the work is never complete

1 view
Skip to first unread message

Raoul Duke

unread,
Jan 9, 2022, 6:48:39 PM1/9/22
to pi...@googlegroups.com
what are the holes in each and every language sdk?

eg. lots fail to handle unicode well at all. 

eg. swift doesn't have a stack collection. 

eg. java's treemap is apparently rare elsewhere. 

eg. haxe is just crap in many ways, maybe since it overly cross platform. 


Raoul Duke

unread,
Jan 9, 2022, 6:49:17 PM1/9/22
to pi...@googlegroups.com
eg concurrency is a four letter word just about everywhere 

Kyle Hayes

unread,
Jan 9, 2022, 7:20:37 PM1/9/22
to pi...@googlegroups.com
I think part of why the SDKs all kinda suck right now is that the bases on which we write programs are in the process of changing.

Concurrency and parallelism are tricky ones.   I think we are in the middle of a shift in how things are done.   OSes provide several primitives that SDKs can build on, but if you try to use the "wrong" model on an OS that does not really support that model, you may find that your performance is abysmal.  

For example, it used to be that there was roughly the Windows model and the POSIX model for IO, everything else was in the noise.   Now you have Linux starting to really diverge with io_uring.  It is a great way to do things but is most definitely not as initially simple as Java's "just use lots of threads and let everything block" model.  Apple has decided that GCD is the way to go for macOS (and I assume iOS).  Since mobile is the largest number of devices for non-embedded programming, that has a lot of weight.

How all that is going to interact with processors with more cores, I am not sure.

But I think that the SDKs are going to come later, not sooner, to the party until at least one combination of OS and concurrency model/IO model grabs the gold ring.  The hardware and the OSes are all trying to make use of lots of cores and layers of smart IO (i.e. network cards are smarter and GPUs are talking directly with MMUs etc.) all while using as little power as possible.  Computing is getting more and more heterogeneous with GPU compute and ML logic peeling away a lot of FLOPS from the main CPUs.

Interesting times.

Best,
Kyle


--
You received this message because you are subscribed to the Google Groups "PiLuD" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pilud+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pilud/CAJ7XQb4Gqw29XfTwFxcF97pq%2BKwh7HG5C4VmGK93_vVKtytcLg%40mail.gmail.com.
Reply all
Reply to author
Forward
0 new messages