I'll probably never need it because nothing uses it
and there's just a handful of games now that require AVX1
long raft
games requiring cpu features? whyy
surely some library theyre using a static binary of
fickle aurora
they're using AVX for better game performance
it comes down to what I was saying before, you can calculate floating numbers in any cpu register, but if you use the specialized FPU registers or the special math extensions, you will have significantly better performance and higher accuracy
however, the software has to support using those extensions
you can't use an if statement in this case, because it's machine code level, you'd have to make a separate version of your game for non-AVX, AVX2, and AVX-512 so you'd need 3 different compiled versions if you were going to offer people a choice of which one to support - or you could pick one that most people can use - most software probably out of worries of userbase just simply don't use any of them ... it is kind of like when we used to have x86, amd64, Itanium2, VAX, Alpha, DEC and all these different architectures with various instruction sets ... now most of the computer world is amd64 (now referred to as x86-64 or x64 as AMD and Intel decided to come to one standard)
fickle aurora
fickle aurora
good example of why compilers can be bad compared to handwritten assembly
the source code for both along with the commands you need to use to compile/assemble them are here https://pastebin.com/nAD86vJP
for the assembler, I just used nasm and for linking I used Link.exe from MSVC Tools 14.29.30037 and libraries were in the Windows Kit 10.0.20348.0 (both you have to install as components in visual studio)
if you want to make it into a DLL instead of executable you can change the linker command to <path-to-link.exe> factorial.obj /dll /export:factorial /nologo /incremental:no /opt:ref /OUT:factorial.dll /MACHINE:X64 <libraries> and then you can call int factorial(int) from another program using the DLL written in assembly
long raft
so allan why do you use windows if youre so anti microsoft? why arent you on linux or something?
gamer reasons?
fickle aurora
not so much anti-microsoft but rather anti-.NET/CIL
the sole purpose of .NET/CIL was to make you buy a new dev kit every year and force the old one to expire
long raft
i have never paid a dollar for anything .net
its open source... MIT license
the dev environments are either free or open source
fickle aurora
you can use all of the .NET libraries from GCC anyway, but at the same time, they are terribly slow and inefficient
long raft
but you complain about microsoft compilers and stuff
fickle aurora
yes because MS switched everything over to CIL
long raft
yea let me just use that web server library 
good luck finding one as fast as whats made in .net
you just hate one particular set of libraries and runtimes... thats weird
most people who dont like .net dont like it because microsoft and thats basically their only reason
fickle aurora
they were originally made long ago to help bring more people into programming because their purpose was to make programming easier and more understandable - but then it morphed into mainstream production when it should have remained as educational
long raft
thats not correct
fickle aurora
same with python, it was originally made for simple scripts to compete with shell scripts, it was never supposed to be mainstream
long raft
it was always meant for production, never for education...
you are incorrect about many things .net, which i imagine is because you havent actually used it
fickle aurora
if you compare VB 6.0 and VB .NET you can clearly see why they made .NET
long raft
.net is not vb
fickle aurora
and VB 6.0 will be used in eternity, I can bet in 20 years I will still see VB 6.0
long raft
i see how making vb.net made basic easier. .net has nothing to do with making programming easier
it does simplify some old terrible interfaces, e.g. COM
fickle aurora
well when I say .NET I am referring to CIL
long raft
thats just incidental stuff - people learn better ways to make things on their second try
fickle aurora
VB 6.0 was the last real compiler microsoft ever made, everything after that compiles to CIL instead of assembly/machine code
long raft
this is nonsense
fickle aurora
have you looked at any of your compiled programs in hex editor yet
long raft
you spend too much time with your head in a hex editor
im aware of what the IL looks like
i dont really care why you dont like .net, i was just curious why you run windows
fickle aurora
linux is equally guilty of copying microsoft by making everything use clang/llvm
so far it's not bad, as long as they don't suddenly decide to force everyone to use it like a runtime
long raft
you truly are in last century's computer science
youre gonna get left behind, allan
llvm is pretty damn sweet
also im sure youve used linux, but as your main OS i think youd like it
fickle aurora
I'll go to the niche area of fixing things people can't fix at the low level
I'll be honest windows does things better, just that I think MS is making bad moves and trying to be too greedy
long raft
but a user does not have much control in windows
fickle aurora
and is this really true
fickle aurora
you can do a lot to secure windows, it's just a lot of extra work
long raft
no im not talking about security. i dont think windows has security problems. and it sure is fast
fickle aurora
if linux wanted to make the desktop market, Linus should have worked on X himself instead of getting the community to make it
long raft
its way too much to ask 1 person to work on
its difficult to commercialize because of the license, so i dont think itll be a common desktop OS in my life
but for a programmer...
unless you make games, most programmers are developing for linux now
fickle aurora
even Linus now admits the GPL was a mistake, he wishes he could have gone back and done things differently
long raft
was it not copyleft enough for him?
fickle aurora
the idea was to allow future development, not to make the software for free as in penniless
the problem is that if you give out source code to allow future development, someone is going to steal it and say they made it, or they can compile it and sell it to undercut you, and now you need to litigate it to make any profit from it, and in the end, a country that does not have intellectual property laws will just actively steal it and distribute it
long raft
if you give out source code, people use it (not steal), and they dont need to claim credit because you gave it away
a lot of money is made supporting software, not just writing it
jesus look how much money redhat brings in
its like theres more money in harry potter merchandise than the book advances probably
pure sierra
@fickle aurorahow hard to make crypto minor ? why fpga not suitable compared with asic ?
fickle aurora
it can probably be done, but how efficient would that be compared to a xeon with plenty of L3 cache, cores and fairly high clock speed
fickle aurora
I'm using a 2013-era xeon to play star citizen and I outperform some brand new builds lol
it is hilarious how bad new CPU models are now
pure sierra
fickle aurora
of course, if you can get a new CPU with the same or better stats than the older xeons, then you will certainly get better performance
fickle aurora
oh that is based on power vs cost efficiency
yeah I can see why xeon would be bad for mining crypto
pure sierra
asic miners are like 12-70k$ right now
but only because they are pegged againts crypto price and roi
fickle aurora
has anyone considered completely changing the software side to run faster in a single process and then use single core with highest clock speed possible, with efficient cooling
pure sierra
not sure but @fickle aurora would be the guy that knows the most about low level efficiency 😆
fickle aurora
I mean multithreading is good, but 4 cars against a single car with a faster engine will not get to the finish line faster
pure sierra
would they not have the same combined work output ?
fickle aurora
so it's only an advantage if you can do future calculations and combine them
pure sierra
but cant the 4 jobs just run in parralel until complete
fickle aurora
yes, assuming that you can do that
pure sierra
some time as running 4 jobs one after other at 4x speed
fickle aurora
sometimes you need to go in sequence
pure sierra
true
long raft
you cannot, because its hashes
pure sierra
cannot what ?
long raft
its all intentionally slow
fickle aurora
if the calculations must be done sequentially, then multithreading is useless
pure sierra
isnt it just randomly generating numbers ?
long raft
you can still multithread but the hashing algorithm is sequential
fickle aurora
you could however run multiple separate calculations
long raft
yea it can scale with CPU
fickle aurora
then you're better off with a board that supports multiple single core cpus with highest possible clock speed and cache
long raft
but the ASICs are orders of magnitude faster at the specific operations
and it doesnt need any floating point
pure sierra
yeah you would need like 500 cpu, just guessing number
fickle aurora
the problem with multicore is that you can't overclock too much due to heat
long raft
bitcoin mining doesnt use like 99% of the CPUs features
pure sierra
i feel like fpga might be best approach as you can make ur own circuit so there is room for innovation
fickle aurora
and instead of RAM you just need cache which is essentially SRAM
long raft
you dont really need innovation as you know the algorithm
cache is not helpful, you cant build rainbow tables
its all just hard integer calculations
fickle aurora
so you just need the clock speed
long raft
you cant get high enough speed out of generalized processors
pure sierra
but could you design a fpga that could perform on par with a asic but fro cheaper
long raft
i cant remember the btc exact specs, but im guessing its all pivoted around a modulo, because division is generally the mathematical trap door
pure sierra
as its not inflated price to peg againts btc price
long raft
but you wouldnt need to - the algorithm doesnt change, and the asics are cheaper to manufacture at bulk
pure sierra
but making asic not possible by me , but i can make fpga ?
long raft
right but if you were doing this at large scale to really make money, youd make an asic
you know 3d printing vs molds
pure sierra
youd likely make more money from reselling asics than mining
but problem is getting it made
long raft
hah in the california gold rush, a chicken cost as much as an ounce of gold
fickle aurora
you might make more money automating farm equipment
long raft
yea sometimes the support industry is bigger...
fickle aurora
there are still farmers paying people 20 bucks an hour to shovel out barns
and then there's IBM hiring masses to produce the stuff that the farmer has to shovel
the only reason IBM is still in existence is the endless government contracts that happily go overbudget and have to be scrapped and reworked constantly because of obvious security holes, the wrong order of starbucks coffee, and things like the wonderful world of java
long raft
C surely has contributed more security exploits than any other language
i mean obviously actionscript/flash but ...
fickle aurora
when they did the Shared Services system for the canadian government, it was a total disaster and now CRA is shut down because of it
I wonder if they filter user input in web text boxes
pure sierra
shard 1 dead
lyric turtle
Found this some time ago: https://www.nandland.com/articles/using-your-fpga-to-mine-for-bitcoins.html
I think it is about 3 years old.
pure sierra
thanks
so if you somehow think you can do it better thhan anyone else you might as well do it in software on your home computer first
lyric turtle
I was thinking about btc mining some years ago but it is to expensive in germany. There are options for cloud mining (some are scam!) where you virtually rent the MH. I'm personally not very much into crypto. Thought about buying rather than mining.
pure sierra
yeah Australia electricity price is really too high too
this in going to be a bit of a problem thhough
gaunt rivet
foxhole needs to be in VR
fickle aurora
yeah I wouldn't rent a server for mining, because you will get a virtual private server in a VM environment with hundreds of other virtual servers on the same machine and they will throttle your server
a lot of dedicated server co's are con artists, my raspberry pi had more computing power than the VPS that some of them were offering
fickle aurora
https://pastebin.com/K9ri3Tx3 another example, 32 bit this time and instead of calling printf, it calls WriteFile and writes to STDOUT which is essentially what printf is, just a wrapper function that uses WriteFile (we can't save space here because we are stuck with using ucrt,lib, vcruntime.lib, msvcrt.lib, and kernel32.Lib as the minimum things we need to link to)
long raft
re: the bitcoin mining, i think if you can do renewable energy and you have low maintenance costs then it can still be viable to mine
but not if youre replacing parts or generators or solar panels often enough to eat into your cost
lyric turtle
Australia would be suitable for solar powered mining but I think getting all the stuff needed is quite expensive there. Most miners are operating from countries rich in natural resources for energy production like China (coal and nuclear), followed by Russia/India (coal and nuclear too) and Iceland (hot vulcanic geothermal energie).
@fickle aurora You are not renting a vps server for mining. You rent calculation time on asics: https://shamining.com based on what you pay you get more power.
long raft
until we get quantum computers
fickle aurora
sounds fair, but remember if they are selling you a service, they are in it for profit for themselves
lyric turtle
Yes of course!
This is what everyone is looking for.
@long raft At least you can try that for free. You can model the crypto algorithm with gates: https://quantum-computing.ibm.com/composer/files/new But you can't use the result, it is an isolated process without blockchain access.
lyric turtle
That is not an emulator! These are real quantum computer. Just register, write your code and wait for execution time!
fickle aurora
made a fully windowed app in 64-bit assembly, only uses 1kb of memory compared to the 30MB+ you would get from .net framework
I think I will go back to using C, populating stack manually is too much work 😂 most of the info online is all for 32 bit which is incompatible, and the x64 calling convention is significantly different from the x86 one ... but not a bad thing to learn
https://pastebin.com/eabZAC27 instructions for assembling and linking are in the paste header
hollow gyro
Hi all, the Wiki lacks stats for the previous war WC85 nobody happens to have data on the deaths for that war end do they?
fierce sky
any Frontend or backend developers with web development experience? Interested in setting up a logistics portal as some of the tools out there are either regiment specific and authenticated for them only or are not very user friendly.
pure sierra
@hollow gyro my site does
hollow gyro
<@130170417519394816> my site does
I don't see them exposed for previous wars.
Could you update WC85 Deaths on: https://foxhole.fandom.com/wiki/World_Conquest
lunar cobalt
Updated it @hollow gyro
quick shoal
Do we have activity logs like the new stats available on the war api yet or do you guys think they won’t add them?
fickle aurora
so much for garbage collection, windows is using 13GB of memory and nothing intensive is even running
summer meteor
so much for garbage collection, windows is using 13GB of memory and nothing inte...
How much RAM does your build have
Coz windows is kinda famous for "if you give me more RAM I will use more RAM"
Then again judging by the chat history you probably know more about this than I do so I'mma just
fickle aurora
I have a lot of ram, it's very minor, but garbage collection is still a bad concept imo
long raft
garbage collection has many advantages
also this idea that youre only doing small things why is windows using so much ram? its the cache. and it improves your performance.
why spend money on ram and then not use it?
fickle aurora
true, it's caching 45GB of mem
pure sierra
Do we have activity logs like the new stats available on the war api yet or do y...
Negative
quick shoal
darn:(
mossy pelican
Windows using 13gb is definitely cache
And most of it will free up if another process needs it
I have 20gb of ram and windows only uses about 3gb (windows 10's ram requirement is 4gb mind you)
long raft
and garbage collected memory takes less space because it defragments it
which could also mean less page faults
fickle aurora
it actually says it's using 44.5GB cached and 13.6GB in use
mossy pelican
How do you even use so many ram?
long raft
probably just disk cache
mossy pelican
Huh
long raft
well theres no reason to unload it
i mean in theory all the data passing through the disk could jsut be cached. i mean thats how i live on zfs. all the ram.
just chillin in volatile memory, catchin cosmic rays
fickle aurora
page files are all completely disabled
fickle aurora
I just love to torture my PC 😂
hollow gyro
I'm surprised the Fozhole War API is not pinned in this channel.
fickle aurora
well the first pin has the link to the API https://github.com/clapfoot/warapi
deft quartz
I have a lot of ram, it's very minor, but garbage collection is still a bad conc...
Try Rust. Its basically a better C/C++. No Garbage Collectors
devout ivy
well floating point math can be done on the general purpose registers, it's just...
Yes and my point is that your statement was false as arm does possess a true hardware FPU
Also your description of how arduino operates was fundamentally incorrect
It actually uses avr-gcc and is not a runtime at all, its nothing but a library of helpers, its otherwise raw C++ in avr-gcc, much like atmel studio uses
fickle aurora
well AVR does not have any FPU and it's not based on ARM ... ARM and AVR are 2 totally different architectures ... also, A32 instruction set for some of ARM cpus do not have an FPU ... AVR is the architecture of atmel microcontrollers, ARM is a whole different platform ... there is also PIC which is another architecture
but you can store a binary version of float in general registers, but it is perhaps a bad idea since the general registers have no method of setting the decimal point, so you would have to store that in another register as a modifier - but you could theoretically do it
and yes arduino is a framework (notice that arduino calls them sketches and not programs?) and you have to link to arduino's libraries, which in turn includes the framework in your binary
fickle aurora
devout ivy
not been used in about a decade mate
ARMv6hf has FPU. ARMv7 A profile its compulsory. ARMv8 it is compulsory on all profiles
devout ivy
Never mentioned AVR
you did
Well aware AVR is something else entirely but I never mentioned AVR outside the context of a basic arduino (though not all arduinos are AVR either, nor are all atmels AVR). Nor did I ever mention floating point in the context of an AVR or arduino
But you did state arduino to be a runtime, which it is not
ARduino for non avr targets of course doesnt use AVR-gcc, though generally uses gcc backends where possible
Only real exception are a small handful of arm targets for arduino are technically threads within mbedOS so use the mbedOS toolchains under the hood, themselves often using gcc - there you're getting closer to a runtime
fickle aurora
so you are saying if I compile this code https://pastebin.com/3y9uLABe in atmel studio using avr-gcc that it will compile? (this is the blink example from arduino's examples) without trying, I know it won't compile, because the first thing it will do is that it will say there's no main() function ... of course, arduino likely uses a wrapper (I haven't been bothered to read their source code), but even then, it's introducing bloat - because as we know calling functions adds overhead due to needing to push variables onto the stack, then saving the current instruction pointer, then going into the function, popping them back off the stack to work with, eventually returning a value, of course more pushing and popping instructions - all when you could have just done most of it in main()
devout ivy
/*
main.cpp - Main loop for Arduino sketches
Copyright (c) 2005-2013 Arduino Team. All right reserved.
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <Arduino.h>
// Declared weak in Arduino.h to allow user redefinitions.
int atexit(void (* /*func*/ )()) { return 0; }
// Weak empty variant initialization function.
// May be redefined by variant files.
void initVariant() __attribute__((weak));
void initVariant() { }
void setupUSB() __attribute__((weak));
void setupUSB() { }
int main(void)
{
init();
initVariant();
#if defined(USBCON)
USBDevice.attach();
#endif
setup();
for (;;) {
loop();
if (serialEventRun) serialEventRun();
}
return 0;
}
https://pastebin.com/SpyGzaY0 so here is my comparison of using Arduino for blinking the LED on pin 13 to making the same function in Atmel Studio with AVR-GCC ... the difference is that the minimum size (without bootloader) that arduino could produce is a 2.55kB hex file and I could produce a 418 byte hex file. Of course, I could have saved even more by making it in assembly
devout ivy
if you really want, you can even observe the build window in arduino and see the gcc invokations. But apparently you're more interested in strawmanning and goalpost shifting
fickle aurora
arduino's hex file was 3.83kB with the bootloader attached
I'm just giving facts, you can compile it and see what it does
devout ivy
No, you're not. You're moving the goalposts having been proven factually wrong about previous statements
fickle aurora
the extra 2132 bytes that arduino adds to an otherwise 418 byte hex file is basically the whole arduino library, even if you don't need it, and it runs through hundreds of extra code that you don't need ... we call it bloatware in programming
devout ivy
Except it doesnt run through it all
2048 bytes are the bootloader, most of those bytes are actually nulled
fickle aurora
the hex file is literally the machine code that it will run through
the chip has to run the entire hex file
devout ivy
It actually doesnt, you are well aware yourself you can jump about
bootloader sector is 2k, so bootloader is 2k but not filled. First executable byte is address 2048
They dont pull any tricks compressing the hex
fickle aurora
without bootloader, arduino's hex file was still 2.55kB
devout ivy
And not all executed as by default optimisation isnt enabled
so libs will be pulled in, but never executed
You can forcibly alter the optimisation level if you wish
Most of the arduino standard lib is incredibly inoptimal versus direct register manipulation. but I never argued that point, and neither did you originally until your previous point about arduino being a runtime was proven otherwise.
Doesn't mean arduino is bad. Its specifically designed to be accessible
Equally I wouldnt go and implement the firmware for some mass produced medical device in it
fickle aurora
it isn't legal to use 328p in medical equipment anyway
fickle aurora
if you do, you break atmel's license agreement in the first place because of the legal agreement you agree to states it must not be used for medical equipment
devout ivy
you can use anything in medical equipment, you just gotta prove/certify it. 328 just isnt precertified
fickle aurora
arduino is also considered a runtime library afaik according to the wikipedia definition, since it's not only a program library, it literally makes a whole user-friendly interface so you don't have to do things like clear the buffer every time you want to send/receive serial data, adding debugging symbols everywhere, validation of logic and data, and handling the directives to the chip for you, such as setting IO configuration
my bad, the 328p is allowed in automobiles according to spec, but there are some other models that I ordered that explicitly prohibited its use in automobiles - and only some chips are licensed for medical equipment - you can't just use any chip there
fickle aurora
And not all executed as by default optimisation isnt enabled
also, optimization has nothing to do with anything other than how the compiler will convert your code into machine code, for example, if you declare a variable but it seems like it is not used in your program, the compiler will omit it - sometimes you need that variable in there, so you use the volatile keyword to prevent the compiler from removing it ... sometimes optimization even makes the end result worse, such as -O3, in some cases that will actually give you worse optimization than -O2, but you have to know how the compiler will convert your source code to understand this
devout ivy
That isnt what volatile does. But you've already tried badly misexplaining enough
fickle aurora
volatile is literally a keyword that tells the compiler "do not optimize this"
devout ivy
Code removal is performed as an optimisation pass, with it all disabled, yes, it will emit unused code frequently
and no, that isnt what volatile does
fickle aurora
Here is another reference taken from the book written by the guys who invented the C language, Brian Kernighan and Dennis Ritchie, ISBN number 0-13-110362-8 page 211 states that volatile keyword prevents compiler optimizations
devout ivy
stand corrected
fickle aurora
it's all good we're here to learn
long raft
you can still use optimizations but none that mess with threading
its just a cache issue
i havent seen volatile used in 20+ years
like union, lol, when disk mentioned using one the other day i wasnt even sure the context
oh jesus remember when people used to inline everything?
god i used to think i was so clever inlining functions i cared about performance
devout ivy
aye when thats one of the most basic optimisations going so most compilers have a decent implementation for doing
so common that some JIT's even do it automatically
devout ivy
Its present as a decorator in .net
long raft
c# has inline?? 
devout ivy
Implemented differently
long raft
time to refactor all my code
devout ivy
but yes, its not super useful
long raft
i dont care about memory i just want speeeeeeed
devout ivy
Nah, modern roslyn is quite good at performance counting method hits and inlining at runtime on hot paths
devout ivy
Some benches, faster
devout ivy
using System;
using System.Globalization;
using System.Runtime.CompilerServices;
public class Utility
{
[MethodImplAttribute(MethodImplOptions.NoInlining)]
public static string GetCalendarName(Calendar cal)
{
return cal.ToString().Replace("System.Globalization.", "").
Replace("Calendar", "");
}
}
or replace NoInlining with AggressiveInlining
fickle aurora
https://pastebin.com/fM8EATg7 this is a very good example code to learn how volatile keyword works. I suggest trying it out, you will really like my example. You have to compile with -O2 optimization flag and try it out, then add volatile keyword to line 45 and then recompile and try again to see how it works
and while you say you havent seen it in 20 years, well, here is a modern example that is still relevant
long raft
its a 10 line program, what about that is modern?
like where would it come up?
like instead of deoptimizing my program and second guessing my CPU i woudl rather use proper thread safety mechanisms
fickle aurora
well you are programming in C#, it's a whole different language
my example is valid for C and C++ - specifically for GCC compiler
long raft
i know its valid but people still wouldnt do it that way now
theyd use a proper semaphore or better suited mechanism than skipping optimizations
fickle aurora
I see the volatile keyword used all the time
fickle aurora
yes it's used in libraries all over the linux platform, which you suggest that I use :D
long raft
the linux platform is a lot of old code
god knows their record on thread safety is rough
fickle aurora
yeah it's probably really bad considering the top 500 servers in the world run that old code
C and C++ will always exist, while python will die a horrible death
long raft
you know there were decades before C existed, right?
it wasnt like the first on the scene
devout ivy
Except python is meant to serve a totally different role to C
meanwhile, even torvalds admits most of the code in linux sucks
fickle aurora
yes everything around it died out but C and C++ have carried on
long raft
its well tested, i wouldnt want to rewrite it, but its not very good C
devout ivy
python has anything but died out
long raft
yes everything around it died out but C and C++ have carried on
COBOL is still very much alive
fickle aurora
well cobol is more similar to assembly
devout ivy
Hell, we've even got efforts underway to improve the linux codebase, by sticking rust in it
long raft
yea i dont think id go back to C id probably learn rust if i had to work low level again
fickle aurora
these days if you want to get into anything useful, C++ is the way forward
long raft
forward to ... what? writing games?
fickle aurora
I will admit, ANSI C is not great, and most of what I do is a combination of C and C++ which is why I often default to calling it C/C++
long raft
i have always called it c+
most c++ programmers think theyre c++ programmers but they are really C programmers
devout ivy
or - are just bad
long raft
they arent bad, c++ is just so huge they have no idea what features theyre not using
if youre not writing obfuscated operator overloads in templates are you even trying?!
youre not even using multiple inheritance?
devout ivy
Good god, multiple inheritance, burn it in fire
long raft
c++ has cleaned up a lot but its not a good language for getting work done, and its not great for writing very low level code, its this middle ground for optimized algorithms that are a hassle to write in C
and you can really write unreadable indecipherable code in c++
devout ivy
I spend a lot of time working with macOS Cocoa UI's. In C#. A framework designed around using multiple inheritance in objective-C. C# doesnt have multiple inheritance
long raft
i was disappointed c# doesnt have multiple inheritance until i realized i have never needed it
fickle aurora
objective-c was good until apple decided to copy windows and use garbage collection
long raft
copy windows? windows invented garbage collection? lol
fickle aurora
I don't know who did or where it came from, I just know that it eats memory like a hog and doesn't let go
long raft
thats not true
devout ivy
That literally isnt what garbage collection is
Objective-C also isnt garbage collected
long raft
as i explained before, garbage collected memory blocks actually use less memory
devout ivy
Intentionally using unmanaged is an option yes
fickle aurora
so are you saying NSGarbageCollector is not a keyword in objective-C?
long raft
but garbage collection improves performance, but as a programmer you have less control over when the delays hit
vague otter
long raft
and it uses less memory in a GC like .net because it compacts the memory, removing fragmented unusable small spaces
devout ivy
NSGarbageCollector is not a keyword. Its a type, for manually adding a collector, much like you can write a garbage collector in your C++ code, Objective-C itself is not collected
fickle aurora
apple pushed xcode to use it long ago, that's why I stopped bothering with objective-c
long raft
allan is a CS luddite, wont use anything from this century
fickle aurora
if you tried to free memory, the compiler would give you errors so I dropped it like a hot potato
long raft
disregards the last 40 years of development and progress as overhead
devout ivy
sounds like a pebkac issue
long raft
also c++ full of garbage collection now
devout ivy
and it uses less memory in a GC like .net because it compacts the memory, removi...
where it can, theres a lot .net cannot compact and that is causing me hell at the moment
long raft
but not really the efficient kind of garbage collection in c++, just the kind that protects forgetful programmers
fickle aurora
you mean the past 20 years of development, yeah that has been going backwards and making code less efficient
long raft
*where it can*, theres a lot .net cannot compact and that is causing me hell at ...
yea large objects are a no go, but they arent causing fragmentation generally
devout ivy
Not just that
long raft
you mean the past 20 years of development, yeah that has been going backwards an...
you are 40 years out of date, not 20
devout ivy
theres a tonne it cant free on the SOH too
Codebase does a lot of managed/unmanaged interop, generally requiring pinned objects
which cannot be compacted
long raft
pinned types?
you didnt want to just allocate them natively?
i suppose its ...the same thing huh
devout ivy
.net 6 has introduced the POH to help alleviate it
long raft
what is POH?
devout ivy
pinned object heap
so theres now a small object heap, large object heap and pinned object heap
devout ivy
items that would go on the small object heap but are marked pinned, now get stuck in the large object heap instead - POH isnt truly its own heap
long raft
is there an advantage of the pinned heap vs just marshalling some heap memory from the OS?
devout ivy
Unsure in all honesty, but in brief tests it at least solves some of the issues we were having
as unfortunately, as soon as the compactor encounters an item it cannot compact, it stops compacting the entire generation
so just 1 pinned object can basically just totally fuck compaction
long raft
i feel like i can get around most pinning these days
long raft
whether span is using it or not, im unclear, but damn i can get some great performance with span
devout ivy
codebase is mostly net 4.7.2, with 1 assembly in 4.8, and a few making it over the line into standard 2.0 as part of the road towards moving to net 6
long raft
ive been working with ffmpeg a lot lately and im getting speeds the same as the C routines available in the library now
fickle aurora
with which language, python?
fickle aurora
well C# is close to C++ in optimization, python on the other hand is like going to disney world on a horse instead of an airplane
long raft
I have to use span to get C speeds
long raft
The naive for loops don't cut it
fickle aurora
while/do while loops are better than for loops
devout ivy
often compile down to same thing now if you're using a compiler from the last decade
long raft
fickle aurora
the reason for loops are slower is because of the conditional statements it has to run through
long raft
i always ++i because its faster than i++ 
fickle aurora
yes you are right ++i is better
long raft
but i want the conditional statement
fickle aurora
in some cases, the goto keyword is better than an if statement or function call
long raft
for performance? usually.
but you have the shame of using a goto statement
this is ...like inlining everything
just let the compiler do its job
fickle aurora
compiler is not always right
long raft
im not worried about individual clocks
no. but its right more often than my primate brain
devout ivy
I'd rather waste a clock than write a goto ever
Sounds like you love unreadable bug prone code
fickle aurora
as this example https://pastebin.com/fM8EATg7 shows that compilers aren't perfect
long raft
i dont think goto is that buggy, its actually pretty easy to debug them
they got a bad rap because it made code difficult to prove
i guess they have a stink of amateur to them
fickle aurora
the bad rap is programmers who don't want to learn more about the language they use
long raft
a lot of them were told never to use goto
devout ivy
nah, more programmers that know just because it was in the language once doesnt mean it should be, but equally you can't delete it
long raft
java did, didnt they?
fickle aurora
MIT certainly never told people never to use goto
devout ivy
not but brian and dennis did
long raft
if break were smarter we wouldnt need goto
fickle aurora
the problem with using a function call over goto is that a function call will push rbp onto the stack even if it's not needed
goto does not push rbp on the stack, so you have to be aware when you use it
long raft
these are generally not performance concerns tho, allan
fickle aurora
you can break a lot of stuff using goto, but again if you can think about what the code will do on the assembly level, you can use it for performance
long raft
theyre very easy to track down tho
balmy tangle
hi, can you tell me if there is a public logistics api to check base resource counts?
long raft
there is not
fickle aurora
well here is a code example of using goto to jump to an error output, so rather than have a separate function for it you can simply just jump to it within main() https://pastebin.com/6jLY5APs
this is the functionality of C that other languages just don't have the ability to do as fast as C can do without intensive overhead of runtime libraries, basic string and character manipulation
long raft
yea but i can write it in one line
so if you ever want to write something more complicated than replacing commas or counting numbers, good luck dumping all of your time into C so you can get a 5% performance increase
also, zero terminated strings are ridiculous in the 21st century
fickle aurora
lol
long raft
seriously, you cant loop unroll it, and youre gonna pay a time penalty just to get the string length, etc
fickle aurora
I'm pretty sure everything still uses null termination in C#
long raft
just because youre using C doesnt mean you have to do things in an old and terrible way
it for sure does not.
even pascal (from before C) wasnt using null terminated strings
fickle aurora
well you can disagree, that's fine, but I'm saying that it's still a thing
long raft
they simply include an integer with the string length
its still a thing ... yes. in C. in old C. its an anti pattern
there are alternative string libraries for C that wont use zero termination
if you care so much about every last ounce of performance ...
You live in a time of abundant processing power and memory, but developer time is still the same
It's time for modern tools to solve modern problems unless you legitimately have memory constraints
fickle aurora
all the modern tools do is automate the things that lower level requires you to do yourself
long raft
Yes and saves the most expensive part- development time
fickle aurora
so like null termination, yeah you don't have to do it, but the compiler is certainly doing it
if there was no null termination at all, your program would cause a buffer overrun on the first syscall
long raft
The library does it the compiler has no knowledge of strings
I have explained strings before, I'm not doing it again
Dont post me some unrelated sample
neat fossil
you're no true coder if you've never messed up the iostream library
long raft
I suppose the compiler turns string literals into asciiz but you dont have to use those and you dont have to terminate an array of characters with a zero, which is a valid character you cant even use in asciiz
fickle aurora
it doesn't have to be in ascii, null is just 0, which is the same 0 in unicode, utf-8, etc
long raft
Such a short sighted string convention
Allan, imagine the first 4 bytes of your string were an integer specifying length and your libraries used that
fickle aurora
the reason it is designated as '\0' in C is that when you use single quotes, you are referencing an ascii character
fickle aurora
and on the CPU level, by the way, ASCII is still the standard for text
long raft
What. No.
fickle aurora
unicode is not supported in real mode
long raft
There are some old interrupts that do, but c doesnt use those
devout ivy
I like how once again Allan has derailed into irrelevance. String encoding and termination are different things
And null termination isn't used a lot outside of c anymore
devout ivy
As derp said, most are length prefixed
tawdry zenith
hi there
fickle aurora
well some will just not see it I guess
devout ivy
If length prefixing made no sense and null termination was just better in all scenarios, then they'd be null terminated everywhere, and oh look they aren't
C regularly does it the wrong way, regularly the worst way
Doesn't mean I hate C, it has a place
long raft
C has many bad conventions
devout ivy
But it is not the best tool for every single scenario ever
fickle aurora
Two of the four text-string processing instructions specify string length explicitly. They use two general-purpose registers (EDX, EAX) to specify the number of valid data elements (either word or byte) in the source operands. The other two instructions specify valid string elements using null termination. A data element is considered valid only if it has a lower index than the least significant null data element. from page 308 of the Intel x64 developers manual
apparently even Intel instruction set uses null termination in machine code
devout ivy
Oh look. Only half use null terminators and the other half use prefixed
In your own quote you proved yourself wrong
devout ivy
2 prefix. 2 null suffix
long raft
Im out of date on c++, does everyone use the stl string library?
I know the standard library isnt using zero termination
devout ivy
Prefixing generally tends to result in it being trivial to perform a bounds check before even starting to parse the string or copy it to the buffer. Both improving performance in the error case and reducing chances of overrun. Additionally makes it trivial to perform a copy operation of known memory length rather than running until a particular character is reached - performance difference there varying on chip architecture and exact implementation remembering even the different intel generations aren't cycle accurate to each other
long raft
also the ability to use the 0 character
and unrolling loops because you have a known length ahead of time
fickle aurora
lol
long raft
all of this you can do with C!
devout ivy
Garbage collected object oriented C is very possible
long raft
object oriented C? 
long raft
imagine working on a C project so big you need to OO it
fickle aurora
it's really bad when programmers think that just because their high level language doesn't do this or that, that their code works the way they think it works
devout ivy
It's never as nice as a language where classes are a genuine language feature. But you can always shoehorn any paradigm into any language. There's even libraries for it in many cases. gObject for c for example
long raft
yea strings are so basic they really should be a basic type, and implemented well
fickle aurora
everything eventually goes down to low level system calls, and those system calls are either done in assembly, C or C++ on all the modern operating systems
long raft
god i just cant imagine doing things in C/c++ anymore, it was so much work to do the most basic crap
its all machine code in the end, allan. and even that is basically byte code reinterpreted by the CPU
this idea that youre working at a native CPU level is false
fickle aurora
there will never be system calls written in C# for example
devout ivy
And once again
long raft
but there are
fickle aurora
but there aren't
long raft
you dont even know C#
devout ivy
Do I have to keep pointing out that there are tools for different jobs
fickle aurora
ntdll is not made in C#
long raft
c# can still invoke C calls
fickle aurora
all the system calls are in there
long raft
in fact it can invoke different types of function calls
fickle aurora
function calls are not system calls
long raft
well they are when its a system library
you know theres an entire layer between machine code and actual instructions that are run, right?
fickle aurora
a system call, for example 0x2c means ExitProcess() on Windows 10 on all the updates (but you can't count on it because Microsoft keeps changing the system call codes)
long raft
are you talking about interrupts?
devout ivy
Most of the .net standard library is pure C#. But file I/O etc is always gonna have a system call
Win32.dll for example on windows
long raft
i have to use interop system calls all the time in c# directly
fickle aurora
you have no choice but to call either NtTerminateProcess() in ntdll or ExitProcess() in kernel32 to exit properly, instead of using code 0x2c cause one update to the next, MS could change the system call tables
long raft
your knowledge of C# and .net is very spotty
devout ivy
On that one he is right derp
fickle aurora
.net is just wrappers for the underlying win32 API
devout ivy
Process.exit and application.exit both result in an eventual call to win32
fickle aurora
programming in .net does not add anything to the OS
long raft
ok but im talking about interop
fickle aurora
everything literally comes down to the win32 API, that's the bottom before it is translated into syscalls which are stored in tables
long raft
do you want to be a professional C programmer, allan?
devout ivy
Arithmetic, doesn't touch it. File IO or networking etc obviously does
fickle aurora
yes basically everything, for example, printf is not a system call, it is actually a call to NtOpenProcess
the standard output is just a file IO on very low level in windows, and that has been there since windows NT when windows became posix compliant
devout ivy
Windows hasn't been POSIX compliant in a very very very long time.
fickle aurora
you have to pass -11 in decimal to specify "stdout" for example
long raft
but in c# you can just define an external function call (using a C style method entrance) and call the system libraries directly
[DllImport()]
theres all the platform invoke stuff
fickle aurora
what are you considering as system libraries?
fickle aurora
which DLL's are you referencing to do this?
long raft
you can do it with any DLL
fickle aurora
well ntdll.dll is the one I know of with most of the low level syscalls, the others like kernel32.dll are higher level wrappers for the syscalls
long raft
of course your portable program becomes windows dependent then
devout ivy
of course your portable program becomes windows dependent then
This is why there's usually Type.cs, Type.Windows.cs and Type.Unix.cs in the .net source code
devout ivy
All as partials. Occasionally split further as Linux and macOS specifically if required
Not everything in .net invokes win32 these days ;)
None of the string handling code touches kernel anymore. It's all pure c#
fickle aurora
well it's wrappers to win32
devout ivy
It's not
long raft
its wrappers to any external dll
devout ivy
There is stuff that is
fickle aurora
the wrappers handle the message translation and everything for you
devout ivy
Stuff that isnt
fickle aurora
it's more complicated than you think
long raft
but also youre not aware of the capabilities of c#
devout ivy
So you literally thing absolutely every single line of c# is a wrapper to win32
You know how absurd that is. In a runtime that isn't even win32 specific
fickle aurora
to make a window, you have to paint the screen, capture messages for the window, issue callbacks, translate messages into instructions on when to repaint the window on the screen, etc ... the higher level libraries do that automatically for you, but I assure you that it exists, otherwise it wouldn't work
devout ivy
That I've repeatedly pointed out, does involve where appropriate
Point out where I said using a GUI didn't invoke kernel
fickle aurora
you are somehow thinking that everything is magic, it is not
long raft
if we were still stuck using c# computing would suck now, itd be 20 years behind where it is now
devout ivy
Point out where I thought something is magic
long raft
still implementing linked lists lol
devout ivy
I even specifically mentioned files as an example of something that invokes win32
Yet strings as something that doesnt
long raft
we will never break through this wall of "C is the best, even tho I don't know other things"
devout ivy
Clearly painting a window will. Don't dispute that
fickle aurora
it has nothing to do with C, we're talking about windows native API
fickle aurora
well at least you can admit it
long raft
lol
long raft
whats the largest program youve written, allan?
fickle aurora
lol
devout ivy
If I'm computing Fourier transforms on captured data from an oscilloscope, am I invoking the kernel
Maybe I might in order to spin up and manage a new thread if appropriate for the workload (unlikely in a 1 dimensional fft) and maybe for a mutex. Otherwise. No. Not during processing.
And no, the corresponding code in . net or even in python, won't invoke the kernel at all for that
long raft
whats the mutex for?
devout ivy
Syncing the thread -if I even make one, and if required. So likely no mutex
long raft
hmm well you can sync a thread without a mutex.
kind of a specific type of thread safety
devout ivy
If I then log that data to a file post computation. Then I'll clearly be invoking the kernel
fickle aurora
if everyone thinks like you guys, in 20 years there will be no computer updates because there won't be anyone who knows how to program for the CPU
long raft
i suspect i have years more C experience than you
perhaps people who think like us are because they know more
fickle aurora
I doubt it
long raft
C is so easy its not like there will ever be a lack of people who know it
devout ivy
You do realise my degree is embedded systems programming. I have programmed in C, for both x86 and assorted mcu's. I have done assembly in msp430, avr, mips and x86. I work in c# as a profession sure. Doesn't mean I don't know C. I have written CPU emulators. I have written an assembler
long raft
yea and ive written x86 assembly for 5 years
fickle aurora
thinking that "C and Asm are outdated" is thinking incorrectly and leading to a major disaster in the future when we will not have anyone left to keep the API up to date
devout ivy
I never said out dated
long raft
oh ill say it. its outdated
devout ivy
I said not the right tool for the job in all scenarios
Because it isn't the right tool for the job
long raft
shit, you could do all the same work in rust
devout ivy
Rust is not quite at a mature 1.0 yet.
long raft
ive never been one to wait for 1.0
devout ivy
That's all I have against it
long raft
i cannot imagine being a programmer these days without knowing modern languages, like java or c# or javascript...
i would never hire that person
theyd spend a year doing the work a fresh college grad could do in weeks
devout ivy
But frankly I wouldn't employ a person that thinks the only way to code is in C, if it means they waste a week writing something in C that could have been written in 5 minutes in python and the python performs the job satisfactorily
long raft
and yet, i think every programmer should know C, youd be so hamstrung without it
C is basically pseudo code now
devout ivy
Equally. I ain't gonna go and have some ridiculous number crunching done in python.
long raft
remember the old days when youd find an algorithm example in delphi or fortran? lol
long raft
Equally. I ain't gonna go and have some ridiculous number crunching done in pyth...
but you could. why not. throw more cheap computing power at it
devout ivy
Eh, libs like scipy I guess, albeit, their main crunching logic is in C
long raft
sure, if its done for you might as well use it
devout ivy
But would again depend on project requirements
long raft
but ive got no problem if my c# program runs half as fast, i can just run it on two cores
its not like im waiting for a slice of time to run my program on a univac
devout ivy
C# generally gets you 90% of the way there these days anyway, and probably with half the effort
fickle aurora
yes it is faster to develop (not faster to run), but thinking that it will replace C for good is wrong thinking
devout ivy
Where did I say it will replace c for good
devout ivy
We had someone at work think C was the only way to do things. So we had a little challenge for fun.
Project Euler, first 10 problems
long raft
it reminds me of the people who think manual transmissions are the be all end all. omg the mileage is better. youre not a real man if you dont use a clutch, etc.
until you break your foot or you want to drive and eat at the same time
or lend someone else your car
devout ivy
My c# solution compiled and executed faster on a Jetson nano with .net core 3 on Ubuntu 18, faster than his first 5 problems ran in C on an i7-7700 also under Ubuntu 18
long raft
because he couldnt implement quicksort without 2 weeks lead time?
devout ivy
Pretty much
long raft
i assume it was his poor choice of algorithms
fickle aurora
ubuntu has other problems, C is not one of them
fickle aurora
ubuntu is one of the worst linux distros
devout ivy
On a 4 core 1.5ghz cortex a57
long raft
im working on a video editor, and it compiles my c# code and runs it on the gpu, the first time i ran it on arm on my jetson nano my brain almost exploded
i wrote it on amd64 on windows and then it runs on a jetson nano lol
with cuda.
i still can barely believe it works
fickle aurora
they still have a bug in the OS where if you read from an SSD, it will try to read from your DVD burner (if you have one) and throw an error
have you tried installing apps from the GUI instead of terminal window? lol
devout ivy
i still can barely believe it works
See it was great when mono got us net framework on Mac and Linux. But mono was very rough round edges. Netcore, droolworthy
fickle aurora
it is so bad, if the GUI is so broken they need to just scrap gnome and make something else the default window manager
fickle aurora
I just use plain linux, like slackware, gentoo or rasbpian
devout ivy
Unfortunately, I rely on mono currently at work
long raft
ooof im so sorry
fickle aurora
I have no preference, but I don't use GUI
long raft
a lot of my stuff i compile now for musl/alpine with .net its such a joy
long raft
I have no preference, but I don't use GUI
well if you think ubuntu is among the worst you must think there is a best
fickle aurora
yes I think macOS is the best linux distro, and before you laugh, I know it's not linux, but to be honest, how many issues does macOS have compared to Ubuntu?
long raft
what was that old linux distro the super linux dorks used to use. gentoo?
i feel like they all migrated to arch
macos full of issues as far as i know
devout ivy
Yeah, WPF+netframework 4.8 on windows, GTK+Mono/.net 4.7.2 BCL on Linux. Xamarin.mac on a Mac, under the hood being a weird hybrid of mono 6 with a modified net 4.8 bcl with standard 2.1 compatability that regular net 4.8 doesn't have
fickle aurora
at least on macOS, the GUI actually works
long raft
and yea ... its not a linux distro, you cant say that
devout ivy
Working with Mac professionally
It's super buggy
Usb stack in particular is horrific
long raft
as much as i hate apple i cant deny their gui is pretty good
devout ivy
Eh, until you mix retina and non retina displays in a single multi monitor setup
fickle aurora
I just stick to windows, it makes no sense to switch to X
devout ivy
At which point it can't decide on whether it wants to anti alias fonts properly and bugs out badly
long raft
my only experience with retina issues is in web development front-end. javascript ... and high resolution is funky
no X is terrible its definitely not the strong suit of linux, im just curious if you dont like ubuntu which alternative do you think is better
devout ivy
It has different font kerning and anti aliasing settings for retina v non retina
long raft
i guess you think theyre all better?
fickle aurora
I have several computers, the mac one never gets booted even though it is nice to use macOS
devout ivy
And if you have multi monitor setups on a Mac. That setting will be wrong for at least 1 display
long raft
im stuck on windows as long as visual studio is exclusive to windows
devout ivy
Resulting in the font rendering as if it's smudged on either display
long raft
i cannot imagine a less capable dev environment
fickle aurora
I'm thinking Fedora is going to be the one I pick after windows 10 goes to crap in a few years
long raft
hmm fedora ... you dont hear much about it these days
devout ivy
Fedora is in my pile of least favourite distros
long raft
i personally hate RHEL/fedora
devout ivy
Above opensuse
long raft
debian was always too concerned with stability for me, too many problems
devout ivy
Reminds me, need to check if the mono package in Fedora is still broken
long raft
i do very much enjoy alpine
devout ivy
It's a fedora maintained package, the rpm direct from xamarin is correct
Had to work around it to get works software to install on fedora
long raft
it seems almost irrelevant now, i dont run anything outside of a container now
the hosts are so ... bare metal im barely concerned with the distro if it can run podman or docker
devout ivy
Opensuse, they've switched package manager to zypper. Zypper does some funny funny things
fickle aurora
I tried to install Fedora a couple years ago and the installer didn't work
devout ivy
The installer is still broken on a 1920*1080p screen
long raft
allan have you used the linux subsystem on windows?
fickle aurora
cygwin?
long raft
no...
long raft
wsl 2 is basically a virtual machine with tighter integrations
devout ivy
Windows subsystem for Linux. Added in win10
long raft
its an easy way to get a linux command line on your linux machine. you can run it as a vm or more like cygwin with the native compilation (wsl 1)
devout ivy
And with 11. Wsl2g
long raft
but the new version of wsl2 with windows update 21h2 has graphics card support
windows 11 or 10
fickle aurora
I tried cygwin long ago on a 200mhz pentium 1 and it ran horribly slow compared to windows 95
long raft
cygwin was ... fine for performance but buggy because the library support
wsl is a feature for windows you can enable, and you can install any linux distro
devout ivy
WSL is more Linux kernel in hyperv. With mapped storage bidirectionally. With WSLg, X forwarding too
long raft
its a great way to get yourself a fully working linux command line on your windows machine
WSL1 is less like a virtual machine
its more like cygwin...
devout ivy
Can do weird and whacky things like launching windows applications from the Linux bash shell
And vice versa
fickle aurora
I already use the command line through ssh, but mostly on FreeBSD
devout ivy
Powershell window on windows, using Linux installed wget, in a windows directory
long raft
its just nice to have linux tools on a windows machine that are reliable and predictable and come from a known distribution
yea ssh is fine but you have to run another computer. its nice to have linux right on your windows machine
freebsd huh?
devout ivy
Fun macOS bug for you guys. Present since apple first introduced USB controllers on PowerPC products
fickle aurora
I am not a fan of the new FreeBSD releases, the project went down the hole since 2010 or so
but the older releases were really good
devout ivy
If it encounters either 2 usb devices enumerating at once. Or 2 usb devices disconnecting at once. Or 1 connecting and 1 disconnecting
long raft
yea bsd has been sadly in decline with the popularity of linux
devout ivy
It can get the usb driver in a deadlock
long raft
and now that linux has zfs i have no reason to stick with freebsd sadly
containerd > jails, linux hardware support is just better
devout ivy
All usb devices on the system cease functioning, on every controller in multi controller systems
long raft
always loved the monolithic kernel of bsd tho
devout ivy
If any usb device (including the one that was unplugged) was an audio device, you also lose the audio stack
As every Mac except the newer M1 based ones uses usb based Bluetooth, you also lose Bluetooth
long raft
its almost inexcusable apple have bugs for so long. its not like theyre supporting a billion combinations of hardware
is boot camp dead for m1 hardware?
devout ivy
Absolutely
long raft
oof breaks my heart
devout ivy
Remember M1 is arm not x64
long raft
i know but windows is on arm
long raft
.net 6 already compiles for m1 which is cool
when i recently updated the tiler people use for foxhole maps, i added an m1 binary
devout ivy
Bootloader isn't just not locked. It's intentionally unlocked and third party OS booting is explicitly allowed
Linux porting is going extremely well
long raft
if i were primarily a linux user id heavily consider buying new apple laptop
i hear that m1 is fast AF
devout ivy
Microsoft have stated they have no interest in porting windows at this time
long raft
i am not surprised
devout ivy
And nobody has so far looked at taking on the challenge in doing it as a community port
Ie, pi style
long raft
no .. itll have to be virtual machines or something
fickle aurora
is that what KDE looks like now
devout ivy
Yes
devout ivy
no .. itll have to be virtual machines or something
Already done
long raft
with video accel?
fickle aurora
it is sad that gnome killed itself
devout ivy
Parallels supports windows 10 for arm
fickle aurora
used to be the best
devout ivy
Dunno what state of acceleration is
long raft
if you cant passthrough GPU then it wont be much use for gaming
or anything really
its not like you cant edit a spreadsheet in osx
long raft
thatd be a deal breaker for me, although if youre a gamer using an apple laptop maybe youre not a serious gamer
devout ivy
No but on the M1 max, there's a huge amount of GPU performance on tap for something that doesn't have a discrete gpu
fickle aurora
the thing is mac has a better chance of getting game developers than linux, because of the GPL problem
long raft
GPL isnt really an issue for game programmers
fickle aurora
GPLv3 will forever be the thorn in linux
fickle aurora
it should have been a requirement that all libraries are ineligible for GPLv3, and require libraries to only be allowed under LGPL
long raft
i cant think of any example where gpl is an issue for game developers
fickle aurora
or at least BSD licensing
long raft
well ... it wouldnt be linux with bsd licensing would it
devout ivy
Should also note. The usb issue above is still present on m1
fickle aurora
LGPL allows linking to libraries, that's where the advantage is
devout ivy
The only difference. The cause of the bug is a race condition. It expects the usb event queue to be processed in a particular time span during the re-enumeration
long raft
licensing is complicated
fickle aurora
you will never get game development with GPLv3
devout ivy
It's harder to exceed the timespan on M1, but possible
fickle aurora
because game devs need to get paid and don't want to throw source code out
fickle aurora
if you link to a library that is under GPL3, it's no longer closed source
fickle aurora
depends on the libraries you link to
long raft
libraries used for games are not gpl3
devout ivy
Gplv3 isn't allowed in linux
long raft
vulkan is apache license, opengl is similar to BSD
there are some gpl sound libraries you can choose not to use
fickle aurora
I wish that were true, lol
long raft
really i see no issue developing software for linux, you wouldnt have to share source
theres countless examples of developers not sharing their code
devout ivy
More so. Never has anyone at the Linux foundation raised it as an issue, despite the increasing popularity of gaming on linux
long raft
one could argue a lot of linux gaming is operating with ...windows binaries lol
devout ivy
There's also the GCC runtime exception
long raft
but yea theres a ton of stuff on steam compiled for linux, no source code
as much as i deeply hate gpl its never really caused me any development problems
agpl is the real monster.
devout ivy
You have permission to propagate a work of Target Code formed by combining the Runtime Library with Independent Modules, even if such propagation would otherwise violate the terms of GPLv3, provided that all Target Code was generated by Eligible Compilation Processes. You may then convey such a combination under terms of your choice, consistent with the licensing of the Independent Modules.
long raft
its so counter to the intent of gpl its disgusting
long raft
```You have permission to propagate a work of Target Code formed by combining th...
lol what do those words mean? are they defined?
target code? "combining"?
"Independent" rofl
devout ivy
Much of linking against the Linux API is already either allowable by the libraries existing license, gplv2 or the GCC exception
That tl;Dr and FAQs down to you can link against quite a few gnu libs in proprietary software in many cases
long raft
sometimes i wonder how significant containers are affecting gpl
like, if i externally link ffmpeg, but then i use the gpl3 version and i make a docker container that installs that one ...
isnt it kind of like the end user doing it?
do i have to provide a dockerfile instead of the image?
how is everyone else posting docker images without violating gpl
devout ivy
End user is also accountable for licence violations
long raft
dockerhub is in theory one big gpl violation
fickle aurora
it's a legal can of worms and convoluted so that you never know the grey area until you put effort into a project and then have to hand over all your work for free to some guy just because you linked to something trivial
long raft
aha allan, theres a solution
if youre forced to release source code, release it as agpl
then nobody can use it
long raft
its stability is probably the biggest reason
the kernel is pretty damn rock solid
i find myself wondering if this is because largely of the vision of a single programmer, and how large can a project grow centered around the ideology of a single person
fickle aurora
well if Linus switches the codebase to C# I'm not sure that it will still be rock solid and efficient
might as well rewrite socket.h in php and see how that goes, it would be a similar move
long raft
devout ivy
Note neither of us ever suggested doing that
But. Rust is already working it's way into the kernel to replace c/C++
long raft
Yes because a safe language would be bad for stability?
Linux is not great because of the language it's written in
devout ivy
Also note that the Linux kernel is also not bug free
fickle aurora
I'm thinking more along the lines that it would become slow like windows and then people would fork the project to another thing for server dev
devout ivy
It's only very recently a major privilege escalation attack was found and thankfully rapidly fixed
long raft
Windows is... not slow, unless you mean specifically explorer.exe
fickle aurora
"prevents all segfaults" sounds good until all the memory leaked from other places get read by malicious code that keeps running indefinitely due to the language keeping it from crashing
crashing is sometimes a good thing, it tells you when you made a mistake
long raft
fickle aurora
I read about rust, wasn't excited about preventing crashes due to bad code
you can already do that in C actually by managing signals
long raft
do you like your car without an airbag and safety belts too?
hmm and thats a 10/10 analogy
like you prefer older cars before they had crumple zones and airbags and only had a lap belt and no anti lock breaks
fickle aurora
crashing due to buffer overflow is actually more like an airbag than preventing the crash and allowing all the sensitive data to be leaked from another process
long raft
because they used less fuel (not really) or drove faster
fickle aurora
it's actually more secure to crash than to not crash, because the crash prevents malicious code from reading memory of another process
long raft
crashing potentially your whole OS just to know you have a small bug somewhere is like thinking seat belts kill more people than they save
fickle aurora
well that is a linux problem
long raft
thanks, i dont need my entire computer to die because somebody started counting at 1
fickle aurora
linux purposely crashes when something goes wrong
long raft
anyhow you can turn that off in windows
you can surely compile a rust program without that
fickle aurora
would you rather have an unbootable system or a system that crashes when something goes badly wrong?
long raft
wait why would it become unbootable?
fickle aurora
if your buffer overflow causes garbage to leak into the kernel and save to system files, you will be left with an unbootable system
long raft
why would it save to system files
thats not like something that would happen accidentally
fickle aurora
I absolutely prefer the computer to crash rather than have the system files corrupted and prevent booting
because program instructions and virtually everything is in memory
long raft
is this some scenario where a virus is doing it?
why would the kernel write over the boot loader
fickle aurora
it doesn't have to be a virus, if random junk is overwritten to the memory location of the kernel files, it can very easily trigger a write instruction
long raft
to write over the kernel thats compiled and compressed as an image?
fickle aurora
it could even be worse if it happens to have elevated privileges and write to other places, like bios
fickle aurora
that can be done
long raft
its not going to do that from a random piece of memory overwriting, that would have to be targetted
thats not going to happen from randomness in any likely reality we are in
fickle aurora
segfault is done on purpose, it's designed that way to protect the system, that is literally a security feature
fickle aurora
like I said you can disable it if you really want to, by capturing the signal and preventing the program from exiting
usually programmers do that so they can return an error message
fickle aurora
I think it's bad to abuse the try {} method of C# to prevent crashes, absolutely
long raft
exceptions are not segfaults
i should point out c++ also has exceptions
its not ... abuse its a conditional statement (logically)
fickle aurora
the intention is for doing things like sending an error msg to the user, recording it in a logfile or something, but if you let the buffer overrun or whatever to continue, it can lead to some serious security holes
fickle aurora
if you did it in openbsd, theo would give you a long lecture about it
long raft
you cant catch a segfault.
ok ...maybe im understanding you. im trying. you would prefer the OS handle reading/writing out of bounds than the program itself?
fickle aurora
long raft
oh i guess people can catch that signal
fickle aurora
in C, it's sigaction(SIGSEGV, &<function for handling seg fault>, NULL);
long raft
im not aware of anybody doing this intentionally, allan
for a particular use case maybe, like a host program that runs another process maybe
so im still confused. you would like every program on a computer to have access to the memory anywhere else on the computer?
fickle aurora
yes because if you don't prevent the program from overrunning the buffer, it is a security issue
long raft
but these languages do prevent you from overrunning the buffer
fickle aurora
of course they do and that's why they are bad, but on the plus side, they are good for making development much smoother
long raft
why is it bad? i thought you agreed it was good to prevent that
fickle aurora
ideally you want to use that to your advantage to debug the program, but you don't want to have your users deal with the security implications of it
long raft
but you wont catch every segfault by just debugging
its the same reason you wont catch integer overflows
youre not testing every single scenario that can happen...
fickle aurora
well integer overflow isn't a thing because integers are a set bit length
unless you mean arrays
long raft
no
fickle aurora
arrays can overflow, you need to terminate them
fickle aurora
integers have a set bit length, so like a regular int is generally 4 bytes, whereas a string can be any number of bytes from 1 byte up to whatever your memory can hold
long raft
answer my 'riddle'
fickle aurora
an array is almost like a string
long raft
i am not talking about arrays
fickle aurora
on the CPU level, there is no distinction between data types except for floats
fickle aurora
but then again, floats are simply just considered floats in a specific register
it depends a lot on the architecture
some will keep it to 0xff and set the sign flag and carry flag
long raft
then do you know what integer overflows are?
fickle aurora
yes there is a flag for overrun
long raft
do you test for integer overflows every time you do an add?
fickle aurora
you can
long raft
but do you
fickle aurora
libraries will handle it for you if you use libraries, otherwise you can check eflags
fickle aurora
usually you test input to ensure you never go above the max value
long raft
you would have to prove your code
testing would not be enough, youd have to prove it by induction
otherwise one could creep in
or you can just ... test it every time
fickle aurora
the thing is, one byte isn't going to cause a problem, because the CPU will only access one memory location for a one byte value
long raft
same thing for buffer overruns