spullara

50 ms·

841.
by spullara 4y ago
All python code generally does is call C/C++ code and you're telling me it is slow to do that as well? Yikes.
842.
by spullara 4y ago
statics don't override
843.
by spullara 4y ago
For Wavefront I got this down to about 1.5 bytes per metric reported (on average across customers) where each metric was a tuple of: name, source, timestamp (s), value (int) and a set of name/value pairs Key to optimizing it was having
844.
by spullara 4y ago
Trying this: Without a terminal, you were never more than a question or a shout away from almost anything you wanted to know It suggests results that don't include terminal (but of course that is a critical word for this quote), select
845.
by spullara 4y ago
People are constantly saying these things about google and when I search for: "Stories set in the Culture in which Things Went Wrong tended to start with humans losing or forgetting or deliberately leaving behind their terminal" I
846.
by spullara 4y ago
When I was a teenager, I got dropped into a shell on a VAX by doing the same thing when it was trying to identify the terminal type.
847.
by spullara 5y ago
This is a good test to see if they have what it takes to survive in the metaverse. /s
848.
by spullara 5y ago
Reminds me a lot of https://sigmacomputing.com
849.
by spullara 5y ago
There is no possible way for someone just out of college to actually understand the tradeoffs. Young people have little to no experience with building a career or what it entails.
850.
by spullara 5y ago
It absolutely will show them because they cost nothing. 0 will show up in the profiler which is how much deallocations cost with GC.
851.
by spullara 5y ago
It is definitely both functional and imperative depending on author :)
852.
by spullara 5y ago
What do you think won't show up in a profiler with GC? With the right profiler I can see the entire heap as well as where every object was allocated and all those deallocations are basically free. With a modern collector like ZGC, max
853.
by spullara 5y ago
"I don't see any truly large, complex, momentous applications or platforms written in Python." Except YouTube LOL. Don't get me wrong, Python is not fast. But every time you run into a bottleneck you just rewrite it in C
854.
by spullara 5y ago
Reference counting absolutely has pauses. Try allocating millions of objects in a block and then exiting that block. Hell, even malloc pauses when fragmentation gets too high.
855.
by spullara 5y ago
You don't consider Scala to have a reasonably expressive type system?
856.
by spullara 5y ago
This was my experience as well at a small engineering school.
857.
by spullara 5y ago
I very carefully batch things / break them up to make it work well since none of the models I am using individually need 15 min for a single frame.
858.
by spullara 5y ago
I have an entire video analytics pipeline running about a dozen containers for inference. Works great.
859.
by spullara 5y ago
Correct. Which is what most of the complaints are about, knowingly or not.
860.
by spullara 5y ago
It would be pretty easy if they just reverted it to reverse chronological and lose the engagement win they get by having it.
861.
by spullara 5y ago
In the US your choice of type of phone (iOS or Android) is extremely correlated with the choice of phone of the people you communicate with.
862.
by spullara 5y ago
The first time you could every truly run two Mac OS programs at the same time (non-cooperatively) was when you were running two copies of ShapeShifter on an Amiga. Pulling down the Mac OS desktop screen to reveal the other Mac OS desktop
863.
by spullara 5y ago
You can get up to 10 Gbit in Chattanooga with a similar system: https://epb.com/
864.
by spullara 5y ago
When did Disney buy TikTok?!
865.
by spullara 5y ago
For it to be available in EU at all I think they will have to list all of them anyway.
866.
by spullara 5y ago
Just make it a $1/min like it was in the old days and people will limit their communication to folks they don't see regularly for only important things.
867.
by spullara 5y ago
Why do these still exist? Is this really how academics find out about the paper and not just read the paper? I'm flabbergasted conferences are like middle school science fairs.
868.
French court revokes Starlink's internet license(space.com)
21 points by spullara 5y ago|10 comments
869.
by spullara 5y ago
Could be something that is just one once per process or the like. But yeah, if you are doing multiple inferences it would be silly to load the model each time.
870.
by spullara 5y ago
Agreed. The results I get for that query seem very relevant and created by people.

More ›